anystyle 1.0.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/HISTORY.md +78 -0
- data/LICENSE +27 -0
- data/README.md +103 -0
- data/lib/anystyle.rb +71 -0
- data/lib/anystyle/dictionary.rb +132 -0
- data/lib/anystyle/dictionary/gdbm.rb +52 -0
- data/lib/anystyle/dictionary/lmdb.rb +67 -0
- data/lib/anystyle/dictionary/marshal.rb +27 -0
- data/lib/anystyle/dictionary/redis.rb +55 -0
- data/lib/anystyle/document.rb +264 -0
- data/lib/anystyle/errors.rb +14 -0
- data/lib/anystyle/feature.rb +27 -0
- data/lib/anystyle/feature/affix.rb +43 -0
- data/lib/anystyle/feature/brackets.rb +32 -0
- data/lib/anystyle/feature/canonical.rb +13 -0
- data/lib/anystyle/feature/caps.rb +20 -0
- data/lib/anystyle/feature/category.rb +70 -0
- data/lib/anystyle/feature/dictionary.rb +16 -0
- data/lib/anystyle/feature/indent.rb +16 -0
- data/lib/anystyle/feature/keyword.rb +52 -0
- data/lib/anystyle/feature/line.rb +39 -0
- data/lib/anystyle/feature/locator.rb +18 -0
- data/lib/anystyle/feature/number.rb +39 -0
- data/lib/anystyle/feature/position.rb +28 -0
- data/lib/anystyle/feature/punctuation.rb +22 -0
- data/lib/anystyle/feature/quotes.rb +20 -0
- data/lib/anystyle/feature/ref.rb +21 -0
- data/lib/anystyle/feature/terminal.rb +19 -0
- data/lib/anystyle/feature/words.rb +74 -0
- data/lib/anystyle/finder.rb +94 -0
- data/lib/anystyle/format/bibtex.rb +63 -0
- data/lib/anystyle/format/csl.rb +28 -0
- data/lib/anystyle/normalizer.rb +65 -0
- data/lib/anystyle/normalizer/brackets.rb +13 -0
- data/lib/anystyle/normalizer/container.rb +13 -0
- data/lib/anystyle/normalizer/date.rb +109 -0
- data/lib/anystyle/normalizer/edition.rb +16 -0
- data/lib/anystyle/normalizer/journal.rb +14 -0
- data/lib/anystyle/normalizer/locale.rb +30 -0
- data/lib/anystyle/normalizer/location.rb +24 -0
- data/lib/anystyle/normalizer/locator.rb +22 -0
- data/lib/anystyle/normalizer/names.rb +88 -0
- data/lib/anystyle/normalizer/page.rb +29 -0
- data/lib/anystyle/normalizer/publisher.rb +18 -0
- data/lib/anystyle/normalizer/pubmed.rb +18 -0
- data/lib/anystyle/normalizer/punctuation.rb +23 -0
- data/lib/anystyle/normalizer/quotes.rb +14 -0
- data/lib/anystyle/normalizer/type.rb +54 -0
- data/lib/anystyle/normalizer/volume.rb +26 -0
- data/lib/anystyle/parser.rb +199 -0
- data/lib/anystyle/support.rb +4 -0
- data/lib/anystyle/support/finder.mod +3234 -0
- data/lib/anystyle/support/finder.txt +75 -0
- data/lib/anystyle/support/parser.mod +15025 -0
- data/lib/anystyle/support/parser.txt +75 -0
- data/lib/anystyle/utils.rb +70 -0
- data/lib/anystyle/version.rb +3 -0
- data/res/finder/bb132pr2055.ttx +6803 -0
- data/res/finder/bb550sh8053.ttx +18660 -0
- data/res/finder/bb599nz4341.ttx +2957 -0
- data/res/finder/bb725rt6501.ttx +15276 -0
- data/res/finder/bc605xz1554.ttx +18815 -0
- data/res/finder/bd040gx5718.ttx +4271 -0
- data/res/finder/bd413nt2715.ttx +4956 -0
- data/res/finder/bd466fq0394.ttx +6100 -0
- data/res/finder/bf668vw2021.ttx +3578 -0
- data/res/finder/bg495cx0468.ttx +7267 -0
- data/res/finder/bg599vt3743.ttx +6752 -0
- data/res/finder/bg608dx2253.ttx +4094 -0
- data/res/finder/bh410qk3771.ttx +8785 -0
- data/res/finder/bh989ww6442.ttx +17204 -0
- data/res/finder/bj581pc8202.ttx +2719 -0
- data/res/parser/bad.xml +5199 -0
- data/res/parser/core.xml +7924 -0
- data/res/parser/gold.xml +2707 -0
- data/res/parser/good.xml +34281 -0
- data/res/parser/stanford-books.xml +2280 -0
- data/res/parser/stanford-diss.xml +726 -0
- data/res/parser/stanford-theses.xml +4684 -0
- data/res/parser/ugly.xml +33246 -0
- metadata +195 -0
@@ -0,0 +1,4271 @@
|
|
1
|
+
title | CULTIVATING THE GROWTH OF COMPLEX ENGINEERED SYSTEMS
|
2
|
+
| USING
|
3
|
+
| EMERGENT BEHAVIOURS OF ENGINEERING PROCESSES
|
4
|
+
blank |
|
5
|
+
|
|
6
|
+
|
|
7
|
+
|
|
8
|
+
text | A DISSERTATION
|
9
|
+
| SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING
|
10
|
+
| AND THE COMMITTEE ON GRADUATE STUDIES
|
11
|
+
| OF STANFORD UNIVERSITY
|
12
|
+
| IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
|
13
|
+
| FOR THE DEGREE OF
|
14
|
+
| DOCTOR OF PHILOSOPHY
|
15
|
+
| IN
|
16
|
+
| MECHANICAL ENGINEERING
|
17
|
+
blank |
|
18
|
+
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
text | Eric Alan Byler
|
22
|
+
| March 2014
|
23
|
+
| © 2014 by Eric Alan Byler. All Rights Reserved.
|
24
|
+
| Re-distributed by Stanford University under license with the author.
|
25
|
+
blank |
|
26
|
+
|
|
27
|
+
|
|
28
|
+
text | This work is licensed under a Creative Commons Attribution-
|
29
|
+
| Noncommercial 3.0 United States License.
|
30
|
+
| http://creativecommons.org/licenses/by-nc/3.0/us/
|
31
|
+
blank |
|
32
|
+
|
|
33
|
+
|
|
34
|
+
|
|
35
|
+
text | This dissertation is online at: http://purl.stanford.edu/bd040gx5718
|
36
|
+
blank |
|
37
|
+
|
|
38
|
+
|
|
39
|
+
|
|
40
|
+
meta | ii
|
41
|
+
text | I certify that I have read this dissertation and that, in my opinion, it is fully adequate
|
42
|
+
| in scope and quality as a dissertation for the degree of Doctor of Philosophy.
|
43
|
+
blank |
|
44
|
+
text | Larry Leifer, Primary Adviser
|
45
|
+
blank |
|
46
|
+
|
|
47
|
+
|
|
48
|
+
text | I certify that I have read this dissertation and that, in my opinion, it is fully adequate
|
49
|
+
| in scope and quality as a dissertation for the degree of Doctor of Philosophy.
|
50
|
+
blank |
|
51
|
+
text | Mark Cutkosky
|
52
|
+
blank |
|
53
|
+
|
|
54
|
+
|
|
55
|
+
text | I certify that I have read this dissertation and that, in my opinion, it is fully adequate
|
56
|
+
| in scope and quality as a dissertation for the degree of Doctor of Philosophy.
|
57
|
+
blank |
|
58
|
+
text | Friedrich Prinz
|
59
|
+
blank |
|
60
|
+
|
|
61
|
+
|
|
62
|
+
|
|
63
|
+
text | Approved for the Stanford University Committee on Graduate Studies.
|
64
|
+
| Patricia J. Gumport, Vice Provost for Graduate Education
|
65
|
+
blank |
|
66
|
+
|
|
67
|
+
|
|
68
|
+
|
|
69
|
+
text | This signature page was generated electronically upon submission of this dissertation in
|
70
|
+
| electronic format. An original signed hard copy of the signature page is on file in
|
71
|
+
| University Archives.
|
72
|
+
blank |
|
73
|
+
|
|
74
|
+
|
|
75
|
+
|
|
76
|
+
meta | iii
|
77
|
+
title | ABSTRACT
|
78
|
+
blank |
|
79
|
+
|
|
80
|
+
text | Design, construction, and operation of very large, complex aerospace systems
|
81
|
+
| (mega-systems) is challenging, with potential cost and schedule over-runs and with
|
82
|
+
| awkward upgrade issues. These challenges are amplified when the engineering
|
83
|
+
| organization is producing multiple mega-systems in parallel. This dissertation applies
|
84
|
+
| information theory, genetic computing, and chaos theory to product development within a
|
85
|
+
| large distributed organization. The goal is design environments that efficiently produce
|
86
|
+
| mega-systems that are better able to evolve within the changing world context.
|
87
|
+
| Complex systems are composed of dynamically interacting pieces that exhibit
|
88
|
+
| emergent behaviours as a result of their growth and/or the passage of time. Modern large
|
89
|
+
| engineered systems (~10,000+ man-years of design time), such as the American Air
|
90
|
+
| Traffic Control System and International Space Station, exhibit complex behaviour over
|
91
|
+
| their life, including design, installation, operation, and growth. The desired capability of
|
92
|
+
| the systems is so complex that it is difficult and time consuming to consistently specify
|
93
|
+
| detail for what is desired. The installation and operation of the complex systems changes
|
94
|
+
| the environment in which they are installed, creating a continuously changing context for
|
95
|
+
| follow-on and adjunct system requirements. Finally, maintenance and growth
|
96
|
+
| characteristics for complex systems, including new adjunct systems, exhibit various
|
97
|
+
| levels of brittleness.
|
98
|
+
| This dissertation develops a new type of model to analyze the design and
|
99
|
+
| construction of mega-systems, based on combining an organization network-model with
|
100
|
+
| an information-transfer link-model. Analysis of previous efforts within the Design
|
101
|
+
| Research and the System Engineering communities shows three major shortcomings:
|
102
|
+
| they assume efficient network connectivity, they assume static networks, and they
|
103
|
+
| assume loss-less data transfer. A new model type is developed and exercised to examine
|
104
|
+
| methods and techniques to improve mega-system design and construction, and then
|
105
|
+
| validated in a three-year experiment implementing an enterprise product delivery process.
|
106
|
+
| Model development is based on data from US aerospace companies and programs
|
107
|
+
| spanning the last 30 years.
|
108
|
+
blank |
|
109
|
+
|
|
110
|
+
|
|
111
|
+
meta | iv
|
112
|
+
text | The design of complex systems is not a linear, mechanistic process, but rather a
|
113
|
+
| multi-dimensional, growth process that can be cultivated over and above the analytical
|
114
|
+
| disciplines of design and optimization. The duality of a complex system designing a
|
115
|
+
| complex system reduces determinism. This dissertation identifies parameters of
|
116
|
+
| information exchange activities that can be adjusted to produce an efficient, successful
|
117
|
+
| outcome, albeit a non-deterministic outcome. This adjustment or shaping is similar to
|
118
|
+
| guided genetic evolution of organizations that produce mega-systems as the product of
|
119
|
+
| their computation, but within a restricted context of specified, but evolving, goals and
|
120
|
+
| constraints.
|
121
|
+
| This new analytical framework shows how to match information flow
|
122
|
+
| characteristics to emergent engineering activities to improve the efficiency and quality of
|
123
|
+
| the design process, and to improve the efficiency and coupling of research and
|
124
|
+
| development activities within the larger organization. Information flow must not be
|
125
|
+
| confused with data transfer. Information flow parameters include information resolution,
|
126
|
+
| variability, time scale, compression, link bandwidth, connectivity maps, and more. For
|
127
|
+
| instance, simply avoiding information aliasing in chaotic channels by matching the flow
|
128
|
+
| to the information aging rate can improve design efficiency 80% and product delivery
|
129
|
+
| efficiency 50% in the overall steady state.
|
130
|
+
| The implication of these results is a radical improvement in procurement of large
|
131
|
+
| systems. However, inherent linkages to government organizations that typically fund
|
132
|
+
| these large scale projects require a dual adaptation approach similar to meta-norms in
|
133
|
+
| game theory. Otherwise, the emergent properties of efficient engineering management
|
134
|
+
| will cause a reversion to a lower level of capability per status quo. Also, in order to
|
135
|
+
| support the greater output rate, mega-system specifications will become somewhat non-
|
136
|
+
| deterministic and delivery will rely inherently on product engineering, requiring a change
|
137
|
+
| in perspective of program management. This is beneficial evolution of the organization
|
138
|
+
| to improve delivery, and the total value delivered to society will accrue at higher
|
139
|
+
| compound growth rates.
|
140
|
+
blank |
|
141
|
+
|
|
142
|
+
|
|
143
|
+
|
|
144
|
+
meta | v
|
145
|
+
title | ACKNOWLEDGEMENTS
|
146
|
+
blank |
|
147
|
+
|
|
148
|
+
|
|
149
|
+
text | I would like to thank Professor Larry Leifer for his support of this effort as
|
150
|
+
| it evolved over time, and for his enthusiasm and generosity sharing the breadth of
|
151
|
+
| the design experience.
|
152
|
+
blank |
|
153
|
+
|
|
154
|
+
text | I would like to thank Tess for her commitment, support, and enjoyment of
|
155
|
+
| the roving, rambling path to get here and beyond.
|
156
|
+
blank |
|
157
|
+
|
|
158
|
+
|
|
159
|
+
|
|
160
|
+
meta | vi
|
161
|
+
title | TABLE OF CONTENTS
|
162
|
+
text | 1 Introduction ................................................................................................................. 1
|
163
|
+
| 1.1 Problem ................................................................................................................ 1
|
164
|
+
| 1.2 Mega Design Communities and Design Organization Model ............................. 3
|
165
|
+
| 1.3 Information Transfer and Link Model ................................................................. 4
|
166
|
+
| 1.4 Results .................................................................................................................. 9
|
167
|
+
| 1.5 Conclusions ........................................................................................................ 13
|
168
|
+
| 1.5.1 Design of Individual Mega-Systems ........................................................... 13
|
169
|
+
| 1.5.2 Company and Design-Organization Implications ....................................... 14
|
170
|
+
| 1.5.3 Government and Societal Implications ....................................................... 15
|
171
|
+
| 2 Network Model Development and Mega-Organizations .......................................... 17
|
172
|
+
| 2.1 Structure of Very Large Aerospace Design Communities ................................. 17
|
173
|
+
| 2.2 Network Characteristics ..................................................................................... 21
|
174
|
+
| 2.3 Abstraction to Meta-Network Model ................................................................. 27
|
175
|
+
| 3 Link Model Development ......................................................................................... 28
|
176
|
+
| 3.1 Model Overview ................................................................................................. 28
|
177
|
+
| 3.2 Design Information ............................................................................................ 31
|
178
|
+
| 3.2.1 Information in Design Communities .......................................................... 31
|
179
|
+
| 3.2.2 Information Modeling ................................................................................. 34
|
180
|
+
| 3.2.3 Product Entropy and Information Packages ............................................... 36
|
181
|
+
| 3.2.4 Information Growth, and Product Evolution .............................................. 39
|
182
|
+
| 3.3 Task Variety ....................................................................................................... 42
|
183
|
+
| 3.3.1 Variety and Information Processing ........................................................... 42
|
184
|
+
| 3.3.2 Variety in Distributed Product Development ............................................. 44
|
185
|
+
| 3.4 Information Transfer in Chaotic Channels ......................................................... 46
|
186
|
+
| 3.4.1 Dynamical Systems and the Logistics Map ................................................ 46
|
187
|
+
| 3.4.2 Information Transfer in Chaotic Channels ................................................. 48
|
188
|
+
| 3.4.3 Multi-scale Information Transfer ................................................................ 51
|
189
|
+
| 3.4.4 Combined Varietal, Organizational, and Temporal Aliasing ..................... 57
|
190
|
+
| 3.4.5 Noise in Chaotic Channels .......................................................................... 59
|
191
|
+
| 3.4.6 Synchronization .......................................................................................... 59
|
192
|
+
blank |
|
193
|
+
meta | vii
|
194
|
+
| 4 Application of Integrated Link/Network Model ....................................................... 60
|
195
|
+
text | 4.1 Matching Scale Dispersion to Time Regimes .................................................... 60
|
196
|
+
| 4.2 Identification of Candidate Links, Nodes, and Chains....................................... 63
|
197
|
+
| 4.3 Network Improvements ...................................................................................... 66
|
198
|
+
| 4.3.1 Link Discontinuity Adjustment................................................................... 67
|
199
|
+
| 4.3.2 Link-Chain Improvement............................................................................ 69
|
200
|
+
| 4.4 Cultivated Changes and Observed Emergent Side-Effects ................................ 72
|
201
|
+
| 5 Performance Observations and Implications ............................................................ 75
|
202
|
+
| 5.1 R&D Applicability Performance ........................................................................ 75
|
203
|
+
| 5.2 Estimating the Amount of Communication you Should Afford ........................ 76
|
204
|
+
| 5.3 Expected Product Changes ................................................................................. 77
|
205
|
+
| 5.3.1 Exporting Products at Delivery Dates......................................................... 78
|
206
|
+
| 5.3.2 Incremental Technology Incorporation....................................................... 79
|
207
|
+
| 5.3.3 Coupled Efficiency ..................................................................................... 80
|
208
|
+
| 5.3.4 Company and Product Goals ...................................................................... 80
|
209
|
+
| 5.4 Limits to Change ................................................................................................ 81
|
210
|
+
| 5.5 Applicability ....................................................................................................... 82
|
211
|
+
| 6 Conclusions ............................................................................................................... 85
|
212
|
+
| 6.1 More Efficient Design of Mega-Systems ........................................................... 85
|
213
|
+
| 6.2 Relying on Emergence and Organizational Computation .................................. 86
|
214
|
+
| 6.3 Government Aerospace, Low-Price Procurement, and Overhead Rates ........... 87
|
215
|
+
| 6.4 Additional Research Vectors .............................................................................. 89
|
216
|
+
| 6.4.1 Organizational Memory .............................................................................. 89
|
217
|
+
| 6.4.2 Engineering Tools ....................................................................................... 89
|
218
|
+
| 6.4.3 Long-Scale Neural Training of Sub-Networks ........................................... 89
|
219
|
+
| 7 Appendix A – Research Background........................................................................ 90
|
220
|
+
| 7.1 Evolutionary Design ........................................................................................... 90
|
221
|
+
| 7.2 Engineering Systems .......................................................................................... 91
|
222
|
+
| 7.3 Flow-Service-Quality Engineering .................................................................... 92
|
223
|
+
| 7.4 Engineering Tools .............................................................................................. 93
|
224
|
+
| 8 References ................................................................................................................. 95
|
225
|
+
blank |
|
226
|
+
meta | viii
|
227
|
+
title | LIST OF FIGURES
|
228
|
+
blank |
|
229
|
+
|
|
230
|
+
text | Figure 1-1 Examples of Mega-Systems ............................................................................. 2
|
231
|
+
| Figure 1-2 Development of Abstracted Design Organization Model ................................ 3
|
232
|
+
| Figure 1-3 Link Information Transfer Model underpins all node connections ................. 5
|
233
|
+
| Figure 1-4 Link Information Transfer Model .................................................................... 6
|
234
|
+
| Figure 1-5 Information Exchange Resonances .................................................................. 8
|
235
|
+
| Figure 1-6 Information Transfer Regimes and Amplification Factors .............................. 9
|
236
|
+
| Figure 1-7 Set of Consistent Guides to Repair Discontinuity ......................................... 10
|
237
|
+
| Figure 1-8 Link-chain Paths of Interest ........................................................................... 12
|
238
|
+
blank |
|
239
|
+
|
|
240
|
+
text | Figure 2-1 Geographic Distribution ................................................................................. 18
|
241
|
+
| Figure 2-2 Conceptual Organizational Structure, Staffing Levels, and Characteristics .. 19
|
242
|
+
| Figure 2-3 Organizational Size Characteristics ............................................................... 20
|
243
|
+
| Figure 2-4 Overlapping Hierarchies ................................................................................ 21
|
244
|
+
| Figure 2-5 Typical Hierarchical Organization ................................................................. 22
|
245
|
+
| Figure 2-6 Control Stability for Various Organizations .................................................. 23
|
246
|
+
| Figure 2-7 Organizational Performance........................................................................... 23
|
247
|
+
| Figure 2-8 Typical and Combined Network Characteristics ........................................... 24
|
248
|
+
| Figure 2-9 Abstracted Organizational Model .................................................................. 27
|
249
|
+
blank |
|
250
|
+
|
|
251
|
+
text | Figure 3-1 Parameters of Link Transfer Model ............................................................... 29
|
252
|
+
| Figure 3-2 Organizations, Information Products, and Time Epochs ............................... 32
|
253
|
+
| Figure 3-3 Product Entropy Flexibility for Related Designs ........................................... 36
|
254
|
+
| Figure 3-4 Product Entropy Levels for different rates of Growth ................................... 38
|
255
|
+
| Figure 3-5 Mutual and Conditional Information ............................................................. 39
|
256
|
+
| Figure 3-6 Design Process Models and Networks ........................................................... 44
|
257
|
+
| Figure 3-7 Meta-Performance Measures of Distributed Product Development .............. 45
|
258
|
+
| Figure 3-8 Bifurcation Diagram....................................................................................... 47
|
259
|
+
| Figure 3-9 Conditional, mutual, and total information for two and three time steps ...... 49
|
260
|
+
| Figure 3-10 Mutual information for larger iterations of time .......................................... 50
|
261
|
+
blank |
|
262
|
+
meta | ix
|
263
|
+
text | Figure 3-11 Values for Combined Organization (OAF) and Variety (VAF) Aliasing .... 55
|
264
|
+
| Figure 3-12 Values for Time Aliasing Factor (TAF)....................................................... 56
|
265
|
+
| Figure 3-13 Network with total Product Information Aliasing Factors (p) ..................... 58
|
266
|
+
blank |
|
267
|
+
|
|
268
|
+
text | Figure 4-1 Mutual, Conditional, and Postdicted (I(y)) Information ................................ 61
|
269
|
+
| Figure 4-2 Mutual information for larger numbers of iterations ..................................... 62
|
270
|
+
| Figure 4-3 Information Transfer Regimes and Amplification Factors ............................ 64
|
271
|
+
| Figure 4-4 Set of Consistent Guides to Repair Discontinuity ......................................... 67
|
272
|
+
| Figure 4-5 Link-chain Paths of Interest ........................................................................... 70
|
273
|
+
| Figure 4-6 Cultivated Organizational Change ................................................................. 74
|
274
|
+
blank |
|
275
|
+
|
|
276
|
+
text | Figure 5-1 Improved Value Over Time ........................................................................... 78
|
277
|
+
blank |
|
278
|
+
|
|
279
|
+
|
|
280
|
+
|
|
281
|
+
meta | x
|
282
|
+
title | GLOSSARY
|
283
|
+
blank |
|
284
|
+
|
|
285
|
+
|
|
286
|
+
text | Agent – an actor that undertakes an action in a design chain, here including individuals,
|
287
|
+
| design groups, project teams, discipline departments, support functions, small
|
288
|
+
| businesses, etc. A node with variety.
|
289
|
+
| Aliasing – loss of understanding of some detail within a set of information. Similar to
|
290
|
+
| straight-line aliasing in graphics or image analysis.
|
291
|
+
| Chaotic channel – an information exchange channel or link, that is susceptible to
|
292
|
+
| dynamic evolution with the ensuing noise injection and dissipation.
|
293
|
+
| Communicativity - is related to the number, speed, and bandwidth of connections between
|
294
|
+
| agents, spanning multiple scales (individual, group, organization) with different
|
295
|
+
| parameters at each scale.
|
296
|
+
| Context Variety– variety needed to acquire contextual information from an external
|
297
|
+
| source to connect the design to external detail.
|
298
|
+
| Directed links – a directed link in graph theory, or here a link through which information
|
299
|
+
| flows only one way.
|
300
|
+
| Encoded – packaged information such that it can be understood by an undefined recipient
|
301
|
+
| at an undefined time in the future.
|
302
|
+
| Evolutionary computing organisms – complex systems that evolve overtime including
|
303
|
+
| genetic populations.
|
304
|
+
| Exchange resonances – states where information package format and timing is consistent
|
305
|
+
| with environmental transmission characteristics.
|
306
|
+
| External norms – in game theory, externally applied forces that affect the pareto solution.
|
307
|
+
| Information content – the amount of information being exchanged.
|
308
|
+
| Information resolution – the resolution or scale of the information similar to image
|
309
|
+
| resolution and engineering tolerances.
|
310
|
+
| Information theoretic – Information modeled in a manner similar to entropy, statistical
|
311
|
+
| mechanics, and/or quantum physics.
|
312
|
+
blank |
|
313
|
+
|
|
314
|
+
meta | xi
|
315
|
+
text | Link – a link on a graph or network connecting two nodes, here especially connections
|
316
|
+
| between groups to exchange information.
|
317
|
+
| Link chains - sets of directed links corresponding to control paths required to focus the
|
318
|
+
| organization on particular future goal states.
|
319
|
+
| Mega-systems - systems whose design and production efforts span on the order of 10
|
320
|
+
| years of development time and require on the order of 10,000 man-years of effort.
|
321
|
+
| Mega-organizations – organizations that have tens of thousands of engineers spread
|
322
|
+
| widely across a large geographic area. Not necessarily with singular control.
|
323
|
+
| Niche construction – in biology or genetics, an internal adjustment to better fit the local
|
324
|
+
| environment for improved performance.
|
325
|
+
| Node – a node on a graph or network, here including individuals, design groups, project
|
326
|
+
| teams, discipline departments, support functions, small businesses, etc.
|
327
|
+
| Non-directed graph – in graph theory having multi-directional links, here especially bi-
|
328
|
+
| directional links connecting two separate nodes.
|
329
|
+
| Post diction - the forward use of information to increase the knowledge of another actor
|
330
|
+
| in a strictly subsequent time state. The inverse of prediction in an information
|
331
|
+
| theory sense.
|
332
|
+
| Resolution Variety – variety needed to add more detailed design information to increase
|
333
|
+
| the fidelity or resolution of the design.
|
334
|
+
| Scale-free network – in network theory, a random network with a degree distribution that
|
335
|
+
| follows a power law, at least asymptotically.
|
336
|
+
| Translation Variety – variety needed to translate or extract information from the current
|
337
|
+
| design to further the design in an additional dimension.
|
338
|
+
| Transmitted – delivering information such that it reaches additional recipients.
|
339
|
+
| Task-bits – quantum information of a task, similar to data bits or letters in a document.
|
340
|
+
| Variety – the diverse set of capabilities needed by an actor or group to execute an
|
341
|
+
| activity. The variety represents the set of capabilities and functionality.
|
342
|
+
blank |
|
343
|
+
|
|
344
|
+
|
|
345
|
+
|
|
346
|
+
text | xii
|
347
|
+
| …complex products
|
348
|
+
| produced by complex organizations
|
349
|
+
| become fragments of complex systems
|
350
|
+
blank |
|
351
|
+
|
|
352
|
+
|
|
353
|
+
|
|
354
|
+
meta | xiii
|
355
|
+
title | 1 Introduction
|
356
|
+
text | Engineers propose to understand canonical mechanisms and prototypical behaviours to
|
357
|
+
| better control the systems development effort. But you cannot control a complex system
|
358
|
+
| in which you are immersed, and within an environmental context that changes before
|
359
|
+
| your controls can respond. You must cultivate and guide these systems at various and
|
360
|
+
| appropriate time scales.
|
361
|
+
blank |
|
362
|
+
|
|
363
|
+
text | This dissertation develops a new type of model to analyze the design and construction of
|
364
|
+
| mega-systems, based on combining an organization network-model with an information-
|
365
|
+
| transfer link-model. This combined model is exercised to examine methods and
|
366
|
+
| techniques to improve mega-system design and construction, and then validated in a
|
367
|
+
| three-year experiment implementing an enterprise product delivery process. Model
|
368
|
+
| development is based on data from US aerospace companies and programs spanning the
|
369
|
+
| last 30 years.
|
370
|
+
blank |
|
371
|
+
|
|
372
|
+
text | Previous efforts within the Design Research and the System Engineering communities to
|
373
|
+
| analyze and improve Engineering Processes that are used to design mega-systems are
|
374
|
+
| summarized in Appendix A. Analysis of these recent efforts shows three major
|
375
|
+
| shortcomings: they assume efficient network connectivity, they assume static networks,
|
376
|
+
| and they assume loss-less data transfer. This dissertation is a first step toward correcting
|
377
|
+
| these deficiencies, and accounting for these very real constraints.
|
378
|
+
blank |
|
379
|
+
title | 1.1 Problem
|
380
|
+
text | Large Complex Mega-Systems are defined here as efforts that span 10 years of
|
381
|
+
| development time (design and production) and require 10,000 man-years of effort.
|
382
|
+
| Typically engineers are drawn from a wide supply chain spanning a large geographic
|
383
|
+
| area. These systems include the NASA Space Shuttle, the International Space Station
|
384
|
+
| (ISS), and planned Mars and asteroid missions (Figure 1-1). Terrestrial examples include
|
385
|
+
| the Air Traffic Control Systems (ATCS), missile defense systems, nuclear power plants,
|
386
|
+
| and others.
|
387
|
+
blank |
|
388
|
+
|
|
389
|
+
|
|
390
|
+
meta | 1
|
391
|
+
text | Figure 1-1 Examples of Mega-Systems
|
392
|
+
| (A) Space shuttles Atlantis and Endeavor at Kennedy Space Center launch complexes 39A and 39B.
|
393
|
+
| (B) Artist’s concept of exploration of Mars’s moon Phobos.
|
394
|
+
blank |
|
395
|
+
|
|
396
|
+
text | Problems arise when developing these systems, likely caused by their immense scale.
|
397
|
+
| Some systems are never built, or take a much longer time than planned before the design
|
398
|
+
| is finalized for production: this is sometimes caused by requirements churn such as for
|
399
|
+
| the ISS and its precursors. Other systems exhibit upgrade brittleness such as the ATCS:
|
400
|
+
| here, because of either the safety requirements, distributed nature, or necessary
|
401
|
+
| incremental approach to upgrades, the follow-on replacement design was unable to close
|
402
|
+
| and had to be restarted several times at great expense. Some products collide with
|
403
|
+
| obsolescence issues since their design process takes so long that some of the originally
|
404
|
+
| included components are no longer available.
|
405
|
+
blank |
|
406
|
+
|
|
407
|
+
text | “About one in three major U.S. Defense Department programs since 1997 have had cost
|
408
|
+
| overruns of as much as 50 percent over their original projections. The overruns, found in
|
409
|
+
| 47 of 134 programs included in a study by the U.S. Government Accountability Office,
|
410
|
+
| were enough to trigger a law [Nunn-McCurdy Act] that requires congressional
|
411
|
+
| notifications and potential termination” [Capaccio 2011].
|
412
|
+
blank |
|
413
|
+
|
|
414
|
+
|
|
415
|
+
|
|
416
|
+
meta | 2
|
417
|
+
text | 1.2 Mega Design Communities and Design Organization Model
|
418
|
+
| Mega-Systems are designed by Mega-Organizations that have tens of thousands of
|
419
|
+
| engineers spread widely across a large geographic area. Organizations are treated here as
|
420
|
+
| evolutionary computing organisms that evolve as they compute. An abstract
|
421
|
+
| organizational network model is developed in Section 2.0 as summarized in Figure 1-2.
|
422
|
+
| Subsequent discussion in Sections 3.0 and 4.0 will add measures of emergent parameters.
|
423
|
+
blank |
|
424
|
+
text | Louisville, CO
|
425
|
+
| Bangor, WA Denver, CO Newtown, PA
|
426
|
+
| • Coherent
|
427
|
+
| • FBM IA&T
|
428
|
+
| Technologies • Navigation Systems Sector
|
429
|
+
| Coord
|
430
|
+
| Sector
|
431
|
+
| Coord
|
432
|
+
| Major
|
433
|
+
| Sector
|
434
|
+
| Coord
|
435
|
+
blank |
|
436
|
+
text | Palo Alto, CA • Commercial Space Teammates
|
437
|
+
blank |
|
438
|
+
|
|
439
|
+
text | • ATC BD PROG
|
440
|
+
| PROG
|
441
|
+
| R&D
|
442
|
+
| PROG
|
443
|
+
| Sector
|
444
|
+
| PROG
|
445
|
+
| PROG
|
446
|
+
| Coord
|
447
|
+
| Exec Exec Exec
|
448
|
+
| Sunnyvale, CA
|
449
|
+
| Valley Forge, PA
|
450
|
+
| Santa Cruz, CA • Reentry Systems Adjacent
|
451
|
+
| Adjacent
|
452
|
+
| BD Executive Senior Adjacent
|
453
|
+
| • Test Support Markets
|
454
|
+
| Adjacent
|
455
|
+
| Markets
|
456
|
+
| Adjacent
|
457
|
+
| Markets
|
458
|
+
| Markets
|
459
|
+
| Director Staff Support Technologies
|
460
|
+
blank |
|
461
|
+
|
|
462
|
+
text | Washington, DC
|
463
|
+
| • FBM Liaison BD Advanced Senior
|
464
|
+
blank |
|
465
|
+
text | Business • Washington Ops Program Research
|
466
|
+
| Staff
|
467
|
+
| Program
|
468
|
+
| Program
|
469
|
+
| Program Develop
|
470
|
+
| Senior
|
471
|
+
| Senior
|
472
|
+
| Manager
|
473
|
+
| Senior
|
474
|
+
| ManagerSenior
|
475
|
+
| Manager
|
476
|
+
| Senior
|
477
|
+
| ManagerSenior
|
478
|
+
| Manager
|
479
|
+
| Manager
|
480
|
+
blank |
|
481
|
+
text | Development Management
|
482
|
+
| Manager
|
483
|
+
| Arlington, VA Senior Senior
|
484
|
+
| Edwards AFB, CA BD
|
485
|
+
| • Targets Prime PROG R&D
|
486
|
+
| Engineer
|
487
|
+
| Subsystem
|
488
|
+
| Subsystem
|
489
|
+
| Subsystem
|
490
|
+
| Subsystem
|
491
|
+
| Subsystem
|
492
|
+
| CRAD
|
493
|
+
| CRAD
|
494
|
+
| CRAD
|
495
|
+
| CRAD
|
496
|
+
| CRAD
|
497
|
+
| Scientist
|
498
|
+
| Subsystem CRAD
|
499
|
+
| • ABL
|
500
|
+
| Subsystem
|
501
|
+
| Huntsville, AL
|
502
|
+
| • Strategic & Missile Defense Tools Engineer Engineers Engineer
|
503
|
+
| Staff
|
504
|
+
| El Segundo, CA • Ares Support
|
505
|
+
| Scientist
|
506
|
+
blank |
|
507
|
+
text | • MILSATCOM Exec Exec Exec
|
508
|
+
| Albuquerque, NM Kings Bay, GA Discipline
|
509
|
+
| • ATC • FBM IA&T Discipline
|
510
|
+
| Director
|
511
|
+
| Discipline
|
512
|
+
| Director
|
513
|
+
| Discipline
|
514
|
+
| Suppliers Small Universities
|
515
|
+
| Director
|
516
|
+
| Kauai, HI White Sands, NM • ATK
|
517
|
+
| Discipline
|
518
|
+
| Director
|
519
|
+
| Discipline
|
520
|
+
| Director
|
521
|
+
| Business
|
522
|
+
blank |
|
523
|
+
|
|
524
|
+
text | • PMRF Courtland, AL Cape Canaveral, FL Discipline
|
525
|
+
| Director
|
526
|
+
| • Missile Range • Targets & Adjacent
|
527
|
+
| • Targets &
|
528
|
+
| BD
|
529
|
+
| Eastern Range
|
530
|
+
| Executive Senior
|
531
|
+
| Sector
|
532
|
+
| Discipline
|
533
|
+
| Adjacent
|
534
|
+
| Director
|
535
|
+
| Director
|
536
|
+
| Stennis, MS Markets Director Staff Coord
|
537
|
+
| Support
|
538
|
+
| ENG
|
539
|
+
| Technologies
|
540
|
+
| Countermeasures Countermeasures • Launch Operations
|
541
|
+
| • Propulsion
|
542
|
+
| Houston, TX Center
|
543
|
+
| ~19,500 employees • Orion
|
544
|
+
| New Orleans, LA BD Advanced Senior
|
545
|
+
| • ~ 8,700 in Silicon Valley
|
546
|
+
| • External Tank Staff
|
547
|
+
| Program
|
548
|
+
| Develop Manager
|
549
|
+
| • ~ 4,700 in Denver area
|
550
|
+
| • ~ 2,000 in New Orleans • Orion
|
551
|
+
| • ~ 1,500 in Cape and V
|
552
|
+
| • ~ 1,100 in Delaware Valley
|
553
|
+
| Senior
|
554
|
+
| Engineer
|
555
|
+
| Subsystem CRAD
|
556
|
+
| Senior
|
557
|
+
| Scientist
|
558
|
+
| 009_03.18.08
|
559
|
+
| • ~ 500 in Alabama
|
560
|
+
blank |
|
561
|
+
text | Staff
|
562
|
+
| Tools Engineer Engineer Engineer
|
563
|
+
| Scientist
|
564
|
+
blank |
|
565
|
+
|
|
566
|
+
|
|
567
|
+
text | Discipline
|
568
|
+
| Director
|
569
|
+
| Operational Prototype
|
570
|
+
| Systems Systems
|
571
|
+
| ENG
|
572
|
+
| Engineering
|
573
|
+
blank |
|
574
|
+
text | Figure 1-2 Development of Abstracted Design Organization Model
|
575
|
+
| (A) A large mega-organization (design community) of 100,000 people is spread across the wide range
|
576
|
+
| of North America in different cities. (B) Organizational groups (nodes) and major relations between
|
577
|
+
| them (links) are identified. (C) The nodes and links are then simplified into a two-dimensional
|
578
|
+
| structure to highlight parameters associated with activity duration, organization hierarchies, and
|
579
|
+
| different engineering capabilities and products.
|
580
|
+
blank |
|
581
|
+
|
|
582
|
+
text | The design-organization network model represents a distributed mega-design-community
|
583
|
+
| of 100,000 people producing a set of mega-systems over time. These people are
|
584
|
+
| assembled into various organizations including design groups, project teams, discipline
|
585
|
+
| departments, support functions, small businesses, etc. Each organizational group, or
|
586
|
+
| node, is composed of many people, and there may be multiple instances of any one
|
587
|
+
| group. The variety of people within each node provides a diverse set of capabilities to
|
588
|
+
| execute the group’s higher level function (see Section 3.3).
|
589
|
+
blank |
|
590
|
+
|
|
591
|
+
meta | 3
|
592
|
+
text | Multiple nodes (groups) may connect to a single node performing a coordinative
|
593
|
+
| function, but each link conveys separate information in different contexts (sections 1.3
|
594
|
+
| and 3.0). Groups may be linked to multiple sets of groups in different fashions for
|
595
|
+
| different purposes such as product design, tool development, financial control, etc. These
|
596
|
+
| overlapping sets represent various sub-groups or sub-functions within the design
|
597
|
+
| community and within the abstract organizational network model.
|
598
|
+
blank |
|
599
|
+
|
|
600
|
+
text | The resulting organizational network model is neither fully-connected nor hierarchically
|
601
|
+
| connected nor even efficiently connected, but rather has scale-free connectivity such that
|
602
|
+
| nodes can have several possible connection paths, even for the same information transfer
|
603
|
+
| activity, but with different efficiencies and restrictions (see Section 2.2).
|
604
|
+
blank |
|
605
|
+
title | 1.3 Information Transfer and Link Model
|
606
|
+
text | Earlier efforts in modeling multi-scale organizations with emergent properties [Byler
|
607
|
+
| 2000] showed the critical nature of multi-scale communications. Following an Adaptive
|
608
|
+
| Non-linear Testing technique [Miller 2007], the modeling validation and verification
|
609
|
+
| indicated massive instability around the parameters defining “communicativity”. This
|
610
|
+
| defined a critical area of investigation explored further in this dissertation.
|
611
|
+
blank |
|
612
|
+
|
|
613
|
+
text | “Communicativity” is related to the number, speed, and bandwidth of connections
|
614
|
+
| between agents. For individual design agents this is related, for example, to the number
|
615
|
+
| of emails sent per day, the number of addressees in an agents personal address book,
|
616
|
+
| number of off-site web-page accesses, typical phone usage, etc. For design groups, this is
|
617
|
+
| related to frequency and size of design reviews, availability and type of online design
|
618
|
+
| storage, geographic separation (bull-pen vs. cubicles), etc. For project swarms (groups,
|
619
|
+
| departments, small businesses), “communicativity” is rate of engineer reassignment, size
|
620
|
+
| and quality of information system network, frequency of training, and whether the project
|
621
|
+
| is owned by the company or someone else (which implies internalization of requirements
|
622
|
+
| and operations).
|
623
|
+
blank |
|
624
|
+
|
|
625
|
+
|
|
626
|
+
|
|
627
|
+
meta | 4
|
628
|
+
text | To address the importance of communicativity, an underpinning information transfer
|
629
|
+
| model for the network links was developed to better include factors addressing
|
630
|
+
| communications parameters as well as organization parameters, functional activities, and
|
631
|
+
| environmental constraints (Figure 1-3).
|
632
|
+
blank |
|
633
|
+
|
|
634
|
+
|
|
635
|
+
text | Agent X Agent Y
|
636
|
+
blank |
|
637
|
+
|
|
638
|
+
text | Information Transfer
|
639
|
+
blank |
|
640
|
+
text | Business Program Research
|
641
|
+
| Development BD PROG
|
642
|
+
| Management R&D
|
643
|
+
blank |
|
644
|
+
text | Chaotic Channel (non-linear)
|
645
|
+
| Exec Exec Exec
|
646
|
+
blank |
|
647
|
+
|
|
648
|
+
|
|
649
|
+
|
|
650
|
+
text | Adjacent BD Executive Senior Adjacent
|
651
|
+
| Markets Director Staff Support Technologies
|
652
|
+
blank |
|
653
|
+
|
|
654
|
+
|
|
655
|
+
text | BD Advanced Senior
|
656
|
+
| Program
|
657
|
+
| Staff Develop Manager
|
658
|
+
blank |
|
659
|
+
|
|
660
|
+
|
|
661
|
+
text | Senior Senior
|
662
|
+
| Subsystem CRAD
|
663
|
+
| Engineer Scientist
|
664
|
+
blank |
|
665
|
+
|
|
666
|
+
|
|
667
|
+
text | Staff
|
668
|
+
| Tools Engineer Engineer Engineer
|
669
|
+
| Scientist
|
670
|
+
blank |
|
671
|
+
|
|
672
|
+
|
|
673
|
+
text | Discipline
|
674
|
+
| Director
|
675
|
+
| Operational Prototype
|
676
|
+
| Systems Systems
|
677
|
+
| ENG
|
678
|
+
| Engineering
|
679
|
+
blank |
|
680
|
+
|
|
681
|
+
text | Figure 1-3 Link Information Transfer Model underpins all node connections
|
682
|
+
| Each link in the network model incorporates a detailed model of a chaotic communications channel.
|
683
|
+
blank |
|
684
|
+
|
|
685
|
+
|
|
686
|
+
text | This link information transfer model applies to all connections in the network model, and
|
687
|
+
| is shown in additional detail in Figure 1-4.
|
688
|
+
blank |
|
689
|
+
|
|
690
|
+
|
|
691
|
+
|
|
692
|
+
meta | 5
|
693
|
+
text | Link Information Transfer Model
|
694
|
+
| “Node” parameters
|
695
|
+
| defined by information encoding and transmitting activities
|
696
|
+
| and by design information content, resolution, & processing
|
697
|
+
| “Link” parameters
|
698
|
+
| defined by relative design role
|
699
|
+
| Agent X and by organizational activities Agent Y
|
700
|
+
| w/ Variety 7 w/ Variety 9
|
701
|
+
| Encoder V6
|
702
|
+
| Transmitter Vn
|
703
|
+
| Transmitter V7
|
704
|
+
blank |
|
705
|
+
text | environment, tools, time
|
706
|
+
blank |
|
707
|
+
|
|
708
|
+
text | Information Transfer
|
709
|
+
blank |
|
710
|
+
|
|
711
|
+
|
|
712
|
+
text | Chaotic Channel (non-linear)
|
713
|
+
| Machining V1 Receiver V1
|
714
|
+
| Cost Analysis V2 Decoder V2
|
715
|
+
| Programmer V3
|
716
|
+
blank |
|
717
|
+
|
|
718
|
+
text | Variety = a set of capabilities when
|
719
|
+
| and functionality Information Parameters (node + info package)
|
720
|
+
| are incompatible with
|
721
|
+
| Agents = can be individuals, Transmission Line Parameters (link + environment)
|
722
|
+
| groups, organizations, “aliasing” effects occur
|
723
|
+
| or other complices
|
724
|
+
blank |
|
725
|
+
text | Figure 1-4 Link Information Transfer Model
|
726
|
+
blank |
|
727
|
+
|
|
728
|
+
text | For the organization model developed here (Section 1.2 and 2.0), each node is an agent
|
729
|
+
| composed of many people (design groups, project swarms, etc.) in the sense defined by
|
730
|
+
| [Petrie 1996]. Agents can be individuals, groups, organizations, or other complices.
|
731
|
+
blank |
|
732
|
+
|
|
733
|
+
text | Each agent performs its function using a variety of people or skills. Producing a
|
734
|
+
| controller card, for instance, may require a machinist, a programmer, and a financial
|
735
|
+
| analyst. Then the product must be packaged and delivered. If one considers the
|
736
|
+
| controller card as inherently containing the design and product information, then the
|
737
|
+
| output could be termed to be encoded and transmitted, similar to packaged and delivered.
|
738
|
+
blank |
|
739
|
+
|
|
740
|
+
text | Depending on the complexity of the product or output of the Agent, it may require more
|
741
|
+
| or less variety of sub-tasks. The complexity of the product and/or of the agent is
|
742
|
+
blank |
|
743
|
+
|
|
744
|
+
meta | 6
|
745
|
+
text | represented by the Variety (Vi) of internal skills. The variety represents the set of
|
746
|
+
| capabilities and functionality.
|
747
|
+
blank |
|
748
|
+
|
|
749
|
+
text | The incremental task executed (item produced) by an agent can be a physical or
|
750
|
+
| ephemeral addition. In any event, information was imposed on the inputs (machining,
|
751
|
+
| assembly sequence, tuning) to increase its usefulness and reduce the entropy of the parts.
|
752
|
+
| The type of information imposed on the inputs, and representing the final product (piece-
|
753
|
+
| part), is modeled by information content, resolution, and processing.
|
754
|
+
blank |
|
755
|
+
|
|
756
|
+
text | The node parameters include the information content and resolution, as well as the
|
757
|
+
| information encoding and transmission. Information is modeled in an information
|
758
|
+
| theoretic sense after [Shannon 1948] and [Bais 2010]
|
759
|
+
blank |
|
760
|
+
|
|
761
|
+
text | I() = -ipilogpi
|
762
|
+
blank |
|
763
|
+
|
|
764
|
+
text | Any new (conditional) information received by agent y from the information sent by
|
765
|
+
| agent x will be
|
766
|
+
blank |
|
767
|
+
|
|
768
|
+
text | I(y|x) = -iPx(xi)jPy|x(yj|xi)log2Py|x(yj|xj)
|
769
|
+
blank |
|
770
|
+
|
|
771
|
+
text | which expresses the fact that the receiver must interpret (decode) the information he
|
772
|
+
| receives based on their previous mutual understanding. Note that information data-bits
|
773
|
+
| can be interpreted similarly with design “task-bits”.
|
774
|
+
blank |
|
775
|
+
|
|
776
|
+
text | The Link parameters are defined both by the relative design roles of the attached nodes
|
777
|
+
| within the organizational activities, and by the environment within which the products
|
778
|
+
| pass and exist. The effect of relative design roles is driven not so much by the different
|
779
|
+
| activities, but rather by how much work must be done to understand or incorporate
|
780
|
+
| (receive and decode) the material being transmitted by the input agent. The effect of the
|
781
|
+
| design environment includes how the information is stored (transmission is not
|
782
|
+
| instantaneous), the amount of time delay between production and use, the types of tools
|
783
|
+
blank |
|
784
|
+
|
|
785
|
+
meta | 7
|
786
|
+
text | used to decode or process the intermediate product, etc. Because of the dynamic nature
|
787
|
+
| of the information and the link, this model is a chaotic channel.
|
788
|
+
blank |
|
789
|
+
|
|
790
|
+
text | Information production, coding, aging, and obsolescence is modeled with the logistic
|
791
|
+
| growth curve [Verhulst 1838]
|
792
|
+
blank |
|
793
|
+
|
|
794
|
+
text | xt+1 = axt (1 - xt)
|
795
|
+
blank |
|
796
|
+
|
|
797
|
+
text | such that information is subject to dynamic effects over time. Iteration of mutual
|
798
|
+
| information over time shows “exchange resonances” (Figure 1-5).
|
799
|
+
blank |
|
800
|
+
|
|
801
|
+
text | When the Information parameters (node and product information package) are
|
802
|
+
| incompatible with Transmission Link parameters (link and environment), aliasing occurs.
|
803
|
+
| The information or product is not completely usable, or has become corrupted, and
|
804
|
+
| additional work will be needed to adapt or understand it. In fact even minor mismatches
|
805
|
+
| create observable aliasing; it is a matter of scale.
|
806
|
+
blank |
|
807
|
+
|
|
808
|
+
text | 7
|
809
|
+
| B
|
810
|
+
| 6
|
811
|
+
| A C
|
812
|
+
| Mutual Information I(x^y)
|
813
|
+
blank |
|
814
|
+
|
|
815
|
+
|
|
816
|
+
|
|
817
|
+
text | 5
|
818
|
+
blank |
|
819
|
+
text | 4
|
820
|
+
| t = 10
|
821
|
+
| 3 t = 20
|
822
|
+
| t = 40
|
823
|
+
| 2
|
824
|
+
blank |
|
825
|
+
text | 1
|
826
|
+
blank |
|
827
|
+
text | 0
|
828
|
+
| 0.0 1.0 2.0 3.0 4.0
|
829
|
+
| a
|
830
|
+
blank |
|
831
|
+
|
|
832
|
+
text | Figure 1-5 Information Exchange Resonances
|
833
|
+
| Information can be exchanged more completely at certain periods of time at the “resonant” nodes A,
|
834
|
+
| B, and C on the graph where a = 1, 3, and 3.5.
|
835
|
+
blank |
|
836
|
+
|
|
837
|
+
|
|
838
|
+
|
|
839
|
+
meta | 8
|
840
|
+
title | 1.4 Results
|
841
|
+
text | Model parameters were developed for each link in the organization model. Multiple
|
842
|
+
| parameters are summarized into a value for product information characteristics ‘p’, and
|
843
|
+
| for transmission characteristics ‘a’, for each link. Several of these are shown in Figure
|
844
|
+
| 1-6 to highlight cases of interest. The product information aliasing factor ‘p’, has been
|
845
|
+
| placed into three bins: (s) small (1-2), (m) medium (3), and (l) large (4-6). The
|
846
|
+
| environmental transmission/transfer rate is represented by ‘a’. (Note that ‘a’ is non-
|
847
|
+
| continuous, and has a reverse scale compared to ‘p’ (see Sections 3.4.2 and 4.2).)
|
848
|
+
blank |
|
849
|
+
|
|
850
|
+
text | BD PROG R&D
|
851
|
+
blank |
|
852
|
+
|
|
853
|
+
|
|
854
|
+
text | Exec Exec Exec
|
855
|
+
blank |
|
856
|
+
text | m=3
|
857
|
+
| Adjacent BD
|
858
|
+
| m=3
|
859
|
+
| a=3 √ Executive
|
860
|
+
| a=1
|
861
|
+
| Senior Adjacent
|
862
|
+
| Markets Director Staff Support Technologies
|
863
|
+
blank |
|
864
|
+
|
|
865
|
+
text | l=5 m=3
|
866
|
+
| BD
|
867
|
+
| Program a=1 a=1 Advanced Senior
|
868
|
+
| Staff m=3 Develop Manager
|
869
|
+
| a=3 √ √
|
870
|
+
| s=2
|
871
|
+
| Senior
|
872
|
+
| Subsystem CRAD
|
873
|
+
| √ a=4 Senior
|
874
|
+
| Tools Engineer m=3 Scientist
|
875
|
+
| a=3 √
|
876
|
+
| s=2
|
877
|
+
| √ a=4 Staff
|
878
|
+
| s=2 Engineer m=3 Engineer Engineer
|
879
|
+
| Scientist
|
880
|
+
| a=1 a=3 √
|
881
|
+
| Discipline
|
882
|
+
| Director
|
883
|
+
blank |
|
884
|
+
|
|
885
|
+
|
|
886
|
+
|
|
887
|
+
text | ENG
|
888
|
+
blank |
|
889
|
+
|
|
890
|
+
|
|
891
|
+
text | Figure 1-6 Information Transfer Regimes and Amplification Factors
|
892
|
+
| Network performance is evaluated by looking for link/information mismatch. The red circles are
|
893
|
+
| points of interest that have been evaluated for consistency. The checkmarks indicate consistent link
|
894
|
+
| parameters and information formatting. The sad-faces indicate mismatches that will be discussed in
|
895
|
+
| subsequent sections.
|
896
|
+
blank |
|
897
|
+
|
|
898
|
+
|
|
899
|
+
text | Potential network improvements can be assessed both at the individual links and nodes,
|
900
|
+
| and over “link-chains”. Between individual nodes, mismatches of Link parameters (p, a)
|
901
|
+
| indicate a discontinuity, or a mismatch of Lyupanov exponents (Section 3.4.1). Two
|
902
|
+
| nodes can improve their combined performance by exchanging information more
|
903
|
+
blank |
|
904
|
+
|
|
905
|
+
|
|
906
|
+
|
|
907
|
+
meta | 9
|
908
|
+
text | efficiently. However, improving information transfer at one link may or may not
|
909
|
+
| improve overall system performance, although there will certainly be local effects.
|
910
|
+
blank |
|
911
|
+
|
|
912
|
+
text | Figure 1-7 shows an analysis of an individual link for the transfer and insertion of
|
913
|
+
| research information into a new mega-system. In this case, new research was not being
|
914
|
+
| efficiently delivered into new system. Analysis of the information transfer characteristics
|
915
|
+
| BD PROG
|
916
|
+
| indicated potential for a five-fold increase in productivity. A complete description R&D
|
917
|
+
| and
|
918
|
+
| analysis is developed in section 4.3.1.
|
919
|
+
blank |
|
920
|
+
|
|
921
|
+
text | Exec Exec Exec
|
922
|
+
blank |
|
923
|
+
|
|
924
|
+
|
|
925
|
+
text | t=m
|
926
|
+
| BD
|
927
|
+
| B Executive a= 1 Senior Adjacent
|
928
|
+
| A
|
929
|
+
| Director Staff Support Technologies
|
930
|
+
| t=m D
|
931
|
+
| A a= 1 C
|
932
|
+
| BD C Advanced Senior
|
933
|
+
| Program
|
934
|
+
| Staff Develop Manager
|
935
|
+
| C D
|
936
|
+
| D
|
937
|
+
blank |
|
938
|
+
text | Senior Senior
|
939
|
+
| Subsystem CRAD
|
940
|
+
| Engineer Scientist
|
941
|
+
| Figure 1-7 Set of Consistent Guides to Repair Discontinuity
|
942
|
+
| This figure zooms into our design community network to focus on two identified discontinuities
|
943
|
+
| Staff is
|
944
|
+
| labeled A. In both of these cases, p = m and a = 1. So although the communication timeframe
|
945
|
+
| Engineer Engineer Engineer
|
946
|
+
| Scientist
|
947
|
+
| “medium”, the information delivered is in the wrong form for direct consumption. Also variety was
|
948
|
+
| expended at some cost to transform the original information. So the resources used to produce the
|
949
|
+
| original information are not being exploited efficiently. For steps B, C, and D, please refer to section
|
950
|
+
| Discipline 4.3.1.
|
951
|
+
| Director
|
952
|
+
blank |
|
953
|
+
text | Our mega-organization is not independent. Rather it is given particular direction and
|
954
|
+
| goals in order to achieve the design and production of mega-systems. For instance, some
|
955
|
+
| ENG information flows only a certain direction, and some information is tightly controlled.
|
956
|
+
| Therefore the network is only a quasi, scale-free network. Imposing external goals and
|
957
|
+
blank |
|
958
|
+
|
|
959
|
+
|
|
960
|
+
meta | 10
|
961
|
+
text | specifications onto our scale-free, multi-hierarchical control structure, creates sets of
|
962
|
+
| directed links (link-chains) corresponding to control paths required to focus the
|
963
|
+
| community on our particular future goal states. These directed link-chains need to be
|
964
|
+
| evaluated to make sure the information arrives timely and with minimal (acceptable)
|
965
|
+
| corruption.
|
966
|
+
blank |
|
967
|
+
|
|
968
|
+
text | Link-chains of interest focus on particular extended pathways that execute higher level
|
969
|
+
| functions, such as translating customer product requirements into research needs.
|
970
|
+
| Evaluation of various link chains can be analyzed both from an a priori interest as well as
|
971
|
+
| from evaluating the scale-free network for paths of overall poor quality.
|
972
|
+
blank |
|
973
|
+
|
|
974
|
+
text | Case (2) in Figure 1-8 shows the delivery of new Research products into large Programs.
|
975
|
+
| This was part of the same analysis of the transfer and insertion of research information
|
976
|
+
| into a new mega-system, but at a larger scale to examine total flow and also to look for
|
977
|
+
| side-effects. This path has a Total Aliasing Factor (TAF) of 15 (see Section 3.4).
|
978
|
+
| Changes from annealing the discontinuities from Figure 1-7 (discussed in Section 4.3.1)
|
979
|
+
| will reduce the link values (aliasing factors) of 5 and 3 to 2 and 2, and improve the
|
980
|
+
| efficiency of incorporating current (and appropriate) technology into the new mega-
|
981
|
+
| system.
|
982
|
+
blank |
|
983
|
+
|
|
984
|
+
|
|
985
|
+
|
|
986
|
+
meta | 11
|
987
|
+
text | BD PROG R&D
|
988
|
+
blank |
|
989
|
+
|
|
990
|
+
|
|
991
|
+
text | Exec 3 Exec Exec
|
992
|
+
blank |
|
993
|
+
text | 1 2 1 1
|
994
|
+
| Adjacent BD Executive
|
995
|
+
| 3 Senior Adjacent
|
996
|
+
| Markets Director Staff Support Technologies
|
997
|
+
| 3 2
|
998
|
+
| 2 2 5
|
999
|
+
| BD
|
1000
|
+
| Staff
|
1001
|
+
| 3 Program
|
1002
|
+
| 5
|
1003
|
+
| Advanced
|
1004
|
+
| Develop
|
1005
|
+
| Senior
|
1006
|
+
| Manager
|
1007
|
+
blank |
|
1008
|
+
text | 2 2 2
|
1009
|
+
| Senior 3 Subsystem CRAD
|
1010
|
+
| 2 Senior 3
|
1011
|
+
| Engineer Scientist
|
1012
|
+
blank |
|
1013
|
+
text | 2 2 3 2 22
|
1014
|
+
| Tools Engineer
|
1015
|
+
| 3 1
|
1016
|
+
| Engineer Engineer
|
1017
|
+
| 2 Staff
|
1018
|
+
| 5 Scientist
|
1019
|
+
| 1
|
1020
|
+
| 2 Discipline
|
1021
|
+
| Director
|
1022
|
+
blank |
|
1023
|
+
|
|
1024
|
+
|
|
1025
|
+
|
|
1026
|
+
text | ENG
|
1027
|
+
blank |
|
1028
|
+
|
|
1029
|
+
text | Figure 1-8 Link-chain Paths of Interest
|
1030
|
+
| External goals and specifications, combined with our quasi, scale-free network, create directed links
|
1031
|
+
| corresponding to control paths required to focus the community on the particular future goal states.
|
1032
|
+
| Numbered blue lines are directed link paths; red numbers are total aliasing factors (TAF) (section
|
1033
|
+
| 3.4.4). These link-chains need to be evaluated to make sure the information arrives timely and un-
|
1034
|
+
| corrupted. Corruption can be caused by aliasing, occurring either during variety compression /
|
1035
|
+
| decoding, or from multi-scale time-aliasing. Case (2), the delivery of new Research products into
|
1036
|
+
| large Programs, has the highest TAF corruption factor of 15.
|
1037
|
+
blank |
|
1038
|
+
|
|
1039
|
+
text | In fact, a case study addressing the delivery of new research products into large programs
|
1040
|
+
| was monitored in parallel wherein the delivery path of design products was adjusted.
|
1041
|
+
| Several interesting behaviours emerged around design ownership, as reflected within
|
1042
|
+
| changed design variety and program variety (section 4.4).
|
1043
|
+
blank |
|
1044
|
+
|
|
1045
|
+
text | While emergent engineering behaviours are interesting, another question is whether this
|
1046
|
+
| improved the value to the company and/or the value delivered to the environment
|
1047
|
+
| (society). These questions might not be definitively answerable, but one metric – the
|
1048
|
+
| inserted value of R&D efforts – rose from less than 5% to almost 80% as measured by
|
1049
|
+
| the number of research projects being fully used in operational programs. The research
|
1050
|
+
blank |
|
1051
|
+
|
|
1052
|
+
|
|
1053
|
+
|
|
1054
|
+
meta | 12
|
1055
|
+
text | projects themselves can be quite successful, but information transfer into a development
|
1056
|
+
| program runs into the problems indicated in terms of time synchronization.
|
1057
|
+
blank |
|
1058
|
+
title | 1.5 Conclusions
|
1059
|
+
text | This dissertation developed a new type of model for extremely large design organizations
|
1060
|
+
| and their environment by combining perspectives from:
|
1061
|
+
| - evolutionary genetic computing organisms,
|
1062
|
+
| - information theoretic (quantum) information production, and
|
1063
|
+
| - non-linear dynamics and chaos theory for information exchange.
|
1064
|
+
| to study some of the behaviours of complex organizations, especially information
|
1065
|
+
| transfer, to show how organizations can be guided to more efficiently and quickly
|
1066
|
+
| produce complex mega-systems.
|
1067
|
+
blank |
|
1068
|
+
|
|
1069
|
+
text | A three-year experiment was performed within a large design organization to see what
|
1070
|
+
| level and type of improvements would actually manifest.
|
1071
|
+
blank |
|
1072
|
+
|
|
1073
|
+
text | The results fall into three categories:
|
1074
|
+
| - production of individual mega-systems,
|
1075
|
+
| - improvement in organizations that develop mega-systems, and
|
1076
|
+
| - extraction of societal benefits from government funded mega-systems.
|
1077
|
+
| These are addressed separately below.
|
1078
|
+
blank |
|
1079
|
+
title | 1.5.1 Design of Individual Mega-Systems
|
1080
|
+
text | Mismatched quality of information transfer among organizational elements causes
|
1081
|
+
| significant inefficiencies. The transfer quality includes time synchronization, method of
|
1082
|
+
| transfer, and other variables. It is not that anything is currently wrong, just that a lot of
|
1083
|
+
| time is wasted waiting for information, translating information, or re-developing the
|
1084
|
+
| context to understand the information. Also, additional time (and resources) may be
|
1085
|
+
| spent waiting for task convergence at a higher level. Matching the characteristics of the
|
1086
|
+
| product design information to the characteristics of the transfer path minimizes aliasing
|
1087
|
+
| effects (dissipation and noise), creating many benefits including:
|
1088
|
+
blank |
|
1089
|
+
|
|
1090
|
+
|
|
1091
|
+
meta | 13
|
1092
|
+
text | - More efficient design and design execution,
|
1093
|
+
| - Improved design agility / currency, and
|
1094
|
+
| - Enablement / easing of “Evolutionary Design” ([Bobinis 2006], [Bar-Yam
|
1095
|
+
| 2003], [Sangiovanni 2006]) at the upper, large-scale limit.
|
1096
|
+
blank |
|
1097
|
+
|
|
1098
|
+
text | Results from a three-year experiment showed that the inserted value of R&D efforts rose
|
1099
|
+
| from less than 5% to almost 80% as measured by the number of research projects being
|
1100
|
+
| fully used in operational programs. The research projects themselves can be quite
|
1101
|
+
| successful, but information transfer into a development program runs into the problems
|
1102
|
+
| indicated in terms of time synchronization.
|
1103
|
+
blank |
|
1104
|
+
title | 1.5.2 Company and Design-Organization Implications
|
1105
|
+
text | Improvements in the performance of the organization will incrementally emerge, given
|
1106
|
+
| the inertia of large organizations and of engineering processes, and will initially manifest
|
1107
|
+
| as quicker mega-system deliveries and an accelerating ability to innovate ahead of other
|
1108
|
+
| companies.
|
1109
|
+
blank |
|
1110
|
+
|
|
1111
|
+
text | “One consequence of encoding information as statistical and time-varying patterns of
|
1112
|
+
| low-level components is that no individual component of the system can perceive or
|
1113
|
+
| communicate the ‘big-picture’ of the state of the system. Instead, information must be
|
1114
|
+
| communicated via spatial and temporal sampling.” [Mitchell 2009] Dynamically, the
|
1115
|
+
| mega-design process is progressing, and any snapshot is an incomplete view in a
|
1116
|
+
| changing context. This implies that the product directors exert little real control on where
|
1117
|
+
| the company is going. Their actual role is simply to monitor the outputs and nascent
|
1118
|
+
| products, and to guide the output (and rate of new product introduction and of upgrades)
|
1119
|
+
| to match the funding (requirements) source.
|
1120
|
+
blank |
|
1121
|
+
|
|
1122
|
+
text | This dynamic perspective implies that design communities can, and should, “construct
|
1123
|
+
| their own [evolutionary] selective constructs designed toward increasing mutual
|
1124
|
+
| information [within itself, and] between the organism and the environment” [Krakauer
|
1125
|
+
| 2011], for better computability or production of designs (cf. niche construction). In this
|
1126
|
+
blank |
|
1127
|
+
|
|
1128
|
+
|
|
1129
|
+
meta | 14
|
1130
|
+
text | case, the management structure ought to also look at higher level abstractions, such as
|
1131
|
+
| information transfer efficiencies, in addition to process control (like design reviews),
|
1132
|
+
| because looking at static variables will not improve the evolutionary adaptation rate.
|
1133
|
+
blank |
|
1134
|
+
title | 1.5.3 Government and Societal Implications
|
1135
|
+
text | The stability of any implemented improvements is somewhat problematic given the
|
1136
|
+
| evolving environment, but especially given the counter-acting efforts of government
|
1137
|
+
| acquisition regulations and nominal management tendencies.
|
1138
|
+
blank |
|
1139
|
+
|
|
1140
|
+
text | Scale, synchronization, and economic nudges may be appropriate for personal scale
|
1141
|
+
| complices such as smartPhone apps, but the economic and scale factors are different from
|
1142
|
+
| government “apps”. A focus on lowest cost procurement, for government “apps” (mega-
|
1143
|
+
| systems), has side effects on the engineering design process. Financially speaking, and
|
1144
|
+
| for more effective mega-system development, the government should modify cost-
|
1145
|
+
| selection constraints to avoid extra effort to achieve a particular instance when any of a
|
1146
|
+
| set of instances is all that is needed. This would necessarily become somewhat non-
|
1147
|
+
| deterministic.
|
1148
|
+
blank |
|
1149
|
+
|
|
1150
|
+
text | Individual program managers will avoid this because it is locally inefficient and also
|
1151
|
+
| uncontrollable. However, according to game theory (prisoner’s dilemma [Morrow
|
1152
|
+
| 1994]), a Pareto optimal solution is better, and can be achieved with “external norms”
|
1153
|
+
| [Mitchell 2009]. The external norms could either be via company processes and
|
1154
|
+
| procedures or, better, by the government’s Federal Acquisition Regulations (FAR).
|
1155
|
+
blank |
|
1156
|
+
|
|
1157
|
+
text | For instance, implementation at the government FAR level could be to assign 1% - 2% of
|
1158
|
+
| the value of programs over a certain size to internal efforts supporting adaptability and
|
1159
|
+
| risk reduction. For instance the Orion spacecraft program internally allocated $100M
|
1160
|
+
| (for a $4B program) to reduce schedule risk occurring from software integration. This is
|
1161
|
+
| the same scale as proposed, and was considered and approved as the correct engineering
|
1162
|
+
| approach. However, it was implemented without adjusting the encompassing process, so
|
1163
|
+
blank |
|
1164
|
+
|
|
1165
|
+
|
|
1166
|
+
|
|
1167
|
+
meta | 15
|
1168
|
+
text | savings could never be recaptured, nor future transitions eased. This fund would reduce
|
1169
|
+
| risk on the delivery of the initial system, and reduce cost of subsequent systems.
|
1170
|
+
blank |
|
1171
|
+
|
|
1172
|
+
text | Such a fund could also be allocated for application to follow-on project development, and
|
1173
|
+
| coordination of supporting technology and product development. Note that this would
|
1174
|
+
| avoid aliasing effects from Federally Funded Research and Development Companies
|
1175
|
+
| (FFRDCs) in their information transfer efforts, and would greatly improve technology
|
1176
|
+
| transfer from government labs to industry and society.
|
1177
|
+
blank |
|
1178
|
+
|
|
1179
|
+
|
|
1180
|
+
|
|
1181
|
+
meta | 16
|
1182
|
+
title | 2 Network Model Development and Mega-Organizations
|
1183
|
+
text | In order to design, integrate, and deliver a mega-system, the organizational unit must
|
1184
|
+
| itself be very large and thus inherently complex, especially when efficient use of capital
|
1185
|
+
| and labor indicate (out of phase) development of multiple mega-systems.
|
1186
|
+
blank |
|
1187
|
+
|
|
1188
|
+
|
|
1189
|
+
title | 2.1 Structure of Very Large Aerospace Design Communities
|
1190
|
+
text | Large aerospace companies typically have on the order of (100,000) employees, of whom
|
1191
|
+
| (50%) are engineers and scientists of various sorts. These organizations typically have a
|
1192
|
+
| large number of entwined organizations to produce and deliver their platforms.
|
1193
|
+
| (Numbers in parentheses refer to the notional parameter values in our abstracted model.)
|
1194
|
+
| (Capitalized terms are symbolic names.)
|
1195
|
+
blank |
|
1196
|
+
|
|
1197
|
+
text | Aerospace companies might be divided into (five) business areas or sectors to address
|
1198
|
+
| different market areas. An upper-level corporate organization might provide common
|
1199
|
+
| resources for business development, common research, common engineering tools and
|
1200
|
+
| process, and other functions.
|
1201
|
+
blank |
|
1202
|
+
|
|
1203
|
+
text | For our purposes we will discuss a medium sized space company (sector) of ($10B)
|
1204
|
+
| annual sales. Companies are divided into lines of business or business units on the scale
|
1205
|
+
| of ($1B) annual sales. The (10) business units are the major components of the company
|
1206
|
+
| and are responsible for producing and delivering the output of the organization.
|
1207
|
+
blank |
|
1208
|
+
|
|
1209
|
+
text | Geographically, the organization will be spread across the nation (Figure 2-1) [courtesy
|
1210
|
+
| of Lockheed Martin Space Systems Company]. Some locations have large numbers of
|
1211
|
+
| people; other sites have smaller numbers. The business units may or may not be
|
1212
|
+
| geographically centralized. If not, there is some communications overhead to coordinate
|
1213
|
+
| the pieces.
|
1214
|
+
blank |
|
1215
|
+
|
|
1216
|
+
|
|
1217
|
+
|
|
1218
|
+
meta | 17
|
1219
|
+
text | Louisville, CO
|
1220
|
+
| Bangor, WA Denver, CO Newtown, PA
|
1221
|
+
| • Coherent
|
1222
|
+
| • FBM IA&T
|
1223
|
+
| Technologies • Navigation Systems
|
1224
|
+
| Palo Alto, CA • Commercial Space
|
1225
|
+
| • ATC
|
1226
|
+
blank |
|
1227
|
+
|
|
1228
|
+
text | Sunnyvale, CA
|
1229
|
+
| Valley Forge, PA
|
1230
|
+
| Santa Cruz, CA • Reentry Systems
|
1231
|
+
| • Test Support
|
1232
|
+
| Washington, DC
|
1233
|
+
| • FBM Liaison
|
1234
|
+
| • Washington Ops
|
1235
|
+
| Arlington, VA
|
1236
|
+
| Edwards AFB, CA • Targets Prime
|
1237
|
+
| • ABL Huntsville, AL
|
1238
|
+
| • Strategic & Missile Defense
|
1239
|
+
| El Segundo, CA • Ares Support
|
1240
|
+
| • MILSATCOM
|
1241
|
+
| Albuquerque, NM Kings Bay, GA
|
1242
|
+
| • ATC • FBM IA&T
|
1243
|
+
| Kauai, HI White Sands, NM • ATK
|
1244
|
+
| • PMRF Courtland, AL Cape Canaveral, FL
|
1245
|
+
| • Missile Range • Targets & • Targets & Eastern Range
|
1246
|
+
| Stennis, MS
|
1247
|
+
| Countermeasures Countermeasures • Launch Operations
|
1248
|
+
| • Propulsion
|
1249
|
+
| Houston, TX Center
|
1250
|
+
| ~19,500 employees • Orion
|
1251
|
+
| New Orleans, LA
|
1252
|
+
| • ~ 8,700 in Silicon Valley
|
1253
|
+
| • External Tank
|
1254
|
+
| • ~ 4,700 in Denver area
|
1255
|
+
| • ~ 2,000 in New Orleans • Orion
|
1256
|
+
| • ~ 1,500 in Cape and V
|
1257
|
+
| • ~ 1,100 in Delaware Valley
|
1258
|
+
| 009_03.18.08
|
1259
|
+
| • ~ 500 in Alabama
|
1260
|
+
blank |
|
1261
|
+
text | Figure 2-1 Geographic Distribution
|
1262
|
+
blank |
|
1263
|
+
|
|
1264
|
+
|
|
1265
|
+
text | Our prototypical medium sized space company has on the order of (20,000) employees.
|
1266
|
+
| The (10,000) engineers frequently reside in an Engineering home-shop organization that
|
1267
|
+
| may span (10) disciplines, each with a Discipline Director. The goal of the Engineering
|
1268
|
+
| home-shops is to have 99.9% of their staff assigned to a Program where they are
|
1269
|
+
| developing deliverable products and profit. Each business unit would then have about
|
1270
|
+
| (1000) engineers and (2000) total staff. Engineering frequently has some small number
|
1271
|
+
| of senior engineering staff who coordinates Tools and processes.
|
1272
|
+
blank |
|
1273
|
+
|
|
1274
|
+
text | Business Development coordinates all the business units to coordinate their outputs to
|
1275
|
+
| optimally take advantage of resources while delivering to the customer requirements.
|
1276
|
+
blank |
|
1277
|
+
|
|
1278
|
+
text | Our typical company will have a Research group which may have (1000) scientists. Of
|
1279
|
+
| these we will assume that (400) work on advanced or specialized products, and (600)
|
1280
|
+
| advance the state of the art in (6) different technical areas. Hopefully their research is
|
1281
|
+
| applied to the next generation products being developed by the business units.
|
1282
|
+
blank |
|
1283
|
+
|
|
1284
|
+
|
|
1285
|
+
|
|
1286
|
+
meta | 18
|
1287
|
+
text | (Figure 2-2) abstracts our model community into a non-directed graph. (Figure 2-3)
|
1288
|
+
| shows typically sized organizational parameters.
|
1289
|
+
blank |
|
1290
|
+
|
|
1291
|
+
text | Sector Sector Sector
|
1292
|
+
| Coord Coord Coord
|
1293
|
+
| Major
|
1294
|
+
| Teammates
|
1295
|
+
blank |
|
1296
|
+
text | BD PROG R&D
|
1297
|
+
| PROG
|
1298
|
+
| PROG
|
1299
|
+
| Sector
|
1300
|
+
| PROG
|
1301
|
+
| PROG
|
1302
|
+
| Coord
|
1303
|
+
| Exec Exec Exec
|
1304
|
+
blank |
|
1305
|
+
|
|
1306
|
+
|
|
1307
|
+
|
|
1308
|
+
text | Adjacent BD Executive Senior Adjacent
|
1309
|
+
| Adjacent
|
1310
|
+
| Markets
|
1311
|
+
| Adjacent Director Staff Support Technologies
|
1312
|
+
| Markets
|
1313
|
+
| Adjacent
|
1314
|
+
| Markets
|
1315
|
+
| Markets
|
1316
|
+
blank |
|
1317
|
+
|
|
1318
|
+
text | BD Advanced Senior
|
1319
|
+
| Senior
|
1320
|
+
| Program
|
1321
|
+
| Program Senior
|
1322
|
+
| Staff Program Develop Manager
|
1323
|
+
| Senior
|
1324
|
+
| ManagerSenior
|
1325
|
+
| Manager
|
1326
|
+
| Senior
|
1327
|
+
| ManagerSenior
|
1328
|
+
| Manager
|
1329
|
+
| Manager
|
1330
|
+
| Manager
|
1331
|
+
blank |
|
1332
|
+
text | Senior Senior
|
1333
|
+
| Subsystem
|
1334
|
+
| Subsystem CRAD
|
1335
|
+
| CRAD
|
1336
|
+
| Engineer Subsystem
|
1337
|
+
| Subsystem CRAD
|
1338
|
+
| CRAD Scientist
|
1339
|
+
| Subsystem
|
1340
|
+
| Subsystem CRAD
|
1341
|
+
| CRAD
|
1342
|
+
| Subsystem
|
1343
|
+
blank |
|
1344
|
+
|
|
1345
|
+
text | Staff
|
1346
|
+
| Tools Engineer Engineers Engineer
|
1347
|
+
| Scientist
|
1348
|
+
blank |
|
1349
|
+
|
|
1350
|
+
|
|
1351
|
+
text | Discipline
|
1352
|
+
| Discipline
|
1353
|
+
| Director
|
1354
|
+
| Discipline
|
1355
|
+
| Director Suppliers Small Universities
|
1356
|
+
| Discipline
|
1357
|
+
| Director Business
|
1358
|
+
| Discipline
|
1359
|
+
| Director
|
1360
|
+
| Discipline
|
1361
|
+
| Director
|
1362
|
+
| Discipline
|
1363
|
+
| Director
|
1364
|
+
| Discipline
|
1365
|
+
| Director
|
1366
|
+
| Sector
|
1367
|
+
| ENG Director
|
1368
|
+
| Coord
|
1369
|
+
blank |
|
1370
|
+
|
|
1371
|
+
|
|
1372
|
+
text | Figure 2-2 Conceptual Organizational Structure, Staffing Levels, and Characteristics
|
1373
|
+
| “Sector Coord” (green arrows) maps to the other five Business Areas as additional, similar layers.
|
1374
|
+
| “Clouds” map to other complex communities such as university research networks, small businesses
|
1375
|
+
| and suppliers, etc. The linkages cross-connect at many places. “Major teammates” represent
|
1376
|
+
| another similar aerospace company of similar size (20,000 employees).
|
1377
|
+
blank |
|
1378
|
+
|
|
1379
|
+
|
|
1380
|
+
|
|
1381
|
+
meta | 19
|
1382
|
+
title | Sector Sector Sector
|
1383
|
+
text | Coord Coord Coord
|
1384
|
+
| Major
|
1385
|
+
| Teammates
|
1386
|
+
blank |
|
1387
|
+
text | BD PROG R&D
|
1388
|
+
| PROG
|
1389
|
+
| PROG
|
1390
|
+
| Sector
|
1391
|
+
| PROG
|
1392
|
+
| PROG
|
1393
|
+
| Coord
|
1394
|
+
| Exec Exec Exec
|
1395
|
+
| 1 1
|
1396
|
+
| 1 1 1
|
1397
|
+
| 8 5 3 3 2
|
1398
|
+
| Adjacent BD 1 1 Executive 1 3 Senior Adjacent
|
1399
|
+
| Adjacent
|
1400
|
+
| Markets
|
1401
|
+
| Adjacent Director Staff Support Technologies
|
1402
|
+
| Markets
|
1403
|
+
| Adjacent 1 240
|
1404
|
+
| Markets
|
1405
|
+
| Markets 1 1
|
1406
|
+
| 4 markets 5 3 1
|
1407
|
+
| 6
|
1408
|
+
| 3 progs
|
1409
|
+
| BD 5 1 Advanced Senior
|
1410
|
+
| Senior 6 techs
|
1411
|
+
| Program
|
1412
|
+
| Program Senior
|
1413
|
+
| Staff Program Develop Manager
|
1414
|
+
| Senior
|
1415
|
+
| Manager
|
1416
|
+
| Senior
|
1417
|
+
| Manager
|
1418
|
+
| Senior
|
1419
|
+
| 1 Manager
|
1420
|
+
| Senior
|
1421
|
+
| Manager
|
1422
|
+
| 1 1 Manager
|
1423
|
+
| 10 7 7 CRADS 10 Manager
|
1424
|
+
| 10 subs
|
1425
|
+
| Senior // 1/discipline Subsystem CRAD // Senior
|
1426
|
+
| Engineer Subsystem
|
1427
|
+
| Subsystem CRAD
|
1428
|
+
| CRAD Scientist
|
1429
|
+
| Subsystem
|
1430
|
+
| Subsystem CRAD
|
1431
|
+
| CRAD 1 10/tech
|
1432
|
+
| 10 Subsystem
|
1433
|
+
| Subsystem CRAD
|
1434
|
+
| 1 1 1 1
|
1435
|
+
| 5/discipline
|
1436
|
+
| 100 100 10 7
|
1437
|
+
| // 100/discipline Engineers // 3 Staff
|
1438
|
+
| Tools Engineer 1000 Engineer 70/tech
|
1439
|
+
| Scientist
|
1440
|
+
| 5 1000 10
|
1441
|
+
| 1
|
1442
|
+
| 1
|
1443
|
+
| Discipline
|
1444
|
+
| Discipline
|
1445
|
+
| Director
|
1446
|
+
| Discipline
|
1447
|
+
| Director Suppliers Small Universities
|
1448
|
+
| Discipline
|
1449
|
+
| Director Business
|
1450
|
+
| Discipline
|
1451
|
+
| Director
|
1452
|
+
| Discipline
|
1453
|
+
| Director
|
1454
|
+
| Discipline
|
1455
|
+
| Director
|
1456
|
+
| Discipline
|
1457
|
+
| Director 10 disciplines
|
1458
|
+
| Sector
|
1459
|
+
| ENG Director
|
1460
|
+
| Coord
|
1461
|
+
blank |
|
1462
|
+
|
|
1463
|
+
text | Figure 2-3 Organizational Size Characteristics
|
1464
|
+
| Numbers in red by ovals indicate typical number of similar groups. Double slashes indicate multiple
|
1465
|
+
| connections. Numbers in black near lines indicate the communications I/O dispersion.
|
1466
|
+
blank |
|
1467
|
+
|
|
1468
|
+
|
|
1469
|
+
|
|
1470
|
+
text | Most Engineering Design research focuses on performance developing a single, limited
|
1471
|
+
| product. We want to look both at a larger product and at the overall organizational
|
1472
|
+
| combination and the integrated product evolution.
|
1473
|
+
blank |
|
1474
|
+
|
|
1475
|
+
|
|
1476
|
+
|
|
1477
|
+
meta | 20
|
1478
|
+
title | 2.2 Network Characteristics
|
1479
|
+
text | In our model of this aerospace organization or community, there are many overlapping
|
1480
|
+
| hierarchies. First there are the five different Business Units that are controlled by upper-
|
1481
|
+
| level corporate organization that supplies common resources. In parallel, there are also
|
1482
|
+
| functional hierarchies such as for Engineering disciplines, Finance, Research, Quality
|
1483
|
+
| Control, etc. Finally, there are additional parallel hierarchies for business management,
|
1484
|
+
| Program control and engineering process control.
|
1485
|
+
blank |
|
1486
|
+
|
|
1487
|
+
text | This amalgamation of connections is entwined with the central corporate organizations to
|
1488
|
+
| support related market efforts. A simple two-level, overlapping hierarchy is shown in
|
1489
|
+
| Figure 2-4. Note that priorities, sequencing, and control issues may arise with multiple
|
1490
|
+
| different control staff.
|
1491
|
+
blank |
|
1492
|
+
|
|
1493
|
+
|
|
1494
|
+
|
|
1495
|
+
text | Figure 2-4 Overlapping Hierarchies
|
1496
|
+
blank |
|
1497
|
+
|
|
1498
|
+
text | In terms of task and staffing control, a simplistic hierarchical organization is shown in
|
1499
|
+
| Figure 2-5. Each level coordinates multiple activities of the level below.
|
1500
|
+
blank |
|
1501
|
+
|
|
1502
|
+
|
|
1503
|
+
|
|
1504
|
+
meta | 21
|
1505
|
+
text | System
|
1506
|
+
blank |
|
1507
|
+
|
|
1508
|
+
text | {
|
1509
|
+
| Design
|
1510
|
+
| Process Type
|
1511
|
+
| (Connectivity)
|
1512
|
+
| SubSystem
|
1513
|
+
| ...
|
1514
|
+
blank |
|
1515
|
+
|
|
1516
|
+
text | Requirements Product
|
1517
|
+
| ...
|
1518
|
+
| SubSubSystem
|
1519
|
+
| Design Coordination ?
|
1520
|
+
| Specifications Components
|
1521
|
+
blank |
|
1522
|
+
text | ...
|
1523
|
+
| Figure 2-5 Typical Hierarchical Organization
|
1524
|
+
| Types of connectivity, both vertical and horizontal, vary by type of engineering design process.
|
1525
|
+
| [Byler 2001]
|
1526
|
+
blank |
|
1527
|
+
|
|
1528
|
+
|
|
1529
|
+
|
|
1530
|
+
text | Hierarchical systems are difficult to scale. Single points of control can’t digest enough
|
1531
|
+
| information to quickly process the next step in the organization’s computation. You need
|
1532
|
+
| more information and more coordination among the larger team for goal-oriented, “feed-
|
1533
|
+
| forward” information to achieve timely efficiency. Also, hierarchies tend to create walls,
|
1534
|
+
| and by creating separate groups, you create (evolutionary) sub-populations which both
|
1535
|
+
| cannot process as much as the whole, but also compete with each other internally.
|
1536
|
+
| Finally, when the sub-populations ‘die’, they lose their knowledge/information; and when
|
1537
|
+
| larger groups “die”, they even lose their assets (like the international space station).
|
1538
|
+
blank |
|
1539
|
+
|
|
1540
|
+
text | For single systems with limited levels of control, both efficient performance and task
|
1541
|
+
| execution stability can be achieved. For larger systems, both stability and performance
|
1542
|
+
| degrade with scale (Figure 2-6 and Figure 2-7). In very large product development
|
1543
|
+
| networks (see section 3.3.2) networks can lose convergence, and tasks never complete.
|
1544
|
+
blank |
|
1545
|
+
|
|
1546
|
+
|
|
1547
|
+
|
|
1548
|
+
meta | 22
|
1549
|
+
text | Figure 2-6 Control Stability for Various Organizations
|
1550
|
+
| Scaling of the norm of the minimal control matrix required to achieve stability for various
|
1551
|
+
| organizations: global (solid, black), local (dashed, black), hierarchy (solid, gray) and multi-hierarchy
|
1552
|
+
| (dashed, gray) as a function of N. The hierarchical organizations are chosen with r = 0:5 and
|
1553
|
+
| branching ratio 2. [Hogg 2001]
|
1554
|
+
blank |
|
1555
|
+
|
|
1556
|
+
|
|
1557
|
+
|
|
1558
|
+
text | Figure 2-7 Organizational Performance
|
1559
|
+
| Effect of noise on the expected value of the largest eigenvalue, for N = 8, _ = 1, f = 0:8 and controls
|
1560
|
+
| with s = 0:7 as a function of _ for asymmetric multiplicative noise with zero mean and standard
|
1561
|
+
| deviation _. The organizations are global (solid, black), local (dashed, black), hierarchy (solid, gray)
|
1562
|
+
| and multi-hierarchy (dashed, gray). For the hierarchical cases, we have r = 0:5 and a branching ratio
|
1563
|
+
| of two. The error bars give the standard error estimate of the mean, based on 1000 samples.
|
1564
|
+
| Eigenvalues above zero indicate situations where the noise makes the system unstable, on average.
|
1565
|
+
| [Hogg 2001]
|
1566
|
+
blank |
|
1567
|
+
|
|
1568
|
+
|
|
1569
|
+
|
|
1570
|
+
meta | 23
|
1571
|
+
text | Cross connections within social systems are natural, and engineering design communities
|
1572
|
+
| have also evolved into ‘neural’ communities. Intuitively, this is good because they are
|
1573
|
+
| better able to compute their products. This is because of the extra information available
|
1574
|
+
| from peer groups (related to ‘a’ as defined in Section 3.0). The quicker computation rates
|
1575
|
+
| let the group/community evolve better because of broader scope and more agility. Thus
|
1576
|
+
| the community organism competes better.
|
1577
|
+
blank |
|
1578
|
+
|
|
1579
|
+
text | For our complex mega-design organization, these networks, in fact, have so much
|
1580
|
+
| secondary interconnections that they approach the connectedness of scale free networks.
|
1581
|
+
| Figure 2-8 shows several network types that underpin our organizational network model.
|
1582
|
+
blank |
|
1583
|
+
text | Scale Free Network
|
1584
|
+
| Abstracted Example of D
|
1585
|
+
| A
|
1586
|
+
| Design Community
|
1587
|
+
| Network
|
1588
|
+
| Sector Sector Sector
|
1589
|
+
| Coord Coord Coord
|
1590
|
+
| Major
|
1591
|
+
| Teammates
|
1592
|
+
blank |
|
1593
|
+
text | BD PROG R&D
|
1594
|
+
| PROG
|
1595
|
+
| PROG
|
1596
|
+
| PROG
|
1597
|
+
| Multi-Dimensional Sector
|
1598
|
+
| Coord
|
1599
|
+
| PROG
|
1600
|
+
blank |
|
1601
|
+
text | Scale Free Network Exec Exec Exec
|
1602
|
+
blank |
|
1603
|
+
|
|
1604
|
+
|
|
1605
|
+
|
|
1606
|
+
text | B Adjacent
|
1607
|
+
| Adjacent
|
1608
|
+
| Markets
|
1609
|
+
| Adjacent
|
1610
|
+
| BD
|
1611
|
+
| Director
|
1612
|
+
| Executive
|
1613
|
+
| Staff
|
1614
|
+
| Senior
|
1615
|
+
| Support
|
1616
|
+
| Adjacent
|
1617
|
+
| Technologies
|
1618
|
+
| Markets
|
1619
|
+
| Adjacent
|
1620
|
+
| Markets
|
1621
|
+
| Markets
|
1622
|
+
blank |
|
1623
|
+
|
|
1624
|
+
text | BD Advanced Senior
|
1625
|
+
| Senior
|
1626
|
+
| Program
|
1627
|
+
| Program Senior
|
1628
|
+
| Staff Program Develop Manager
|
1629
|
+
| Senior
|
1630
|
+
| ManagerSenior
|
1631
|
+
| Manager
|
1632
|
+
| Senior
|
1633
|
+
| ManagerSenior
|
1634
|
+
| Manager
|
1635
|
+
| Manager
|
1636
|
+
| Manager
|
1637
|
+
blank |
|
1638
|
+
text | Senior Senior
|
1639
|
+
| Subsystem
|
1640
|
+
| Subsystem CRAD
|
1641
|
+
| CRAD
|
1642
|
+
| Engineer Subsystem
|
1643
|
+
| Subsystem CRAD
|
1644
|
+
| CRAD Scientist
|
1645
|
+
| Subsystem
|
1646
|
+
| Subsystem CRAD
|
1647
|
+
| CRAD
|
1648
|
+
| Subsystem
|
1649
|
+
blank |
|
1650
|
+
|
|
1651
|
+
text | Staff
|
1652
|
+
| Tools Engineer Engineers Engineer
|
1653
|
+
| Scientist
|
1654
|
+
blank |
|
1655
|
+
|
|
1656
|
+
|
|
1657
|
+
text | Hierarchical Cluster Network Discipline
|
1658
|
+
| Discipline
|
1659
|
+
| Director
|
1660
|
+
| Discipline
|
1661
|
+
| Director Suppliers Small Universities
|
1662
|
+
| Discipline
|
1663
|
+
| Director Business
|
1664
|
+
| Discipline
|
1665
|
+
| Director
|
1666
|
+
| Discipline
|
1667
|
+
| C Director
|
1668
|
+
| Discipline
|
1669
|
+
| Director
|
1670
|
+
| Discipline
|
1671
|
+
| Sector Director
|
1672
|
+
| ENG Director
|
1673
|
+
| Coord
|
1674
|
+
blank |
|
1675
|
+
|
|
1676
|
+
|
|
1677
|
+
|
|
1678
|
+
text | Multi-Dimensional
|
1679
|
+
| Information Transfer Links
|
1680
|
+
| (details in section 3.0)
|
1681
|
+
blank |
|
1682
|
+
|
|
1683
|
+
|
|
1684
|
+
|
|
1685
|
+
text | Figure 2-8 Typical and Combined Network Characteristics
|
1686
|
+
| (A) shows a simple scale free network that spans the continental United States. Central nodes (red)
|
1687
|
+
| connect to leaf nodes (green) but there are multiple paths available between any two nodes. (B)
|
1688
|
+
| shows a much larger network, within which there are several types of connectivity among the nodes.
|
1689
|
+
| (C) is a hierarchical cluster network within which the hierarchies are not linear, but rather clusters
|
1690
|
+
| themselves. (D) Our organization has communication connectivity like a scale-free network, but also
|
1691
|
+
| includes directed links and geographic bandwidth restrictions like a cluster hierarchy. The multi-
|
1692
|
+
| dimensional transfer links are discussed in section 3.0
|
1693
|
+
blank |
|
1694
|
+
|
|
1695
|
+
|
|
1696
|
+
|
|
1697
|
+
meta | 24
|
1698
|
+
text | In the context of network theory a scale-free ideal network is a random network with a
|
1699
|
+
| degree distribution a power law, at least asymptotically. That is, the fraction P(k) of
|
1700
|
+
| nodes in the network having k connections to other nodes goes for large values of k as
|
1701
|
+
blank |
|
1702
|
+
|
|
1703
|
+
text | P(k) ~ k -
|
1704
|
+
blank |
|
1705
|
+
|
|
1706
|
+
text | where is a parameter whose value is typically in the range 2 < < 3.
|
1707
|
+
blank |
|
1708
|
+
|
|
1709
|
+
text | The most notable characteristic in a scale-free network is the relative commonness of
|
1710
|
+
| vertices with a degree that greatly exceeds the average. The highest-degree nodes are
|
1711
|
+
| often called "hubs". The scale-free property also strongly correlates with the network's
|
1712
|
+
| robustness to failure.
|
1713
|
+
blank |
|
1714
|
+
|
|
1715
|
+
text | Another important characteristic of scale-free networks is the clustering coefficient
|
1716
|
+
| distribution, which decreases as the node degree increases. This distribution also follows
|
1717
|
+
| a power law. This implies that the low-degree nodes belong to very dense sub-graphs and
|
1718
|
+
| those sub-graphs are connected to each other through hubs.
|
1719
|
+
blank |
|
1720
|
+
|
|
1721
|
+
text | Consider a social network in which nodes are people and links are acquaintance
|
1722
|
+
| relationships between people. It is easy to see that people tend to form communities, i.e.,
|
1723
|
+
| small groups in which everyone knows everyone (one can think of such community as a
|
1724
|
+
| complete graph). In addition, the members of a community also have a few acquaintance
|
1725
|
+
| relationships to people outside that community. Some people, however, are connected to
|
1726
|
+
| a large number of communities (e.g., celebrities, politicians). Those people may be
|
1727
|
+
| considered the hubs responsible for the small-world phenomenon. [Wikkipedia]
|
1728
|
+
blank |
|
1729
|
+
|
|
1730
|
+
text | Our mega-organization is not independent and therefore not completely scale-free.
|
1731
|
+
| Rather it is given particular direction and goals in order to achieve the design and
|
1732
|
+
| production of mega-systems. For instance, some information paths flow only a certain
|
1733
|
+
| direction, and some information is tightly controlled. Therefore the network is only a
|
1734
|
+
| quasi, scale-free network. Imposing external goals and specifications onto our scale-free,
|
1735
|
+
blank |
|
1736
|
+
|
|
1737
|
+
meta | 25
|
1738
|
+
text | multi-hierarchical control structure, creates sets of directed links (link-chains)
|
1739
|
+
| corresponding to control paths required to focus the community on our particular future
|
1740
|
+
| goal states. These directed link-chains need to be correctly addressed.
|
1741
|
+
blank |
|
1742
|
+
|
|
1743
|
+
|
|
1744
|
+
|
|
1745
|
+
meta | 26
|
1746
|
+
title | 2.3 Abstraction to Meta-Network Model
|
1747
|
+
text | (Figure 2-9) has abstracted the many actors from Section 2.1 into a simplified model to
|
1748
|
+
| which we will attach our organizational parameters in Section 3.0.
|
1749
|
+
blank |
|
1750
|
+
|
|
1751
|
+
|
|
1752
|
+
text | Business Program Research
|
1753
|
+
| Development BD PROG
|
1754
|
+
| Management R&D
|
1755
|
+
blank |
|
1756
|
+
|
|
1757
|
+
|
|
1758
|
+
text | Exec Exec Exec
|
1759
|
+
blank |
|
1760
|
+
|
|
1761
|
+
|
|
1762
|
+
|
|
1763
|
+
text | Adjacent BD Executive Senior Adjacent
|
1764
|
+
| Markets Director Staff Support Technologies
|
1765
|
+
blank |
|
1766
|
+
|
|
1767
|
+
|
|
1768
|
+
text | BD Advanced Senior
|
1769
|
+
| Program
|
1770
|
+
| Staff Develop Manager
|
1771
|
+
blank |
|
1772
|
+
|
|
1773
|
+
|
|
1774
|
+
text | Senior Senior
|
1775
|
+
| Subsystem CRAD
|
1776
|
+
| Engineer Scientist
|
1777
|
+
blank |
|
1778
|
+
|
|
1779
|
+
|
|
1780
|
+
text | Staff
|
1781
|
+
| Tools Engineer Engineer Engineer
|
1782
|
+
| Scientist
|
1783
|
+
blank |
|
1784
|
+
|
|
1785
|
+
|
|
1786
|
+
text | Discipline
|
1787
|
+
| Director
|
1788
|
+
| Operational Prototype
|
1789
|
+
| Systems Systems
|
1790
|
+
| ENG
|
1791
|
+
| Engineering
|
1792
|
+
| Figure 2-9 Abstracted Organizational Model
|
1793
|
+
blank |
|
1794
|
+
|
|
1795
|
+
text | In our abstracted organizational model, the information from Figure 2-3 has been
|
1796
|
+
| collapsed into two dimensions for display clarity and in order to focus on the
|
1797
|
+
| multidimensional parameters. The links are multi-dimensional themselves, including
|
1798
|
+
| characteristics on communication I/O dispersion, types of information carried, etc.
|
1799
|
+
| Likewise, the links are not necessarily singular stages, but can encompass multiple steps.
|
1800
|
+
| Extraction and development of these organizational parameters and link characteristics
|
1801
|
+
| are discussed in detail in Section 3.0.
|
1802
|
+
blank |
|
1803
|
+
|
|
1804
|
+
|
|
1805
|
+
|
|
1806
|
+
meta | 27
|
1807
|
+
title | 3 Link Model Development
|
1808
|
+
blank |
|
1809
|
+
title | 3.1 Model Overview
|
1810
|
+
text | For the organization model developed here (Section 2.0), each node is an agent composed
|
1811
|
+
| of many people. Agents can be individuals, design groups, project swarms, or other
|
1812
|
+
| complices. These agents are peer-to-peer agents in a weak sense as defined by Petrie
|
1813
|
+
| [Petrie 1996]. In this sense they are true agents that provide generation of new
|
1814
|
+
| information, but they are weak in that they do not use standard protocols for
|
1815
|
+
| communication. Each agent’s protocols (methods) are particular to that set of agents.
|
1816
|
+
blank |
|
1817
|
+
|
|
1818
|
+
text | Each agent performs its function using a variety of people or skills. Producing a
|
1819
|
+
| controller card, for instance, may require a machinist, a programmer, and a financial
|
1820
|
+
| analyst. Then the product must be packaged and delivered. If one considers the
|
1821
|
+
| controller card as inherently containing the design and product information, then the
|
1822
|
+
| output could be termed to be encoded and transmitted, similar to packaged and delivered.
|
1823
|
+
blank |
|
1824
|
+
|
|
1825
|
+
|
|
1826
|
+
|
|
1827
|
+
meta | 28
|
1828
|
+
text | Link Information Transfer Model
|
1829
|
+
| “Node” parameters
|
1830
|
+
| defined by information encoding and transmitting activities
|
1831
|
+
| and by design information content, resolution, & processing
|
1832
|
+
| “Link” parameters
|
1833
|
+
| defined by relative design role
|
1834
|
+
| Agent X and by organizational activities Agent Y
|
1835
|
+
| w/ Variety 7 w/ Variety 9
|
1836
|
+
| Encoder V6
|
1837
|
+
| Transmitter Vn
|
1838
|
+
| Transmitter V7
|
1839
|
+
blank |
|
1840
|
+
text | environment, tools, time
|
1841
|
+
blank |
|
1842
|
+
|
|
1843
|
+
text | Information Transfer
|
1844
|
+
blank |
|
1845
|
+
|
|
1846
|
+
|
|
1847
|
+
text | Chaotic Channel (non-linear)
|
1848
|
+
| Machining V1 Receiver V1
|
1849
|
+
| Cost Analysis V2 Decoder V2
|
1850
|
+
| Programmer V3
|
1851
|
+
blank |
|
1852
|
+
|
|
1853
|
+
text | Variety = a set of capabilities when
|
1854
|
+
| and functionality Information Parameters (node + info package)
|
1855
|
+
| are incompatible with
|
1856
|
+
| Agents = can be individuals, Transmission Line Parameters (link + environment)
|
1857
|
+
| groups, organizations, “aliasing” effects occur
|
1858
|
+
| or other complices
|
1859
|
+
blank |
|
1860
|
+
text | Figure 3-1 Parameters of Link Transfer Model
|
1861
|
+
blank |
|
1862
|
+
|
|
1863
|
+
text | Depending on the complexity of the product or output of the Agent, it may require more
|
1864
|
+
| or less variety of sub-tasks. Thus the complexity of the product and/or of the agent can
|
1865
|
+
| be represented by the Variety (Vi) of internal skills. The variety represents the set of
|
1866
|
+
| capabilities and functionality.
|
1867
|
+
blank |
|
1868
|
+
|
|
1869
|
+
text | The item produced by an agent can be a physical or ephemeral item. In any event
|
1870
|
+
| information was imposed on the inputs (machining, assembly sequence, tuning) to
|
1871
|
+
| increase its usefulness and reduce the entropy of the parts. The type of information
|
1872
|
+
| imposed on the inputs, and representing the final product (piece-part), can be modeled by
|
1873
|
+
| information content, resolution, and processing.
|
1874
|
+
blank |
|
1875
|
+
|
|
1876
|
+
|
|
1877
|
+
|
|
1878
|
+
meta | 29
|
1879
|
+
text | The node parameters include the information content and resolution, as well as the
|
1880
|
+
| information encoding and transmission. Information is modeled in an information
|
1881
|
+
| theoretic sense after [Shannon 1948] and [Bais 2010]
|
1882
|
+
blank |
|
1883
|
+
|
|
1884
|
+
text | I() = -ipilogpi
|
1885
|
+
blank |
|
1886
|
+
|
|
1887
|
+
text | Any new (conditional) information received by agent y from the information sent by
|
1888
|
+
| agent x will be
|
1889
|
+
blank |
|
1890
|
+
|
|
1891
|
+
text | I(y|x) = -iPx(xi)jPy|x(yj|xi)log2Py|x(yj|xj)
|
1892
|
+
blank |
|
1893
|
+
|
|
1894
|
+
text | which expresses the fact that the receiver must interpret (decode) the information he
|
1895
|
+
| receives based on their previous mutual understanding. Note that information data-bits
|
1896
|
+
| can be interpreted similarly with design “task-bits”.
|
1897
|
+
blank |
|
1898
|
+
|
|
1899
|
+
text | The Link parameters are defined both by the relative design roles of the attached nodes
|
1900
|
+
| within the organizational activities, and by the environment within which the products
|
1901
|
+
| pass and exist. The effect of relative design roles is driven not so much by the different
|
1902
|
+
| activities, but rather by how much work must be done to understand or incorporate
|
1903
|
+
| (receive and decode) the material being transmitted by the input agent. The effect of the
|
1904
|
+
| design environment includes how the information is stored (transmission is not
|
1905
|
+
| instantaneous), the amount of time delay between production and use, the types of tools
|
1906
|
+
| used to decode or process the intermediate product, etc. Because of the dynamic nature
|
1907
|
+
| of the information and the link, this model is a chaotic channel.
|
1908
|
+
blank |
|
1909
|
+
|
|
1910
|
+
text | Information production, aging, and obsolescence is modeled with the logistic growth
|
1911
|
+
| curve [Verhulst 1838]
|
1912
|
+
blank |
|
1913
|
+
|
|
1914
|
+
text | xt+1 = axt (1 - xt)
|
1915
|
+
blank |
|
1916
|
+
|
|
1917
|
+
|
|
1918
|
+
|
|
1919
|
+
meta | 30
|
1920
|
+
text | such that information is subject to dynamic effects over time. Iteration of conditional
|
1921
|
+
| information over time shows “exchange resonances” (see Section 3.4.2).
|
1922
|
+
blank |
|
1923
|
+
|
|
1924
|
+
text | When the Information parameters (node and product information package) are
|
1925
|
+
| incompatible with Transmission Link parameters (link and environment), aliasing occurs.
|
1926
|
+
| The information or product is not completely usable or has become corrupted, and
|
1927
|
+
| additional work will be needed to adapt or understand it. In fact even minor mismatches
|
1928
|
+
| create observable aliasing; it is a matter of scale.
|
1929
|
+
blank |
|
1930
|
+
|
|
1931
|
+
|
|
1932
|
+
title | 3.2 Design Information
|
1933
|
+
blank |
|
1934
|
+
title | 3.2.1 Information in Design Communities
|
1935
|
+
text | For our quasi-scale-free engineering design community (Section 2.0), “information”
|
1936
|
+
| includes the material acquired and developed to further the knowledge base, the product
|
1937
|
+
| design, and the integration, assembly and test of the output products. This includes point
|
1938
|
+
| design descriptions, experimental results, performance measurements, reports,
|
1939
|
+
| spreadsheets, technologies, CNC tapes, test sequences, manufacturing tolerances,
|
1940
|
+
| fabrication techniques, etc.
|
1941
|
+
blank |
|
1942
|
+
|
|
1943
|
+
text | This “information” accumulates over time and is applied to design and produce new
|
1944
|
+
| products. Our engineering community actually produces a portfolio of products
|
1945
|
+
| including evolutionary products, new products in current markets, current products in
|
1946
|
+
| new markets, and occasionally new products for new markets. There is an ongoing,
|
1947
|
+
| evolving set of core competencies within each design community that, when infused with
|
1948
|
+
| new information, allow the product stream to adapt to the changing market
|
1949
|
+
| (environment). This design community produces some number of mega-systems in
|
1950
|
+
| parallel efforts (but offset in time).
|
1951
|
+
blank |
|
1952
|
+
|
|
1953
|
+
|
|
1954
|
+
|
|
1955
|
+
meta | 31
|
1956
|
+
text | Business Program Research
|
1957
|
+
| Development BD PROG
|
1958
|
+
| Management R&D
|
1959
|
+
blank |
|
1960
|
+
|
|
1961
|
+
|
|
1962
|
+
|
|
1963
|
+
text | 1 year Exec Exec Exec
|
1964
|
+
blank |
|
1965
|
+
text | 3 years
|
1966
|
+
| Adjacent BD Executive Senior Adjacent
|
1967
|
+
| Markets Director Staff Support Technologies
|
1968
|
+
blank |
|
1969
|
+
|
|
1970
|
+
|
|
1971
|
+
text | BD Advanced Senior
|
1972
|
+
| Program
|
1973
|
+
| Staff Develop Manager
|
1974
|
+
blank |
|
1975
|
+
text | 2 years
|
1976
|
+
| Senior Senior
|
1977
|
+
| Subsystem CRAD
|
1978
|
+
| Engineer Scientist
|
1979
|
+
blank |
|
1980
|
+
|
|
1981
|
+
text | 12 years
|
1982
|
+
| Staff
|
1983
|
+
| Tools Engineer Engineer Engineer
|
1984
|
+
| Scientist
|
1985
|
+
blank |
|
1986
|
+
|
|
1987
|
+
|
|
1988
|
+
|
|
1989
|
+
text | 4 years Discipline Operational Prototype 1 year
|
1990
|
+
| Director
|
1991
|
+
blank |
|
1992
|
+
|
|
1993
|
+
|
|
1994
|
+
|
|
1995
|
+
text | ENG
|
1996
|
+
| Systems
|
1997
|
+
| Engineering
|
1998
|
+
blank |
|
1999
|
+
text | Figure 3-2 Organizations, Information Products, and Time Epochs
|
2000
|
+
blank |
|
2001
|
+
|
|
2002
|
+
text | In Figure 3-2, there are several major groups whose development products are roughly
|
2003
|
+
| summarized below (capitalized terms symbolically represent their definition). Programs
|
2004
|
+
| create the major output of our design community – an Operational System that takes 12
|
2005
|
+
| years to design, produce, integrate and test. Programs are controlled by Program
|
2006
|
+
| Management; this group coordinates several large projects, and is driven by a (typical)
|
2007
|
+
| three-year Long Range Plan (LRP). The internal information packages within Program
|
2008
|
+
| Management support the activity around successfully executing LRPs (not the program).
|
2009
|
+
| Creation of LRPs is guided by Business Development who operates on a one-year cycle,
|
2010
|
+
| based on accounting cycles including the government budgetary process. Research
|
2011
|
+
| groups typically operate on one-year cycles also, based on an internal budget allocation
|
2012
|
+
| cycle. Research reports are published and filed annually, and other internal information
|
2013
|
+
| products support these efforts. Research groups frequently develop new Prototype
|
2014
|
+
| Systems to demonstrate new capabilities to Program Management. These might take two
|
2015
|
+
| years to develop.
|
2016
|
+
blank |
|
2017
|
+
|
|
2018
|
+
text | To produce the product stream, the actors within the engineering community must
|
2019
|
+
| exchange information in a timely and focused manner. (Note that faster information
|
2020
|
+
| exchange is frequently better for quicker growth, unless the information transfer is so
|
2021
|
+
| quick it is misunderstood, or not yet broadly applicable.)
|
2022
|
+
blank |
|
2023
|
+
|
|
2024
|
+
meta | 32
|
2025
|
+
text | Actors (and groups of actors) within our design community also transfer part of the
|
2026
|
+
| evolving context, in addition to official design documents or drawings, as a broad
|
2027
|
+
| synchronization method. Thus, the information exchanged within the community is the
|
2028
|
+
| combination of newly created technology pieces and components, and the updated
|
2029
|
+
| pointers/relations to additional sources/actors.
|
2030
|
+
blank |
|
2031
|
+
|
|
2032
|
+
text | “In larger organizations, the complexity is vastly different and the chance for deep
|
2033
|
+
| personal co-commitment is diminished. Ad hoc exchanges within one’s personal
|
2034
|
+
| network do happen, but systematic knowledge sharing usually requires brokering through
|
2035
|
+
| people or technology. Along with brokering come issues such as knowledge
|
2036
|
+
| identification, verification, and validation [Leifer 1999].”
|
2037
|
+
blank |
|
2038
|
+
|
|
2039
|
+
text | “[All organizations] can benefit from the implementation of a central knowledge capture
|
2040
|
+
| and re-use infrastructure, a core technology for sharing and learning from each other,
|
2041
|
+
| peer-to-peer. In large institutions (greater than 500 people) this infrastructure will have to
|
2042
|
+
| be explicit and managed separately from the design team support infrastructure.” [Leifer]
|
2043
|
+
blank |
|
2044
|
+
|
|
2045
|
+
text | A separate explicit infrastructure however becomes problematic, especially on projects
|
2046
|
+
| with long time scales (~10 years). Issues arise that interfere with having a consistent
|
2047
|
+
| structure. The portfolio of projects across a large organization use different engineering
|
2048
|
+
| tools. This will always be true because of a mismatch of time frequencies between new
|
2049
|
+
| tools and new projects, and the fact you can’t and don’t want to update your tools all at
|
2050
|
+
| once on a preordained schedule. Also actors bring different sets of tool knowledge with
|
2051
|
+
| them to each activity. It actually takes some finite time for a team to get synchronized on
|
2052
|
+
| common tools and data structures. Much training is completed on the job, with no
|
2053
|
+
| separate time allotment and no separate instructors, with the result that tools are not
|
2054
|
+
| always used as they were designed, nor to their full potential. Finally, retrieval of old
|
2055
|
+
| knowledge becomes problematic and time consuming, and may become moot as different
|
2056
|
+
| technologies advance at different rates, opening up new solution spaces for consideration.
|
2057
|
+
blank |
|
2058
|
+
|
|
2059
|
+
|
|
2060
|
+
|
|
2061
|
+
meta | 33
|
2062
|
+
text | Another issue arising from large organizations that affects knowledge retention and
|
2063
|
+
| retrieval is that engineering team members do not always follow a project from
|
2064
|
+
| conception through design, manufacture, production, maintenance, and growth.
|
2065
|
+
| Typically, engineers focus on one stage of design. As a result the information associated
|
2066
|
+
| with a project bifurcates, with one instance progressing with the product as the product
|
2067
|
+
| “design” and “context”, and the other instance going with the engineer or data base to a
|
2068
|
+
| new product at a similar stage as “knowledge”.
|
2069
|
+
blank |
|
2070
|
+
|
|
2071
|
+
|
|
2072
|
+
title | 3.2.2 Information Modeling
|
2073
|
+
text | Information is modeled similarly to entropy, and transcends its basis in statistical
|
2074
|
+
| mechanics “to describe any system with states {i} and a given probability distribution
|
2075
|
+
| {pi}. Credit for realizing this is usually given to Shannon [Shannon 1948], although
|
2076
|
+
| antecedents include Szilard, Nyquist, and Hartley.”[Bais 2010] Treating information this
|
2077
|
+
| way is useful to analyze and design communication and encryption systems.
|
2078
|
+
blank |
|
2079
|
+
|
|
2080
|
+
text | Shannon proposed that by analogy to the entropy S, information can be defined as
|
2081
|
+
blank |
|
2082
|
+
|
|
2083
|
+
text | H = -ipilogpi
|
2084
|
+
blank |
|
2085
|
+
|
|
2086
|
+
text | A microstate pi is one particular instantiation of all the degrees of freedom of our system.
|
2087
|
+
| If there are N particles (or bits of information), and each bit can have 2 values, then there
|
2088
|
+
| are 2N possible microstates. The space of possible microstates is called the phase space.
|
2089
|
+
blank |
|
2090
|
+
|
|
2091
|
+
text | Information is usually written as I() to emphasize its dependence on the scale of
|
2092
|
+
| resolution of the measurements (and without the Boltzman constant)
|
2093
|
+
blank |
|
2094
|
+
|
|
2095
|
+
text | I() = -ipilogpi
|
2096
|
+
blank |
|
2097
|
+
|
|
2098
|
+
text | This () can be used to define the information dimension
|
2099
|
+
blank |
|
2100
|
+
|
|
2101
|
+
|
|
2102
|
+
|
|
2103
|
+
meta | 34
|
2104
|
+
text | D = lim( ->0) ( I() / |log | )
|
2105
|
+
blank |
|
2106
|
+
|
|
2107
|
+
text | which becomes the fractal dimension with sufficient resolution.
|
2108
|
+
blank |
|
2109
|
+
|
|
2110
|
+
text | In information theory it is common to take logarithms in base two. Units of information
|
2111
|
+
| are called bits as opposed to ‘nats’ for entropy with the conversion that 1 nat = 1.443 bits.
|
2112
|
+
blank |
|
2113
|
+
|
|
2114
|
+
text | Each information package can have its information content “I” measured. The
|
2115
|
+
| information content is a higher level abstraction from the data contained in the package.
|
2116
|
+
blank |
|
2117
|
+
|
|
2118
|
+
text | Information exchange between actors can be modeled with information theory. Here we
|
2119
|
+
| will distinguish information transfer from information acquisition or creation (as in a lab
|
2120
|
+
| or designer’s mind).
|
2121
|
+
blank |
|
2122
|
+
|
|
2123
|
+
text | If we assume the information being transferred can be represented with i bits of
|
2124
|
+
| resolution r, then the piece of information being transferred by actor x, I(x), can be
|
2125
|
+
| represented as
|
2126
|
+
| I(x) = -iPx(xi)log2Px(xi) (1)
|
2127
|
+
blank |
|
2128
|
+
|
|
2129
|
+
text | The new (conditional) information received by actor y from the information sent by actor
|
2130
|
+
| x will be
|
2131
|
+
| I(y|x) = -iPx(xi)jPy|x(yj|xi)log2Py|x(yj|xj) (2)
|
2132
|
+
blank |
|
2133
|
+
|
|
2134
|
+
text | This equation expresses the fact that the receiver must interpret (decode) the information
|
2135
|
+
| he receives based on their previous mutual understanding.
|
2136
|
+
blank |
|
2137
|
+
|
|
2138
|
+
|
|
2139
|
+
|
|
2140
|
+
meta | 35
|
2141
|
+
title | 3.2.3 Product Entropy and Information Packages
|
2142
|
+
text | In discussing information content (I(x)) as opposed to information transfer, the entropy
|
2143
|
+
| (information) level of a possible conceptual design is very high in that there are a very
|
2144
|
+
| large number of ways the design could be implemented; different functions included;
|
2145
|
+
| different materials used; different reliabilities; etc. The entropy level of a completed
|
2146
|
+
| product is close to zero in that the design is frozen. This “zero” entropy is a relative
|
2147
|
+
| value since there is still myriad flexibility in subsequent product use, in product
|
2148
|
+
| combinations, etc. (Figure 3-3).
|
2149
|
+
blank |
|
2150
|
+
|
|
2151
|
+
|
|
2152
|
+
|
|
2153
|
+
text | Figure 3-3 Product Entropy Flexibility for Related Designs
|
2154
|
+
| Entropy (s) is very large when a design problem starts. As decisions are made, materials chosen,
|
2155
|
+
| and functionality included, entropy is reduced until there is no flexibility (s = 0) in the final
|
2156
|
+
| product. At an earlier preliminary design stage (s > 0), there is flexibility in the design to create
|
2157
|
+
| different possible variants. The set of possible variants is a Hilbert space.
|
2158
|
+
blank |
|
2159
|
+
|
|
2160
|
+
|
|
2161
|
+
|
|
2162
|
+
text | Likewise the entropy of a new technology or technique is very large. This new material
|
2163
|
+
| or new processor could be used in thousands of different products; the process could be
|
2164
|
+
| refined through cost reduction, or not, which would drive how and where it would be
|
2165
|
+
| used. Of course, in this sense the entropy will likely grow as the ideas expand.
|
2166
|
+
blank |
|
2167
|
+
|
|
2168
|
+
|
|
2169
|
+
meta | 36
|
2170
|
+
text | The entropy evolution of a product over the design life is an interesting phenomenon. If
|
2171
|
+
| you have a product that you want to incrementally improve, you would need to step back
|
2172
|
+
| some number of “computation” steps to find a sufficiently flexible design space (with
|
2173
|
+
| more entropy) that allows you to move forward with a slightly different instantiation
|
2174
|
+
| using other technology for your new product.
|
2175
|
+
blank |
|
2176
|
+
|
|
2177
|
+
text | In quantum information theory, the physical state is described by a wave function that
|
2178
|
+
| can be thought of as a vector in an abstract, multi-dimensional state, called a Hilbert
|
2179
|
+
| space. Hilbert space replaces the concept of phase space in classical mechanics. An
|
2180
|
+
| important difference is that many quantum information quantities cannot be measured
|
2181
|
+
| simultaneously. Also, as time passes and the q-bits become less deterministic, the Hilbert
|
2182
|
+
| space rapidly diverges away from what you have earlier defined.
|
2183
|
+
blank |
|
2184
|
+
|
|
2185
|
+
text | Note that another interesting phenomenon is observed in product families. Here you
|
2186
|
+
| would like to have an intermediate shelf (stage) that has enough flexibility to be able to
|
2187
|
+
| evolve in several directions, or allows a time-phased introduction corresponding to a
|
2188
|
+
| time-phased improvement of your particular core technology. The essential conundrum
|
2189
|
+
| is to have the right amount of breadth so that you can easily arrive at your next step, but
|
2190
|
+
| not too much breadth which can be confusing and hard to consider (Figure 3-4). This
|
2191
|
+
| “shelf” stage must be stored and documented.
|
2192
|
+
blank |
|
2193
|
+
|
|
2194
|
+
|
|
2195
|
+
|
|
2196
|
+
meta | 37
|
2197
|
+
text | ln(x)
|
2198
|
+
blank |
|
2199
|
+
text | t2
|
2200
|
+
blank |
|
2201
|
+
text | Design Space
|
2202
|
+
| t1
|
2203
|
+
| Q = f(Mt)
|
2204
|
+
blank |
|
2205
|
+
text | S>f
|
2206
|
+
| S >> f S = f (relative)
|
2207
|
+
blank |
|
2208
|
+
|
|
2209
|
+
|
|
2210
|
+
|
|
2211
|
+
text | Figure 3-4 Product Entropy Levels for different rates of Growth
|
2212
|
+
| Intermediate shelves (stages) have design flexibility (s > 0) to be able to evolve in several directions, or
|
2213
|
+
| allow a time-phased introduction corresponding to a time-phased improvement of your particular
|
2214
|
+
| core technology. Since the Hilbert space rapidly diverges away from what you have earlier defined,
|
2215
|
+
| the essential conundrum is to have the right amount of breadth so that you can easily arrive at your
|
2216
|
+
| next step, but not too much breadth which can be confusing and hard to consider
|
2217
|
+
blank |
|
2218
|
+
|
|
2219
|
+
|
|
2220
|
+
|
|
2221
|
+
text | Controlling the entropy level (information level) is one of the functions of organizational
|
2222
|
+
| computation.
|
2223
|
+
blank |
|
2224
|
+
|
|
2225
|
+
|
|
2226
|
+
|
|
2227
|
+
meta | 38
|
2228
|
+
title | 3.2.4 Information Growth, and Product Evolution
|
2229
|
+
blank |
|
2230
|
+
|
|
2231
|
+
text | The degree to which the state of an actor (or group) at one point in time determines the
|
2232
|
+
| state of a different actor at a subsequent point in time is driven by the mutual information
|
2233
|
+
| between these observations, and the conditional information between them. The mutual
|
2234
|
+
| information I(x/\y) = I(x) intersect I(y) is the common set of knowledge that allows them
|
2235
|
+
| to understand each other and includes the range from basic engineering equations to
|
2236
|
+
| current catalog numbers. When Actor X discovers new information I(xi+1) > I(xi) and
|
2237
|
+
| wants to transfer this information to Actor Y, he will encode it for transfer based on
|
2238
|
+
| mutual information I(xi/\y) and conditional information I(y|xi).
|
2239
|
+
blank |
|
2240
|
+
|
|
2241
|
+
text | For two correlated events x and y, the relations between the information I(x), conditional
|
2242
|
+
| information I(x|y), joint information I(x,y), and mutual information I(x/\y) are illustrated
|
2243
|
+
| in (Figure 3-5) [Metzler 2003].
|
2244
|
+
blank |
|
2245
|
+
|
|
2246
|
+
|
|
2247
|
+
|
|
2248
|
+
text | Figure 3-5 Mutual and Conditional Information
|
2249
|
+
blank |
|
2250
|
+
|
|
2251
|
+
|
|
2252
|
+
|
|
2253
|
+
meta | 39
|
2254
|
+
text | For engineering design we are interested mainly in post-diction. Post-diction is the
|
2255
|
+
| forward use of information to increase the knowledge of another actor in a strictly
|
2256
|
+
| subsequent time state. The initiator’s message is based on his current state, and the
|
2257
|
+
| receiver interprets it according to what the receiver knows. (Note that this is the inverse
|
2258
|
+
| direction to prediction (coding) wherein one could compute the desired resolution or
|
2259
|
+
| scale that the sender should use.) The mutual information I(y/\x) is simply the difference
|
2260
|
+
| between input information I(x) and the conditional information I(x|y) or I(y) and I(y|x).
|
2261
|
+
| The quality of the information received by Actor Y (at a subsequent time i + n) will be a
|
2262
|
+
| function of the accuracy of coding I(x|y) given I(x/\y). Thus
|
2263
|
+
blank |
|
2264
|
+
|
|
2265
|
+
text | I(yi+1) = I(xi/\y) + I(xi+1|y)f(I(xi/\y)) (3)
|
2266
|
+
blank |
|
2267
|
+
|
|
2268
|
+
text | Mutual information I(xi/\yi) is not static, and actors’ mutual information will decrease
|
2269
|
+
| over time if not refreshed or synchronized. This can be thought of as the actors evolving
|
2270
|
+
| separately, starting to use different materials, using different design lingo [Leifer 1999],
|
2271
|
+
| or even working on different programs or products.
|
2272
|
+
blank |
|
2273
|
+
|
|
2274
|
+
text | Given our logarithmic definition of information, we note that mutual information decays
|
2275
|
+
| to zero on the order(log r) time steps, where r is the resolution of the information.
|
2276
|
+
| Statically, without refresh, I(x/\y) will naturally approach 0 on orders similar to product
|
2277
|
+
| life, material life, and/or component life. This is similar to having a design use a
|
2278
|
+
| particular Digital Signal Processor (DSP) that goes out of production and must now be
|
2279
|
+
| replaced with a different technology. For our community, this is on the order of 5 to 12
|
2280
|
+
| years for aerospace type products. This will be discussed more later.
|
2281
|
+
blank |
|
2282
|
+
|
|
2283
|
+
text | When I(x/\y) reaches zero, it creates a wall within the complex (design community)
|
2284
|
+
| where information is not flowing. This has several effects. For one, it reduces the
|
2285
|
+
| computational efficiency of the organization. For another, it effectively creates separate,
|
2286
|
+
| competing sub-complexes. In a small scale industry this could be good as it maintains a
|
2287
|
+
| vibrant population of new small businesses. For our very large systems within our
|
2288
|
+
| constrained resources, this is not good. A third side effect that may occur as a result of
|
2289
|
+
blank |
|
2290
|
+
|
|
2291
|
+
|
|
2292
|
+
meta | 40
|
2293
|
+
text | losing mutual information coherence is that the group that is cut off may die (or become
|
2294
|
+
| irrelevant). At this point it would be consuming resources without a useful output or, if
|
2295
|
+
| recognized, it would have to be reabsorbed with the concomitant costs of re-training.
|
2296
|
+
blank |
|
2297
|
+
|
|
2298
|
+
|
|
2299
|
+
|
|
2300
|
+
meta | 41
|
2301
|
+
title | 3.3 Task Variety
|
2302
|
+
blank |
|
2303
|
+
title | 3.3.1 Variety and Information Processing
|
2304
|
+
text | Variety is defined as the set of actions needed by an actor to execute an activity, where an
|
2305
|
+
| actor can be an individual, a group of individuals, or a group of groups. The set of
|
2306
|
+
| actions includes information gathering, information creation and processing, distribution
|
2307
|
+
| of resultant information, self control, and environmental interaction with other actors.
|
2308
|
+
| Executing a task requires an actor to have sufficient variety to perform all of the
|
2309
|
+
| component activity required to execute the task.
|
2310
|
+
blank |
|
2311
|
+
|
|
2312
|
+
text | It is conventional to measure variety, like information, in logarithmic units so that the
|
2313
|
+
| total variety of a set of independent components V=log(M) is the sum of the variety of
|
2314
|
+
| the components, V=Nv, where v=log(m).
|
2315
|
+
blank |
|
2316
|
+
|
|
2317
|
+
text | The key to task analysis of the variety of any system is that each of the actors has a limit
|
2318
|
+
| on its variety – the logarithm of the number of distinguishable states. Components can be
|
2319
|
+
| individuals that act in performing tasks, or individuals that manage or coordinate tasks or
|
2320
|
+
| serve as communications channels. [Bar-Yam 2004b]
|
2321
|
+
blank |
|
2322
|
+
|
|
2323
|
+
text | Assume that the design system is composed of a number of subsystems, N, that are
|
2324
|
+
| variously coordinated to respond to external contexts. The number of possible actions
|
2325
|
+
| that the system can take, M, is not more than mN, the product of the possible actions of
|
2326
|
+
| each part, m. We further constrain the problem of effective function by assuming that
|
2327
|
+
| effective actions require a sufficient variety at each scale of action corresponding to the
|
2328
|
+
| requirements for action at that scale. At every scale the variety of the system must be
|
2329
|
+
| larger than the variety necessary for the task. [Bar-Yam 2004b]
|
2330
|
+
blank |
|
2331
|
+
|
|
2332
|
+
text | Assuming a simple coordination mechanism such that the system is partitioned into
|
2333
|
+
| groups that are fully coordinated and that different groups are independent of each other,
|
2334
|
+
| then the variety of actions of each group is the same as the variety of actions of any
|
2335
|
+
| individual of that group, and the scale of action is just the number of individuals in that
|
2336
|
+
blank |
|
2337
|
+
|
|
2338
|
+
meta | 42
|
2339
|
+
text | group. For the entire system the variety at scale k is D(k)=vn(k) where n(k) is the
|
2340
|
+
| number of different k-member fully coordinated groups needed to perform the entire task,
|
2341
|
+
| which therefore at a minimum requires N=kn(k) components to perform. The total
|
2342
|
+
| variety of the task is proportional to the total number of subsets of any scale V=D(k).
|
2343
|
+
| [Bar-Yam 2004b]
|
2344
|
+
blank |
|
2345
|
+
|
|
2346
|
+
text | With these assumptions, given a predetermined number of components N, the system
|
2347
|
+
| can, in the extreme, perform a task of scale N, with variety equal to that of one
|
2348
|
+
| component, or a task of scale one with variety N times as great.
|
2349
|
+
blank |
|
2350
|
+
|
|
2351
|
+
text | More generally the equation Nv=kD(k) can be considered a constraint on the possible
|
2352
|
+
| behaviour patterns (sum rule) of a system due to different mechanisms of organization. It
|
2353
|
+
| is often convenient to think about the variety of a system, V(k), that has a scale k or
|
2354
|
+
| larger, as this is the set of possible actions that can have at least that scale,
|
2355
|
+
blank |
|
2356
|
+
text | V(k) = Σ D(k’) (summed for k’ = k,N and 1 < k < N)
|
2357
|
+
blank |
|
2358
|
+
|
|
2359
|
+
text | Then the total variety of the system is V(1), and the sum rule can be written as:
|
2360
|
+
blank |
|
2361
|
+
text | Σ V(k) = Nv (summed for k = 1,N)
|
2362
|
+
blank |
|
2363
|
+
|
|
2364
|
+
text | This equation “describes the existence of a tradeoff between variety at different scales.
|
2365
|
+
| Increasing the variety at one scale by changing the organizational form must come at the
|
2366
|
+
| expense of variety at other scales.” [Bar-Yam 2004 p.19]
|
2367
|
+
blank |
|
2368
|
+
|
|
2369
|
+
text | Hierarchical systems have performance limits. Scale-free organizations avoid these
|
2370
|
+
| limits by adding (essentially) communication paths which avoid variety limits, but the
|
2371
|
+
| price is loss of controllability. The balance of controllability and efficiency is important,
|
2372
|
+
| especially for Mega-Systems, and provides a distinguishing characteristic from
|
2373
|
+
| Evolutionary Engineering which has been primarily developed for larger populations
|
2374
|
+
| without considering costs of scale, or even availability at scale.
|
2375
|
+
blank |
|
2376
|
+
|
|
2377
|
+
|
|
2378
|
+
|
|
2379
|
+
meta | 43
|
2380
|
+
title | 3.3.2 Variety in Distributed Product Development
|
2381
|
+
text | Performance of distributed complex product development networks has been analyzed to
|
2382
|
+
| discover the role of various types and mixes of variety in system development. For
|
2383
|
+
| instance [Braha 2007] studied relations between input and output tasks, design iterations,
|
2384
|
+
| and convergence in large distributed task networks for product design of complex
|
2385
|
+
| systems such as vehicles, hospitals, and pharmaceutical facilities. This analysis was
|
2386
|
+
| based on detailed design process models and task networks as shown in Figure 3-6.
|
2387
|
+
blank |
|
2388
|
+
|
|
2389
|
+
|
|
2390
|
+
|
|
2391
|
+
text | Figure 3-6 Design Process Models and Networks
|
2392
|
+
| This PD task network consists of 1,245 directed information flows between 466 development tasks.
|
2393
|
+
| Each task is assigned to one or more actors (“design teams” or “engineers”) who are responsible for
|
2394
|
+
| it. [Braha 2007] Input information can be ‘external’, intra-disciplinary’, or ‘cross-disciplinary’.
|
2395
|
+
| Output information moves toward the right. The primary diagram shows a closeup of three tasks
|
2396
|
+
| with 17 information streams; the inset shows the overall (quasi-scale-free) network.
|
2397
|
+
blank |
|
2398
|
+
|
|
2399
|
+
|
|
2400
|
+
meta | 44
|
2401
|
+
text | Some of the interesting results from these studies showed that limits on input versus
|
2402
|
+
| output variety had emergent effects on design convergence (Figure 3-7). In this case, one
|
2403
|
+
| would apply special attention to tasks with a large number of outputs.
|
2404
|
+
blank |
|
2405
|
+
|
|
2406
|
+
|
|
2407
|
+
|
|
2408
|
+
text | Figure 3-7 Meta-Performance Measures of Distributed Product Development
|
2409
|
+
| Research results show that limits on input versus output variety has emergent effects on design
|
2410
|
+
| convergence. In this case, one would apply special attention to tasks with a large number of outputs.
|
2411
|
+
| [Braha 2007]
|
2412
|
+
blank |
|
2413
|
+
|
|
2414
|
+
text | However, a major limit on these types of analysis is an assumption of homogenous
|
2415
|
+
| communication channels. Also these links are directed and have a single defined
|
2416
|
+
| information package. Another limit is that the transfer of information was assumed to be
|
2417
|
+
| 100% complete so there was no degradation due to scale. These limits will be addressed
|
2418
|
+
| in Section 4.0.
|
2419
|
+
blank |
|
2420
|
+
|
|
2421
|
+
|
|
2422
|
+
meta | 45
|
2423
|
+
title | 3.4 Information Transfer in Chaotic Channels
|
2424
|
+
blank |
|
2425
|
+
title | 3.4.1 Dynamical Systems and the Logistics Map
|
2426
|
+
text | Dynamical systems evolve over time. The composition/capabilities evolve as a function
|
2427
|
+
| of the previous state. One model of interest representing this behaviour is the logistic
|
2428
|
+
| equation. “The logistic equation (sometimes called the Verhulst model, logistic map, or
|
2429
|
+
| logistic growth curve) is a model of population growth first published by Pierre Verhulst
|
2430
|
+
| [Verhulst 1838]. The Verhulst model is continuous in time, but a modification of the
|
2431
|
+
| continuous equation to a discrete quadratic recurrence equation is also known as the
|
2432
|
+
| logistic model.” [http://mathworld.wolfram.com/LogisticEquation.html] The logistic
|
2433
|
+
| map is the name given to one particularly useful way of expressing the logistic model:
|
2434
|
+
blank |
|
2435
|
+
|
|
2436
|
+
text | nt+1 = (birthrate – deathrate)[knt – nt2]/k
|
2437
|
+
blank |
|
2438
|
+
|
|
2439
|
+
text | where nt is the population at the current generation and k is the carrying capacity. To
|
2440
|
+
| derive the logistic map from this model, let
|
2441
|
+
| xt = nt/k, and
|
2442
|
+
| a = (information creation rate – information annihilation rate)
|
2443
|
+
| to get the map posed as a simple iteration:
|
2444
|
+
blank |
|
2445
|
+
|
|
2446
|
+
text | xt+1 = axt (1 - xt) (4)
|
2447
|
+
blank |
|
2448
|
+
|
|
2449
|
+
text | Note that xt is a fraction of the carrying capacity (from the population model). For us,
|
2450
|
+
| this is the amount of the original information that is still understandable over the passage
|
2451
|
+
| of time as t --> t + 1. The variable ‘a’ represents the rate of net information production.
|
2452
|
+
blank |
|
2453
|
+
|
|
2454
|
+
text | The logistic map can be used to study the behaviour of complex systems over time as
|
2455
|
+
| they evolve. For our model of mega-design communities, the link performance between
|
2456
|
+
| two groups will evolve over time, causing the relationship of the two groups to evolve in
|
2457
|
+
| response. As each link in our quasi-scale-free community evolves according to its own
|
2458
|
+
| characteristics, the overall design community will evolve in a doubly complex fashion.
|
2459
|
+
blank |
|
2460
|
+
|
|
2461
|
+
meta | 46
|
2462
|
+
text | The logistic map is frequently called the chaos model, and discussions of information
|
2463
|
+
| transfer in complex systems are sometimes discussed as information transfer in chaotic
|
2464
|
+
| channels.
|
2465
|
+
blank |
|
2466
|
+
|
|
2467
|
+
text | One representation of the behaviour of our dynamical system is shown in (Figure 3-8)
|
2468
|
+
| [Mitchell 1992], which shows how the output in these systems quickly become
|
2469
|
+
| unpredictable (chaotic). The bifurcation points, where one path expands into two, are
|
2470
|
+
| strange attractors and represent areas of local stability.
|
2471
|
+
blank |
|
2472
|
+
|
|
2473
|
+
|
|
2474
|
+
text | B C
|
2475
|
+
| A
|
2476
|
+
blank |
|
2477
|
+
|
|
2478
|
+
|
|
2479
|
+
|
|
2480
|
+
text | Figure 3-8 Bifurcation Diagram
|
2481
|
+
| The bifurcation points, where one path expands into two, are strange attractors and represent
|
2482
|
+
| areas of local stability. In this diagram the areas of local stability are (A) 3.0, (B) 3.45, and (C) 3.55.
|
2483
|
+
blank |
|
2484
|
+
|
|
2485
|
+
|
|
2486
|
+
|
|
2487
|
+
text | “Note that for an N dimensional dynamical system there are N Lyapunov exponents. The
|
2488
|
+
| positive Lyapunov exponents + measure the rates of exponential divergence, and the
|
2489
|
+
| negative ones - the rates of convergence.” [Bais 2010] The sum of the positive
|
2490
|
+
| Lyapunov exponents (the metric entropy) and the information dimension can be used to
|
2491
|
+
| estimate the amount of time that postdictions remain valid.
|
2492
|
+
blank |
|
2493
|
+
|
|
2494
|
+
|
|
2495
|
+
meta | 47
|
2496
|
+
title | 3.4.2 Information Transfer in Chaotic Channels
|
2497
|
+
blank |
|
2498
|
+
|
|
2499
|
+
text | Earlier (Section 3.2), information was modeled in a general, bi-directional sense. Now
|
2500
|
+
| we want to look at the time sequence of information and how it behaves as it moves into
|
2501
|
+
| subsequent states. Conditional information transfer was defined in equation (2). Actor X
|
2502
|
+
| starts with some particular, local information iPx(xi), so that our local conditional
|
2503
|
+
| information becomes
|
2504
|
+
blank |
|
2505
|
+
|
|
2506
|
+
text | Il(y|x) = jPy|x(yj|xi)log2Py|x(yj|xj) (5)
|
2507
|
+
blank |
|
2508
|
+
|
|
2509
|
+
text | We can define a variable for the rate of information change d = dy/dx which represents
|
2510
|
+
| how the information is changed by the environment during a transfer (or over time). This
|
2511
|
+
| includes both to the dissipation rate and to the divergence rate.
|
2512
|
+
blank |
|
2513
|
+
|
|
2514
|
+
text | Assuming sufficiently fine resolution, the conditional information at a subsequent time
|
2515
|
+
| from a given initial state can be solved to be:
|
2516
|
+
blank |
|
2517
|
+
|
|
2518
|
+
text | < Il(y|x)>o = log2d + 1/(2dln2) (6)
|
2519
|
+
blank |
|
2520
|
+
|
|
2521
|
+
text | Now we subject the conditional information (equation (6)) to evolution in the logistics
|
2522
|
+
| map (equation (4)). This can be computed by numerical integration (avoiding the
|
2523
|
+
| singularity at d = 1). Results are shown in Figure 3-9 (note a = d). Even after only 3
|
2524
|
+
| iterations, it can be seen that I(y/\x) is no longer monotonic in a – several maxima of
|
2525
|
+
| conserved information can be seen. I(y|x) does not depend strongly on r, whereas the
|
2526
|
+
| other quantities include an additive term of log2r [Metzler 2003].
|
2527
|
+
blank |
|
2528
|
+
|
|
2529
|
+
|
|
2530
|
+
|
|
2531
|
+
meta | 48
|
2532
|
+
text | t=2 t=3
|
2533
|
+
blank |
|
2534
|
+
|
|
2535
|
+
|
|
2536
|
+
|
|
2537
|
+
text | Figure 3-9 Conditional, mutual, and total information for two and three time steps
|
2538
|
+
| “Here, I(y|x) represents the uncertainty generated by stretching and compressing; I(x) – I(y)
|
2539
|
+
| represents the uncertainty through shrinking of phase space; and I(x) – I(y/\x) is the average
|
2540
|
+
| information necessary to reconstruct x from knowledge of y. The mutual information I(y/\x) is the
|
2541
|
+
| amount of information about the initial state retained after the mapping [Metzler 2003].
|
2542
|
+
blank |
|
2543
|
+
|
|
2544
|
+
|
|
2545
|
+
|
|
2546
|
+
text | For intermediate time iterations (3 < t < 100), Figure 3-10 shows the mutual information
|
2547
|
+
| at various intermediate times (note a = d). Several large peaks are apparent. The wider
|
2548
|
+
| peaks in Figure 3-10 at a = 1, 3, and 3.54 correspond to the bifurcation points. There, the
|
2549
|
+
| Lyapunov exponent is 0 and deviations from the fixed point or cycle decay like a power
|
2550
|
+
| law rather than exponentially. These broader peaks indicate information transfer regimes
|
2551
|
+
| that are still working effectively, as long as you can choose the correct formatting.
|
2552
|
+
blank |
|
2553
|
+
|
|
2554
|
+
|
|
2555
|
+
|
|
2556
|
+
meta | 49
|
2557
|
+
text | B C
|
2558
|
+
| A
|
2559
|
+
blank |
|
2560
|
+
|
|
2561
|
+
|
|
2562
|
+
|
|
2563
|
+
text | Figure 3-10 Mutual information for larger iterations of time
|
2564
|
+
| The wider peaks at a = 1, 3, and 3.54 correspond to the bifurcation points. There, the Lyapunov
|
2565
|
+
| exponent is 0 and deviations from the fixed point or cycle decay like a power law rather than
|
2566
|
+
| exponentially [Metzler 2003]. These broader peaks (A, B, and C) indicate information transfer
|
2567
|
+
| regimes that are still working effectively, as long as you can choose the correct formatting.
|
2568
|
+
blank |
|
2569
|
+
|
|
2570
|
+
text | “The narrow peaks (e.g. near a = 1.3, 1.6, and 2.5) occur when the fixed point is very
|
2571
|
+
| close to the boundary between two bins, such that small deviations from the fixed point
|
2572
|
+
| lead to ambiguities in the outcome. They change positions if the binning [resolution] is
|
2573
|
+
| chosen differently.” [Metzler 2003] This is similar to data aliasing caused by information
|
2574
|
+
| transfer between groups within our design community. These effects will be described in
|
2575
|
+
| Section 4.3 - Section 4.6.
|
2576
|
+
blank |
|
2577
|
+
|
|
2578
|
+
|
|
2579
|
+
|
|
2580
|
+
meta | 50
|
2581
|
+
title | 3.4.3 Multi-scale Information Transfer
|
2582
|
+
blank |
|
2583
|
+
|
|
2584
|
+
text | Regarding information transfer between two actors, if the message received is completely
|
2585
|
+
| uncertain, the message is lost – this is often the case in chaotic systems. On the other
|
2586
|
+
| hand if all received messages appear to be the same, the recipient cannot determine what
|
2587
|
+
| message he was supposed to receive – this can occur in dissipative systems.
|
2588
|
+
blank |
|
2589
|
+
|
|
2590
|
+
text | “Dissipation and chaos have a common aspect: in both cases the dynamics makes a
|
2591
|
+
| connection between large and small scales. Chaotic dynamics amplify small differences
|
2592
|
+
| in the initial states until they reach macroscopic proportions, whereas dissipative
|
2593
|
+
| dynamics shrink differences until they vanish below the threshold of perception.” [Bais
|
2594
|
+
| 2010]
|
2595
|
+
blank |
|
2596
|
+
|
|
2597
|
+
text | For our discussion, we will customize the following definitions of variables:
|
2598
|
+
blank |
|
2599
|
+
|
|
2600
|
+
text | a = = the (designed) spread rate of information. (Too high and too much non-
|
2601
|
+
| synchronized information results in chaos; too low and dissipation (friction)
|
2602
|
+
| slows information exchange so much that there is no growth from synergy.) (This
|
2603
|
+
| is similar to the amplification parameter in logistics map terminology which is a
|
2604
|
+
| combination of growth rate and death rate.) “a” is a controllable or cultivatable
|
2605
|
+
| parameter.
|
2606
|
+
blank |
|
2607
|
+
|
|
2608
|
+
text | r = = resolution. Information packets have behaviors related to their size divided
|
2609
|
+
| by the resolution of the information. Packetization of the information at some
|
2610
|
+
| resolution is related to the size of the producing group: individual researchers or
|
2611
|
+
| other producers that have high r create so much information that the channel
|
2612
|
+
| becomes clogged. Clumping the information (to a higher scale) lets it spread
|
2613
|
+
| better because it rejects noise.
|
2614
|
+
blank |
|
2615
|
+
|
|
2616
|
+
|
|
2617
|
+
|
|
2618
|
+
meta | 51
|
2619
|
+
text | t = = time. We will calculate the individual time scales based on a static
|
2620
|
+
| dissipation time of 5 years, or 60 months or 250 weeks. In order for our scale to
|
2621
|
+
| span orders of magnitude, we will work in weeks.
|
2622
|
+
blank |
|
2623
|
+
|
|
2624
|
+
text | X = = Sender.
|
2625
|
+
| Y = = Receiver.
|
2626
|
+
blank |
|
2627
|
+
|
|
2628
|
+
text | Organizations have both peer and hierarchical levels of “production” that affect
|
2629
|
+
| information transfer and consumption. We will next define multi-scale, and then three
|
2630
|
+
| types of multi-scale parameters: organizational, task, and time scales.
|
2631
|
+
blank |
|
2632
|
+
|
|
2633
|
+
text | A) Multi-scale Defined [Bar-Yam 2004b]
|
2634
|
+
| Multi-scale effects can be pictured by imagining information at different scales
|
2635
|
+
| analogously to multi-scale real numbers. “On one hand, each real number can be
|
2636
|
+
| represented by one point on the real axis – it is a one-dimensional quantity. On the other
|
2637
|
+
| hand, in the usual descriptions (decimal, binary, etc.), real numbers are represented by a
|
2638
|
+
| set of integers that stand for different scales – the scales of 1s, 10s, 100s, etc. This makes
|
2639
|
+
| sense because it reflects what happens when some x is added to x: the digits of x are
|
2640
|
+
| strongly affected for all scales finer than the scale of x, and weakly affected for coarser
|
2641
|
+
| scales. There are two [aspects] with the usual representations: first, they are
|
2642
|
+
| discontinuous: a small change on a fine scale will have no effect at all on coarser scales
|
2643
|
+
| most of the time, but a dramatic effect in rare cases (such as when 0.001 is added to
|
2644
|
+
| 0.999). This is a necessary side effect of using discrete (integer) representations on each
|
2645
|
+
| scale. Second, they do not lend themselves to simple modification under multiplication:
|
2646
|
+
| multiplication is basically a convolution of the representations of the two factors.
|
2647
|
+
| Whereas multiplications with numbers that have a simple representation in the chosen
|
2648
|
+
| base gives a shift (e.g. multiplication with 10 in the decimal representation just shifts the
|
2649
|
+
| decimal point), all other factors lead to changes throughout the scales.”
|
2650
|
+
blank |
|
2651
|
+
|
|
2652
|
+
|
|
2653
|
+
|
|
2654
|
+
meta | 52
|
2655
|
+
title | B) Organization Scales
|
2656
|
+
text | For our engineering community, one instantiation of scale that can be easily recognized is
|
2657
|
+
| levels in an organizational hierarchy. Information being passed up and down the chain
|
2658
|
+
| gets compressed or expanded for delivery. Months’ worth of testing is compressed into a
|
2659
|
+
| report that the group delivers to their manager. The manager uses certain meta-aspects at
|
2660
|
+
| his level (budget spent, Intellectual Property (IP) filed, etc.) but has no visceral under-
|
2661
|
+
| standing of whether the new component is easy or hard to use, and whether it provides
|
2662
|
+
| the improved performance easily or only with great tweaking. The report may be filed in
|
2663
|
+
| an IP repository. Later, a design group looking for a new component accesses the reports
|
2664
|
+
| and tries to understand it enough to determine if they should use it in their product.
|
2665
|
+
blank |
|
2666
|
+
|
|
2667
|
+
text | So vertical-organization information-transfer either dissipates information (upward), or it
|
2668
|
+
| adds noise (chaos) (downward) (unless accompanied by additional (decompression)
|
2669
|
+
| information).
|
2670
|
+
blank |
|
2671
|
+
|
|
2672
|
+
text | An Organizational Aliasing Factor (OAF) is defined, and estimated to be on the order of
|
2673
|
+
| the relative sizes of the hierarchical groups. For instance, if a manager has ten
|
2674
|
+
| subordinates, then:
|
2675
|
+
| OAF = O( scale)
|
2676
|
+
| = O(10) – O(1)
|
2677
|
+
| =1
|
2678
|
+
| This factor is similar to limits on organizational controllability [Hogg 2001].
|
2679
|
+
blank |
|
2680
|
+
|
|
2681
|
+
text | C) Variety Scales
|
2682
|
+
| Cross-organization information transfer can ostensibly retain the information scale and
|
2683
|
+
| relay information at very high bandwidths without loss, although again the transfers are
|
2684
|
+
| subject to transmission effects. Cross-organization transfers are often not peer to peer,
|
2685
|
+
| and are frequently dis-intermediated by other functional groups.
|
2686
|
+
blank |
|
2687
|
+
|
|
2688
|
+
text | These agents are peer-to-peer agents in a weak sense as defined by Petrie [Petrie 1996].
|
2689
|
+
| In this sense they are true agents that provide generation of new information, but they are
|
2690
|
+
blank |
|
2691
|
+
|
|
2692
|
+
meta | 53
|
2693
|
+
text | weak in that they do not use standard protocols for communication. Each agent’s
|
2694
|
+
| protocols (methods) are particular to that set of agents.
|
2695
|
+
| Variety scale differences in peer to peer relations involve the definition of mutual
|
2696
|
+
| information needed for the information transfer. These scale differences can arise from
|
2697
|
+
| three variables: resolution, translation, and context.
|
2698
|
+
blank |
|
2699
|
+
|
|
2700
|
+
text | Resolution effects occur when the information needs to be further developed to add more
|
2701
|
+
| detailed design information.
|
2702
|
+
blank |
|
2703
|
+
|
|
2704
|
+
text | Translation effects occur when the type of information needs to be transformed. For
|
2705
|
+
| example, when a prototype’s key development focus was improving and characterizing a
|
2706
|
+
| new control system possibly using new lightweight components, and it now needs to be
|
2707
|
+
| characterized for material properties for reliability and radiation effects. The original
|
2708
|
+
| (controls) engineers may not have documented the reference information needed to start
|
2709
|
+
| to collect the information and samples for material environmental testing.
|
2710
|
+
blank |
|
2711
|
+
|
|
2712
|
+
text | Context effects involve the environmental linkages and might include introducing
|
2713
|
+
| subcontractors and transferring between points of contact within the subcontractors
|
2714
|
+
| organization.
|
2715
|
+
blank |
|
2716
|
+
|
|
2717
|
+
text | We define a Variety Aliasing Factor VAF = ln(V1/V2). Here Variety Vi is estimated to be
|
2718
|
+
| linear with the number of people in the group, although other studies [BarYam 2004a]
|
2719
|
+
| have addressed fractional variety with non-linear additive effects. VAF has been
|
2720
|
+
| computed for each link in our scale-free design community as shown in (Figure 3-11).
|
2721
|
+
| Organizational aliasing factors (OAF) are included with variety aliasing factors (VAF)
|
2722
|
+
| since each half of an explicit organizational hierarchy should be designed to have the
|
2723
|
+
| correct agents to compress / decode the information packages.
|
2724
|
+
blank |
|
2725
|
+
|
|
2726
|
+
|
|
2727
|
+
|
|
2728
|
+
meta | 54
|
2729
|
+
text | Figure 3-11 Values for Combined Organization (OAF) and Variety (VAF) Aliasing
|
2730
|
+
| Values in red are the combined OAF and VAF based on the measured organizational values shown
|
2731
|
+
| in the Mega-Organization model in Section 2.1
|
2732
|
+
blank |
|
2733
|
+
|
|
2734
|
+
|
|
2735
|
+
text | At this level of modeling, we can ignore actor interactions as interposed by variety
|
2736
|
+
| [BarYam 2004a] regarding whether they are simply coordinated, fully coordinated, or
|
2737
|
+
| independent.
|
2738
|
+
blank |
|
2739
|
+
|
|
2740
|
+
title | D) Time Scales
|
2741
|
+
text | Information in our mega-design community is produced by certain groups for certain
|
2742
|
+
| purposes. Information can be produced quickly or it may take quite some time.
|
2743
|
+
| Consumption rates will be different for different groups and, within groups, will be
|
2744
|
+
| different for different types of information. Information exchanges may be scheduled
|
2745
|
+
| processes (like quarterly reports), they may be continuous (as may occur within a lab), or
|
2746
|
+
| they may be ad hoc.
|
2747
|
+
blank |
|
2748
|
+
|
|
2749
|
+
text | During the early stages of design of a very large product, the design engineers may talk
|
2750
|
+
| with laboratory staff frequently to discover the possible span of solutions that are
|
2751
|
+
blank |
|
2752
|
+
|
|
2753
|
+
|
|
2754
|
+
meta | 55
|
2755
|
+
text | implementable. However, after the PDR (Preliminary Design Review), the design point
|
2756
|
+
| has become conceptually fixed and the designers switch from talking with scientists to
|
2757
|
+
| talking with component vendors in order to develop a detailed design. The product team
|
2758
|
+
| will not talk with the laboratory scientists until after detailed design, manufacturing,
|
2759
|
+
| integration, test, launch, and initial operations. There may be a 12 year gap before the
|
2760
|
+
| design engineers again talk with the research team. Now, assuming the laboratory is not
|
2761
|
+
| lazy, they have been producing research reports for each of the last 12 years (600 weeks)
|
2762
|
+
| which have been ignored by the engineers. This is multi-scale aliasing (Figure 3-12).
|
2763
|
+
blank |
|
2764
|
+
|
|
2765
|
+
text | BD PROG R&D
|
2766
|
+
blank |
|
2767
|
+
|
|
2768
|
+
|
|
2769
|
+
|
|
2770
|
+
text | 1 year Exec Exec Exec
|
2771
|
+
blank |
|
2772
|
+
text | 3 years
|
2773
|
+
| Adjacent BD 3 Executive
|
2774
|
+
| 3 Senior Adjacent
|
2775
|
+
| Markets Director Staff Support Technologies
|
2776
|
+
| 1.5
|
2777
|
+
| BD
|
2778
|
+
| 12 4 Advanced Senior
|
2779
|
+
| Program
|
2780
|
+
| Staff Develop Manager
|
2781
|
+
blank |
|
2782
|
+
text | 2 years
|
2783
|
+
| Senior 3 Subsystem CRAD
|
2784
|
+
| Senior
|
2785
|
+
| Engineer Scientist
|
2786
|
+
blank |
|
2787
|
+
|
|
2788
|
+
text | 12 years 2
|
2789
|
+
| Staff
|
2790
|
+
| Tools Engineer Engineer Engineer
|
2791
|
+
| Scientist
|
2792
|
+
| 3
|
2793
|
+
| 4 years Discipline
|
2794
|
+
| 1 year
|
2795
|
+
| Director
|
2796
|
+
blank |
|
2797
|
+
|
|
2798
|
+
|
|
2799
|
+
|
|
2800
|
+
text | ENG
|
2801
|
+
blank |
|
2802
|
+
|
|
2803
|
+
text | Figure 3-12 Values for Time Aliasing Factor (TAF)
|
2804
|
+
| Numbers in black are the event lifetimes for each of the red-boxed major organizational elements.
|
2805
|
+
| For instance, ‘Programs’ take 12 years to develop their mega-system; ‘Advanced Development’ takes
|
2806
|
+
| 2 years to develop a prototype. TAFs (red numbers), are calculated on links that cross major
|
2807
|
+
| boundaries, and are based on the relative values of the two intersecting time epochs.
|
2808
|
+
blank |
|
2809
|
+
|
|
2810
|
+
text | Of course the situation is not quite so simple as there are other mega-products within the
|
2811
|
+
| design community that are at different stages of development. So assuming a reasonably
|
2812
|
+
| large set of programs, there are likely other efforts that were listening during the
|
2813
|
+
| intervening 12 years. But unfortunately with large systems we do not have a normal
|
2814
|
+
blank |
|
2815
|
+
|
|
2816
|
+
|
|
2817
|
+
|
|
2818
|
+
meta | 56
|
2819
|
+
text | distribution of programs, and, operating within a quasi-scale-free organization, there are
|
2820
|
+
| other timing constraints.
|
2821
|
+
blank |
|
2822
|
+
|
|
2823
|
+
text | So the opportunity for multi-scale time aliasing is large for organizations dealing with
|
2824
|
+
| large products. There are dis-intermediating organizations that are set up to overcome
|
2825
|
+
| this time aliasing such as groups that create annual research reports from the continuous
|
2826
|
+
| science production, and groups that match business requirements to technical capabilities.
|
2827
|
+
| These groups have been inserted onto information transfer paths to reduce multi-scale
|
2828
|
+
| aliasing. Balance and effectiveness of these anti-aliasing functions can be analyzed
|
2829
|
+
| within our network.
|
2830
|
+
blank |
|
2831
|
+
|
|
2832
|
+
|
|
2833
|
+
title | 3.4.4 Combined Varietal, Organizational, and Temporal Aliasing
|
2834
|
+
text | The aliasing factors (AF) from individual scale parameters:
|
2835
|
+
| TAF = t1/t2
|
2836
|
+
| OAF = O(scale)
|
2837
|
+
| VAF = ln(V1/V2)
|
2838
|
+
| are combined according to an efficiency equation for combined link activity. The total
|
2839
|
+
| product information aliasing factor ‘p’, for each link, becomes:
|
2840
|
+
| p = π(T,O,V) = (t1/t2)*(O(scale))*(ln(v1/v2))
|
2841
|
+
blank |
|
2842
|
+
|
|
2843
|
+
text | Reference (Figure 3-13). These values will be used in subsequent performance analyses.
|
2844
|
+
blank |
|
2845
|
+
|
|
2846
|
+
|
|
2847
|
+
|
|
2848
|
+
meta | 57
|
2849
|
+
text | Figure 3-13 Network with total Product Information Aliasing Factors (p)
|
2850
|
+
| for combined Time, Organization, and Variety Factors
|
2851
|
+
blank |
|
2852
|
+
|
|
2853
|
+
|
|
2854
|
+
|
|
2855
|
+
meta | 58
|
2856
|
+
title | 3.4.5 Noise in Chaotic Channels
|
2857
|
+
text | Scale factors cause received/stored information to not equal transmitted/produced
|
2858
|
+
| information. Is it noise or is it compression or aliasing? At the level of abstraction we
|
2859
|
+
| are discussing here, it does not really matter. What matters is that we have a measure of
|
2860
|
+
| the turbulence structure in the link so we can smoothly elide.
|
2861
|
+
blank |
|
2862
|
+
|
|
2863
|
+
text | For individual link smoothing (improvement), there are questions about what kind of
|
2864
|
+
| noise and dissipation is injected, and how can it be controlled, reduced, or avoided?
|
2865
|
+
| Inversely, how can we amplify “good” information exchange? For example, synthesizing
|
2866
|
+
| information into a report or a meeting format compresses it. In order to decompress the
|
2867
|
+
| report properly, or at least better, we may want to add the engineer’s names and phone
|
2868
|
+
| numbers, or add more hotlinks to the supporting sources.
|
2869
|
+
blank |
|
2870
|
+
|
|
2871
|
+
text | The time length between preparation and consumption also causes dissipation. For
|
2872
|
+
| instance, a lab engineer currently remembers how to debug his system, but 6 months later
|
2873
|
+
| he would have to refer to his documentation and think for a while and re-analyze his
|
2874
|
+
| system. This may not be adjustable, except possibly by deciding whether to use the
|
2875
|
+
| prototype system before the inventor has time to forget.
|
2876
|
+
blank |
|
2877
|
+
title | 3.4.6 Synchronization
|
2878
|
+
text | Organizations are kept internally synchronized according to communication among the
|
2879
|
+
| actors. The rate, scale, and granularity of synchronization create a characteristic that is
|
2880
|
+
| similar to the evolution rate in a computing organism. A tightly coupled organization can
|
2881
|
+
| steer more quickly to adapt to changes. One could imagine that steering a big ship is not
|
2882
|
+
| as hard as steering a line of barges with a tug at one-end through as s-curve. Despite the
|
2883
|
+
| inertia, at least it is tightly coupled.
|
2884
|
+
blank |
|
2885
|
+
|
|
2886
|
+
|
|
2887
|
+
|
|
2888
|
+
meta | 59
|
2889
|
+
title | 4 Application of Integrated Link/Network Model
|
2890
|
+
blank |
|
2891
|
+
title | 4.1 Matching Scale Dispersion to Time Regimes
|
2892
|
+
text | We now interpret and expand the results of previous sections with a view to the problem
|
2893
|
+
| of optimal computation (communication) within our organization network. Given the
|
2894
|
+
| time scales of product delivery, tool evolution, and staff relocations defined earlier, how
|
2895
|
+
| do we select the size of groups, their relationships, and the type of information transfer
|
2896
|
+
| and storage (memory), such that we get “low-noise” information spread and growth
|
2897
|
+
| within the time frame of evolutionary design?
|
2898
|
+
blank |
|
2899
|
+
|
|
2900
|
+
text | Our goal is to foster a system whose dynamics are somewhat stable to perturbations and
|
2901
|
+
| slow compared to the time t. Remember that we are solving the problem of post diction,
|
2902
|
+
| not prediction, which is what gives us flexibility for genetic evolution and emergence
|
2903
|
+
| within this design organization. This is very significant. Prediction implies accurately
|
2904
|
+
| arriving at a predetermined optimal point. Determining the future specification is time-
|
2905
|
+
| consuming, and arriving at the point requires extra (control) resources. Post diction
|
2906
|
+
| implies efficiently using your (past) resources to move generally toward your goal.
|
2907
|
+
blank |
|
2908
|
+
|
|
2909
|
+
text | The information being received by agent Y (I(y)) is less than the original information I(x)
|
2910
|
+
| because of the dynamics of the information carrying capacity of the channel (Figure 4-1).
|
2911
|
+
| The received information is a function of I(x^y) and I(x|y). There are certain exchange
|
2912
|
+
| resonances where the information scale matches the transmission scale, and the
|
2913
|
+
| information is not distorted.
|
2914
|
+
blank |
|
2915
|
+
|
|
2916
|
+
|
|
2917
|
+
|
|
2918
|
+
meta | 60
|
2919
|
+
| 7
|
2920
|
+
blank |
|
2921
|
+
text | 6 Good Information
|
2922
|
+
| 5
|
2923
|
+
| Postdiction
|
2924
|
+
| Information (bits)
|
2925
|
+
blank |
|
2926
|
+
text | 4 I(x)
|
2927
|
+
blank |
|
2928
|
+
text | 3 I(y)
|
2929
|
+
| I(y^x)
|
2930
|
+
| 2
|
2931
|
+
| I(y|x)
|
2932
|
+
| 1
|
2933
|
+
blank |
|
2934
|
+
text | 0
|
2935
|
+
| 0.0 1.0 2.0 3.0 4.0
|
2936
|
+
| aa
|
2937
|
+
blank |
|
2938
|
+
text | Figure 4-1 Mutual, Conditional, and Postdicted (I(y)) Information
|
2939
|
+
| Information evolution after three time steps. Despite the rapid loss of mutual information I(y^x)
|
2940
|
+
| within the first several time steps, there is still enough conditional information I(y|x) in the higher ‘a’
|
2941
|
+
| regions to allow good information postdiction I(y) near a=4.
|
2942
|
+
blank |
|
2943
|
+
text | “For very short times (t < 3), Figure 4-1 shows that agent Y has the best chances of
|
2944
|
+
| postdicting the initial state in the chaotic regime near a = 4. “The loss of information
|
2945
|
+
| through chaos is not as significant as that through phase space shrinking in the low-a-
|
2946
|
+
| regime.” [Metzler 2003] These activities are similar to a technical symposium, where the
|
2947
|
+
| information exchange benefits are a massive (iconic) exchange of new techniques and
|
2948
|
+
| results. Information transfer is better when a range of engineers attend a conference
|
2949
|
+
| rather than ordering a copy of the proceedings and mailing it around to a distribution list
|
2950
|
+
| (given that cost is a separate variable).
|
2951
|
+
blank |
|
2952
|
+
|
|
2953
|
+
text | “In the limit of very long times, when the system has converged to its attractors, the only
|
2954
|
+
| regime where information about the initial state persists is the bifurcation regimes. What
|
2955
|
+
| value of “a” gives optimal transmission depends on the resolution: the recipient has to be
|
2956
|
+
| able to resolve all branches of the cycle to make full use of the remaining information.”
|
2957
|
+
| [Metzler 2003] This implies the information to be transferred must be completely
|
2958
|
+
| defined in a static form with all assumptions and requirements documented; with the
|
2959
|
+
| trade space documented; performance results parameterized; and consistent with future
|
2960
|
+
| engineering and IT tools. Thus for large t (t > 100), we need to have a = 1. In other
|
2961
|
+
blank |
|
2962
|
+
|
|
2963
|
+
|
|
2964
|
+
meta | 61
|
2965
|
+
text | words, all details need to be documented in long-term memory as otherwise there will be
|
2966
|
+
| insufficient knowledge to interpret the information.
|
2967
|
+
blank |
|
2968
|
+
|
|
2969
|
+
text | The Saturn V rocket used to launch the moon missions is an excellent example. Even
|
2970
|
+
| though the rocket was produced for ten years during the 1960s, some items were not
|
2971
|
+
| documented, such that when the space program was being restarted in the 1980s and
|
2972
|
+
| large lift capacity was needed again, the Saturn V could not be reproduced, and a whole
|
2973
|
+
| new design cycle had to be started anew. In fact, forty years later, we are still in the
|
2974
|
+
| midst of a ten-year process to design and qualify the new Space Launch System (SLS).
|
2975
|
+
blank |
|
2976
|
+
|
|
2977
|
+
text | For intermediate times (10 < t < 80)), at high “a”, chaotic dynamics eliminates all
|
2978
|
+
| information about the initial state; so does fast convergence to a single fixed point for
|
2979
|
+
| small “a”. As (Figure 4-2) shows, the quality of agent Y’s mutual information I(x/\y) is
|
2980
|
+
| greatest if “a” is close to one of the bifurcation points ( approximately 3 or 3.5), where
|
2981
|
+
| convergence follows a power law rather than an exponential. (Remember from Section
|
2982
|
+
| 3.4.1 that this is where the solutions change form.)
|
2983
|
+
blank |
|
2984
|
+
|
|
2985
|
+
text | Available/current link transfer rates must match information package delivery type
|
2986
|
+
| (resolution), delivery rate, and decoding capability.
|
2987
|
+
| 7
|
2988
|
+
| B
|
2989
|
+
| 6
|
2990
|
+
| A C
|
2991
|
+
| Mutual Information I(x^y)
|
2992
|
+
blank |
|
2993
|
+
|
|
2994
|
+
|
|
2995
|
+
|
|
2996
|
+
text | 5
|
2997
|
+
blank |
|
2998
|
+
text | 4
|
2999
|
+
| t = 10
|
3000
|
+
| 3 t = 20
|
3001
|
+
| t = 40
|
3002
|
+
| 2
|
3003
|
+
blank |
|
3004
|
+
text | 1
|
3005
|
+
blank |
|
3006
|
+
text | 0
|
3007
|
+
| 0.0 1.0 2.0 3.0 4.0
|
3008
|
+
| a
|
3009
|
+
blank |
|
3010
|
+
|
|
3011
|
+
|
|
3012
|
+
text | Figure 4-2 Mutual information for larger numbers of iterations
|
3013
|
+
| The loss of mutual information I(x^y) becomes so significant over longer times that there are only
|
3014
|
+
| a few “resonant” regions that allow good information transfer. Information can be exchanged
|
3015
|
+
| more completely at certain periods of time in certain formats at the resonant nodes B and C on the
|
3016
|
+
| graph where a = 3, and 3.5. Node A has low conditional information, and is unrecoverable despite
|
3017
|
+
| reasonable mutual information.
|
3018
|
+
blank |
|
3019
|
+
meta | 62
|
3020
|
+
title | 4.2 Identification of Candidate Links, Nodes, and Chains
|
3021
|
+
text | Network performance is evaluated by looking for link/information mismatch. In (Figure
|
3022
|
+
| 4-3), the red circles are evaluated for consistency. The checkmarks indicate consistent
|
3023
|
+
| link parameters and information formatting. The sad-faces indicate mismatches that will
|
3024
|
+
| be discussed in subsequent sections.
|
3025
|
+
blank |
|
3026
|
+
|
|
3027
|
+
text | The product information aliasing factor ‘p’, related to the combination of all the time and
|
3028
|
+
| variety scale factors (Figure 3-13), can be placed into three bins:
|
3029
|
+
| s – small (1-2) (chaotic)
|
3030
|
+
| m – medium (3)
|
3031
|
+
| l – large (4-6) (dissipative)
|
3032
|
+
blank |
|
3033
|
+
|
|
3034
|
+
text | These time regimes address the rate of change of I(x/\y) which is essentially the
|
3035
|
+
| differential growth rate (in different directions as well as vector magnitudes) of I(x) and
|
3036
|
+
| I(y). See (Figure 4-3). The ‘small’ total-aliasing region should be in the broadband
|
3037
|
+
| transfer regime (a=4). ‘Large’ (l) and ‘medium’ (m) total-aliasing factors represent
|
3038
|
+
| dissipative regions that require defined communications.
|
3039
|
+
blank |
|
3040
|
+
|
|
3041
|
+
text | To modify or adjust the aliasing, you could change the number of people on the link, or
|
3042
|
+
| change the percent of people in the transceivers (delta variety ‘receivers’) (V).
|
3043
|
+
blank |
|
3044
|
+
|
|
3045
|
+
|
|
3046
|
+
|
|
3047
|
+
meta | 63
|
3048
|
+
text | BD PROG R&D
|
3049
|
+
blank |
|
3050
|
+
|
|
3051
|
+
|
|
3052
|
+
text | Exec Exec Exec
|
3053
|
+
blank |
|
3054
|
+
text | m=3
|
3055
|
+
| Adjacent BD
|
3056
|
+
| m=3
|
3057
|
+
| a=3 √ Executive
|
3058
|
+
| a=1 Senior Adjacent
|
3059
|
+
| Markets Director Staff Support Technologies
|
3060
|
+
blank |
|
3061
|
+
|
|
3062
|
+
text | l=5 m=3
|
3063
|
+
| BD
|
3064
|
+
| Program a=1 a=1 Advanced Senior
|
3065
|
+
| Staff m=3 Develop Manager
|
3066
|
+
| a=3 √ √
|
3067
|
+
| s=2
|
3068
|
+
| Senior
|
3069
|
+
| Subsystem CRAD
|
3070
|
+
| √ a=4 Senior
|
3071
|
+
| Tools Engineer m=3 Scientist
|
3072
|
+
| a=3 √
|
3073
|
+
| s=2
|
3074
|
+
| √ a=4 Staff
|
3075
|
+
| s=2 Engineer m=3 Engineer Engineer
|
3076
|
+
| Scientist
|
3077
|
+
| a=1 a=3 √
|
3078
|
+
| Discipline
|
3079
|
+
| Director
|
3080
|
+
blank |
|
3081
|
+
|
|
3082
|
+
|
|
3083
|
+
|
|
3084
|
+
text | ENG
|
3085
|
+
blank |
|
3086
|
+
|
|
3087
|
+
|
|
3088
|
+
text | Figure 4-3 Information Transfer Regimes and Amplification Factors
|
3089
|
+
| Network performance is evaluated by looking for link/information mismatch. The red circles are
|
3090
|
+
| points of interest that have been evaluated for consistency. The checkmarks indicate consistent link
|
3091
|
+
| parameters and information formatting. The sad-faces indicate mismatches that will be discussed in
|
3092
|
+
| subsequent sections.
|
3093
|
+
blank |
|
3094
|
+
|
|
3095
|
+
|
|
3096
|
+
|
|
3097
|
+
meta | 64
|
3098
|
+
text | The (environmental) transmission/transfer rate of information is represented by ‘a’. This
|
3099
|
+
| is proportional to f(birthrate – deathrate), and is similar to the net information creation
|
3100
|
+
| rate or circulation rate (recall section 3.4.1).
|
3101
|
+
blank |
|
3102
|
+
|
|
3103
|
+
text | The birthrate could be about 1 month in lab; or about 2 years in contracted research and
|
3104
|
+
| development (CR&D) if you assume information is transferred only when the prototype
|
3105
|
+
| is delivered. This also implies 3months for technology management reports and one year
|
3106
|
+
| for internal research and development (IR&D) management planning. The deathrate is
|
3107
|
+
| somewhat constant and could be proportional to engineering re-training rate (~5yrs).
|
3108
|
+
| Here, retraining might imply new tools, etc.
|
3109
|
+
blank |
|
3110
|
+
|
|
3111
|
+
text | In (Figure 4-3), on the link between “BD Director” and “Prog Executive Staff”, a = 3 as
|
3112
|
+
| there are monthly program coordination meetings and customer meetings every four
|
3113
|
+
| weeks. Between “R&D Senior Support” and “Prog Executive Staff”, a = 1 representing
|
3114
|
+
| the annual funding cycle, and the annual report. On the two links between “R&D” and
|
3115
|
+
| “CR&D Prototyping”, the teams are working shoulder to shoulder with constant
|
3116
|
+
| information exchange and a = 4. The “Operational design team” has monthly design
|
3117
|
+
| reviews with the “Engineering process control staff” for a = 3; in “Engineering Tool
|
3118
|
+
| Development”, it is rare to have a faster update rate than once per year or 50 weeks, so a
|
3119
|
+
| = 1. Finally, the delivery of the “CR&D Prototype” to the “Prog Executive Staff” only
|
3120
|
+
| occurs every 100 weeks, so a = 1.
|
3121
|
+
blank |
|
3122
|
+
|
|
3123
|
+
text | The resolution ‘r’ which represents the granularity of the information package, whether
|
3124
|
+
| delivered face-to-face or in a document, is not a driver. There is only the requirement
|
3125
|
+
| that you need a larger ‘r’ for larger ‘a’.
|
3126
|
+
blank |
|
3127
|
+
|
|
3128
|
+
|
|
3129
|
+
|
|
3130
|
+
meta | 65
|
3131
|
+
title | 4.3 Network Improvements
|
3132
|
+
text | Network improvements can be assessed both at the individual links and nodes, and over
|
3133
|
+
| “link-chains”. Between individual nodes, mismatches of Link parameters (p, a) indicate
|
3134
|
+
| a discontinuity, or a mismatch of Lyupanov exponents. Two nodes can improve their
|
3135
|
+
| combined performance by exchanging information more efficiently. However, fixing one
|
3136
|
+
| link may or may not improve overall system performance, although there will certainly
|
3137
|
+
| be local effects.
|
3138
|
+
blank |
|
3139
|
+
|
|
3140
|
+
text | Link-chains of interest focus on particular extended pathways that execute higher level
|
3141
|
+
| functions such as translating customer product requirements into research needs.
|
3142
|
+
| Evaluation of various link chains can be analyzed both from an a priori interest as well as
|
3143
|
+
| from evaluating the scale-free network for paths of poor quality.
|
3144
|
+
blank |
|
3145
|
+
|
|
3146
|
+
text | We will first address Link improvements, then Chain improvements in Section 4.3.1 and
|
3147
|
+
| Section 4.3.2, respectively.
|
3148
|
+
blank |
|
3149
|
+
|
|
3150
|
+
|
|
3151
|
+
|
|
3152
|
+
meta | 66
|
3153
|
+
title | 4.3.1 Link Discontinuity Adjustment
|
3154
|
+
text | (Figure 4-4) zooms into our design community network to focus on two identified
|
3155
|
+
| discontinuities labeled A (sad faces in Figure 4-3). In both of these cases, t = m and a =
|
3156
|
+
| 1. The communication timeframe is “medium”, but the information delivered is in the
|
3157
|
+
| BD PROG
|
3158
|
+
| wrong form (a=1) for direct consumption. Also variety was expended at some costR&D
|
3159
|
+
| to
|
3160
|
+
| transform the original information (through ‘Executive Staff’ at B), so the resources used
|
3161
|
+
| to produce the original information are not being exploited efficiently.
|
3162
|
+
| Exec Exec Exec
|
3163
|
+
blank |
|
3164
|
+
|
|
3165
|
+
|
|
3166
|
+
text | t=m
|
3167
|
+
| BD
|
3168
|
+
| B Executive a= 1 Senior Adjacent
|
3169
|
+
| A
|
3170
|
+
| Director Staff Support Technologies
|
3171
|
+
| t=m D
|
3172
|
+
| A a= 1 C
|
3173
|
+
| BD C Advanced Senior
|
3174
|
+
| Program
|
3175
|
+
| Staff Develop Manager
|
3176
|
+
| C D
|
3177
|
+
| D
|
3178
|
+
blank |
|
3179
|
+
text | Senior Senior
|
3180
|
+
| Subsystem CRAD
|
3181
|
+
| Engineer Scientist
|
3182
|
+
| Figure 4-4 Set of Consistent Guides to Repair Discontinuity
|
3183
|
+
| This figure zooms into our design community network to focus on two identified discontinuities
|
3184
|
+
| Staff is
|
3185
|
+
| labeled A. In both of these cases, p = m and a = 1. So although the communication timeframe
|
3186
|
+
| Engineer Engineer Engineer
|
3187
|
+
| “medium”, the information delivered is in the wrong form for direct consumption. Scientist
|
3188
|
+
blank |
|
3189
|
+
text | Adjusting the information transfer regime ‘a’ to 3 for the two links labeled A would
|
3190
|
+
| Discipline improve the performance of this system by matching the link parameters. However, this
|
3191
|
+
| Director increase in bandwidth (changing ‘a’ from 1 to 3 on two links) must be absorbed by the
|
3192
|
+
| consuming agent B in order to maintain overall continuity.
|
3193
|
+
blank |
|
3194
|
+
|
|
3195
|
+
text | ENG The potential information overload at B (from increasing the bandwidth handled) must be
|
3196
|
+
| digested by appropriately increasing the Variety (to decode, process, and re-encode).
|
3197
|
+
| This could be managed by adding more staff within agent B (Executive Staff), but this is
|
3198
|
+
blank |
|
3199
|
+
|
|
3200
|
+
|
|
3201
|
+
meta | 67
|
3202
|
+
text | bad for several reasons: first, there are two links so you would need to add two additional
|
3203
|
+
| sets; second, the difference in variety is 2 (ln[v1/v2]) which would require significant
|
3204
|
+
| additional staff to overcome those magnitudes; and third, the overall link chain would
|
3205
|
+
| still be problematic (see section 4.3.2).
|
3206
|
+
blank |
|
3207
|
+
|
|
3208
|
+
text | A better adjustment would be to reconnect the “Advanced Development Group” D with
|
3209
|
+
| two new links at C . This matches the agent’s time scales (t = m), has a smaller change in
|
3210
|
+
| variety (∆V = O(1)), and addresses overall link chain performance.
|
3211
|
+
blank |
|
3212
|
+
|
|
3213
|
+
text | The final step is to add more Variety at all three D so that the new information flow paths
|
3214
|
+
| can be absorbed.
|
3215
|
+
blank |
|
3216
|
+
|
|
3217
|
+
|
|
3218
|
+
|
|
3219
|
+
meta | 68
|
3220
|
+
title | 4.3.2 Link-Chain Improvement
|
3221
|
+
text | Link-chains within the extended design community organization can be arbitrarily long.
|
3222
|
+
| This is not necessarily an adverse condition, as long as the information is not corrupted
|
3223
|
+
| and as long as the information arrives in a timely fashion. In fact, having a quasi-free
|
3224
|
+
| scale organization implies some limits on chain lengths (because of the cross
|
3225
|
+
| connections).
|
3226
|
+
blank |
|
3227
|
+
|
|
3228
|
+
text | However, by having imposed our external goals/specifications, and our multi-hierarchical
|
3229
|
+
| control structure, there are directed links corresponding to control paths required to focus
|
3230
|
+
| the community on our particular future goal states. These link-chains need to be
|
3231
|
+
| evaluated to make sure the information arrives timely and un-corrupted.
|
3232
|
+
blank |
|
3233
|
+
|
|
3234
|
+
text | Five link chains of interest are computed below. The link-chain “performance” value is a
|
3235
|
+
| linear sum of the included links, each of which is composed of variety and other multi-
|
3236
|
+
| scale terms. For labeling, the start- and end- Agents are listed, and a representative name
|
3237
|
+
| is added in between. Higher values imply more chances for information corruption.
|
3238
|
+
blank |
|
3239
|
+
|
|
3240
|
+
text | (1) ENG -> tools selection and deployment -> PROG
|
3241
|
+
| 2 + 2 + 3 + 3 = 10
|
3242
|
+
| (2) RES -> technology development -> PROG
|
3243
|
+
| 2 + 2 + 2 + 1 + 3 + 5 = 15
|
3244
|
+
| (3) BD -> customer requested research -> RES
|
3245
|
+
| 2 + 3 + 3 + 1 + 2 + 2 = 13
|
3246
|
+
| (4) PROG -> follow-on platforms -> PROG
|
3247
|
+
| 12 + 3 = 15
|
3248
|
+
| (5) RES -> prototype development and deployment -> LOB
|
3249
|
+
| 2 + 2 + 2 + 3 + 5 = 15
|
3250
|
+
blank |
|
3251
|
+
|
|
3252
|
+
text | (Figure 4-5) shows these link-chain paths (directed graphs) within our scale-free design
|
3253
|
+
| community network.
|
3254
|
+
blank |
|
3255
|
+
|
|
3256
|
+
|
|
3257
|
+
|
|
3258
|
+
meta | 69
|
3259
|
+
text | BD PROG R&D
|
3260
|
+
blank |
|
3261
|
+
|
|
3262
|
+
|
|
3263
|
+
text | Exec 3 Exec Exec
|
3264
|
+
blank |
|
3265
|
+
text | 1 2 1 1
|
3266
|
+
| Adjacent BD Executive
|
3267
|
+
| 3 Senior Adjacent
|
3268
|
+
| Markets Director Staff Support Technologies
|
3269
|
+
| 3 2
|
3270
|
+
| 2 2 5
|
3271
|
+
| BD
|
3272
|
+
| Staff
|
3273
|
+
| 3 Program
|
3274
|
+
| 5
|
3275
|
+
| Advanced
|
3276
|
+
| Develop
|
3277
|
+
| Senior
|
3278
|
+
| Manager
|
3279
|
+
blank |
|
3280
|
+
text | 2 2 2
|
3281
|
+
| Senior 3 Subsystem CRAD
|
3282
|
+
| 2 Senior 3
|
3283
|
+
| Engineer Scientist
|
3284
|
+
blank |
|
3285
|
+
text | 2 2 3 2 22
|
3286
|
+
| Tools Engineer
|
3287
|
+
| 3 1
|
3288
|
+
| Engineer Engineer
|
3289
|
+
| 2 Staff
|
3290
|
+
| 5 Scientist
|
3291
|
+
| 1
|
3292
|
+
| 2 Discipline
|
3293
|
+
| Director
|
3294
|
+
blank |
|
3295
|
+
|
|
3296
|
+
|
|
3297
|
+
|
|
3298
|
+
text | ENG
|
3299
|
+
blank |
|
3300
|
+
|
|
3301
|
+
text | Figure 4-5 Link-chain Paths of Interest
|
3302
|
+
| External goals and specifications, combined with our quasi, scale-free network, create directed links
|
3303
|
+
| corresponding to control paths required to focus the community on the particular future goal states.
|
3304
|
+
| Numbered blue lines are directed link paths; red numbers are total aliasing factors (TAF) (section
|
3305
|
+
| 3.4.4). These link-chains need to be evaluated to make sure the information arrives timely and un-
|
3306
|
+
| corrupted. Corruption can be caused by aliasing, occurring either during variety compression /
|
3307
|
+
| decoding, or from multi-scale time-aliasing. Case (2), the delivery of new Research products into
|
3308
|
+
| large Programs, has the highest TAF corruption factor of 15.
|
3309
|
+
blank |
|
3310
|
+
|
|
3311
|
+
text | Case (2), the delivery of new Research products into large Programs, has the highest
|
3312
|
+
| corruption factor of 15. Changes from annealing the discontinuities discussed earlier in
|
3313
|
+
| Section 4.3.1 will reduce the link values (aliasing factors) of 5 and 3 to 2 and 2.
|
3314
|
+
blank |
|
3315
|
+
|
|
3316
|
+
text | In addition to making links more efficient, one can also remove links entirely by
|
3317
|
+
| implementing organizational or process changes. An example of this, with emergent side
|
3318
|
+
| effects, will be discussed in Section 4.4.
|
3319
|
+
blank |
|
3320
|
+
|
|
3321
|
+
text | Another perspective, when looking at an extended chain as opposed to an individual link,
|
3322
|
+
| is to adjust link transfer rates (p * a) to all be similar. This similarity indicates similar
|
3323
|
+
| efficiencies through the process so that one poor link doesn’t downgrade the values of the
|
3324
|
+
blank |
|
3325
|
+
|
|
3326
|
+
|
|
3327
|
+
meta | 70
|
3328
|
+
text | attached links. This can be thought of as ensuring the transfer rate of the program staff
|
3329
|
+
| (variety providers) is on the same order as the learning rate of the program engineers who
|
3330
|
+
| are applying the information to the product, although this doesn’t address whether “V” is
|
3331
|
+
| correct.
|
3332
|
+
blank |
|
3333
|
+
|
|
3334
|
+
|
|
3335
|
+
|
|
3336
|
+
meta | 71
|
3337
|
+
title | 4.4 Cultivated Changes and Observed Emergent Side-Effects
|
3338
|
+
text | One limit in many organizations is how internal and external research is incorporated into
|
3339
|
+
| defined programs delivering products. Delivering major products that are pre-specified
|
3340
|
+
| in contracts does not need any new research, so program managers would rather control
|
3341
|
+
| internal technology resources to solve contract problems rather than create more new
|
3342
|
+
| technologies.
|
3343
|
+
blank |
|
3344
|
+
|
|
3345
|
+
text | This is exactly what is seen comparing a 90/5/5 portfolio to a 50/30/20 portfolio in the
|
3346
|
+
| time distribution of R&D allocation. Here, the numbers represent the percent of funds
|
3347
|
+
| devoted to “current-technology” / “product-development&qualification-efforts” /
|
3348
|
+
| “research-efforts”.
|
3349
|
+
blank |
|
3350
|
+
|
|
3351
|
+
text | Because program managers are not receiving any immediate value from the longer term
|
3352
|
+
| investments they try to reallocate all (90%) the R&D money internal to their product
|
3353
|
+
| development and production. Since the program managers control the source of income,
|
3354
|
+
| they have a large say in how internal funds are allocated. This is more efficient for their
|
3355
|
+
| individual program, but puts the company into a weaker long-term competitive posture
|
3356
|
+
| (slower evolution rate).
|
3357
|
+
blank |
|
3358
|
+
|
|
3359
|
+
text | A 50/30/20 distribution places more emphasis on research and development and enables a
|
3360
|
+
| more forward-leaning product offering that adapts more quickly and takes advantage of
|
3361
|
+
| recent advances in technology and new lower cost methods.
|
3362
|
+
blank |
|
3363
|
+
|
|
3364
|
+
text | For the network model of our particular design community, looking at product
|
3365
|
+
| development, the organization had evolved into a 60/5/35 investment position; value-
|
3366
|
+
| delivery of research into programs was close to zero (separate from experiential
|
3367
|
+
| development side-effects); and product evolution was stale. The significant research
|
3368
|
+
| effort was not being transferred anyplace.
|
3369
|
+
blank |
|
3370
|
+
|
|
3371
|
+
text | In this particular case we discover:
|
3372
|
+
blank |
|
3373
|
+
|
|
3374
|
+
|
|
3375
|
+
meta | 72
|
3376
|
+
text | (6) LOB -> product development -> PROG
|
3377
|
+
meta | 2 + 3 + 5 = 10
|
3378
|
+
text | This implies there are some impediments to transition new technologies into qualified
|
3379
|
+
| products for the new mega-systems.
|
3380
|
+
blank |
|
3381
|
+
|
|
3382
|
+
text | To increase innovation of the income-producing product stream, the investment position
|
3383
|
+
| was changed to 60/30/10 to increase product transition (30%), with the understanding (or
|
3384
|
+
| assumption) that many of the side-benefits accruing from a robust R&D program would
|
3385
|
+
| be lost (amount of research reduced from 35% to 10%). The implementation method was
|
3386
|
+
| to supply internal development funds directly to the Engineering subsystems to improve
|
3387
|
+
| their products for the set of Programs that were being developed.
|
3388
|
+
blank |
|
3389
|
+
|
|
3390
|
+
text | To maintain organizational coherence, new communication pathways had to be
|
3391
|
+
| developed. Allocation of the development funds to subsystem engineering groups was
|
3392
|
+
| equivalent to doubling their task diversity. There were now two roles for these
|
3393
|
+
| engineering groups: subassembly improvement with links to Research, and subassembly
|
3394
|
+
| integration with links to the Programs.
|
3395
|
+
blank |
|
3396
|
+
|
|
3397
|
+
text | A snapshot of the new information flow network is shown in (Figure 4-6) on the left.
|
3398
|
+
| The right side shows the diagram with an updated arrangement. This evolved version can
|
3399
|
+
| now be evaluated for future cultivation (the blue portion is new).
|
3400
|
+
blank |
|
3401
|
+
|
|
3402
|
+
|
|
3403
|
+
|
|
3404
|
+
meta | 73
|
3405
|
+
text | Figure 4-6 Cultivated Organizational Change
|
3406
|
+
| Adding resources and responsibility to ‘Engineering’ (Senior Engineers and Engineers) allowed them
|
3407
|
+
| to increase their variety both to expand their design work, and also to work the transitions into the
|
3408
|
+
| Program and from Research. Allocation of the development funds to subsystem engineering groups
|
3409
|
+
| was equivalent to doubling their task diversity. Two new links for these engineering groups were
|
3410
|
+
| required: subassembly improvement with links to Research, and subassembly integration with links
|
3411
|
+
| to the Programs. A second equivalent diagram is shown on the right. Either can be used depending
|
3412
|
+
| on the purpose and resolution of the subsequent analysis.
|
3413
|
+
blank |
|
3414
|
+
|
|
3415
|
+
|
|
3416
|
+
|
|
3417
|
+
text | Two interesting side-effects emerged. One of these might have been predicted in that the
|
3418
|
+
| Program Subsystem actor now had fewer responsibilities. In essence, the actor could
|
3419
|
+
| discard the variety needed to lead a new design effort, and focus on system integration.
|
3420
|
+
blank |
|
3421
|
+
|
|
3422
|
+
text | Another more interesting result was the effect this had on the Engineering Subsystem
|
3423
|
+
| teams. Previously they had no responsibility for their products other than to deliver their
|
3424
|
+
| design within the particular program requirements. Now, as they were tasked to improve
|
3425
|
+
| the status quo and coordinate its insertion into the program, their engineering behaviors
|
3426
|
+
| emerged into product ownership with the related activity that they desired to see ‘their’
|
3427
|
+
| product evolve and get inserted into additional programs. This was a significant
|
3428
|
+
| evolution of the company engineering culture which might be subsequently applied to
|
3429
|
+
| other evolutionary threads as indicated by the many variations of product strategy.
|
3430
|
+
blank |
|
3431
|
+
|
|
3432
|
+
text | Note that the resulting information transfer formats are still evolving.
|
3433
|
+
blank |
|
3434
|
+
|
|
3435
|
+
|
|
3436
|
+
|
|
3437
|
+
meta | 74
|
3438
|
+
title | 5 Performance Observations and Implications
|
3439
|
+
text | A large corporation was modeled as an adaptive computing organism that creates new
|
3440
|
+
| assets (mega-platforms). We are trying to make this (organizational) computing more
|
3441
|
+
| efficient so we can get more output so we can achieve better survivability (for the
|
3442
|
+
| organization) and better social value (for the environment). This section identifies areas
|
3443
|
+
| of potential benefits and changes.
|
3444
|
+
blank |
|
3445
|
+
|
|
3446
|
+
|
|
3447
|
+
title | 5.1 R&D Applicability Performance
|
3448
|
+
text | While emergent engineering behaviours are interesting, another question is whether this
|
3449
|
+
| improved the value to the company and/or the value delivered to the environment
|
3450
|
+
| (society). These questions may not be answerable, but one metric – the inserted value of
|
3451
|
+
| R&D efforts – rose from less than 5% to almost 80% as measured by the number of
|
3452
|
+
| research projects being fully used in an operational program.
|
3453
|
+
blank |
|
3454
|
+
|
|
3455
|
+
text | How well are typical research-expenditures used? This is a good question since there are
|
3456
|
+
| currently no good metrics and there is currently no generally accepted technique. People
|
3457
|
+
| have been looking for a long time, but nothing seems to work. This may be because the
|
3458
|
+
| desired answer doesn’t match the computed answer. Common methods of calculation
|
3459
|
+
| give extremely low values (<5% (?!)), indicating very low rates of return on invested
|
3460
|
+
| funds.
|
3461
|
+
blank |
|
3462
|
+
|
|
3463
|
+
text | In fact, it is very hard to discover any direct relation between research investments and
|
3464
|
+
| new product delivery. Many people would agree that the return on investment is mostly
|
3465
|
+
| in the experience gained by the engineers.
|
3466
|
+
blank |
|
3467
|
+
|
|
3468
|
+
text | Following the framework developed in Section 2.0 through Section 4.0 indicates that less
|
3469
|
+
| than 20% of R&D is directly useful (in the aerospace community). The research projects
|
3470
|
+
| themselves can be quite successful, but information transfer to a development program
|
3471
|
+
| runs into the problems indicated in terms of time synchronization. Reports may be
|
3472
|
+
blank |
|
3473
|
+
|
|
3474
|
+
|
|
3475
|
+
meta | 75
|
3476
|
+
text | prepared every year at the end of each research project, but the program engineers only
|
3477
|
+
| look to see what new technologies are available at the start of a new program. A
|
3478
|
+
| particular system and/or upgrade may start anew once every fifteen years (twelve years
|
3479
|
+
| development, plus three years of initial operations). Even within a business or product
|
3480
|
+
| area that averages three large systems in concurrent development, the average time
|
3481
|
+
| between new starts would be five years. Thus the research results for four years are put
|
3482
|
+
| on the shelf, and only in the fifth year do application engineers research new approaches
|
3483
|
+
| for their program (1/5 = 20%).
|
3484
|
+
blank |
|
3485
|
+
|
|
3486
|
+
|
|
3487
|
+
text | 5.2 Estimating the Amount of Communication you Should Afford
|
3488
|
+
| When you change the information transfer rate a, you must also adjust the variety at both
|
3489
|
+
| the receiving and transmitting agents. Otherwise the agents will suffer an information
|
3490
|
+
| overload, and you will still have inefficient information transfer. It may be that the cost
|
3491
|
+
| of the extra staff needed to increase the variety of the agent is more than the value of the
|
3492
|
+
| improved information transfer.
|
3493
|
+
blank |
|
3494
|
+
|
|
3495
|
+
text | The value of the improved information transfer is the benefit minus the cost.
|
3496
|
+
| Value = Benefit – Cost = Benefit – (changes in variety at all nodes)
|
3497
|
+
blank |
|
3498
|
+
|
|
3499
|
+
text | In the simplest example, when passing information from one peer actor to another,
|
3500
|
+
| instead of vertical information transfers (up and down) through a common management
|
3501
|
+
| actor, you should simply arrange a peer transfer link. Here you wouldn’t have to pay for
|
3502
|
+
| a compression and a decompression task – you would simply pay for the variety to make
|
3503
|
+
| the original product understandable to the new actor. Assuming all encoding/decoding
|
3504
|
+
| tasks are the same and equal V, you would delete an up-link (2V) (one encode, one
|
3505
|
+
| receive) and a down-link (2V) (one transmit, and one decode), and replace it with a single
|
3506
|
+
| peer link (2V) (one transmit, one receive). So the change in value would be - (- 4 + 2)
|
3507
|
+
| which equals +2, or a 50% improvement.
|
3508
|
+
blank |
|
3509
|
+
|
|
3510
|
+
|
|
3511
|
+
|
|
3512
|
+
meta | 76
|
3513
|
+
text | For the earlier example in Section 4.4.1, for the short term evaluation, the amount (cost)
|
3514
|
+
| of variety added at D (three places) must be less than or equal to the value of recovered
|
3515
|
+
| resources at A (2x) and D (two places) minus some inefficiencies and overhead costs.
|
3516
|
+
| Here the value would be +3 that is, again, about a 50% improvement. This value change
|
3517
|
+
| may be different for “translation” and “context” effects (cf. Section 3.4.3). More detail
|
3518
|
+
| could be developed by using a fine-grained organizational model with fractional values
|
3519
|
+
| for all actors and all types of variety.
|
3520
|
+
blank |
|
3521
|
+
|
|
3522
|
+
text | Average adjustments of variety will be close to a net-zero effect, compared to the goal
|
3523
|
+
| benefit, so variety cost savings will neither help nor hinder the justification for any
|
3524
|
+
| planned improvement in communication. This is reasonable since the improvements are
|
3525
|
+
| designed to improve a meta-level of activity for the organization.
|
3526
|
+
blank |
|
3527
|
+
|
|
3528
|
+
|
|
3529
|
+
title | 5.3 Expected Product Changes
|
3530
|
+
text | Increased information transfer regarding new technologies and validated prototypes will
|
3531
|
+
| enable more rapid design of aerospace products and a more rapid growth of the aerospace
|
3532
|
+
| environment through two pathways: Delivery Date Export, and Incremental Technology
|
3533
|
+
| Incorporation. A representative cumulative effect is shown in (Figure 5-1). The x-axis is
|
3534
|
+
| time, measured in years, and the y-axis is capacity or social value, measured in some
|
3535
|
+
| combination of dollars, performance, quality, or similar metric. The bottom curve is the
|
3536
|
+
| status quo showing growth in social value in 12 year steps. A current system of this type
|
3537
|
+
| might be the GPS satellite network or the Air Traffic Control system. The top curve
|
3538
|
+
| shows the potential value of a system evolution using system engineering methods such
|
3539
|
+
| as evolutionary design [Bobinis 2006] or flow-service-quality engineering [Linger 2002]
|
3540
|
+
| to improve the efficiency and rate of delivery. Methods to cultivate the incremental
|
3541
|
+
| pathways (within these methods) are discussed below. This is consistent with
|
3542
|
+
| Evolutionary Design practices.
|
3543
|
+
blank |
|
3544
|
+
|
|
3545
|
+
|
|
3546
|
+
|
|
3547
|
+
meta | 77
|
3548
|
+
text | Accelerates
|
3549
|
+
| because of
|
3550
|
+
| Value synergistic
|
3551
|
+
| {Performance, benefits
|
3552
|
+
| Quality, $}
|
3553
|
+
| Larger benefits under
|
3554
|
+
| the curve when
|
3555
|
+
| integrated over time
|
3556
|
+
blank |
|
3557
|
+
text | Can pull out instantiations of
|
3558
|
+
| current version as Keeps getting
|
3559
|
+
| requirements dictate slower because
|
3560
|
+
| complexity of
|
3561
|
+
| upfront
|
3562
|
+
| Performance Slip requirements
|
3563
|
+
| Schedule
|
3564
|
+
| Slip
|
3565
|
+
blank |
|
3566
|
+
text | 3yrs
|
3567
|
+
blank |
|
3568
|
+
|
|
3569
|
+
text | 12yrs
|
3570
|
+
blank |
|
3571
|
+
|
|
3572
|
+
|
|
3573
|
+
text | Time (years)
|
3574
|
+
blank |
|
3575
|
+
|
|
3576
|
+
text | Figure 5-1 Improved Value Over Time
|
3577
|
+
| The x-axis is time, measured in years, and the y-axis is capacity or social value, measured in some
|
3578
|
+
| combination of dollars, performance, quality, or similar metric. The bottom curve is the status quo
|
3579
|
+
| showing growth in social value in 12 year steps. The top curve shows the potential value of a system
|
3580
|
+
| evolution when smaller time steps are enabled due to greater efficiency and agility. This is similar to
|
3581
|
+
| compound interest.
|
3582
|
+
blank |
|
3583
|
+
|
|
3584
|
+
|
|
3585
|
+
|
|
3586
|
+
title | 5.3.1 Exporting Products at Delivery Dates
|
3587
|
+
text | Exporting products at predefined delivery dates is similar to design-to-schedule and
|
3588
|
+
| design-to-cost efforts. However, instead of pre-defining a goal and then forcing your
|
3589
|
+
| design to achieve it, you define a set of interfaces and boundary conditions, and deliver
|
3590
|
+
| within those constraints. This requires some inefficient margins (initially), but as you
|
3591
|
+
| reuse the base design for subsequent products, you can focus on improving the various
|
3592
|
+
| performance parameters without spending time on the interface. So within a couple of
|
3593
|
+
| iterations you will have improved total performance with less cost.
|
3594
|
+
blank |
|
3595
|
+
|
|
3596
|
+
|
|
3597
|
+
|
|
3598
|
+
meta | 78
|
3599
|
+
text | These ideas have been incorporated into chip design in the electronics industry. Chip
|
3600
|
+
| automated design tools had to use inefficient circuit designs to enable automated layout
|
3601
|
+
| and automated test. But the lower cost, compared to hand layouts, enabled greater sales,
|
3602
|
+
| and with subsequent miniaturization outgrew the pre-allocated “inefficient” margins.
|
3603
|
+
blank |
|
3604
|
+
|
|
3605
|
+
text | Engineers can, and will, allow for design growth with exported products at delivery dates,
|
3606
|
+
| and can incorporate the same testing protocols and support equipment, so it will be easy
|
3607
|
+
| to integrate. This decouples a lot of the simultaneous development and feedback
|
3608
|
+
| otherwise required to ensure everything fits, as in Section 3.3.2.
|
3609
|
+
blank |
|
3610
|
+
|
|
3611
|
+
text | Allowing engineers to apply half their time to looking for new technologies and
|
3612
|
+
| implementation methods, instead of optimizing their current design, allows for this
|
3613
|
+
| quicker growth, despite the initial margins needed.
|
3614
|
+
blank |
|
3615
|
+
|
|
3616
|
+
text | In (Figure 5-1), this increases the step height of subsequent deliveries, and improves even
|
3617
|
+
| further the compound growth curve.
|
3618
|
+
blank |
|
3619
|
+
|
|
3620
|
+
|
|
3621
|
+
title | 5.3.2 Incremental Technology Incorporation
|
3622
|
+
text | Program managers and senior engineers, if they use this same technique of improved
|
3623
|
+
| information flow, will have a more current understanding of related systems and can
|
3624
|
+
| define, and more quickly incorporate, consistent and grow-able requirements. Also they
|
3625
|
+
| will have a current understanding of current subsystems from related efforts where they
|
3626
|
+
| could modify an on-the-shelf design and integrate it into their system.
|
3627
|
+
blank |
|
3628
|
+
|
|
3629
|
+
text | There will be some overhead associated with incremental/evolutionary systems, but since
|
3630
|
+
| the subsequent new technologies will likely use less power and weight, they can be
|
3631
|
+
| inserted on the second platform and actually increase performance past the design margin
|
3632
|
+
| on the first vehicle.
|
3633
|
+
blank |
|
3634
|
+
|
|
3635
|
+
|
|
3636
|
+
|
|
3637
|
+
meta | 79
|
3638
|
+
title | 5.3.3 Coupled Efficiency
|
3639
|
+
text | These two effects (Delivery Date Export, and Incremental Technology Incorporation)
|
3640
|
+
| improve the overall Evolutionary Design approach , as compared to classic centralized
|
3641
|
+
| (computed) control, by avoiding large inefficiencies associated with unitary hardware
|
3642
|
+
| instantiation (primary, and 2nd and 3rd tier suppliers) by multiple organizations.
|
3643
|
+
blank |
|
3644
|
+
|
|
3645
|
+
|
|
3646
|
+
title | 5.3.4 Company and Product Goals
|
3647
|
+
text | Design of organizations and processes can affect the growth vector of the larger
|
3648
|
+
| community by emplacing the proper “a” factor in desired areas. For instance, if one
|
3649
|
+
| believes robotic exploration will be a driver of aerospace systems (instead of human
|
3650
|
+
| exploration), then the laboratories/groups associated with autonomy activities need new
|
3651
|
+
| channels formed and trained.
|
3652
|
+
blank |
|
3653
|
+
|
|
3654
|
+
|
|
3655
|
+
|
|
3656
|
+
meta | 80
|
3657
|
+
title | 5.4 Limits to Change
|
3658
|
+
text | Given this new perspective, and the apparent small scale of individual changes, how far
|
3659
|
+
| can you go? There are certainly limits to improvement, most basically from the external
|
3660
|
+
| environment’s ability to incorporate changes at a stable rate.
|
3661
|
+
blank |
|
3662
|
+
|
|
3663
|
+
text | Another restriction, or basic conceptual limit, is that there is no right answer and no
|
3664
|
+
| optimal solution. In fact trying to find an overall solution is diametrically opposite from
|
3665
|
+
| the direction taken here: that you try to improve the organization to more easily
|
3666
|
+
| incorporate engineer’s capabilities. And, although many engineers say that “better is the
|
3667
|
+
| enemy of good enough”, here you are not waiting to become perfect before you move
|
3668
|
+
| forward, you are independently improving engineering efficiency in parallel with
|
3669
|
+
| delivering your products.
|
3670
|
+
blank |
|
3671
|
+
|
|
3672
|
+
text | Likewise, there are costs associated with change. Changes in organizations and in
|
3673
|
+
| policies and procedures take time and effort to implement. You probably can only
|
3674
|
+
| change one link at a time. Also, if you change all the links at once, you would lose
|
3675
|
+
| ‘communicativity’ and your organization would collapse.
|
3676
|
+
blank |
|
3677
|
+
|
|
3678
|
+
text | Finally, you don’t want to change too fast. In this case you will have incurred the costs
|
3679
|
+
| of change without having looked ahead to see where external changes (and your changes)
|
3680
|
+
| have affected the environment in which you are working. You may want to reprioritize
|
3681
|
+
| your cultivation so you don’t overshoot.
|
3682
|
+
blank |
|
3683
|
+
|
|
3684
|
+
|
|
3685
|
+
|
|
3686
|
+
meta | 81
|
3687
|
+
title | 5.5 Applicability
|
3688
|
+
text | At what scale does this dissonant behaviour manifest? It was stated earlier that mega-
|
3689
|
+
| systems require on the order of 10,000 man-years of design time; that the total involved
|
3690
|
+
| set of people may number toward 100,000; and that the company size is around $10B
|
3691
|
+
| orders per year.
|
3692
|
+
blank |
|
3693
|
+
|
|
3694
|
+
text | Depending on the resolution, our proto-typical example organization has 22 nodes at the
|
3695
|
+
| top, virtual-lens level, at least 480 major agents at the composed summary level, and over
|
3696
|
+
| 5,000 working groups at the various individual program levels. From a bottom-up
|
3697
|
+
| perspective, there could be 200,000 information pathways given the number of people
|
3698
|
+
| involved and the required task variety. The exact number is problematic, both to actually
|
3699
|
+
| measure or record people, but also to identify intermediate groups and information
|
3700
|
+
| transfer paths.
|
3701
|
+
blank |
|
3702
|
+
|
|
3703
|
+
text | The work of [Braha 2007] indicates that 460 nodes and 1200 information transfer
|
3704
|
+
| activities is at the top end of what is currently analyzable, but is already at a statistically
|
3705
|
+
| significant population scale to quantify extracted behaviours with reasonable statistical
|
3706
|
+
| validity.
|
3707
|
+
blank |
|
3708
|
+
|
|
3709
|
+
text | From a control theory perspective, one could calculate stability metrics based on
|
3710
|
+
| feedback loops, task duration, measurement delays, etc. [Hogg 2001] showed that
|
3711
|
+
| hierarchies with more than 6 levels and/or with a span greater than 8 have limited
|
3712
|
+
| throughput, but of course this runs into the same analytic limits as the detailed task
|
3713
|
+
| analysis methods. [Park 1995] showed problematic convergence without additional
|
3714
|
+
| ‘parallel’ activities; no size values were given, but current computational limits were
|
3715
|
+
| listed even on cabling tasks.
|
3716
|
+
blank |
|
3717
|
+
|
|
3718
|
+
text | A different approach to defining applicability is to analyze where and why the aliasing
|
3719
|
+
| effects enter the design and production process, and then see how these accumulate into
|
3720
|
+
| increasing levels of significance.
|
3721
|
+
blank |
|
3722
|
+
|
|
3723
|
+
meta | 82
|
3724
|
+
text | Simple organizations and simple activity networks have high ‘a’ (3.5 – 4.0). Information
|
3725
|
+
| transfer is chaotic in the sense that it is immediate, complete, and very high bandwidth.
|
3726
|
+
| Imagine a very small company with a team of 5 working in a single room to design and
|
3727
|
+
| produce a single part for a machine. Each knows what the other is doing, they can
|
3728
|
+
| communicate verbally, and all the information is in the room including the supplier
|
3729
|
+
| catalogs, CNC settings, account ledger, etc. So development and production is straight
|
3730
|
+
| forward. The major risk for information loss is if one of the people is sick for a day:
|
3731
|
+
| since nothing was documented, you would have to wait for his return, or simply
|
3732
|
+
| regenerate the information.
|
3733
|
+
blank |
|
3734
|
+
|
|
3735
|
+
text | As the company size increases, there are more people, more rooms, and even more
|
3736
|
+
| buildings. Task assignments must be better planned and information flow must be better
|
3737
|
+
| documented. Things are less chaotic and more structured. Information is becoming
|
3738
|
+
| somewhat static, and must now be delivered explicitly. ‘a’ is approaching 3.0 or less.
|
3739
|
+
| Depending on how busy the shop is, some production tasks may get temporarily lost; or
|
3740
|
+
| possibly an invoice not given to finance to be paid. But these are small effects, and given
|
3741
|
+
| that the teams are well organized, have good processes, and understand the overall effort,
|
3742
|
+
| no big problems arise. It is not quite as efficient as before since now you need design
|
3743
|
+
| reviews, manufacturing readiness reviews, quality control, etc., but now you can address
|
3744
|
+
| larger efforts, delivering entire machines and getting the higher profits due the extra
|
3745
|
+
| complexity.
|
3746
|
+
blank |
|
3747
|
+
|
|
3748
|
+
text | At some point the time synchronization becomes problematic. Possibly because the next
|
3749
|
+
| use of information is indeterminate at the time of production, possibly because of
|
3750
|
+
| physical separation, or because of a number of reasons. Efforts to not lose information
|
3751
|
+
| become important: documenting fabrication improvements to feedback to design;
|
3752
|
+
| creating reusable components and product families; transferring consistent customer
|
3753
|
+
| requirements to different sub-organizations. This is the dissipative zone (a ~1.5) and care
|
3754
|
+
| must be taken to avoid slowing the design and production process, and becoming a
|
3755
|
+
| lethargic organization.
|
3756
|
+
blank |
|
3757
|
+
|
|
3758
|
+
|
|
3759
|
+
meta | 83
|
3760
|
+
text | So as you move from right to left in Figure 4-2 (high ‘a’ to low ‘a’), the growth of the
|
3761
|
+
| organization and processes provides more opportunities for information aliasing – both
|
3762
|
+
| from noise injection and from dissipation. Any organization, no matter what its size, sees
|
3763
|
+
| these effects. The bigger the organization, the bigger the size of the effects, and the more
|
3764
|
+
| valuable it is to adjust for smoother dynamics.
|
3765
|
+
blank |
|
3766
|
+
|
|
3767
|
+
text | Note that the longer time dynamics (low ‘a’) occur when trying to use resources across
|
3768
|
+
| individual mega-systems. In trying to even out the use of product resources (knowledge,
|
3769
|
+
| capital equipment), problems can be injected. This implies there is (at least) one, or a set
|
3770
|
+
| of activities that are single use within the design of a mega-system. When these are
|
3771
|
+
| applied to the next megasystem, their system context is essentially not cleared. This
|
3772
|
+
| might imply an additional type of aliasing, sort of like a feed-forward echo.
|
3773
|
+
blank |
|
3774
|
+
|
|
3775
|
+
|
|
3776
|
+
|
|
3777
|
+
meta | 84
|
3778
|
+
title | 6 Conclusions
|
3779
|
+
text | This dissertation developed a new type of model for extremely large design organizations
|
3780
|
+
| and their environment by combining perspectives from:
|
3781
|
+
| - evolutionary genetic computing organisms,
|
3782
|
+
| - information theoretic (quantum) information production, and
|
3783
|
+
| - non-linear dynamics and chaos theory for information exchange.
|
3784
|
+
| to study some of the behaviours of complex organizations, especially information
|
3785
|
+
| transfer, to show how organizations can be guided to more efficiently and quickly
|
3786
|
+
| produce complex mega-systems.
|
3787
|
+
blank |
|
3788
|
+
|
|
3789
|
+
text | A three-year experiment was performed within a large design organization to see what
|
3790
|
+
| level and type of improvements would actually manifest.
|
3791
|
+
blank |
|
3792
|
+
|
|
3793
|
+
text | The results fall into three categories:
|
3794
|
+
| - production of individual mega-systems,
|
3795
|
+
| - improvement in organizations that develop mega-systems, and
|
3796
|
+
| - extraction of societal benefits from government funded mega-systems.
|
3797
|
+
| These are addressed separately below.
|
3798
|
+
blank |
|
3799
|
+
|
|
3800
|
+
|
|
3801
|
+
title | 6.1 More Efficient Design of Mega-Systems
|
3802
|
+
text | Ignoring the quality of information transfer among organizational elements causes
|
3803
|
+
| significant inefficiencies. The transfer quality includes time synchronization, method of
|
3804
|
+
| transfer, and other variables. It is not that anything is currently wrong, just that a lot of
|
3805
|
+
| time is wasted waiting for information, translating information, or re-developing the
|
3806
|
+
| context to understand the information, and additional time (and resources) may be spent
|
3807
|
+
| waiting for task convergence. Matching the characteristics of the product design
|
3808
|
+
| information to the characteristics of the transfer path minimizes aliasing effects
|
3809
|
+
| (dissipation and noise), creating many benefits including:
|
3810
|
+
| - More efficient design and design execution,
|
3811
|
+
blank |
|
3812
|
+
|
|
3813
|
+
|
|
3814
|
+
meta | 85
|
3815
|
+
text | - Improved design agility / currency, and
|
3816
|
+
| - Enablement / easing of “Evolutionary Design” ([Bobinis 2006], [Bar-Yam
|
3817
|
+
| 2003], [Sangiovanni 2006]) at the upper, large-scale limit.
|
3818
|
+
blank |
|
3819
|
+
|
|
3820
|
+
text | Results from a three-year experiment showed that the inserted value of R&D efforts rose
|
3821
|
+
| from less than 5% to almost 80% as measured by the number of research projects being
|
3822
|
+
| fully used in operational programs. The research projects themselves can be quite
|
3823
|
+
| successful, but information transfer into a development program runs into the problems
|
3824
|
+
| indicated in terms of time synchronization.
|
3825
|
+
blank |
|
3826
|
+
|
|
3827
|
+
|
|
3828
|
+
title | 6.2 Relying on Emergence and Organizational Computation
|
3829
|
+
text | Improvements in the performance of the organization will incrementally emerge, given
|
3830
|
+
| the inertia of large organizations and of engineering processes, and will initially manifest
|
3831
|
+
| as quicker deliveries and an accelerating ability to innovate ahead of other companies.
|
3832
|
+
blank |
|
3833
|
+
|
|
3834
|
+
text | “One consequence of encoding information as statistical and time-varying patterns of
|
3835
|
+
| low-level components is that no individual component of the system can perceive or
|
3836
|
+
| communicate the ‘big-picture’ of the state of the system. Instead, information must be
|
3837
|
+
| communicated via spatial and temporal sampling.” [Mitchell 2009] Dynamically, the
|
3838
|
+
| mega-design process is progressing, and any snapshot is an incomplete view in a
|
3839
|
+
| changing context. This implies that the product directors can’t really direct or control
|
3840
|
+
| where the company is going: their role should only be to monitor the outputs and nascent
|
3841
|
+
| products, and guide the output (and rate of new product introduction and of upgrades) to
|
3842
|
+
| match the funding (requirements) source. This would balance control and efficiency.
|
3843
|
+
blank |
|
3844
|
+
|
|
3845
|
+
text | “Recently [in the field of biology] there has been a renewed effort to provide a
|
3846
|
+
| macroscopic theory for the origins of selection mechanisms. … This effort goes under
|
3847
|
+
| the name of niche construction. … [Here], the selective filter is partially constructed by
|
3848
|
+
| the organism itself, and exerts an influence over a locally reproducing population. … In
|
3849
|
+
| this way, adaptations lift themselves up by their own bootstraps.” [Krakauer 2011].
|
3850
|
+
blank |
|
3851
|
+
|
|
3852
|
+
|
|
3853
|
+
meta | 86
|
3854
|
+
text | This dynamic perspective implies that design communities can, and should, “construct
|
3855
|
+
| their own [evolutionary] selective constructs designed toward increasing mutual
|
3856
|
+
| information [within itself, and] between the organism and the environment” [Krakauer
|
3857
|
+
| 2011], for better computability or production of designs. In this case, the management
|
3858
|
+
| structure ought to also look at higher level abstractions, such as information transfer
|
3859
|
+
| efficiencies, in addition to process control (like design reviews), because looking at static
|
3860
|
+
| variables will not improve the evolutionary adaptation rate.
|
3861
|
+
blank |
|
3862
|
+
title | 6.3 Government Aerospace, Low-Price Procurement, and
|
3863
|
+
text | Overhead Rates
|
3864
|
+
| The stability of any implemented improvements is somewhat problematic given the
|
3865
|
+
| evolving environment, but especially given the counter-acting efforts of government
|
3866
|
+
| acquisition regulations and nominal management tendencies.
|
3867
|
+
blank |
|
3868
|
+
|
|
3869
|
+
text | Scale, synchronization, and economic nudges may be appropriate for personal scale
|
3870
|
+
| complices such as smartPhone apps, but the economic and scale factors are different from
|
3871
|
+
| government “apps”. A focus on lowest cost procurement, for government “apps” (mega-
|
3872
|
+
| systems), has side effects on the engineering design process. Financially speaking, and
|
3873
|
+
| for more effective mega-system development, the government should modify cost-
|
3874
|
+
| selection constraints to avoid extra effort to achieve a particular instance when any of a
|
3875
|
+
| set of instances is all that is needed. This would necessarily become somewhat non-
|
3876
|
+
| deterministic.
|
3877
|
+
blank |
|
3878
|
+
|
|
3879
|
+
text | Individual program managers will avoid this because it is locally inefficient and also
|
3880
|
+
| uncontrollable. However, according to game theory (prisoner’s dilemma [Morrow
|
3881
|
+
| 1994]), a Pareto optimal solution is better, and can be achieved with “external norms”
|
3882
|
+
| [Mitchell 2009]. The external norms could either be via company processes and
|
3883
|
+
| procedures or, better, by the government’s Federal Acquisition Regulations (FAR).
|
3884
|
+
blank |
|
3885
|
+
|
|
3886
|
+
|
|
3887
|
+
|
|
3888
|
+
meta | 87
|
3889
|
+
text | For instance, implementation at the government FAR level could be to assign 1% - 2% of
|
3890
|
+
| the value of programs over a certain size to internal efforts supporting adaptability and
|
3891
|
+
| risk reduction. For instance the Orion spacecraft program internally allocated $100M
|
3892
|
+
| (for a $4B program) to reduce schedule risk occurring from software integration. This is
|
3893
|
+
| the same scale as proposed, and was considered and approved as the correct engineering
|
3894
|
+
| approach. However, it was implemented without adjusting the encompassing process, so
|
3895
|
+
| savings could never be recaptured, nor future transitions eased. This fund would reduce
|
3896
|
+
| risk on the delivery of the initial system, and reduce cost of subsequent systems.
|
3897
|
+
blank |
|
3898
|
+
|
|
3899
|
+
text | Such a fund could also be allocated for application to follow-on project development, and
|
3900
|
+
| coordination of supporting technology and product development. Note that this would
|
3901
|
+
| avoid aliasing effects from Federally Funded Research and Development Companies
|
3902
|
+
| (FFRDCs) in their information transfer efforts, and would greatly improve technology
|
3903
|
+
| transfer from government labs to industry and society.
|
3904
|
+
blank |
|
3905
|
+
|
|
3906
|
+
|
|
3907
|
+
|
|
3908
|
+
meta | 88
|
3909
|
+
title | 6.4 Additional Research Vectors
|
3910
|
+
text | Several areas of this model could be extended based on related supporting theoretical
|
3911
|
+
| areas, including organizational memory, the effects of different engineering tools, and
|
3912
|
+
| evaluation of sub-networks in large communities. These are briefly discussed below.
|
3913
|
+
blank |
|
3914
|
+
title | 6.4.1 Organizational Memory
|
3915
|
+
text | Including memory as a communication channel with a “long” time constant is a straight-
|
3916
|
+
| forward addition. Efficient internal communication is similar to having some global
|
3917
|
+
| memory with time constant t0. This provides our organism (design community) the best
|
3918
|
+
| combination of a Von Neumann architecture and a cellular automata.
|
3919
|
+
blank |
|
3920
|
+
title | 6.4.2 Engineering Tools
|
3921
|
+
text | An area for future investigation as it is indicated by this analysis. Engineering design
|
3922
|
+
| tools play a large role in communicating knowledge and techniques.
|
3923
|
+
blank |
|
3924
|
+
title | 6.4.3 Long-Scale Neural Training of Sub-Networks
|
3925
|
+
text | Adjusting variety takes time like a single neuron in a network with different time
|
3926
|
+
| constants. The existence of Bifurcative channels in large communities needs additional
|
3927
|
+
| structure as the evolutionary time scale of the channels does not match that of the actors.
|
3928
|
+
blank |
|
3929
|
+
|
|
3930
|
+
|
|
3931
|
+
|
|
3932
|
+
meta | 89
|
3933
|
+
title | 7 Appendix A – Research Background
|
3934
|
+
text | Design Community Responses to Recognition of Complexity
|
3935
|
+
| There are two inter-twined facets of large scale engineering organizations: the large
|
3936
|
+
| organization itself, required to design, build, and operate the product, and the processes
|
3937
|
+
| they follow to produce their mega-products. Mega-organizations were covered in
|
3938
|
+
| Section 1. Here we address the internal design processes that guide the organizational
|
3939
|
+
| information development and transfer.
|
3940
|
+
blank |
|
3941
|
+
title | 7.1 Evolutionary Design
|
3942
|
+
text | Evolutionary design includes functions of complex systems that allow us to discover,
|
3943
|
+
| measure, and implement design activities during the entire life cycle of a system. It
|
3944
|
+
| assumes that we cannot know all the functions of deployed complex systems, and that an
|
3945
|
+
| evolutionary approach to design, is required. Several unique design methods were added
|
3946
|
+
| to the system engineering repertoire, including incremental engineering by which
|
3947
|
+
| portions of the system are redesigned and operate in parallel to other parts of the system.
|
3948
|
+
| Bar-Yam suggests that spiral development is an instantiation, (although simplified), of
|
3949
|
+
| evolutionary acquisition and design. “Recent extensions that adopt some complex
|
3950
|
+
| systems ideas include spiral development and evolutionary acquisition and adaptive
|
3951
|
+
| programming. A complex systems perspective provides a large conceptual framework –
|
3952
|
+
| evolution-from which to understand how incremental change can enable rapid
|
3953
|
+
| innovation.” [Bar-Yam 2003].
|
3954
|
+
blank |
|
3955
|
+
|
|
3956
|
+
text | Evolutionary design implies that since we cannot iteratively account for all functions and
|
3957
|
+
| processes of complex systems, we must provide a design process which allows us to
|
3958
|
+
| inductively determine system behaviors and processes, after the system is fielded. Spiral
|
3959
|
+
| design engineering, is really just a traditional design methodology which is repeated
|
3960
|
+
| recursively over the entire life cycle of the system. But it does account for all the
|
3961
|
+
| interfaces, transactions, and controls of a complex system that cannot be known during
|
3962
|
+
| initial design. Through this process, we can apply traditional methods in changing small
|
3963
|
+
| parts of complex systems, during its life cycle, rather than trying to change the “whole
|
3964
|
+
blank |
|
3965
|
+
|
|
3966
|
+
|
|
3967
|
+
meta | 90
|
3968
|
+
text | system”. And, that through the process of acquiring knowledge of self-organizing
|
3969
|
+
| systems, parts of the system may be known and tested, and replaced through traditional
|
3970
|
+
| planning, specification and implementation. Evolutionary design implies that on-going
|
3971
|
+
| discovery and re-design is a life cycle function of the system. The system must have a
|
3972
|
+
| means of monitoring its own “state”. An example of this, might be the modern
|
3973
|
+
| application of prognostics in complex platforms, where system states can be collected,
|
3974
|
+
| analyzed, and understood to the extent that we are provided the “patterns of system
|
3975
|
+
| behavior” to be able to predict and/or design out system failures, failure modes, and use
|
3976
|
+
| cases that the system may not have been designed for [sic]. [Bobinis 2006]
|
3977
|
+
blank |
|
3978
|
+
title | 7.2 Engineering Systems
|
3979
|
+
text | Rhodes and Hasting, in their paper The Case for Evolving Systems Engineering as a
|
3980
|
+
| Field within Engineering Systems, propose a more holistic systems engineering practice
|
3981
|
+
| which not only encompasses the boundaries of the system under development, but also a
|
3982
|
+
| broader context which includes the environment the system must co-exist in, including
|
3983
|
+
| the evolution of the system in the broader context.
|
3984
|
+
| “When taking the classical view, a major fault cited with Systems engineering is that it
|
3985
|
+
| does not take an adequate holistic perspective and is too introspective. Additionally, it is
|
3986
|
+
| often viewed as taking too much of a top-down approach…Further, Systems Engineering
|
3987
|
+
| is sometimes criticized for focusing too extensively on requirements and not enough on
|
3988
|
+
| system properties and behaviors.” [Rhodes 2004]. It recognizes that the Systems
|
3989
|
+
| Engineering discipline, ontology, and associated epistemology must be revised to account
|
3990
|
+
| for increasingly complex and self-organizing systems. It alludes to a process which
|
3991
|
+
| includes not only traditional engineering disciplines, but also a cross disciplinary
|
3992
|
+
| approach that includes the ability to understand the “human system”, the “environment”,
|
3993
|
+
| and the “community” as part of the context of the overall system, as “determinate
|
3994
|
+
| interfaces”.
|
3995
|
+
blank |
|
3996
|
+
|
|
3997
|
+
|
|
3998
|
+
|
|
3999
|
+
meta | 91
|
4000
|
+
title | 7.3 Flow-Service-Quality Engineering
|
4001
|
+
text | Carnegie Mellon, in response to study commissioned by the U.S. Air Force, has
|
4002
|
+
| developed a methodology to address large scale network centric systems design. “These
|
4003
|
+
| systems are characterized by changing and often unknown boundaries and components,
|
4004
|
+
| constantly varying function and usage, and complexities of pervasive asynchronous
|
4005
|
+
| operations. Their complexity challenges human intellectual control…”[Linger 2002 p.v].
|
4006
|
+
blank |
|
4007
|
+
|
|
4008
|
+
text | They attack the problem through the design of a “system services” function for these
|
4009
|
+
| types of systems as the “unifying”, and specifiable “function” of any system of systems.
|
4010
|
+
| It is the function wherein the myriad of users, systems, and subsystems interact, and
|
4011
|
+
| perform their unique transactions (interfaces).
|
4012
|
+
blank |
|
4013
|
+
|
|
4014
|
+
text | System Services is the primary hub for their analysis and methodology of control. It
|
4015
|
+
| directs, analyzes, stores, and measures all the transactional processes of the complex
|
4016
|
+
| system. It starts at a functional analysis of user based operations and interfaces the
|
4017
|
+
| system services function needs to support all users (human or other systems). By
|
4018
|
+
| mapping the interchanges with the supported “users”, the system services function can
|
4019
|
+
| analyze system, as well as sub-system, states. It also defines the standards and protocols
|
4020
|
+
| for the entire system and serves as the “super hub”.
|
4021
|
+
blank |
|
4022
|
+
|
|
4023
|
+
text | What is important here is not the architecture promoted in this study, but the fact that
|
4024
|
+
| there is both the recognition that all subsystem functions cannot be accounted for in the
|
4025
|
+
| design, and that the primary hub “system services” has specific sub-functions that
|
4026
|
+
| account for growth (new nodes), and reorganization (once added, the types of information
|
4027
|
+
| shared across nodes). It’s ability to account for nodes as they are added, the type of
|
4028
|
+
| requests of these unique nodes, the mapping and storage of system topology, in near real-
|
4029
|
+
| time, and the realization that as unique functional nodes are added to the system, they
|
4030
|
+
| must be accounted for, monitored, and measured, as they function within the system.
|
4031
|
+
blank |
|
4032
|
+
|
|
4033
|
+
text | “FSQ engineering provides systematic, scale-free semantic structures for requirements,
|
4034
|
+
| specification, design, verification and implementation…User flows of services and
|
4035
|
+
blank |
|
4036
|
+
|
|
4037
|
+
meta | 92
|
4038
|
+
text | quality attributes permit system development in terms of user views of services, as
|
4039
|
+
| opposed to strictly functional decomposition or object-based composition.” [Linger 2002
|
4040
|
+
| p.37]
|
4041
|
+
blank |
|
4042
|
+
|
|
4043
|
+
|
|
4044
|
+
title | 7.4 Engineering Tools
|
4045
|
+
text | Some studies that examined the engineering process have focused on tools developed to
|
4046
|
+
| improve product design by engineering teams [Leifer 1999]. Here the tools fall into three
|
4047
|
+
| classes: interpersonal communication, knowledge capture, and knowledge retrieval.
|
4048
|
+
| Productivity gains have been shown when design teams can begin from a better starting
|
4049
|
+
| point as a result of having access to learned information from previous designs.
|
4050
|
+
| “In larger organizations, the complexity is vastly different and the chance for deep
|
4051
|
+
| personal co-commitment is diminished. Ad hoc exchanges within one’s personal
|
4052
|
+
| network do happen, but systematic knowledge sharing usually requires brokering through
|
4053
|
+
| people or technology. Along with brokering come issues such as knowledge
|
4054
|
+
| identification, verification, and validation [Leifer 1999].”
|
4055
|
+
blank |
|
4056
|
+
|
|
4057
|
+
text | “[All organizations] can benefit from the implementation of a central knowledge capture
|
4058
|
+
| and re-use infrastructure, a core technology for sharing and learning from each other,
|
4059
|
+
| peer-to-peer. In large institutions (greater that 500 people) this infrastructure will have to
|
4060
|
+
| be explicit and managed separately from the design team support infrastructure [Leifer
|
4061
|
+
| 1999].”
|
4062
|
+
blank |
|
4063
|
+
|
|
4064
|
+
text | A separate explicit infrastructure however becomes problematic, especially on projects
|
4065
|
+
| with long time scales (~10 years). Issues arise that interfere with having a consistent
|
4066
|
+
| structure. The portfolio of projects across a large organization use different engineering
|
4067
|
+
| tools. This will always be true because of a mismatch of time frequencies between new
|
4068
|
+
| tools and new projects, and the fact you can’t and don’t want to update your tools all at
|
4069
|
+
| once on a preordained schedule. Also engineers bring different sets of tool knowledge
|
4070
|
+
| with them to new projects. It actually takes some finite time for a team to get
|
4071
|
+
| synchronized on common tools and data structures. Also most training is done on the job
|
4072
|
+
blank |
|
4073
|
+
|
|
4074
|
+
|
|
4075
|
+
meta | 93
|
4076
|
+
text | with no separate time allotment and no separate instructors, with the result that tools are
|
4077
|
+
| not always used as they were designed, nor to their full potential. Finally, retrieval of old
|
4078
|
+
| knowledge becomes problematic and time consuming, and may become moot as different
|
4079
|
+
| technologies advance at different rates, opening up new solution spaces for consideration.
|
4080
|
+
blank |
|
4081
|
+
|
|
4082
|
+
text | Another issue arising from large organizations that affects knowledge retention and
|
4083
|
+
| retrieval is that engineering team members do not usually follow a project from
|
4084
|
+
| conception through design, manufacture, production, maintenance, and growth.
|
4085
|
+
| Typically, engineers focus on one stage of design. Therefore the information associated
|
4086
|
+
| with a project bifurcates, with one instance progressing with the product as the product
|
4087
|
+
| “design” and “context”, and the other instance going with the engineer or data base to a
|
4088
|
+
| new product at a similar stage as “knowledge”.
|
4089
|
+
blank |
|
4090
|
+
|
|
4091
|
+
|
|
4092
|
+
|
|
4093
|
+
meta | 94
|
4094
|
+
title | 8 References
|
4095
|
+
ref | [Bais 2010] “The Physics of Information”, F. Alexander Bais and J. Doyne Farmer, SFI
|
4096
|
+
| Working Paper 07-08-029 to be included in “The Handbook on the Philosophy of
|
4097
|
+
| Information, to be published by Elsevier.
|
4098
|
+
blank |
|
4099
|
+
ref | [Barabasi 2003] Linked, Albert-Laszlo Barabasi, Plume, 2003.
|
4100
|
+
blank |
|
4101
|
+
ref | [Bar-Yam 2002] “Large Scale Engineering and Evolutionary Change: Useful Concepts
|
4102
|
+
| for Implementation of FORCEnet,” Bar-Yam, Yaneer, FORCEnet and the 21st
|
4103
|
+
| Century Warrior, USN Strategic Studies Group, Contract F30602-02-C-0158,
|
4104
|
+
| 2002.
|
4105
|
+
blank |
|
4106
|
+
ref | [Bar-Yam 2003] “When Systems Engineering Fails---Toward Complex Systems
|
4107
|
+
| Engineering”; Yaneer Bar-Yam, IEEE International Conference, 5-8 Oct. 2003,
|
4108
|
+
| Volume: 2, on page: 2022 vol.2
|
4109
|
+
blank |
|
4110
|
+
ref | [Bar-Yam 2004a] “A Mathematical Theory of Strong Emergence Using Multiscale
|
4111
|
+
| Variety,” Bar-Yam, Yaneer, in Complexity, Wiley Periodicals, vol. 9, no. 6,
|
4112
|
+
| 2004.
|
4113
|
+
blank |
|
4114
|
+
ref | [Bar-Yam 2004b] “About Engineering Complex Systems: Multiscale Analysis and
|
4115
|
+
| Evolutionary Engineering”, Yaneer Bar-Yam; ESOA 2004, LNCS 3464, pp. 16-
|
4116
|
+
| 31, 2005.
|
4117
|
+
blank |
|
4118
|
+
ref | [Bar-Yam 2005] “About Engineering Complex Systems: Multiscale Analysis and
|
4119
|
+
| Evolutionary Engineering,” Bar-Yam, Yaneer, ESOA 2004, LNCS3464, S.
|
4120
|
+
| Brueckner et al (Eds), Springer-Verlag, 2005.
|
4121
|
+
blank |
|
4122
|
+
ref | [Bobinis 2006] “Complex Self-Organizing Systems And Their Effect on the Systems
|
4123
|
+
| Engineering Paradigm”, Joe Bobinis, 2006.
|
4124
|
+
blank |
|
4125
|
+
ref | [Bonissone 2006] “Evolutionary Algorithms plus Domain Knowledge equals Real-World
|
4126
|
+
| Evolutionary Computation,” Bonissone, Piero P., Subbu, Raj, Eklund, Neil, and
|
4127
|
+
blank |
|
4128
|
+
|
|
4129
|
+
|
|
4130
|
+
|
|
4131
|
+
meta | 95
|
4132
|
+
ref | Kiehl, Thomas R., IEEE Transactions on Evolutionary Computation, vol. 10, no.
|
4133
|
+
| 3, June 2006.
|
4134
|
+
blank |
|
4135
|
+
ref | [Braha 2007] “The Statistical Mechanics of Complex Product Development: Empirical
|
4136
|
+
| and Analytical Results,” Dan Braha and Yaneer Bar-Yam, Management Science,
|
4137
|
+
| Special Issue on Complex Systems; Volume 53, number 7, July 2007.
|
4138
|
+
blank |
|
4139
|
+
ref | [Byler 2000] “Cultivating the Growth of Complex Systems Using Emergent Behaviours
|
4140
|
+
| of Engineering Processes,” International Conference on Complex Systems:
|
4141
|
+
| Control and Modeling; Russian Academy of Sciences; Samara Russia, June
|
4142
|
+
| 2000.
|
4143
|
+
blank |
|
4144
|
+
ref | [Capaccio 2011] Tony Capaccio, "Cost Overruns Plague One-Third Of Pentagon's
|
4145
|
+
| Biggest Programs”, Bloomberg Government, March 29, 2011.
|
4146
|
+
blank |
|
4147
|
+
ref | [Clouse 1999] Clouse, Raja Laifa and Waldron, Manjula. “Effect of History to Achieve
|
4148
|
+
| Self Adaptation”. Center for Integrated Design, Ohio State University,
|
4149
|
+
| manuscript in progress 1999.
|
4150
|
+
blank |
|
4151
|
+
ref | [Crutchfield 2011] James P. Crutchfield and Jon Machta, “Introduction to Focus Issue on
|
4152
|
+
| Randomness, Structure, and Causality: Measures of Complexity from Theory to
|
4153
|
+
| Applications,” Chaos, An Interdisciplinary Journal on Nonlinear Science;
|
4154
|
+
| American Institute of Physics, 2011.
|
4155
|
+
blank |
|
4156
|
+
ref | [de Smith 2008] “Simulation/Modelling (s/m) Systems for Agent-Based Modelling,” de
|
4157
|
+
| Smith, Goodchild, Longley, Geospatial Analysis – a Comprehensive Guide, 2nd
|
4158
|
+
| edition, 2008.
|
4159
|
+
blank |
|
4160
|
+
ref | [Flack 2011] Jessica C. Flack and David C. Krakauer, “Challenges for Complexity
|
4161
|
+
| Measures: A perspective from social dynamics and collective social
|
4162
|
+
| computation,” Chaos, An Interdisciplinary Journal on Nonlinear Science;
|
4163
|
+
| American Institute of Physics, 2011.
|
4164
|
+
blank |
|
4165
|
+
ref | [Flecker 2011] Benjamin Flecker, Wesley Alford, John M. Beggs, Paul L. Williams, and
|
4166
|
+
| Randall D. Beer, “Partial Information Decomposition as a Spatio-temporal Filter,”
|
4167
|
+
| Chaos, An Interdisciplinary Journal on Nonlinear Science; American Institute of
|
4168
|
+
| Physics, 2011.
|
4169
|
+
blank |
|
4170
|
+
|
|
4171
|
+
meta | 96
|
4172
|
+
ref | [Hofstadter] Hofstadter, Douglas. “Godel, Escher, Bach: an Eternal Golden Braid”, Basic
|
4173
|
+
| Books, 1979.
|
4174
|
+
blank |
|
4175
|
+
ref | [Hogg 2001] “Controlling Smart Matter”, Tad Hogg and Bernardo A. Huberman; Xerox
|
4176
|
+
| Palo Alto Research Center, June 1, 2001
|
4177
|
+
blank |
|
4178
|
+
ref | [Holland 1998] Holland, John B. “Emergence: From Chaos to Order”, Addison-Wesley,
|
4179
|
+
| 1998.
|
4180
|
+
blank |
|
4181
|
+
ref | [Howley 1999] Howley, Brian. Discussions on Design Agent Models at the Stanford
|
4182
|
+
| Center for Design Research, 1999.
|
4183
|
+
blank |
|
4184
|
+
ref | [Kleene 1956] Kleene, S.C., “Representation of Events in Nerve Nets and Finite
|
4185
|
+
| Automata” // Automata Studies, Shannon & McCarthy eds. Princeton University
|
4186
|
+
| Press, 1956.
|
4187
|
+
blank |
|
4188
|
+
ref | [Krakauer 2011] David C. Krakauer, “Darwinian Demons, evolutionary complexity, and
|
4189
|
+
| information maximization,” Chaos, An Interdisciplinary Journal on Nonlinear
|
4190
|
+
| Science; American Institute of Physics, 2011.
|
4191
|
+
blank |
|
4192
|
+
ref | [Leifer 1999] Leifer, Liang, Cannon, Mabogunje, Yen, and Yang. “Experience Capture
|
4193
|
+
| and Reuse for Complex System Design-Deployment” // Proceedings of the
|
4194
|
+
| International Conference on Complex Systems: Control and Modeling Problems,
|
4195
|
+
| Myasnikov, Kuznetsov, and Vittikh eds., Samara, Russia, 1999.
|
4196
|
+
blank |
|
4197
|
+
ref | [Linger 2002] “Flow–Service–Quality (FSQ) Engineering: Foundations for Network
|
4198
|
+
| System Analysis and Development”; Richard Linger, Mark Pleszkoch,
|
4199
|
+
| Gwendolyn Walton, & Alan Hevner, Carnegie Mellon University in performance
|
4200
|
+
| of Federal Government contract number F19628-00-C-0003, 2002, pg. v.
|
4201
|
+
blank |
|
4202
|
+
ref | [Metzler 2003] Metzler, Bar-Yam, & Kardar, “Information Flow through a Chaotic
|
4203
|
+
| Channel: Prediction and Postdiction at Finite Resolution”, 2003; section IV A, pg
|
4204
|
+
| 9.
|
4205
|
+
blank |
|
4206
|
+
ref | [Metzler 2005] “Multiscale Complexity of Correlated Gaussians,” Metzler, Richard, and
|
4207
|
+
| Yaneer, Bar-Yam, Physical Review E71 046114, The American Physical Society,
|
4208
|
+
| 2005.
|
4209
|
+
blank |
|
4210
|
+
|
|
4211
|
+
|
|
4212
|
+
meta | 97
|
4213
|
+
ref | [Miller 2007] John H. Miller and Scott E. Page, “Complex Adaptive Systems, an
|
4214
|
+
| Introduction to Computational Models of Social Life”, Princeton University
|
4215
|
+
| Press, 2007, pp 72-73.
|
4216
|
+
blank |
|
4217
|
+
ref | [Minar 1996] Minar, Burkhart, Langton, and Askenazi. “The SWARM Simulation
|
4218
|
+
| System: A Toolkit for building Multi-Agent Simulations”,
|
4219
|
+
| <http://www.swarm.org>, June, 1996.
|
4220
|
+
blank |
|
4221
|
+
ref | [Mitchell 1992] Complexity: The Emerging Science at the Edge of Order and Chaos, M.
|
4222
|
+
| Mitchell Waldrop, Simon & Schuster, 1992.
|
4223
|
+
blank |
|
4224
|
+
ref | [Mitchell 2009] “Complexity: A Guided Tour”, Melanie Mitchell; p180; Oxford
|
4225
|
+
| University Press, 2009.
|
4226
|
+
blank |
|
4227
|
+
ref | [Morrow 1994] Game Theory for Political Scientists, James D. Morrow, Princeton
|
4228
|
+
| University Press, 1994.
|
4229
|
+
blank |
|
4230
|
+
ref | [Omerod 1999] Ormerod, Paul. “Butterfly Economics: A New General Theory of Social
|
4231
|
+
| and Economic Behavior”, Pantheon, 1999.
|
4232
|
+
blank |
|
4233
|
+
ref | [Park 1995] Park, Hisup. “Modeling of Collaborative Design Processes for Agent-
|
4234
|
+
| Assisted Product Design”, Stanford Doctoral Thesis, March 1995.
|
4235
|
+
blank |
|
4236
|
+
ref | [Petrie 1996] Petrie, Charles J. “Agent-Based Engineering, the Web, and Intelligence” //
|
4237
|
+
| IEEE Expert, December 1996.
|
4238
|
+
blank |
|
4239
|
+
ref | [Railsback 2006] “Agent-based Simulation Platforms: Review and Development
|
4240
|
+
| Recommendations,” Railsback, Steven F.,Lytinen, Steven L., and Jackson,
|
4241
|
+
| Stephen K., in press at Simlation, Santa Fe Institute, 2006.
|
4242
|
+
blank |
|
4243
|
+
ref | [Rhodes 2004] “The Case for Evolving Systems Engineering as a Field within
|
4244
|
+
| Engineering Systems”; Donna Rhodes & Daniel Hastings, MIT Engineering
|
4245
|
+
| Systems Symposium, March 2004, pg. 3.
|
4246
|
+
blank |
|
4247
|
+
ref | [Ruffini 2007] Information, Complexity, Brains, and Reality (Kolmogorov Manifesto),
|
4248
|
+
| Giulio Ruffini, Starlab Technical Note TN00054, 2007.
|
4249
|
+
blank |
|
4250
|
+
|
|
4251
|
+
|
|
4252
|
+
|
|
4253
|
+
meta | 98
|
4254
|
+
ref | [Sangiovanni 2006] “A framework for modeling the distributed deployment of
|
4255
|
+
| synchronous designs”, Formal methods in systems design - an international
|
4256
|
+
| journal, Springer-Verlag, Vol. 28, No. 2.
|
4257
|
+
blank |
|
4258
|
+
ref | [Shannon 1948] C. E. Shannon, The Bell System Technical Journal 27, 379 (1948).
|
4259
|
+
blank |
|
4260
|
+
ref | [Strader 1998] Strader, Lin, and Shaw. “Simulation of Order Fulfillment in Divergent
|
4261
|
+
| Assembly Supply Chains” // Journal of Artificial Societies and Social Simulation,
|
4262
|
+
| vol. 1, no. 2, <http://www.soc.surrey.ac.uk/JASSS/1/2/5.html> March 1998.
|
4263
|
+
blank |
|
4264
|
+
ref | [Verhulst 1838] “Notice sur la loi que la population poursuit dans son accroissement”,
|
4265
|
+
| Correspondance mathematique et physique 10: 113-121, 1838.
|
4266
|
+
blank |
|
4267
|
+
|
|
4268
|
+
|
|
4269
|
+
|
|
4270
|
+
meta | 99
|
4271
|
+
blank |
|