wawi 0.0.1__py3-none-any.whl → 0.0.5__py3-none-any.whl

wawi/wind.py

@@ -0,0 +1,1109 @@
+import numpy as np
+from scipy.interpolate import interp1d
+from .modal import statespace, iteig, restructure_as_ref, iteig_naive
+from .tools import print_progress as pp
+from scipy.special import jv as besselj, yv as bessely
+from .general import rodrot, blkdiag
+from .plot import plot_ads
+
+conv_text=''' 
+-----------------------------------------------------
+|                                                   |
+| ~ ~ ~~~ ~  ~~ ~ /^^^^^^^^^^^^\ 88ooo...  .  .  .  |
+| ~ ~ ~ ~~ ~ ~   ~\____________/  88ooo¨¨¨¨  ¨¨     |
+|                   CONVERGED!                      |
+-----------------------------------------------------
+'''
+
+beaufort_dict = {
+    'calm': [0, 0.5],
+    'light air': [0.5, 1.5],
+    'light breeze': [1.6, 3.3],
+    'gentle breeze': [3.4, 5.5],
+    'moderate breeze': [5.6, 7.9],
+    'fresh breeze': [8, 10.7],
+    'strong breeze': [10.8, 13.8],
+    'moderate gale': [13.9, 17.1],
+    'gale': [17.2, 20.7],
+    'strong gale': [20.8, 24.4],
+    'storm': [24.5, 28.4],
+    'violent storm': [28.5, 32.6],
+    'hurricane': [32.7, np.inf]
+}
+
+def get_beaufort(U0):
+    return [key for key in beaufort_dict if inrange(U0, beaufort_dict[key])][0]
+
+def inrange(num, rng):
+    return num<=np.max(rng) and num>=np.min(rng)
+
+class LoadCoefficients:
+    keys = ['Cd', 'Cm', 'Cl', 'dCd', 'dCm', 'dCl']
+    
+    def __repr__(self):
+        return 'LoadCoefficients (Cd, Cl, Cm, dCd, dCl, dCm)'
+
+    def __str__(self):
+        return f'Cd:{self.Cd}, dCd:{self.dCd}, Cl:{self.Cl}, dCl:{self.dCl}, Cm:{self.Cm}, dCm:{self.dCm}'
+    
+    def __init__(self, Cd=None, dCd=None, Cl=None, dCl=None, Cm=None, dCm=None, fill_empty=True):
+        self.Cd = Cd
+        self.dCd = dCd
+        self.Cl = Cl
+        self.dCl = dCl
+        self.Cm = Cm
+        self.dCm = dCm
+        
+        if fill_empty:
+            self.fill_empty_with_zeros()
+        
+    def fill_empty_with_zeros(self):
+        for key in self.keys:
+            if getattr(self, key) is None:
+                setattr(self, key, 0)
+                
+    def to_dict(self):
+        return {key: getattr(self, key) for key in self.keys}
+
+class ADs:
+    ad_keys = ['P1', 'P2', 'P3', 'P4', 'P5', 'P6', 
+              'H1', 'H2', 'H3', 'H4', 'H5', 'H6', 
+              'A1', 'A2', 'A3', 'A4', 'A5', 'A6']
+
+    P1, P2, P3, P4, P5, P6 = None, None, None, None, None, None
+    H1, H2, H3, H4, H5, H6 = None, None, None, None, None, None
+    A1, A2, A3, A4, A5, A6 = None, None, None, None, None, None
+
+    def __init__(self, ad_type='not specified', 
+                 P1=None, P2=None, P3=None, P4=None, P5=None, P6=None,
+                 H1=None, H2=None, H3=None, H4=None, H5=None, H6=None,
+                 A1=None, A2=None, A3=None, A4=None, A5=None, A6=None):
+        
+        self.type = ad_type
+        
+        self.P1 = P1
+        self.P2 = P2
+        self.P3 = P3
+        self.P4 = P4
+        self.P5 = P5
+        self.P6 = P6
+        
+        self.H1 = H1
+        self.H2 = H2
+        self.H3 = H3
+        self.H4 = H4
+        self.H5 = H5
+        self.H6 = H6
+        
+        self.A1 = A1
+        self.A2 = A2
+        self.A3 = A3
+        self.A4 = A4
+        self.A5 = A5
+        self.A6 = A6
+            
+    def plot(self, v=np.arange(0,5,0.01), **kwargs):
+        return plot_ads(self.to_dict(), v, **kwargs)
+        
+        
+    def to_dict(self):
+        return {key: getattr(self, key) for key in self.ad_keys}
+    
+    def evaluate_all(self, v):
+        AD_evaluated = dict()
+        for key in self.ad_keys:
+            AD_evaluated[key] = getattr(self, key)(v)
+
+        return AD_evaluated
+    
+    
+    def evaluate(self, key, v):
+        AD_evaluated = getattr(self, key)(v)
+
+        return AD_evaluated
+    
+def flatplate_ads():
+
+    ad_dict = dict()  
+    
+    def F(v):
+        J1 = besselj(1, 0.5/v)
+        Y1 = bessely(1, 0.5/v)
+        J0 = besselj(0, 0.5/v)
+        Y0 = bessely(0, 0.5/v)
+        
+        a = J1 + Y0
+        b = Y1 - J0
+        c = a**2 + b**2
+        
+        return (J1*a + Y1*b)/c
+        
+    def G(v):
+        J1 = besselj(1, 0.5/v)
+        Y1 = bessely(1, 0.5/v)
+        J0 = besselj(0, 0.5/v)
+        Y0 = bessely(0, 0.5/v)
+        
+        a = J1 + Y0
+        b = Y1 - J0
+        c = a**2 + b**2       
+        return -(J1*J0 + Y1*Y0)/c
+    
+    ad_dict['H1'] = lambda v: -2*np.pi*F(v)*v
+    ad_dict['H2'] = lambda v: np.pi/2*(1+F(v)+4*G(v)*v)*v
+    ad_dict['H3'] = lambda v: 2*np.pi*(F(v)*v-G(v)/4)*v
+    ad_dict['H4'] = lambda v: np.pi/2*(1+4*G(v)*v)
+    ad_dict['H5'] = lambda v: 0*v
+    ad_dict['H6'] = lambda v: 0*v
+
+    ad_dict['A1'] = lambda v: -np.pi/2*F(v)*v
+    ad_dict['A2'] = lambda v: -np.pi/8*(1-F(v)-4*G(v)*v)*v
+    ad_dict['A3'] = lambda v: np.pi/2*(F(v)*v-G(v)/4)*v
+    ad_dict['A4'] = lambda v: np.pi/2*G(v)*v   
+    ad_dict['A5'] = lambda v: 0*v
+    ad_dict['A6'] = lambda v: 0*v
+
+    ad_dict['P1'] = lambda v: 0*v
+    ad_dict['P2'] = lambda v: 0*v
+    ad_dict['P3'] = lambda v: 0*v
+    ad_dict['P4'] = lambda v: 0*v
+    ad_dict['P4'] = lambda v: 0*v
+    ad_dict['P5'] = lambda v: 0*v
+    ad_dict['P6'] = lambda v: 0*v
+    
+    return ad_dict
+
+
+def quasisteady_ads(D, B, load_coefficients):
+    # Assuming load coeffs are normalized wrt. both D (Cd) and B (Cl and Cm) and ADs are 
+    # normalized using B only.
+
+    if type(load_coefficients)==dict: 
+        Cd = load_coefficients['Cd']
+        dCd = load_coefficients['dCd']
+        Cl = load_coefficients['Cl']
+        dCl = load_coefficients['dCl']
+        Cm = load_coefficients['Cm']
+        dCm = load_coefficients['dCm']
+    else:
+        Cd, dCd = load_coefficients.Cd, load_coefficients.dCd
+        Cl, dCl = load_coefficients.Cl, load_coefficients.dCl
+        Cm, dCm = load_coefficients.Cm, load_coefficients.dCm
+
+    ad_dict = dict()
+    ad_dict['P1'], ad_dict['P2'], ad_dict['P3'] = lambda v: -2*Cd*D/B*v, lambda v: 0*v, lambda v: dCd*D/B*v**2
+    ad_dict['P4'], ad_dict['P5'], ad_dict['P6'] = lambda v: 0*v, lambda v: (Cl-dCd*D/B)*v, lambda v: 0*v
+
+    ad_dict['H1'], ad_dict['H2'], ad_dict['H3'] = lambda v: -(dCl+Cd*D/B)*v, lambda v: 0*v, lambda v: dCl*v**2
+    ad_dict['H4'], ad_dict['H5'], ad_dict['H6'] = lambda v: 0*v, lambda v: -2*Cl*v, lambda v: 0*v
+
+    ad_dict['A1'], ad_dict['A2'], ad_dict['A3'] = lambda v: -dCm*v, lambda v: 0*v, lambda v: dCm*v**2
+    ad_dict['A4'], ad_dict['A5'], ad_dict['A6'] = lambda v: 0*v, lambda v: -2*Cm*v, lambda v: 0*v
+
+    return ad_dict
+
+
+def compute_aero_matrices(U, AD, B, elements, T_wind, phi, 
+                          omega_reduced=None, print_progress=False, rho=1.225):
+    
+    if omega_reduced is None:
+        omega_reduced = np.linspace(0.015, 2.0, 75)
+    
+    n_modes = phi.shape[1]
+
+    Kae = np.zeros([n_modes, n_modes, len(omega_reduced)])
+    Cae = np.zeros([n_modes, n_modes, len(omega_reduced)])
+    
+    for element_ix, element in enumerate(elements):
+        
+        if callable(U):
+            U_el_glob = U(element.get_cog())
+        else:
+            U_el_glob = U*1
+            
+        U_el = normal_wind(T_wind, element.T0, U=U_el_glob)
+
+        v = U_el/(B*omega_reduced)
+            
+        for k, v_k in enumerate(v):
+            k_aero, c_aero = element_aero_mats(B, omega_reduced[k], 
+                                               AD.evaluate_all(v_k), 
+                                               element.L, T=element.T0, 
+                                               phi=phi[element.global_dofs, :], rho=rho)
+
+            Kae[:, :, k] = Kae[:, :, k] + k_aero
+            Cae[:, :, k] = Cae[:, :, k] + c_aero
+        
+        if print_progress:
+            pp(element_ix+1, len(elements), sym='=', postfix=' ESTABLISHING WIND EXCITATION')
+            print('')
+    
+    Cae = interp1d(omega_reduced, Cae, kind='quadratic', fill_value='extrapolate', bounds_error=False) 
+    Kae = interp1d(omega_reduced, Kae, kind='quadratic', fill_value='extrapolate', bounds_error=False)
+
+
+    return Kae, Cae
+
+
+def compute_aero_matrices_sets(U, AD, B, elements, T_wind, phi_dict, 
+                          omega_reduced=None, omega=None, print_progress=False, sets=None):
+    
+    if sets is None:
+         sets = elements.keys()
+
+    if omega is None:
+        return_as_function = True
+    else:
+        first_is_zero = omega[0]==0.0
+        if first_is_zero:
+            omega = omega[1:]
+    
+    if omega_reduced is None:
+        omega_reduced = np.logspace(np.log10(0.01), np.log10(2), 100)     #standard values should be reasonable in most typical cases - change later!
+    
+    first_key = [str(key) for key in sets][0]
+    n_modes = np.shape(phi_dict[first_key])[1]
+
+    Kae = np.zeros([n_modes, n_modes, len(omega_reduced)])
+    Cae = np.zeros([n_modes, n_modes, len(omega_reduced)])
+    
+    for set_name in sets:
+        B_set = B[set_name]
+        AD_set = AD[set_name]
+        phi = phi_dict[set_name]
+        elements_set = elements[set_name]
+
+        for element_ix, element in enumerate(elements_set):
+            T_el = element.T0
+            U_el = normal_wind(T_wind, T_el, U=U)
+            v = U_el/(B_set*omega_reduced)
+            
+            dof_range = np.hstack([element.nodes[0].global_dofs, element.nodes[1].global_dofs])
+
+            for k, v_k in enumerate(v):
+                k_aero, c_aero = element_aero_mats(B_set, omega_reduced[k], AD_set.evaluate_all(v_k), element.L, T=T_el, phi=phi[dof_range, :])
+                Kae[:, :, k] += k_aero
+                Cae[:, :, k] += c_aero
+            
+            if print_progress:
+                pp(element_ix+1, len(elements_set), sym='>', postfix=f' finished with set "{set_name}".')
+
+        if print_progress:
+            print('')
+
+        Cae = interp1d(omega_reduced, Cae, kind='quadratic',fill_value='extrapolate') 
+        Kae = interp1d(omega_reduced, Kae, kind='quadratic', fill_value='extrapolate')
+
+        if return_as_function:
+            return Kae, Cae
+        else:
+            Cae = Cae(omega)
+            Kae = Kae(omega)
+            
+            if first_is_zero:
+                Cae = np.insert(Cae, 0, Cae[:,:,0]*0, axis=2)
+                Kae = np.insert(Kae, 0, Kae[:,:,0]*0, axis=2)
+
+            return Kae, Cae
+
+def mvregress_ads(beta):
+    ad_dict = dict()
+    ad_keys = ['P1', 'P2', 'P3', 'P4', 'P5', 'P6', 
+            'H1', 'H2', 'H3', 'H4', 'H5', 'H6', 
+            'A1', 'A2', 'A3', 'A4', 'A5', 'A6']
+    
+    for key in ad_keys:
+        ad_dict[key] = lambda v, key=key: 0
+        
+    #TODO: FINALIZE, NOT FINISHED
+
+    return ad_dict
+
+
+def f_rf_fun_legacy(a, d, v):
+    N = len(a)
+    f = 0j
+    for l in range(0, 3):
+        f = f + a[l] * (1j/v)**l
+
+    for l in range(0, N-3):
+        f = f + a[l+2]*(1j/v) / ((1j/v + d[l]))
+    
+    f = f*v**2
+    return f
+
+
+def f_rf_fun(a, d, v):
+    N = len(a) 
+    f = np.array(a[0])*0j
+    
+    for l in range(0, 3):
+        f = f + a[l] * (1j/v)**l
+
+    for l in range(0, N-3):
+        f = f + a[l+2]*(1j/v) / ((1j/v + d[l]))
+    
+    f = f*v**2    
+
+    return f
+
+
+def rf_ads(a, d):
+    # B assumed to be implicitly included in RF factors
+    ad_dict = dict()
+    ad_keys = ['P1', 'P2', 'P3', 'P4', 'P5', 'P6', 
+               'H1', 'H2', 'H3', 'H4', 'H5', 'H6', 
+               'A1', 'A2', 'A3', 'A4', 'A5', 'A6']
+    
+    imag_component_ad = ['P1', 'P2', 'P5', 'H1', 'H2', 'H5', 'A1', 'A2', 'A5']
+    
+    position_dict = {'P1': [0,0], 'P2': [0,2], 'P3': [0,2], 'P4': [0,0], 'P5': [0,1], 'P6': [0,1],
+                 'H1': [1,1], 'H2': [1,2], 'H3': [1,2], 'H4': [1,1], 'H5': [1,0], 'H6': [1,0],
+                 'A1': [2,1], 'A2': [2,2], 'A3': [2,2], 'A4': [2,1], 'A5': [2,0], 'A6': [2,0]}
+    
+    for key in ad_keys:
+        row = position_dict[key][0]
+        col = position_dict[key][1]
+        a_key = [ai[row, col] for ai in a]
+
+        if key in imag_component_ad:
+            ad_dict[key] = lambda v, a=a_key: np.imag(f_rf_fun_legacy(a, d, v))
+        else:
+            ad_dict[key] = lambda v, a=a_key: np.real(f_rf_fun_legacy(a, d, v))
+
+    return ad_dict
+
+
+def distribute_to_dict(prefix, array, count_start=1):
+    array_dict = dict()
+    for ix,array_i in enumerate(array):
+        key = prefix + str(ix+count_start)
+        array_dict[key] = array_i
+        
+    return array_dict
+        
+
+def distribute_multi_to_dict(prefixes, arrays):
+    array_dict = dict()
+    
+    for prefix_ix, prefix in enumerate(prefixes):
+        for ix, array_i in enumerate(arrays[prefix_ix]):
+            key = prefix + str(ix+1)
+            array_dict[key] = array_i
+            
+    return array_dict
+
+
+def unwrap_rf_parameters(parameters):
+    keys = list(parameters.keys())
+    a_ixs = np.where([word.startswith('a') for word in keys])[0]
+    d_ixs  = np.where([word.startswith('d') for word in keys])[0]
+    a_nums = np.array([int(string.split('a')[1]) for string in np.array(keys)[a_ixs]])
+    d_nums = np.array([int(string.split('d')[1]) for string in np.array(keys)[d_ixs]])
+    
+    a = [np.zeros([3,3])]*(max(a_nums))
+    d = [0]*(max(d_nums))
+    
+    for a_num in a_nums:
+        a[a_num-1] = np.array(parameters['a%i' %a_num])
+    
+    for d_num in d_nums:
+        d[d_num-1] = parameters['d%i' %d_num]
+    
+    d = np.array(d)
+    return a,d
+
+
+def normal_wind(T_g2wi, T_g2el, U=1.0):
+    T_wi2el = T_g2el @ T_g2wi.T
+    e_wind_local = (T_wi2el @ np.array([1, 0, 0])[np.newaxis,:].T).flatten()
+    
+    Un = U * np.sqrt(e_wind_local[1]**2+e_wind_local[2]**2)
+    return Un
+
+
+def el_mat_generic(Ayy,Ayz,Ayt,Azy,Azz,Azt,Aty,Atz,Att,L):
+    mat = np.zeros([12,12])
+
+    mat[0:6, 0:6] = np.array([
+        [0,         0,          0,          0,          0,              0           ],
+        [0,         156*Ayy,    156*Ayz,    147*Ayt,    -22*L*Ayz,      22*L*Ayy    ],
+        [0,         156*Azy,    156*Azz,    147*Azt,    -22*L*Azz,      22*L*Azy    ],
+        [0,         147*Aty,    147*Atz,    140*Att,     -21*L*Atz,      21*L*Aty   ],
+        [0,         -22*L*Azy,  -22*L*Azz,  -21*L*Azt,  4*L**2*Azz,     -4*L**2*Azy ],
+        [0,         22*L*Ayy,   22*L*Ayz,   21*L*Ayt,   -4*L**2*Ayz,    4*L**2*Ayy  ],
+    ])
+
+    mat[0:6, 6:12] = np.array([
+        [0,         0,          0,          0,          0,              0            ],
+        [0,         54*Ayy,    54*Ayz,      63*Ayt,     13*L*Ayz,       -13*L*Ayy    ],
+        [0,         54*Azy,    54*Azz,      63*Azt,     13*L*Azz,       -13*L*Azy    ],
+        [0,         63*Aty,    63*Atz,      70*Att,     14*L*Atz,       -14*L*Aty    ],
+        [0,         -13*L*Azy,  -13*L*Azz,  -14*L*Azt,  -3*L**2*Azz,     3*L**2*Azy  ],
+        [0,         13*L*Ayy,   13*L*Ayz,   14*L*Ayt,   3*L**2*Ayz,     -3*L**2*Ayy  ],
+    ])
+
+    mat[6:12, 0:6] = np.array([
+        [0,         0,          0,          0,          0,              0            ],
+        [0,         54*Ayy,    54*Ayz,      63*Ayt,     -13*L*Ayz,       13*L*Ayy    ],
+        [0,         54*Azy,    54*Azz,      63*Azt,     -13*L*Azz,       13*L*Azy    ],
+        [0,         63*Aty,    63*Atz,      70*Att,     -14*L*Atz,       14*L*Aty    ],
+        [0,         13*L*Azy,  13*L*Azz,    14*L*Azt,   -3*L**2*Azz,     3*L**2*Azy  ],
+        [0,         -13*L*Ayy, -13*L*Ayz,   -14*L*Ayt,   3*L**2*Ayz,     -3*L**2*Ayy ],
+    ])
+
+    mat[6:12,6:12] = np.array([
+        [0,         0,          0,          0,          0,              0               ],
+        [0,         156*Ayy,    156*Ayz,    147*Ayt,    22*L*Ayz,      -22*L*Ayy        ],
+        [0,         156*Azy,    156*Azz,    147*Azt,    22*L*Azz,      -22*L*Azy        ],
+        [0,         147*Aty,    147*Atz,    140*Att,    21*L*Atz,      -21*L*Aty        ],
+        [0,         22*L*Azy,   22*L*Azz,   21*L*Azt,    4*L**2*Azz,   -4*L**2*Azy      ],
+        [0,         -22*L*Ayy,   -22*L*Ayz,   -21*L*Ayt,   -4*L**2*Ayz,    4*L**2*Ayy   ],
+    ])
+
+    return mat
+
+def element_aero_mats(B, omega, ad_dict, L, T=None, phi=None, rho=1.225):
+    # Called for selected reduced velocity, specified by omega value (implicitly mean wind).
+    # Corresponding values of P,H and A are used for given mean wind velocity.
+
+    # Stiffness
+    Ayy = 1/2*rho*B**2*omega**2*ad_dict['P4']
+    Ayz = 1/2*rho*B**2*omega**2*ad_dict['P6']
+    Ayt = -1/2*rho*B**2*omega**2*B*ad_dict['P3']
+
+    Azy = 1/2*rho*B**2*omega**2*ad_dict['H6']
+    Azz = 1/2*rho*B**2*omega**2*ad_dict['H4']
+    Azt = -1/2*rho*B**2*omega**2*B*ad_dict['H3']
+
+    Aty = -1/2*rho*B**2*omega**2*B*ad_dict['A6']
+    Atz = -1/2*rho*B**2*omega**2*B*ad_dict['A4']
+    Att = 1/2*rho*B**2*omega**2*B**2*ad_dict['A3']
+
+    k_aero = L/420 * el_mat_generic(Ayy,Ayz,Ayt,Azy,Azz,Azt,Aty,Atz,Att,L)
+
+    
+    # Damping
+    Ayy = 1/2*rho*B**2*omega*ad_dict['P1']
+    Ayz = 1/2*rho*B**2*omega*ad_dict['P5']
+    Ayt = -1/2*rho*B**2*omega*B*ad_dict['P2']
+
+    Azy = 1/2*rho*B**2*omega*ad_dict['H5']
+    Azz = 1/2*rho*B**2*omega*ad_dict['H1']    
+    Azt = -1/2*rho*B**2*omega*B*ad_dict['H2']
+    
+    Aty = -1/2*rho*B**2*omega*B*ad_dict['A5']
+    Atz = -1/2*rho*B**2*omega*B*ad_dict['A1']
+    Att = 1/2*rho*B**2*omega*B**2*ad_dict['A2']
+
+    c_aero = L/420 * el_mat_generic(Ayy,Ayz,Ayt,Azy,Azz,Azt,Aty,Atz,Att,L)
+
+    if (T is None and phi is None)!=True:
+        if T is not None:     #if no transformation matrix is given, a local matrix is output
+            if np.shape(T)[0]==6:
+                T = np.kron(np.eye(2), T)   #two times 6dof matrix, block diagonal
+            if np.shape(T)[0]==3:
+                T = np.kron(np.eye(4), T)   #four times 3dof matrix, block diagonal
+            elif np.shape(T)[0]!=12:
+                raise ValueError('Wrong size of T (should be 3x3, 6x6 or 12x12')
+        else:
+            T = np.eye(12)
+        
+        if phi is not None:
+            T = T @ phi
+            
+        k_aero = T.T @ k_aero @ T
+        c_aero = T.T @ c_aero @ T
+    
+    return k_aero, c_aero
+
+
+# Spectra
+def kaimal_auto(omega, Lx, A, sigma, V):
+    f = omega/(2*np.pi)
+    fhat = f*Lx/V
+    S = (sigma**2*(A*fhat)/(1+(1.5*A*fhat))**(5/3))/f
+
+    return S/(2*np.pi)
+
+def von_Karman_auto(omega, Lx, sigma, V):
+   
+    A1 = [
+        0.0,
+        0.0,
+        755.2,
+        ]
+    
+    A2 = [
+        70.8,
+        0.0,
+        283.2,
+        ]
+    
+    rr = [
+        5/6,
+        11/6,
+        11/6,
+        ]
+    
+    f = omega/(2*np.pi)
+    fhat = f*Lx/V
+    S = (sigma**2*( (4*fhat)*(1+A1*fhat**2)  )/ (1+A2*fhat**2)**(rr))/f
+    
+    return S/(2*np.pi)
+
+def generic_kaimal_matrix(omega, nodes, T_wind, A, sigma, C, Lx, U, options=None):
+    # Adopted from MATLAB version. `nodes` is list with beef-nodes.        
+    V = np.zeros(len(nodes))      # Initialize vector with mean wind in all nodes
+    Su = np.zeros([len(nodes), len(nodes)])     # One-point spectra for u component in all nodes
+    Sv = np.zeros([len(nodes), len(nodes)])     # One-point spectra for v component in all nodes
+    Sw = np.zeros([len(nodes), len(nodes)])     # One-point spectra for w component in all nodes
+    xyz = np.zeros([len(nodes), 3])  # Nodes in wind coordinate system
+
+    if options is None:
+        options = {
+            'spectra_type': 'Kaimal'
+            }
+
+    for node_ix, node in enumerate(nodes):
+        xyz[node_ix,:] = (T_wind @ node.coordinates).T #Transform node coordinates to the wind coordinate system
+        V[node_ix] = U(node.coordinates) # Mean wind velocity in the nodes
+
+        if 'spectra_type' in options:
+            if options['spectra_type'] == 'vonKarman':
+                Su[node_ix,:], Sv[node_ix,:], Sw[node_ix,:] = von_Karman_auto(omega, Lx, sigma, V[node_ix]) 
+            elif options['spectra_type'] == 'Kaimal':
+                Su[node_ix,:], Sv[node_ix,:], Sw[node_ix,:] = kaimal_auto(omega, Lx, A, sigma, V[node_ix]) # One point spectra for u component in all nodes
+        else: # use Kaimal (default)
+            Su[node_ix,:], Sv[node_ix,:], Sw[node_ix,:] = kaimal_auto(omega, Lx, A, sigma, V[node_ix])
+
+    x = xyz[:, 0]
+    y = xyz[:, 1]
+    z = xyz[:, 2]
+    
+    dxdx = x[np.newaxis,:] - x[np.newaxis,:].T # Matrix with all distances between nodes in x direction
+    dydy = y[np.newaxis,:] - y[np.newaxis,:].T # Matrix with all distances between nodes in y direction
+    dzdz = z[np.newaxis,:] - z[np.newaxis,:].T # Matrix with all distances between nodes in z direction
+    
+    invV = 2/(V[np.newaxis,:]+V[np.newaxis,:].T) # Inverse mean wind velocity for all combination of nodes
+
+    Suu = np.sqrt(Su)*np.sqrt(Su).T*np.exp(
+            -invV*omega/(2*np.pi)*np.sqrt(
+                (C[0,0]*dxdx)**2 + (C[1,0]*dydy)**2 + (C[2,0]*dzdz)**2)
+        )
+    
+    Svv = np.sqrt(Sv)*np.sqrt(Sv).T*np.exp(
+            -invV*omega/(2*np.pi)*np.sqrt(         
+                (C[0,1]*dxdx)**2 + (C[1,1]*dydy)**2 + (C[2,1]*dzdz)**2)
+        )
+    
+    Sww = np.sqrt(Sw)*np.sqrt(Sw).T*np.exp(
+            -invV*omega/(2*np.pi)*np.sqrt( 
+                (C[0,2]*dxdx)**2 + (C[1,2]*dydy)**2 + (C[2,2]*dzdz)**2)
+        )
+    
+    SvSv = np.zeros([3*len(nodes), 3*len(nodes)]) # Cross sectral density matrix containing all the turbulence components
+    SvSv[0::3, 0::3] = Suu
+    SvSv[1::3, 1::3] = Svv
+    SvSv[2::3, 2::3] = Sww
+    
+    return SvSv
+
+
+def loadmatrix_fe(V, load_coefficients, rho, B, D, admittance=None):
+
+    if admittance is None :
+        admittance = lambda omega_k: np.ones( (4,3) )
+
+    Cd = load_coefficients['Cd']
+    dCd = load_coefficients['dCd']
+    Cl = load_coefficients['Cl']
+    dCl = load_coefficients['dCl']
+    Cm = load_coefficients['Cm']
+    dCm = load_coefficients['dCm']
+    
+    # Equation 7 from Oiseth, 2010
+    BqBq = lambda omega_k: 1/2*rho*V*B*admittance(omega_k*B/V/2/np.pi)*np.array([[0,            0,          0],
+                                                                                [0,     2*D/B*Cd,          (D/B*dCd-Cl)],
+                                                                                [0,         2*Cl,          (dCl+D/B*Cd)],
+                                                                                [0,      -2*B*Cm,          -B*dCm]])
+                                          
+    return BqBq
+
+def loadmatrix_fe_static(V, load_coefficients, rho, B, D ):
+    
+    Cd = load_coefficients['Cd']
+    Cl = load_coefficients['Cl']
+    Cm = load_coefficients['Cm']
+        
+    BqBq = 1/2*rho*V**2*B*np.array([[ 0 ],
+                                    [ D/B*Cd ],
+                                    [ Cl  ],
+                                    [ -B*Cm ]])
+    return BqBq
+
+def loadvector(T_el, Bq, T_wind, L, static = False):
+
+    G = np.zeros([12,4])
+    G[0,0] = L/2
+    G[1,1] = L/2
+    G[2,2] = L/2
+    G[3,3] = L/2
+    G[6,0] = L/2
+    G[7,1] = L/2
+    G[8,2] = L/2
+    G[9,3] = L/2
+    G[4,2] = -L**2/12
+    G[5,1] = L**2/12
+    G[10,2] = L**2/12
+    G[11,1] = -L**2/12
+ 
+    # Transform from wind coordinates to local element coordinates 
+    
+    T = T_el @ T_wind.T   
+    
+    T_full = blkdiag(T_el, 4)     # Block diagonal - repeated 4 times to transform both trans and rot DOFs at each node (2+2)
+    
+    # T_full.T transforms L-->G
+    if static is False:
+        R =  T_full.T @ G @ Bq @ T 
+    else:
+        R =  T_full.T @ G @ Bq 
+        
+    R1 =  R[0:6]         # Element node 1
+    R2 =  R[6:12]        # Element node 2
+
+
+    return R1, R2
+
+
+def windaction(omega, S, load_coefficients, elements, T_wind, 
+               phi, B, D, U, omega_reduced=None, rho=1.225, print_progress=True,
+               section_lookup=None, nodes=None, admittance=None):
+    
+    if nodes is None:
+        nodes = list(set([a for b in [el.nodes for el in elements] for a in b]))
+
+    n_dofs = 6
+    
+    # Ensure that first omega value is not 0 when using logspace omega axis
+    if omega_reduced is None:
+        if np.min(omega) == 0:
+            omega_sorted = np.sort(omega)
+            omega_start = omega_sorted[1]
+        else:
+            omega_start = np.min(omega)    
+        
+        omega_reduced = np.logspace(np.log10(omega_start), np.log10(np.max(omega)), num=50) # A log frequency axis that is used to obtain the cross-spectral density matrix
+        
+    genSqSq_reduced = np.zeros([phi.shape[1], phi.shape[1], len(omega_reduced)]) # Initialize the cross-spectral density matrix
+    
+    # Establish RG matrix (common for all freqs)
+    
+
+    if section_lookup is None:
+        lc_fun = lambda el: load_coefficients
+        B_fun = lambda el: B
+        D_fun = lambda el: D
+        admittance_fun = lambda el: admittance
+    else:
+        def get_sec(el):
+            for key in section_lookup:
+                if el in section_lookup[key]:    
+                    return key 
+
+        lc_fun = lambda el: load_coefficients[get_sec(el)]
+        B_fun = lambda el: B[get_sec(el)]
+        D_fun = lambda el: D[get_sec(el)]
+    
+    if admittance is None: # omit the frequency loop if ADmittance is not included - faster !
+        RG = np.zeros([len(nodes)*n_dofs, 3])
+        for el in elements:            
+            node1_dofs = el.nodes[0].global_dofs
+            node2_dofs = el.nodes[1].global_dofs
+
+            mean_wind = U(el.get_cog())         
+            Vn = normal_wind(T_wind, el.T0)*mean_wind        # Find the normal wind
+            BqBq = loadmatrix_fe(Vn, lc_fun(el), rho, B_fun(el), D_fun(el))
+            R1, R2 = loadvector(el.T0, BqBq(1), T_wind, el.L) # Obtain the load vector for each element
+
+            RG[node1_dofs, :] = RG[node1_dofs, :] + R1   # Add the contribution from the element (end 1) to the system
+            RG[node2_dofs, :] = RG[node2_dofs, :] + R2   # Add the contribution from the element (end 2) to the system
+
+            # Make block matrix
+            RG_block = np.zeros([6*len(nodes), 3*len(nodes)])
+            
+            for node in nodes:
+                ix = node.index
+                n = np.r_[6*ix:6*ix+6]
+                m = np.r_[3*ix:3*ix+3]
+                RG_block[np.ix_(n,m)] = RG[n,:]     #verified with MATLAB version for beam example
+
+            for k, omega_k in enumerate(omega_reduced): 
+                if print_progress:
+                    pp(k+1, len(omega_reduced), sym='=', postfix=' ESTABLISHING WIND EXCITATION')
+                    print('')
+                    
+                phiT_RG_block = phi.T @ RG_block
+                genSqSq_reduced[:, :, k] = phiT_RG_block @ S(omega_k) @ phiT_RG_block.T  # to modal coordinates
+
+    else: # admittance is given - triple loop (the old way, slower)
+        admittance_fun = lambda el: admittance[get_sec(el)]
+
+        for k, omega_k in enumerate(omega_reduced): 
+            if print_progress:
+                pp(k+1, len(omega_reduced), sym='=', postfix=' ESTABLISHING WIND EXCITATION')
+                print('')
+        
+            # Establish RG matrix 
+            RG = np.zeros([len(nodes)*n_dofs, 3])
+            
+            for el in elements:            
+                node1_dofs = el.nodes[0].global_dofs
+                node2_dofs = el.nodes[1].global_dofs
+
+                mean_wind = U(el.get_cog())         
+                Vn = normal_wind(T_wind, el.T0)*mean_wind        # Find the normal wind
+                BqBq = loadmatrix_fe(Vn, lc_fun(el), rho, B_fun(el), D_fun(el), admittance=admittance_fun(el))
+                R1, R2 = loadvector(el.T0, BqBq(omega_k), T_wind, el.L) # Obtain the load vector for each element
+
+                RG[node1_dofs, :] = RG[node1_dofs, :] + R1   # Add the contribution from the element (end 1) to the system
+                RG[node2_dofs, :] = RG[node2_dofs, :] + R2   # Add the contribution from the element (end 2) to the system
+                
+                
+            # Make block matrix
+            RG_block = np.zeros([6*len(nodes), 3*len(nodes)])
+            
+            for node in nodes:
+                ix = node.index
+                n = np.r_[6*ix:6*ix+6]
+                m = np.r_[3*ix:3*ix+3]
+                RG_block[np.ix_(n,m)] = RG[n,:]     #verified with MATLAB version for beam example
+                
+            phiT_RG_block = phi.T @ RG_block
+            genSqSq_reduced[:, :, k] = phiT_RG_block @ S(omega_k) @ phiT_RG_block.T  # to modal coordinates
+
+
+    # Interpolate results to full frequency axis
+    genSqSq = interp1d(omega_reduced, genSqSq_reduced, kind='quadratic', axis=2, fill_value=0, bounds_error=False)
+    
+    return genSqSq
+
+def windaction_static(load_coefficients, elements, T_wind, 
+               phi, B, D, U, rho=1.225, print_progress=True,
+               section_lookup=None, nodes=None):
+    
+    if nodes is None:
+        nodes = list(set([a for b in [el.nodes for el in elements] for a in b]))
+
+    n_dofs = 6
+    
+    if section_lookup is None:
+        lc_fun = lambda el: load_coefficients
+        B_fun = lambda el: B
+        D_fun = lambda el: D
+    else:
+        def get_sec(el):
+            for key in section_lookup:
+                if el in section_lookup[key]:    
+                    return key 
+
+        lc_fun = lambda el: load_coefficients[get_sec(el)]
+        B_fun = lambda el: B[get_sec(el)]
+        D_fun = lambda el: D[get_sec(el)]
+            
+        # Establish RG matrix 
+        RG = np.zeros([len(nodes)*n_dofs])
+        
+        for el in elements:            
+            node1_dofs = el.nodes[0].global_dofs
+            node2_dofs = el.nodes[1].global_dofs
+
+            mean_wind = U(el.get_cog())         
+            Vn = normal_wind(T_wind, el.T0)*mean_wind        # Find the normal wind
+            BqBq = loadmatrix_fe_static(Vn, lc_fun(el), rho, B_fun(el), D_fun(el))
+            R1, R2 = loadvector(el.T0, BqBq, T_wind, el.L, static = True) # Obtain the load vector for each element
+
+            RG[node1_dofs] = RG[node1_dofs] + R1[:,0]   # Add the contribution from the element (end 1) to the system
+            RG[node2_dofs] = RG[node2_dofs] + R2[:,0]    # Add the contribution from the element (end 2) to the system
+            
+        # Make block matrix
+        RG_block = np.zeros([6*len(nodes)])
+        
+        for node in nodes:
+            ix = node.index
+            n = np.r_[6*ix:6*ix+6]
+            RG_block[np.ix_(n)] = RG[n]     #verified with MATLAB version for beam example
+            
+               
+        genSqSq = phi.T @ RG_block 
+    
+    return genSqSq
+
+def K_from_ad(ad, V, w, B, rho):
+    if w==0:
+        k = np.zeros([3,3])
+    else:
+        v = V / (B*w) # reduced velocity  
+        
+        k = (0.5*rho*B**2*w**2 * 
+                np.vstack([[ad['P4'](v), ad['P6'](v), -B*ad['P3'](v)],
+                           [ad['H6'](v), ad['H4'](v), -B*ad['H3'](v)],
+                           [-B*ad['A6'](v), -B*ad['A4'](v), B**2*ad['A3'](v)]]))
+        
+        
+    return k
+
+
+def C_from_ad(ad, V, w, B, rho):
+    if w==0:
+        c = np.zeros([3,3])
+    else:
+        v = V / (B*w) # reduced velocity    
+    
+        c = (0.5*rho*B**2*w * 
+            np.vstack([[ad['P1'](v), ad['P5'](v), -B*ad['P2'](v)],
+                      [ad['H5'](v), ad['H1'](v), -B*ad['H2'](v)],
+                      [-B*ad['A5'](v), -B*ad['A1'](v), B**2*ad['A2'](v)]]))  
+    
+    return c
+
+
+def phi_aero_sum(mat, phi, x):
+    n_modes = phi.shape[1]
+    n_points = len(x)
+    
+    mat_int = np.zeros([n_modes, n_modes, n_points])
+    
+    for p in range(n_points):
+        phi_point = phi[p*6+1:p*6+4, :]
+        mat_int[:, :, p] = phi_point.T @ mat @ phi_point
+
+    mat = np.trapz(mat_int, x=x, axis=2)
+
+    return mat
+
+
+def function_sum(fun, const, fun_factor=1):
+    def fsum(x):
+        if fun is None:
+            return const
+        else:
+            return fun(x)*fun_factor + const
+    
+    return fsum
+
+
+def get_aero_cont_adfun(ad_dict_fun, V, B, rho, phi, x):
+    def K(w):
+        n_modes = phi.shape[1]
+        n_points = len(x)
+        
+        mat_int = np.zeros([n_modes, n_modes, n_points])
+        
+        for p in range(n_points):
+            phi_point = phi[p*6+1:p*6+4, :]
+            kae = K_from_ad(ad_dict_fun(x[p]), V, w, B, rho)
+            mat_int[:, :, p] = phi_point.T @ kae @ phi_point
+
+        return np.trapz(mat_int, x=x, axis=2)
+    
+    
+    def C(w):
+        n_modes = phi.shape[1]
+        n_points = len(x)
+        
+        mat_int = np.zeros([n_modes, n_modes, n_points])
+        
+        for p in range(n_points):
+            phi_point = phi[p*6+1:p*6+4, :]
+            kae = C_from_ad(ad_dict_fun(x[p]), V, w, B, rho)
+            mat_int[:, :, p] = phi_point.T @ kae @ phi_point
+
+        return np.trapz(mat_int, x=x, axis=2)
+    
+        
+    return K, C
+
+
+def get_aero_cont_addict(ad_dict, V, B, rho, phi, x):
+    def K(w):
+        kae = K_from_ad(ad_dict, V, w, B, rho)
+        return phi_aero_sum(kae, phi, x)
+    
+    def C(w):
+        cae = C_from_ad(ad_dict, V, w, B, rho)
+        return phi_aero_sum(cae, phi, x)
+        
+    return K, C
+
+
+def itflutter_cont(Ms, Cs, Ks, phi, x, ad_dict, B, V=0.0, rho=1.225, dV=1, 
+              overshoot_factor=0.5, itmax={}, omega_ref=None,
+              tol={}, print_progress=True, keep_all=False, track_by_psi=True):
+    
+    if callable(ad_dict):
+        get_aero = get_aero_cont_adfun
+    else:
+        get_aero = get_aero_cont_addict
+
+    itmax_ = {'V':50, 'f': 15}
+    itmax_.update(**itmax)
+    itmax = dict(itmax_)
+
+    tol_ = {'V': 1e-3, 'f': 1e-4}
+    tol_.update(**tol)
+    tol = tol_
+    
+    res = dict()
+    res['V'] = []
+    res['lambd'] = []
+    res['critical_mode'] = []
+    res['critical_psi'] = []
+    
+    converged = False
+    psi_prev = None
+    
+    if omega_ref is None:
+        A = statespace(Ks, Cs, Ms)
+        lambd_ref, psi = np.linalg.eig(A)  
+        omega_initial = np.sort(np.abs(np.imag(lambd_ref)))[::2]
+        omega_ref = omega_initial[0]
+        
+    for it_vel in range(itmax['V']):
+        Kae, Cae = get_aero(ad_dict, V, B, rho, phi, x)
+        getK = function_sum(Kae, Ks, fun_factor=-1)
+        getC = function_sum(Cae, Cs, fun_factor=-1)
+        getM = function_sum(None, Ms, fun_factor=-1)
+
+        lambd, psi, not_converged = iteig(getK, getC, getM, tol=tol['f'], 
+                                          keep_full=True, mac_min=0.0, itmax=itmax['f'])
+
+        if len(not_converged)>0:
+            lambd[not_converged] = -np.inf + 0j
+            if print_progress:
+                if len(not_converged)<10:
+                    nc_modes = 'index '+ ', '.join([str(i) for i in not_converged])
+                else:
+                    nc_modes = '>10'
+                print(f'** Non-converged modes ({nc_modes}) from iterative eigensolution disregarded! **')
+                
+        if it_vel!=0 and track_by_psi:
+            ixs, __, __, __ = restructure_as_ref(psi_prev, psi)
+            
+            psi = psi[:, ixs]
+            lambd = lambd[ixs] 
+        
+        psi_prev = psi*1
+            
+        critical_mode = np.argmax(np.real(lambd))
+        real_lambd = np.max(np.real(lambd)) 
+        critical_omega = np.abs(np.imag(lambd[critical_mode]))
+        
+        if keep_all or real_lambd<=0:
+            res['critical_mode'].append(critical_mode)
+            res['lambd'].append(lambd)
+            res['V'].append(V)
+            res['critical_psi'].append(psi[:,critical_mode])
+        
+        if dV < tol['V'] and real_lambd<=0:
+            converged = True
+            if print_progress:
+                print(conv_text)
+                print(f'Flutter estimated to occur at V = {V:.2f} m/s ({critical_omega:.2f} rad/s) ==>  v = {V/(B*critical_omega):.2f})\n')
+            
+            break
+        elif real_lambd<0:
+            if print_progress:
+                print(f'Increasing velocity V = {V:.2f} --> {V+dV:.2f}.')
+            V = V + dV
+        else:
+            if print_progress:
+                print(f'Overshot. Reducing velocity V = {V:.2f} --> {V-dV/2:.2f}. Reducing step size dV = {dV:.2f} --> {dV/2:.2f}')
+                
+            dV = overshoot_factor*dV  # adjusting the velocity increment, and step backwards   
+            V = V - dV
+
+    if not converged and print_progress:
+        print('Not able to converge within specified maximum iterations for specified tolerance criteria.')
+    
+    res = {key: np.array(res[key]) for key in ['critical_mode', 'critical_psi', 'V', 'lambd']}
+    
+    return res
+
+
+
+def itflutter_cont_naive(Ms, Cs, Ks, phi, x, ad_dict, B, V=0.0, rho=1.225, dV=1, 
+              overshoot_factor=0.5, itmax={}, tol={}, print_progress=True):
+        
+        
+    if callable(ad_dict):
+        get_aero = get_aero_cont_adfun
+    else:
+        get_aero = get_aero_cont_addict
+    
+    itmax_ = {'V':50, 'f': 15}
+    itmax_.update(**itmax)
+    itmax = itmax_
+    
+    tol_ = {'V': 1e-3, 'f': 1e-4}
+    tol_.update(**tol)
+    tol = tol_
+    
+    res = dict()
+    res['V'] = []
+    res['lambd'] = []
+    res['critical_mode'] = []
+    res['critical_psi'] = []
+    
+    converged = False
+        
+    for it_vel in range(itmax['V']):
+        Kae, Cae = get_aero(ad_dict, V, B, rho, phi, x)
+        getK = function_sum(Kae, Ks, fun_factor=-1)
+        getC = function_sum(Cae, Cs, fun_factor=-1)
+        getM = function_sum(None, Ms, fun_factor=-1)
+
+        lambd, psi = iteig_naive(getK, getC, getM, tol=tol['f'], itmax=itmax['f'])
+        
+        complex_ix = np.imag(lambd) != 0
+        
+        critical_mode = np.argmax(np.real(lambd[complex_ix]))
+        critical_mode = np.where(complex_ix)[0][critical_mode]
+        
+        real_lambd = np.max(np.real(lambd)) 
+        critical_omega = np.abs(np.imag(lambd[critical_mode]))
+        
+        if real_lambd<=0:
+            res['critical_mode'].append(critical_mode)
+            res['lambd'].append(lambd)
+            res['V'].append(V)
+            res['critical_psi'].append(psi[:,critical_mode])
+        
+        if dV < tol['V'] and real_lambd<=0:
+            
+            converged = True
+            if print_progress:
+
+                print(conv_text)
+                print(f'Flutter estimated to occur at V = {V:.2f} m/s ({critical_omega:.2f} rad/s) ==>  v = {V/(B*critical_omega):.2f})\n')
+            
+            break
+        elif real_lambd<=0:
+            if print_progress:
+                print(f'Increasing velocity V = {V:.2f} --> {V+dV:.2f}.')
+            V = V + dV
+        else:
+            if print_progress:
+                print(f'Overshot. Reducing velocity V = {V:.2f} --> {V-dV/2:.2f}. Reducing step size dV = {dV:.2f} --> {dV/2:.2f}')
+                
+            dV = overshoot_factor*dV  # adjusting the velocity increment, and step backwards   
+            V = V - dV
+
+    if not converged and print_progress:
+        print('Not able to converge within specified maximum iterations for specified tolerance criteria.')
+    
+    res = {key: np.array(res[key]) for key in ['critical_mode', 'critical_psi', 'V', 'lambd']}
+    
+    return res
+