import numpy as nm
from sfepy.linalg import dot_sequences
from sfepy.homogenization.utils import iter_sym
from sfepy.terms.terms import Term, terms
from sfepy.terms.terms_th import THTerm, ETHTerm
import sfepy.mechanics.membranes as membranes
## expr = """
## e = 1/2 * (grad( vec( u ) ) + grad( vec( u ) ).T)
## D = map( D_sym )
## s = D * e
## div( s )
## """
## """
## e[i,j] = 1/2 * (der[j]( u[i] ) + der[i]( u[j] ))
## map =
## D[i,j,k,l]
## s[i,j] = D[i,j,k,l] * e[k,l]
## """
[docs]
class LinearElasticTerm(Term):
r"""
General linear elasticity term, with :math:`D_{ijkl}` given in
the usual matrix form exploiting symmetry: in 3D it is :math:`6\times6`
with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in 2D it is
:math:`3\times3` with the indices ordered as :math:`[11, 22, 12]`. Can be
evaluated. Can use derivatives.
:Definition:
.. math::
\int_{\Omega} D_{ijkl}\ e_{ij}(\ul{v}) e_{kl}(\ul{u})
:Arguments 1:
- material : :math:`D_{ijkl}`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
:Arguments 2:
- material : :math:`D_{ijkl}`
- parameter_1 : :math:`\ul{w}`
- parameter_2 : :math:`\ul{u}`
"""
name = 'dw_lin_elastic'
arg_types = (('material', 'virtual', 'state'),
('material', 'parameter_1', 'parameter_2'))
arg_shapes = {'material' : 'S, S', 'virtual' : ('D', 'state'),
'state' : 'D', 'parameter_1' : 'D', 'parameter_2' : 'D'}
modes = ('weak', 'eval')
## symbolic = {'expression': expr,
## 'map' : {'u' : 'state', 'D_sym' : 'material'}}
[docs]
def get_fargs(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(state)
if mode == 'weak':
if diff_var is None:
strain = self.get(state, 'cauchy_strain')
fmode = 0
else:
strain = nm.array([0], ndmin=4, dtype=nm.float64)
fmode = 1
return 1.0, strain, mat, vg, fmode
elif mode == 'eval':
strain1 = self.get(virtual, 'cauchy_strain')
strain2 = self.get(state, 'cauchy_strain')
return 1.0, strain1, strain2, mat, vg
else:
raise ValueError('unsupported evaluation mode in %s! (%s)'
% (self.name, mode))
[docs]
def get_eval_shape(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(state)
return (n_el, 1, 1, 1), state.dtype
[docs]
def set_arg_types(self):
if self.mode == 'weak':
self.function = terms.dw_lin_elastic
else:
self.function = terms.d_lin_elastic
[docs]
class LinearElasticIsotropicTerm(LinearElasticTerm):
r"""
Isotropic linear elasticity term.
:Definition:
.. math::
\int_{\Omega} D_{ijkl}\ e_{ij}(\ul{v}) e_{kl}(\ul{u})\\ \mbox{ with } \\
D_{ijkl} = \mu (\delta_{ik} \delta_{jl}+\delta_{il} \delta_{jk}) +
\lambda \ \delta_{ij} \delta_{kl}
:Arguments:
- material_1: :math:`\lambda`
- material_2: :math:`\mu`
- virtual/parameter_1: :math:`\ul{v}`
- state/parameter_2: :math:`\ul{u}`
"""
name = 'dw_lin_elastic_iso'
arg_types = (('material_1', 'material_2', 'virtual', 'state'),
('material_1', 'material_2', 'parameter_1', 'parameter_2'))
arg_shapes = {'material_1' : '1, 1', 'material_2' : '1, 1',
'virtual' : ('D', 'state'), 'state' : 'D',
'parameter_1' : 'D', 'parameter_2' : 'D'}
geometries = ['2_3', '2_4', '3_4', '3_8']
[docs]
def get_fargs(self, lam, mu, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.mechanics.matcoefs import stiffness_from_lame
mat = stiffness_from_lame(self.region.dim, lam, mu)[:, :, 0, 0, :, :]
return LinearElasticTerm.get_fargs(self, mat, virtual, state,
mode=mode, term_mode=term_mode,
diff_var=diff_var, **kwargs)
[docs]
def get_eval_shape(self, mat1, mat2, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
return LinearElasticTerm.get_eval_shape(self, None, None, state)
[docs]
class SDLinearElasticTerm(Term):
r"""
Sensitivity analysis of the linear elastic term.
:Definition:
.. math::
\int_{\Omega} \hat{D}_{ijkl}\ e_{ij}(\ul{v}) e_{kl}(\ul{u})
.. math::
\hat{D}_{ijkl} = D_{ijkl}(\nabla \cdot \ul{\Vcal})
- D_{ijkq}{\partial \Vcal_l \over \partial x_q}
- D_{iqkl}{\partial \Vcal_j \over \partial x_q}
:Arguments:
- material : :math:`D_{ijkl}`
- parameter_w : :math:`\ul{w}`
- parameter_u : :math:`\ul{u}`
- parameter_mv : :math:`\ul{\Vcal}`
"""
name = 'ev_sd_lin_elastic'
arg_types = ('material', 'parameter_w', 'parameter_u',
'parameter_mv')
arg_shapes = {'material' : 'S, S',
'parameter_w' : 'D', 'parameter_u' : 'D',
'parameter_mv' : 'D'}
geometries = ['2_3', '2_4', '3_4', '3_8']
function = staticmethod(terms.d_sd_lin_elastic)
[docs]
def get_fargs(self, mat, par_w, par_u, par_mv,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(par_u)
grad_w = self.get(par_w, 'grad')
grad_u = self.get(par_u, 'grad')
grad_mv = self.get(par_mv, 'grad')
return 1.0, grad_w, grad_u, grad_mv, mat, vg
[docs]
def get_eval_shape(self, mat, par_w, par_u, par_mv,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(par_u)
return (n_el, 1, 1, 1), par_u.dtype
[docs]
class LinearElasticTHTerm(THTerm):
r"""
Fading memory linear elastic (viscous) term. Can use derivatives.
:Definition:
.. math::
\int_{\Omega} \left [\int_0^t
\Hcal_{ijkl}(t-\tau)\,e_{kl}(\ul{u}(\tau)) \difd{\tau}
\right]\,e_{ij}(\ul{v})
:Arguments:
- ts : :class:`TimeStepper` instance
- material : :math:`\Hcal_{ijkl}(\tau)`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
"""
name = 'dw_lin_elastic_th'
arg_types = ('ts', 'material', 'virtual', 'state')
arg_shapes = {'material' : '.: N, S, S',
'virtual' : ('D', 'state'), 'state' : 'D'}
function = staticmethod(terms.dw_lin_elastic)
[docs]
def get_fargs(self, ts, mats, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(state)
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(state)
if mode == 'weak':
if diff_var is None:
def iter_kernel():
for ii, mat in enumerate(mats):
strain = self.get(state, 'cauchy_strain',
step=-ii)
mat = nm.tile(mat, (n_el, n_qp, 1, 1))
yield ii, (ts.dt, strain, mat, vg, 0)
fargs = iter_kernel
else:
strain = nm.array([0], ndmin=4, dtype=nm.float64)
mat = nm.tile(mats[0], (n_el, n_qp, 1, 1))
fargs = ts.dt, strain, mat, vg, 1
return fargs
else:
raise ValueError('unsupported evaluation mode in %s! (%s)'
% (self.name, mode))
[docs]
class LinearElasticETHTerm(ETHTerm):
r"""
This term has the same definition as dw_lin_elastic_th, but assumes an
exponential approximation of the convolution kernel resulting in much
higher efficiency. Can use derivatives.
:Definition:
.. math::
\int_{\Omega} \left [\int_0^t
\Hcal_{ijkl}(t-\tau)\,e_{kl}(\ul{u}(\tau)) \difd{\tau}
\right]\,e_{ij}(\ul{v})
:Arguments:
- ts : :class:`TimeStepper` instance
- material_0 : :math:`\Hcal_{ijkl}(0)`
- material_1 : :math:`\exp(-\lambda \Delta t)` (decay at :math:`t_1`)
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
"""
name = 'dw_lin_elastic_eth'
arg_types = ('ts', 'material_0', 'material_1', 'virtual', 'state')
arg_shapes = {'material_0' : 'S, S', 'material_1' : '1, 1',
'virtual' : ('D', 'state'), 'state' : 'D'}
function = staticmethod(terms.dw_lin_elastic)
[docs]
def get_fargs(self, ts, mat0, mat1, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _, key = self.get_mapping(state, return_key=True)
if diff_var is None:
strain = self.get(state, 'cauchy_strain')
key += tuple(self.arg_names[ii] for ii in [1, 2, 4])
data = self.get_eth_data(key, state, mat1, strain)
fargs = (ts.dt, data.history + data.values, mat0, vg, 0)
else:
aux = nm.array([0], ndmin=4, dtype=nm.float64)
fargs = (ts.dt, aux, mat0, vg, 1)
return fargs
[docs]
class LinearPrestressTerm(Term):
r"""
Linear prestress term, with the prestress :math:`\sigma_{ij}` given either
in the usual vector form exploiting symmetry: in 3D it has 6 components
with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in 2D it has
3 components with the indices ordered as :math:`[11, 22, 12]`, or in the
matrix (possibly non-symmetric) form. Can be evaluated.
:Definition:
.. math::
\int_{\Omega} \sigma_{ij} e_{ij}(\ul{v})
:Arguments 1:
- material : :math:`\sigma_{ij}`
- virtual : :math:`\ul{v}`
:Arguments 2:
- material : :math:`\sigma_{ij}`
- parameter : :math:`\ul{u}`
"""
name = 'dw_lin_prestress'
arg_types = (('material', 'virtual'),
('material', 'parameter'))
arg_shapes = [{'material' : 'S, 1', 'virtual' : ('D', None),
'parameter' : 'D'},
{'material' : 'D, D'}]
modes = ('weak', 'eval')
[docs]
def get_fargs(self, mat, virtual,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(virtual)
sh = mat.shape
is_nonsym = sh[2] == sh[3] == vg.dim and not(vg.dim == 1)
if is_nonsym:
mat = mat.reshape(sh[:2] + (vg.dim**2, 1))
if mode == 'weak':
return mat, vg
else:
if is_nonsym:
strain = self.get(virtual, 'grad').transpose((0,1,3,2))
nel, nqp, nr, nc = strain.shape
strain = strain.reshape((nel, nqp, nr*nc, 1))
else:
strain = self.get(virtual, 'cauchy_strain')
fmode = {'eval' : 0, 'el_avg' : 1, 'qp' : 2}.get(mode, 1)
return strain, mat, vg, fmode
[docs]
def get_eval_shape(self, mat, virtual,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(virtual)
if mode != 'qp':
n_qp = 1
return (n_el, n_qp, 1, 1), virtual.dtype
[docs]
def d_lin_prestress(self, out, strain, mat, vg, fmode):
aux = dot_sequences(mat, strain, mode='ATB')
if fmode == 2:
out[:] = aux
status = 0
else:
status = vg.integrate(out, aux, fmode)
return status
[docs]
def set_arg_types(self):
if self.mode == 'weak':
self.function = terms.dw_lin_prestress
else:
self.function = self.d_lin_prestress
[docs]
class LinearStrainFiberTerm(Term):
r"""
Linear (pre)strain fiber term with the unit direction vector
:math:`\ul{d}`.
:Definition:
.. math::
\int_{\Omega} D_{ijkl} e_{ij}(\ul{v}) \left(d_k d_l\right)
:Arguments:
- material_1 : :math:`D_{ijkl}`
- material_2 : :math:`\ul{d}`
- virtual : :math:`\ul{v}`
"""
name = 'dw_lin_strain_fib'
arg_types = ('material_1', 'material_2', 'virtual')
arg_shapes = {'material_1' : 'S, S', 'material_2' : 'D, 1',
'virtual' : ('D', None)}
function = staticmethod(terms.dw_lin_strain_fib)
[docs]
def get_fargs(self, mat1, mat2, virtual,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(virtual)
omega = nm.empty(mat1.shape[:3] + (1,), dtype=nm.float64)
for ii, (ir, ic) in enumerate(iter_sym(mat2.shape[2])):
omega[..., ii, 0] = mat2[..., ir, 0] * mat2[..., ic, 0]
return mat1, omega, vg
[docs]
class CauchyStrainTerm(Term):
r"""
Evaluate Cauchy strain tensor.
It is given in the usual vector form exploiting symmetry: in 3D it has 6
components with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in
2D it has 3 components with the indices ordered as :math:`[11, 22,
12]`. The last three (non-diagonal) components are doubled so that it is
energetically conjugate to the Cauchy stress tensor with the same storage.
Supports 'eval', 'el_avg' and 'qp' evaluation modes.
:Definition:
.. math::
\int_{\cal{D}} \ull{e}(\ul{w})
:Arguments:
- parameter : :math:`\ul{w}`
"""
name = 'ev_cauchy_strain'
arg_types = ('parameter',)
arg_shapes = {'parameter' : 'D'}
integration = ('cell', 'facet_extra')
[docs]
@staticmethod
def function(out, strain, vg, fmode):
if fmode == 2:
out[:] = strain
status = 0
else:
status = terms.de_cauchy_strain(out, strain, vg.cmap, fmode)
return status
[docs]
def get_fargs(self, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(parameter)
strain = self.get(parameter, 'cauchy_strain')
fmode = {'eval' : 0, 'el_avg' : 1, 'qp' : 2}.get(mode, 1)
return strain, vg, fmode
[docs]
def get_eval_shape(self, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(parameter)
if mode != 'qp':
n_qp = 1
return (n_el, n_qp, dim * (dim + 1) // 2, 1), parameter.dtype
[docs]
class CauchyStressTerm(Term):
r"""
Evaluate Cauchy stress tensor.
It is given in the usual vector form exploiting symmetry: in 3D it has 6
components with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in
2D it has 3 components with the indices ordered as :math:`[11, 22, 12]`.
Supports 'eval', 'el_avg' and 'qp' evaluation modes.
:Definition:
.. math::
\int_{\cal{D}} D_{ijkl} e_{kl}(\ul{w})
:Arguments:
- material : :math:`D_{ijkl}`
- parameter : :math:`\ul{w}`
"""
name = 'ev_cauchy_stress'
arg_types = ('material', 'parameter')
arg_shapes = {'material' : 'S, S', 'parameter' : 'D'}
integration = ('cell', 'facet_extra')
[docs]
@staticmethod
def function(out, coef, strain, mat, vg, fmode):
if fmode == 2:
out[:] = dot_sequences(mat, strain)
status = 0
else:
status = terms.de_cauchy_stress(out, strain, mat, vg.cmap, fmode)
if coef is not None:
out *= coef
return status
[docs]
def get_fargs(self, mat, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(parameter)
strain = self.get(parameter, 'cauchy_strain')
fmode = {'eval' : 0, 'el_avg' : 1, 'qp' : 2}.get(mode, 1)
return None, strain, mat, vg, fmode
[docs]
def get_eval_shape(self, mat, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(parameter)
if mode != 'qp':
n_qp = 1
return (n_el, n_qp, dim * (dim + 1) // 2, 1), parameter.dtype
[docs]
class CauchyStressTHTerm(CauchyStressTerm, THTerm):
r"""
Evaluate fading memory Cauchy stress tensor.
It is given in the usual vector form exploiting symmetry: in 3D it has 6
components with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in
2D it has 3 components with the indices ordered as :math:`[11, 22, 12]`.
Supports 'eval', 'el_avg' and 'qp' evaluation modes.
:Definition:
.. math::
\int_{\Omega} \int_0^t \Hcal_{ijkl}(t-\tau)\,e_{kl}(\ul{w}(\tau))
\difd{\tau}
:Arguments:
- ts : :class:`TimeStepper` instance
- material : :math:`\Hcal_{ijkl}(\tau)`
- parameter : :math:`\ul{w}`
"""
name = 'ev_cauchy_stress_th'
arg_types = ('ts', 'material', 'parameter')
arg_shapes = {'material' : '.: N, S, S', 'parameter' : 'D'}
[docs]
def get_fargs(self, ts, mats, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(state)
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(state)
fmode = {'eval' : 0, 'el_avg' : 1, 'qp' : 2}.get(mode, 1)
def iter_kernel():
for ii, mat in enumerate(mats):
strain = self.get(state, 'cauchy_strain',
step=-ii)
mat = nm.tile(mat, (n_el, n_qp, 1, 1))
yield ii, (ts.dt, strain, mat, vg, fmode)
return iter_kernel
[docs]
def get_eval_shape(self, ts, mats, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
out = CauchyStressTerm.get_eval_shape(self, mats, parameter, mode,
term_mode, diff_var, **kwargs)
return out
[docs]
class CauchyStressETHTerm(CauchyStressTerm, ETHTerm):
r"""
Evaluate fading memory Cauchy stress tensor.
It is given in the usual vector form exploiting symmetry: in 3D it has 6
components with the indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in
2D it has 3 components with the indices ordered as :math:`[11, 22, 12]`.
Assumes an exponential approximation of the convolution kernel resulting in
much higher efficiency.
Supports 'eval', 'el_avg' and 'qp' evaluation modes.
:Definition:
.. math::
\int_{\Omega} \int_0^t \Hcal_{ijkl}(t-\tau)\,e_{kl}(\ul{w}(\tau))
\difd{\tau}
:Arguments:
- ts : :class:`TimeStepper` instance
- material_0 : :math:`\Hcal_{ijkl}(0)`
- material_1 : :math:`\exp(-\lambda \Delta t)` (decay at :math:`t_1`)
- parameter : :math:`\ul{w}`
"""
name = 'ev_cauchy_stress_eth'
arg_types = ('ts', 'material_0', 'material_1', 'parameter')
arg_shapes = {'material_0' : 'S, S', 'material_1' : '1, 1',
'parameter' : 'D'}
[docs]
def get_fargs(self, ts, mat0, mat1, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _, key = self.get_mapping(state, return_key=True)
strain = self.get(state, 'cauchy_strain')
key += tuple(self.arg_names[1:])
data = self.get_eth_data(key, state, mat1, strain)
fmode = {'eval' : 0, 'el_avg' : 1, 'qp' : 2}.get(mode, 1)
return ts.dt, data.history + data.values, mat0, vg, fmode
[docs]
def get_eval_shape(self, ts, mat0, mat1, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
out = CauchyStressTerm.get_eval_shape(self, mat0, parameter, mode,
term_mode, diff_var, **kwargs)
return out
[docs]
class NonsymElasticTerm(Term):
r"""
Elasticity term with non-symmetric gradient. The indices of matrix
:math:`D_{ijkl}` are ordered as
:math:`[11, 12, 13, 21, 22, 23, 31, 32, 33]` in 3D and as
:math:`[11, 12, 21, 22]` in 2D.
:Definition:
.. math::
\int_{\Omega} \ull{D} \nabla\ul{u} : \nabla\ul{v}
:Arguments 1:
- material : :math:`\ull{D}`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
:Arguments 2:
- material : :math:`\ull{D}`
- parameter_1 : :math:`\ul{w}`
- parameter_2 : :math:`\ul{u}`
"""
name = 'dw_nonsym_elastic'
arg_types = (('material', 'virtual', 'state'),
('material', 'parameter_1', 'parameter_2'))
arg_shapes = {'material' : 'D2, D2', 'virtual' : ('D', 'state'),
'state' : 'D', 'parameter_1' : 'D', 'parameter_2' : 'D'}
modes = ('weak', 'eval')
geometries = ['2_3', '2_4', '3_4', '3_8']
[docs]
def get_fargs(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
vg, _ = self.get_mapping(state)
if mode == 'weak':
if diff_var is None:
grad = self.get(state, 'grad').transpose((0,1,3,2))
nel, nqp, nr, nc = grad.shape
grad = grad.reshape((nel,nqp,nr*nc,1))
fmode = 0
else:
grad = nm.array([0], ndmin=4, dtype=nm.float64)
fmode = 1
return grad, mat, vg, fmode
elif mode == 'eval':
grad1 = self.get(virtual, 'grad').transpose((0,1,3,2))
grad2 = self.get(state, 'grad').transpose((0,1,3,2))
nel, nqp, nr, nc = grad1.shape
return 1.0,\
grad1.reshape((nel,nqp,nr*nc,1)),\
grad2.reshape((nel,nqp,nr*nc,1)),\
mat, vg
else:
raise ValueError('unsupported evaluation mode in %s! (%s)'
% (self.name, mode))
[docs]
def get_eval_shape(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(state)
return (n_el, 1, 1, 1), state.dtype
[docs]
def set_arg_types(self):
if self.mode == 'weak':
self.function = terms.dw_nonsym_elastic
else:
self.function = terms.d_lin_elastic
def _build_wave_strain_op(vec, bf):
dim = len(vec)
if dim == 2:
n0, n1 = vec
nmat = nm.array([[n0, 0],
[0, n1],
[n1, n0]], dtype=nm.float64)
else:
n0, n1, n2 = vec
nmat = nm.array([[n0, 0, 0],
[0, n1, 0],
[0, 0, n2],
[n1, n0, 0],
[n2, 0, n0],
[0, n2, n1]], dtype=nm.float64)
out = nm.einsum('ik,cqkj->cqij', nmat, bf)
return out
from sfepy.base.compat import block
def _build_cauchy_strain_op(bfg):
dim = bfg.shape[2]
if dim == 2:
g1, g2 = bfg[..., 0:1, :], bfg[..., 1:2, :]
zz = nm.zeros_like(g1)
out = block([[g1, zz],
[zz, g2],
[g2, g1]])
else:
g1, g2, g3 = bfg[..., 0:1, :], bfg[..., 1:2, :], bfg[..., 2:3, :]
zz = nm.zeros_like(g1)
out = block([[g1, zz, zz],
[zz, g2, zz],
[zz, zz, g3],
[g2, g1, zz],
[g3, zz, g1],
[zz, g3, g2]])
return out
[docs]
class ElasticWaveTerm(Term):
r"""
Elastic dispersion term involving the wave strain :math:`g_{ij}`,
:math:`g_{ij}(\ul{u}) = \frac{1}{2}(u_i \kappa_j + \kappa_i u_j)`, with the
wave vector :math:`\ul{\kappa}`. :math:`D_{ijkl}` is given in the usual
matrix form exploiting symmetry: in 3D it is :math:`6\times6` with the
indices ordered as :math:`[11, 22, 33, 12, 13, 23]`, in 2D it is
:math:`3\times3` with the indices ordered as :math:`[11, 22, 12]`.
:Definition:
.. math::
\int_{\Omega} D_{ijkl}\ g_{ij}(\ul{v}) g_{kl}(\ul{u})
:Arguments:
- material_1 : :math:`D_{ijkl}`
- material_2 : :math:`\ul{\kappa}`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
"""
name = 'dw_elastic_wave'
arg_types = ('material_1', 'material_2', 'virtual', 'state')
arg_shapes = {'material_1' : 'S, S', 'material_2' : '.: D',
'virtual' : ('D', 'state'), 'state' : 'D'}
geometries = ['2_3', '2_4', '3_4', '3_8']
[docs]
@staticmethod
def function(out, out_qp, geo, fmode):
status = geo.integrate(out, out_qp)
return status
[docs]
def get_fargs(self, mat, kappa, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.discrete.variables import create_adof_conn, expand_basis
geo, _ = self.get_mapping(state)
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(virtual)
ebf = expand_basis(geo.bf, dim)
mat = Term.tile_mat(mat, n_el)
gmat = _build_wave_strain_op(kappa, ebf)
if diff_var is None:
econn = state.field.get_econn('cell', self.region)
adc = create_adof_conn(nm.arange(state.n_dof, dtype=nm.int32),
econn, n_c, 0)
vals = state()[adc]
# Same as nm.einsum('qij,cj->cqi', gmat[0], vals)[..., None]
aux = dot_sequences(gmat, vals[:, None, :, None])
out_qp = dot_sequences(gmat, dot_sequences(mat, aux), 'ATB')
fmode = 0
else:
out_qp = dot_sequences(gmat, dot_sequences(mat, gmat), 'ATB')
fmode = 1
return out_qp, geo, fmode
[docs]
class ElasticWaveCauchyTerm(Term):
r"""
Elastic dispersion term involving the wave strain :math:`g_{ij}`,
:math:`g_{ij}(\ul{u}) = \frac{1}{2}(u_i \kappa_j + \kappa_i u_j)`, with the
wave vector :math:`\ul{\kappa}` and the elastic strain :math:`e_{ij}`.
:math:`D_{ijkl}` is given in the usual matrix form exploiting symmetry: in
3D it is :math:`6\times6` with the indices ordered as :math:`[11, 22, 33,
12, 13, 23]`, in 2D it is :math:`3\times3` with the indices ordered as
:math:`[11, 22, 12]`.
:Definition:
.. math::
\int_{\Omega} D_{ijkl}\ g_{ij}(\ul{v}) e_{kl}(\ul{u})\\
\int_{\Omega} D_{ijkl}\ g_{ij}(\ul{u}) e_{kl}(\ul{v})
:Arguments 1:
- material_1 : :math:`D_{ijkl}`
- material_2 : :math:`\ul{\kappa}`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
:Arguments 2:
- material_1 : :math:`D_{ijkl}`
- material_2 : :math:`\ul{\kappa}`
- state : :math:`\ul{u}`
- virtual : :math:`\ul{v}`
"""
name = 'dw_elastic_wave_cauchy'
arg_types = (('material_1', 'material_2', 'virtual', 'state'),
('material_1', 'material_2', 'state', 'virtual'))
arg_shapes = {'material_1' : 'S, S', 'material_2' : '.: D',
'virtual' : ('D', 'state'), 'state' : 'D'}
geometries = ['2_3', '2_4', '3_4', '3_8']
modes = ('ge', 'eg')
[docs]
@staticmethod
def function(out, out_qp, geo, fmode):
status = geo.integrate(out, out_qp)
return status
[docs]
def get_fargs(self, mat, kappa, gvar, evar,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.discrete.variables import create_adof_conn, expand_basis
geo, _ = self.get_mapping(evar)
n_el, n_qp, dim, n_en, n_c = self.get_data_shape(gvar)
ebf = expand_basis(geo.bf, dim)
mat = Term.tile_mat(mat, n_el)
gmat = _build_wave_strain_op(kappa, ebf)
emat = _build_cauchy_strain_op(geo.bfg)
if diff_var is None:
avar = evar if self.mode == 'ge' else gvar
econn = avar.field.get_econn('cell', self.region)
adc = create_adof_conn(nm.arange(avar.n_dof, dtype=nm.int32),
econn, n_c, 0)
vals = avar()[adc]
if self.mode == 'ge':
# Same as aux = self.get(avar, 'cauchy_strain'),
aux = dot_sequences(emat, vals[:, None, :, None])
out_qp = dot_sequences(gmat, dot_sequences(mat, aux), 'ATB')
else:
aux = dot_sequences(gmat, vals[:, None, :, None])
out_qp = dot_sequences(emat, dot_sequences(mat, aux), 'ATB')
fmode = 0
else:
if self.mode == 'ge':
out_qp = dot_sequences(gmat, dot_sequences(mat, emat), 'ATB')
else:
out_qp = dot_sequences(emat, dot_sequences(mat, gmat), 'ATB')
fmode = 1
return out_qp, geo, fmode
[docs]
class LinearSpringTerm(Term):
r"""Linear spring element.
:Definition:
.. math::
\ul{f}^{(i)} = - \ul{f}^{(j)} = k (\ul{u}^{(j)} - \ul{u}^{(i)})\\
\quad \forall \mbox{ elements } T_K^{i,j}\\
\mbox{ in a region connecting nodes } i, j
:Arguments 1:
- material : :math:`k`
- virtual : :math:`\ul{v}`
- state : :math:`\ul{u}`
"""
name = 'dw_lin_spring'
arg_types = ('material', 'virtual', 'state')
arg_shapes = {'material': '1, 1', 'virtual': ('D', 'state'), 'state': 'D'}
integration_order = 0
geometries = ['1_2', '2_1_2', '3_1_2']
[docs]
@staticmethod
def function(out, stiffness, vec, diff_var):
dim = out.shape[-2] // 2
if diff_var is None:
aux = nm.array([-1, 1])
for k in nm.arange(dim) * 2:
du = (vec[:, k] - vec[:, k + 1])[:, None]
out[:, 0, k:(k + 2), 0] = aux * du
out *= stiffness
else:
eye = nm.eye(2 * dim, 2 * dim, dtype=nm.float64)
eye.shape = (1, 1) + eye.shape
out[...] = - stiffness * eye
for k in nm.arange(dim) * 2:
out[..., k, k + 1] = out[..., k + 1, k] = 1
return 0
[docs]
def get_fargs(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
if diff_var is None:
from sfepy.discrete.variables import create_adof_conn
econn = virtual.field.get_econn('cell', self.region)
_, _, _, _, n_c = self.get_data_shape(virtual)
adc = create_adof_conn(nm.arange(state.n_dof, dtype=nm.int32),
econn, n_c, 0)
return mat, state()[adc], diff_var
else:
return mat, None, diff_var
[docs]
class LinearTrussTerm(Term):
r"""
Evaluate internal force in the element direction. To be used with
'el_avg' or 'qp' evaluation modes which give the same results.
The material parameter :math:`EA` is equal to
Young modulus times element coss-section. The internal force
is given by :math:`F^{(i)} = -F^{(j)} = EA / l (U^{(j)} - U^{(i)})`,
where :math:`l` is the element length and :math:`U`, :math:`F` are
the nodal displacements and the nodal forces in the element direction.
:Definition:
.. math::
F^{(i)} = -F^{(j)} = EA / l (U^{(j)} - U^{(i)})\\
\quad \forall \mbox{ elements } T_K^{i,j}\\
\mbox{ in a region connecting nodes } i, j
:Arguments:
- material : :math:`EA`
- parameter : :math:`\ul{w}`
"""
name = 'dw_lin_truss'
arg_types = ('material', 'virtual', 'state')
arg_shapes = {'material': '1, 1', 'virtual': ('D', 'state'),
'state': 'D'}
integration_order = 0
geometries = ['1_2', '2_1_2', '3_1_2']
[docs]
@staticmethod
def function(out, mat, vec, mtx_t, length, diff_var):
dim = mtx_t.shape[-1]
if diff_var is None:
if dim == 1:
du = vec[:, [1]] - vec[:, [0]]
elif dim == 2:
du = vec[:, [1, 3]] - vec[:, [0, 2]]
elif dim == 3:
du = vec[:, [1, 3, 5]] - vec[:, [0, 2, 4]]
dx = nm.matmul(mtx_t.transpose((0, 2, 1)), du[..., None])[:, 0, :]
Fe = nm.zeros((2 * dim,), dtype=nm.float64)
Fe[0] = -1
Fe[1] = 1
out[...] = (Fe * dx)[:, None, :, None] * mat
if mtx_t is not None:
membranes.transform_asm_vectors(out, mtx_t)
else:
Ke = nm.zeros((2 * dim, 2 * dim), dtype=nm.float64)
Ke[0, 0] = Ke[1, 1] = 1
Ke[1, 0] = Ke[0, 1] = -1
out[...] = Ke * mat
if mtx_t is not None:
membranes.transform_asm_matrices(out, mtx_t)
return 0
[docs]
@staticmethod
def get_mtx_t_and_length(coors, dx=None):
from sfepy.linalg import norm_l2_along_axis as norm
dim = coors.shape[-1]
if dx is None:
dx = coors[:, 1, :] - coors[:, 0, :]
mtx_t = nm.zeros((coors.shape[0], dim, dim), dtype=nm.float64)
length = norm(dx)[:, None]
v1 = dx / length
mtx_t[:, :, 0] = v1
if dim == 1:
pass
elif dim == 2:
mtx_t[:, 0, 1] = -v1[:, 1]
mtx_t[:, 1, 1] = v1[:, 0]
elif dim == 3:
v2 = nm.zeros_like(v1)
for k in [0, 2, 1]:
v2_ = nm.zeros((1, dim), dtype=nm.float64)
v2_[0, k] = 1.
dot = nm.abs(nm.sum(v2_ * v1, axis=1))
v2[(1. - dot) > 1e-12] = v2_
v3 = nm.cross(v1, v2)
v2 = nm.cross(v3, v1)
v2 = v2 / norm(v2)[:, None]
v3 = v3 / norm(v3)[:, None]
mtx_t[:, :, 1] = v2
mtx_t[:, :, 2] = v3
else:
raise ValueError(f'unsupported space dimension {dim}!')
return mtx_t, length
[docs]
def get_fargs(self, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.discrete.variables import create_adof_conn
econn = virtual.field.get_econn('cell', self.region)
coors = virtual.field.get_coor()[econn]
mtx_t, length = self.get_mtx_t_and_length(coors)
mat = mat / length[..., None, None]
if diff_var is None:
_, _, _, _, n_c = self.get_data_shape(virtual)
adc = create_adof_conn(nm.arange(state.n_dof, dtype=nm.int32),
econn, n_c, 0)
return mat, state()[adc], mtx_t, length, diff_var
else:
return mat, None, mtx_t, length, diff_var
[docs]
class LinearTrussInternalForceTerm(Term):
r"""
Evaluate internal force in the element direction. To be used with
'el_avg' or 'qp' evaluation modes which give the same results.
The material parameter :math:`EA` is equal to
Young modulus times element coss-section. The internal force
is given by :math:`F^{(i)} = -F^{(j)} = EA / l (U^{(j)} - U^{(i)})`,
where :math:`l` is the element length and :math:`U`, :math:`F` are
the nodal displacements and the nodal forces in the element direction.
:Definition:
.. math::
F = EA / l (U^{(j)} - U^{(i)})\\
\quad \forall \mbox{ elements } T_K^{i,j}\\
\mbox{ in a region connecting nodes } i, j
:Arguments:
- material : :math:`EA`
- parameter : :math:`\ul{w}`
"""
name = 'ev_lin_truss_force'
arg_types = ('material', 'parameter')
arg_shapes = {'material': '1, 1', 'parameter': 'D'}
integration_order = 0
geometries = ['1_2', '2_1_2', '3_1_2']
[docs]
@staticmethod
def function(out, mat, vec, mtx_t):
dim = mtx_t.shape[-1]
if dim == 1:
du = vec[:, [1]] - vec[:, [0]]
elif dim == 2:
du = vec[:, [1, 3]] - vec[:, [0, 2]]
elif dim == 3:
du = vec[:, [1, 3, 5]] - vec[:, [0, 2, 4]]
dx = nm.matmul(mtx_t.transpose((0, 2, 1)), du[..., None])[:, 0, :]
out[...] = dx[:, None, None, :] * mat
return 0
[docs]
def get_fargs(self, mat, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.discrete.variables import create_adof_conn
econn = parameter.field.get_econn('cell', self.region)
coors = parameter.field.get_coor()[econn]
mtx_t, length = LinearTrussTerm.get_mtx_t_and_length(coors)
mat = mat / length[..., None, None]
_, _, _, _, n_c = self.get_data_shape(parameter)
adc = create_adof_conn(nm.arange(parameter.n_dof, dtype=nm.int32),
econn, n_c, 0)
return mat, parameter()[adc], mtx_t
[docs]
def get_eval_shape(self, mat, parameter,
mode=None, term_mode=None, diff_var=None, **kwargs):
n_el, _, _, _, _ = self.get_data_shape(parameter)
return (n_el, 1, 1, 1), parameter.dtype
[docs]
class LinearDSpringTerm(LinearTrussTerm):
r"""
Linear spring element with the stiffness transformed into
the element direction.
:Definition:
.. math::
f^{(i)}_k = -f^{(j)}_k = K_{kl} (u^{(j)}_l - u^{(i)}_l)\\
\quad \forall \mbox{ elements } T_K^{i,j}\\
\mbox{ in a region connecting nodes } i, j
:Arguments:
- opt_material : :math:`\ul{d}`
- material : :math:`\ul{k}`
- virtual: :math:`\ul{v}`
- state: :math:`\ul{u}`
Stiffness matrix
:math:`\ul{K} = \ul{T(\ul{d})}^T \ul{K(\ul{k})} \ul{T(\ul{d})}`
is defined by 6 components
:math:`\ul{k} = [k_{u1}, k_{u2}, k_{u3}, k_{r1}, k_{r2}, k_{r3}]` in 3D
and by 3 components :math:`\ul{k} = [k_{u1}, k_{u2}, k_{r1}]`,
where :math:`k_{ui}` is the stiffness for the displacement DOF
and :math:`r_{ui}` is for the rotational DOF. Note that the components of
:math:`\ul{k}` are in the local coordinates system specified by a given
direction :math:`\ul{d}` or by the vector
:math:`\ul{d} = \ul{x}^{(j)} - \ul{x}^{(i)}` for non-coincidental end nodes.
The stiffness parameter :math:`\ul{K}` can also be defined as a 6x6 matrix
in 3D or a 3x3 matrix in 2D.
"""
name = 'dw_lin_dspring'
arg_types = ('opt_material', 'material', 'virtual', 'state')
arg_shapes = [{'opt_material': 'D, 1', 'material': 'D, 1',
'virtual': ('D', 'state'), 'state': 'D'},
{'material': 'D, D'}, {'opt_material': None}]
integration_order = 0
geometries = ['1_2', '2_1_2', '3_1_2']
[docs]
@staticmethod
def function(out, mat, vec, mtx_t, diff_var):
nel, _, dim = mtx_t.shape
ndof = mat.shape[2]
ntr = 2 * dim
ke = nm.zeros((nel, 2 * ndof, 2 * ndof), dtype=nm.float64)
for k in range(ndof):
for j in range(ndof):
ke[:, 2*k, 2*j] = ke[:, 2*k + 1, 2*j + 1] = mat[:, 0, k, j,]
ke[:, 2*k + 1, 2*j] = ke[:, 2*k, 2*j + 1] = -mat[:, 0, k, j]
if diff_var is None:
trans_vec = membranes.transform_asm_vectors
vec_loc = vec.copy()[..., None]
trans_vec(vec_loc[:, None, :ntr ,:], mtx_t.transpose((0, 2, 1)))
fe = dot_sequences(ke, vec_loc)
out[...] = fe[:, None, ...]
trans_vec(out[:, :, :ntr ,:], mtx_t)
else:
out[...] = ke[:, None, ...]
membranes.transform_asm_matrices(out[..., :ntr, :ntr], mtx_t)
return 0
[docs]
def get_fargs(self, dvec, mat, virtual, state,
mode=None, term_mode=None, diff_var=None, **kwargs):
from sfepy.discrete.variables import create_adof_conn
econn = virtual.field.get_econn('cell', self.region)
coors = virtual.field.get_coor()[econn]
if dvec is not None:
dvec = dvec[:, 0, :, 0]
mtx_t, _ = self.get_mtx_t_and_length(coors, dvec)
if mat.shape[-1] == 1:
n = mat.shape[-2]
mat_ = nm.zeros(mat.shape[:-1] + (mat.shape[-2],), dtype=mat.dtype)
ii = nm.arange(n)
mat_[..., ii, ii] = mat[..., 0]
mat = mat_
if diff_var is None:
_, _, _, _, n_c = self.get_data_shape(virtual)
adc = create_adof_conn(nm.arange(state.n_dof, dtype=nm.int32),
econn, n_c, 0)
return mat, state()[adc], mtx_t, diff_var
else:
return mat, None, mtx_t, diff_var
[docs]
class LinearDRotSpringTerm(LinearDSpringTerm):
name = 'dw_lin_dspring_rot'
arg_types = ('opt_material', 'material', 'virtual', 'state')
arg_shapes = [{'opt_material': 'D, 1', 'material': 'S, 1',
'virtual': ('S', 'state'), 'state': 'S'},
{'material': 'S, S'}, {'opt_material': None}]