def project(v, V, dx_, bcs=[], nm=None):
w = TestFunction(V)
Pv = TrialFunction(V)
a, L = as_ufl(0), as_ufl(0)
zerofnc = Function(V)
for n in range(len(dx_)):
# check if we have passed in a list of functions or a function
if isinstance(v, list):
fnc = v[n]
else:
fnc = v
if not isinstance(fnc, constantvalue.Zero):
a += inner(w, Pv) * dx_[n]
L += inner(w, fnc) * dx_[n]
else:
a += inner(w, Pv) * dx_[n]
L += inner(w, zerofnc) * dx_[n]
# solve linear system for projection
function = Function(V, name=nm)
lp = LinearProblem(a, L, bcs=bcs, u=function)
lp.solve()
return function
def function_arg(self, g):
'''Set the value of this boundary condition.'''
if isinstance(g, function.Function) and g.function_space() != self._function_space:
raise RuntimeError("%r is defined on incompatible FunctionSpace!" % g)
if not isinstance(g, expression.Expression):
try:
# Bare constant?
as_ufl(g)
except UFLException:
try:
# List of bare constants? Convert to UFL expression
g = as_ufl(as_tensor(g))
if g.ufl_shape != self._function_space.shape:
raise ValueError("%r doesn't match the shape of the function space." % (g,))
except UFLException:
raise ValueError("%r is not a valid DirichletBC expression" % (g,))
if isinstance(g, expression.Expression) or has_type(as_ufl(g), SpatialCoordinate):
if isinstance(g, expression.Expression):
self._expression_state = g._state
try:
g = function.Function(self._function_space).interpolate(g)
# Not a point evaluation space, need to project onto V
except NotImplementedError:
g = projection.project(g, self._function_space)
self._function_arg = g
self._currently_zeroed = False
def function_arg(self, g):
'''Set the value of this boundary condition.'''
if isinstance(g, function.Function) and g.function_space() != self._function_space:
raise RuntimeError("%r is defined on incompatible FunctionSpace!" % g)
if not isinstance(g, expression.Expression):
try:
# Bare constant?
as_ufl(g)
except UFLException:
try:
# List of bare constants? Convert to UFL expression
g = as_ufl(as_tensor(g))
if g.ufl_shape != self._function_space.shape:
raise ValueError("%r doesn't match the shape of the function space." % (g,))
except UFLException:
raise ValueError("%r is not a valid DirichletBC expression" % (g,))
if isinstance(g, expression.Expression) or has_type(as_ufl(g), SpatialCoordinate):
if isinstance(g, expression.Expression):
self._expression_state = g._state
try:
g = function.Function(self._function_space).interpolate(g)
# Not a point evaluation space, need to project onto V
except NotImplementedError:
g = projection.project(g, self._function_space)
self._function_arg = g
self._currently_zeroed = False
def get_transition_bounds(self):
'''Returns dictionary of `(lower, upper)` bounds for energy
transitions. See :py:meth:`get_transition_lower_bounds` for more.
'''
lowers = self.get_transition_lower_bounds()
upper_bound = ufl.as_ufl(self.inf_upper_bound.m_as(self.energy_unit))
r = {}
for k, lower in lowers.items():
lst = []
for k2, lower2 in lowers.items():
if k == k2: continue
# FIXME: this is WRONG if lower is exactly equal to lower2
lst.append(
dolfin.conditional(
ufl.as_ufl(lower2.m) < lower.m, upper_bound, lower2.m))
if not lst:
lst.append(upper_bound)
upper = self._ufl_minimum(lst)
r[k] = (lower, upper * lower.units)
return r
def function_arg(self, g):
'''Set the value of this boundary condition.'''
if isinstance(g, firedrake.Function):
if g.function_space() != self.function_space():
raise RuntimeError("%r is defined on incompatible FunctionSpace!" % g)
self._function_arg = g
elif isinstance(g, ufl.classes.Zero):
if g.ufl_shape and g.ufl_shape != self.function_space().ufl_element().value_shape():
raise ValueError(f"Provided boundary value {g} does not match shape of space")
# Special case. Scalar zero for direct Function.assign.
self._function_arg = ufl.zero()
elif isinstance(g, ufl.classes.Expr):
if g.ufl_shape != self.function_space().ufl_element().value_shape():
raise RuntimeError(f"Provided boundary value {g} does not match shape of space")
try:
self._function_arg = firedrake.Function(self.function_space())
self._function_arg_update = firedrake.Interpolator(g, self._function_arg).interpolate
except (NotImplementedError, AttributeError):
# Element doesn't implement interpolation
self._function_arg = firedrake.Function(self.function_space()).project(g)
self._function_arg_update = firedrake.Projector(g, self._function_arg).project
else:
try:
g = as_ufl(g)
self._function_arg = g
except UFLException:
try:
# Recurse to handle this through interpolation.
self.function_arg = as_ufl(as_tensor(g))
except UFLException:
raise ValueError(f"{g} is not a valid DirichletBC expression")
def test_imag(self):
z0 = Zero()
z1 = as_ufl(1.0)
z2 = as_ufl(1j)
z3 = ComplexValue(1+1j)
assert Imag(z2) == z1
assert Imag(z3) == z1
assert Imag(z1) == z0
def test_latex_formatting_of_literals():
# Test literals
assert expr2latex(ufl.as_ufl(2)) == "2"
assert expr2latex(ufl.as_ufl(3.14)) == '3.14'
assert expr2latex(ufl.as_ufl(0)) == "0"
# These are actually converted to int before formatting:
assert expr2latex(ufl.Identity(2)[0, 0]) == "1"
assert expr2latex(ufl.Identity(2)[0, 1]) == "0"
assert expr2latex(ufl.Identity(2)[1, 0]) == "0"
assert expr2latex(ufl.Identity(2)[1, 1]) == "1"
assert expr2latex(ufl.PermutationSymbol(3)[1, 2, 3]) == "1"
assert expr2latex(ufl.PermutationSymbol(3)[2, 1, 3]) == "-1"
assert expr2latex(ufl.PermutationSymbol(3)[1, 1, 3]) == "0"
def evaluate_active_stress_ode(self, t):
# take care of Frank-Starling law (fiber stretch-dependent contractility)
if self.have_frank_starling:
amp_old_, na = [], 0
for n in range(self.num_domains):
if self.mat_active_stress[n] and self.actstress[na].frankstarling:
# old fiber stretch (needed for Frank-Starling law)
if self.mat_growth[n]: lam_fib_old = self.ma[n].fibstretch_e(self.ki.C(self.u_old), self.theta_old, self.fib_func[0])
else: lam_fib_old = self.ki.fibstretch(self.u_old, self.fib_func[0])
amp_old_.append(self.actstress[na].amp(t-self.dt, lam_fib_old, self.amp_old))
else:
amp_old_.append(as_ufl(0))
amp_old_proj = project(amp_old_, self.Vd_scalar, self.dx_)
self.amp_old.vector.ghostUpdate(addv=PETSc.InsertMode.ADD, mode=PETSc.ScatterMode.REVERSE)
self.amp_old.interpolate(amp_old_proj)
tau_a_, na = [], 0
for n in range(self.num_domains):
if self.mat_active_stress[n]:
# fiber stretch (needed for Frank-Starling law)
if self.actstress[na].frankstarling:
if self.mat_growth[n]: lam_fib = self.ma[n].fibstretch_e(self.ki.C(self.u), self.theta, self.fib_func[0])
else: lam_fib = self.ki.fibstretch(self.u, self.fib_func[0])
else:
lam_fib = as_ufl(1)
tau_a_.append(self.actstress[na].tau_act(self.tau_a_old, t, self.dt, lam_fib, self.amp_old))
na+=1
else:
tau_a_.append(as_ufl(0))
# project and interpolate to quadrature function space
tau_a_proj = project(tau_a_, self.Vd_scalar, self.dx_)
self.tau_a.vector.ghostUpdate(addv=PETSc.InsertMode.ADD, mode=PETSc.ScatterMode.REVERSE)
self.tau_a.interpolate(tau_a_proj)
def assign(self, expr, subset=None):
r"""Set the :class:`Function` value to the pointwise value of
expr. expr may only contain :class:`Function`\s on the same
:class:`.FunctionSpace` as the :class:`Function` being assigned to.
Similar functionality is available for the augmented assignment
operators `+=`, `-=`, `*=` and `/=`. For example, if `f` and `g` are
both Functions on the same :class:`.FunctionSpace` then::
f += 2 * g
will add twice `g` to `f`.
If present, subset must be an :class:`pyop2.Subset` of this
:class:`Function`'s ``node_set``. The expression will then
only be assigned to the nodes on that subset.
"""
expr = ufl.as_ufl(expr)
if isinstance(expr, ufl.classes.Zero):
self.dat.zero(subset=subset)
return self
elif (isinstance(expr, Function)
and expr.function_space() == self.function_space()):
expr.dat.copy(self.dat, subset=subset)
return self
from firedrake import assemble_expressions
assemble_expressions.evaluate_expression(
assemble_expressions.Assign(self, expr), subset)
return self
def g(self, lam):
amp_min = self.params['amp_min']
amp_max = self.params['amp_max']
lam_threslo = self.params['lam_threslo']
lam_maxlo = self.params['lam_maxlo']
lam_threshi = self.params['lam_threshi']
lam_maxhi = self.params['lam_maxhi']
# Diss Hirschvogel eq. 2.107
# TeX: g(\lambda_{\mathrm{myo}}) = \begin{cases} a_{\mathrm{min}}, & \lambda_{\mathrm{myo}} \leq \hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,lo}}, \\ a_{\mathrm{min}}+\frac{1}{2}\left(a_{\mathrm{max}}-a_{\mathrm{min}}\right)\left(1-\cos \frac{\pi(\lambda_{\mathrm{myo}}-\hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,lo}})}{\hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,lo}}-\hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,lo}}}\right), & \hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,lo}} \leq \lambda_{\mathrm{myo}} \leq \hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,lo}}, \\ a_{\mathrm{max}}, & \hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,lo}} \leq \lambda_{\mathrm{myo}} \leq \hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,hi}}, \\ a_{\mathrm{min}}+\frac{1}{2}\left(a_{\mathrm{max}}-a_{\mathrm{min}}\right)\left(1-\cos \frac{\pi(\lambda_{\mathrm{myo}}-\hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,hi}})}{\hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,hi}}-\hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,hi}}}\right), & \hat{\lambda}_{\mathrm{myo}}^{\mathrm{thres,hi}} \leq \lambda_{\mathrm{myo}} \leq \hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,hi}}, \\ a_{\mathrm{min}}, & \lambda_{\mathrm{myo}} \geq \hat{\lambda}_{\mathrm{myo}}^{\mathrm{max,hi}} \end{cases}
return conditional(
le(lam, lam_threslo), amp_min,
conditional(
And(ge(lam, lam_threslo), le(lam, lam_maxlo)), amp_min + 0.5 *
(amp_max - amp_min) * (1. - cos(pi * (lam - lam_threslo) /
(lam_maxlo - lam_threslo))),
conditional(
And(ge(lam, lam_maxlo), le(lam, lam_threshi)), amp_max,
conditional(
And(ge(lam, lam_threshi), le(lam, lam_maxhi)),
amp_min + 0.5 * (amp_max - amp_min) *
(1. - cos(pi * (lam - lam_maxhi) /
(lam_maxhi - lam_threshi))),
conditional(ge(lam, lam_maxhi), amp_min, as_ufl(0))))))
def handle_conditional(v, si, deps, SV_factors, FV, sv2fv, e2fi):
fac0 = SV_factors[deps[0]]
fac1 = SV_factors[deps[1]]
fac2 = SV_factors[deps[2]]
assert not fac0, "Cannot have argument in condition."
if not (fac1 or fac2): # non-arg ? non-arg : non-arg
# Record non-argument subexpression
sv2fv[si] = add_to_fv(v, FV, e2fi)
factors = noargs
else:
f0 = FV[sv2fv[deps[0]]]
f1 = FV[sv2fv[deps[1]]]
f2 = FV[sv2fv[deps[2]]]
# Term conditional(c, argument, non-argument) is not legal unless non-argument is 0.0
assert fac1 or isinstance(f1, Zero)
assert fac2 or isinstance(f2, Zero)
assert () not in fac1
assert () not in fac2
z = as_ufl(0.0)
# In general, can decompose like this:
# conditional(c, sum_i fi*ui, sum_j fj*uj) -> sum_i conditional(c, fi, 0)*ui + sum_j conditional(c, 0, fj)*uj
mas = sorted(set(fac1.keys()) | set(fac2.keys()))
factors = {}
for k in mas:
fi1 = fac1.get(k)
fi2 = fac2.get(k)
f1 = z if fi1 is None else FV[fi1]
f2 = z if fi2 is None else FV[fi2]
factors[k] = add_to_fv(conditional(f0, f1, f2), FV, e2fi)
return factors
def handle_conditional(v, fac, sf, F):
fac0 = fac[0]
fac1 = fac[1]
fac2 = fac[2]
assert not fac0, "Cannot have argument in condition."
if not (fac1 or fac2): # non-arg ? non-arg : non-arg
raise RuntimeError("No arguments")
else:
f0 = sf[0]
f1 = sf[1]
f2 = sf[2]
# Term conditional(c, argument, non-argument) is not legal unless non-argument is 0.0
assert fac1 or isinstance(f1, Zero)
assert fac2 or isinstance(f2, Zero)
assert () not in fac1
assert () not in fac2
z = as_ufl(0.0)
# In general, can decompose like this:
# conditional(c, sum_i fi*ui, sum_j fj*uj) -> sum_i conditional(c, fi, 0)*ui + sum_j conditional(c, 0, fj)*uj
mas = sorted(set(fac1.keys()) | set(fac2.keys()))
factors = {}
for k in mas:
fi1 = fac1.get(k)
fi2 = fac2.get(k)
f1 = z if fi1 is None else F.nodes[fi1]['expression']
f2 = z if fi2 is None else F.nodes[fi2]['expression']
factors[k] = graph_insert(F, conditional(f0, f1, f2))
return factors
def test_real(self):
z0 = Zero()
z1 = as_ufl(1.0)
z2 = ComplexValue(1j)
z3 = ComplexValue(1+1j)
assert Real(z1) == z1
assert Real(z3) == z1
assert Real(z2) == z0
def xtest_cpp2_compile_scalar_literals():
M = as_ufl(0) * dx
code = compile_form(M, 'unittest')
print('\n', code)
expected = 'TODO'
assert code == expected
M = as_ufl(3) * dx
code = compile_form(M, 'unittest')
print('\n', code)
expected = 'TODO'
assert code == expected
M = as_ufl(1.03) * dx
code = compile_form(M, 'unittest')
print('\n', code)
expected = 'TODO'
assert code == expected
def robin_bcs(self, v, v_old):
w, w_old = as_ufl(0), as_ufl(0)
for r in self.bc_dict['robin']:
if r['type'] == 'dashpot':
if r['dir'] == 'xyz':
for i in range(len(r['id'])):
ds_ = ds(
subdomain_data=self.io.mt_b1,
subdomain_id=r['id'][i],
metadata={'quadrature_degree': self.quad_degree})
w += self.vf.deltaP_ext_robin_dashpot(
v, r['visc'], ds_)
w_old += self.vf.deltaP_ext_robin_dashpot(
v_old, r['visc'], ds_)
elif r['dir'] == 'normal': # reference normal
for i in range(len(r['id'])):
ds_ = ds(
subdomain_data=self.io.mt_b1,
subdomain_id=r['id'][i],
metadata={'quadrature_degree': self.quad_degree})
w += self.vf.deltaP_ext_robin_dashpot_normal(
v, r['visc'], ds_)
w_old += self.vf.deltaP_ext_robin_dashpot_normal(
v_old, r['visc'], ds_)
else:
raise NameError("Unknown dir option for Robin BC!")
else:
raise NameError("Unknown type option for Robin BC!")
return w, w_old
def test_conditional(mode, compile_args):
cell = ufl.triangle
element = ufl.FiniteElement("Lagrange", cell, 1)
u, v = ufl.TrialFunction(element), ufl.TestFunction(element)
x = ufl.SpatialCoordinate(cell)
condition = ufl.Or(ufl.ge(ufl.real(x[0] + x[1]), 0.1),
ufl.ge(ufl.real(x[1] + x[1]**2), 0.1))
c1 = ufl.conditional(condition, 2.0, 1.0)
a = c1 * ufl.inner(ufl.grad(u), ufl.grad(v)) * ufl.dx
x1x2 = ufl.real(x[0] + ufl.as_ufl(2) * x[1])
c2 = ufl.conditional(ufl.ge(x1x2, 0), 6.0, 0.0)
b = c2 * ufl.conj(v) * ufl.dx
forms = [a, b]
compiled_forms, module = ffcx.codegeneration.jit.compile_forms(
forms,
parameters={'scalar_type': mode},
cffi_extra_compile_args=compile_args)
form0 = compiled_forms[0][0].create_cell_integral(-1)
form1 = compiled_forms[1][0].create_cell_integral(-1)
ffi = cffi.FFI()
c_type, np_type = float_to_type(mode)
A1 = np.zeros((3, 3), dtype=np_type)
w1 = np.array([1.0, 1.0, 1.0], dtype=np_type)
c = np.array([], dtype=np.float64)
coords = np.array([0.0, 0.0, 1.0, 0.0, 0.0, 1.0], dtype=np.float64)
form0.tabulate_tensor(
ffi.cast('{type} *'.format(type=c_type), A1.ctypes.data),
ffi.cast('{type} *'.format(type=c_type), w1.ctypes.data),
ffi.cast('{type} *'.format(type=c_type), c.ctypes.data),
ffi.cast('double *', coords.ctypes.data), ffi.NULL, ffi.NULL, 0)
expected_result = np.array([[2, -1, -1], [-1, 1, 0], [-1, 0, 1]],
dtype=np_type)
assert np.allclose(A1, expected_result)
A2 = np.zeros(3, dtype=np_type)
w2 = np.array([1.0, 1.0, 1.0], dtype=np_type)
coords = np.array([0.0, 0.0, 1.0, 0.0, 0.0, 1.0], dtype=np.float64)
form1.tabulate_tensor(
ffi.cast('{type} *'.format(type=c_type), A2.ctypes.data),
ffi.cast('{type} *'.format(type=c_type), w2.ctypes.data),
ffi.cast('{type} *'.format(type=c_type), c.ctypes.data),
ffi.cast('double *', coords.ctypes.data), ffi.NULL, ffi.NULL, 0)
expected_result = np.ones(3, dtype=np_type)
assert np.allclose(A2, expected_result)
def xtest_literal_zero_compilation():
uexpr = as_ufl(0)
lines, finals = compile_expression_lines(uexpr)
expected_lines = []
expected_finals = ['0']
assert lines == expected_lines
assert finals == expected_finals
def xtest_literal_float_compilation():
uexpr = as_ufl(2.56)
lines, finals = compile_expression_lines(uexpr)
expected_lines = []
expected_finals = ['2.56']
assert lines == expected_lines
assert finals == expected_finals
def test_complex_algebra(self):
z1 = ComplexValue(1j)
z2 = ComplexValue(1+1j)
# Remember that ufl.algebra functions return ComplexValues, but ufl.mathfunctions return complex Python scalar
# Any operations with a ComplexValue and a complex Python scalar promote to ComplexValue
assert z1*z2 == ComplexValue(-1+1j)
assert z2/z1 == ComplexValue(1-1j)
assert pow(z2, z1) == ComplexValue((1+1j)**1j)
assert sqrt(z2) * as_ufl(1) == ComplexValue(cmath.sqrt(1+1j))
assert ((sin(z2) + cosh(z2) - atan(z2)) * z1) == ComplexValue((cmath.sin(1+1j) + cmath.cosh(1+1j) - cmath.atan(1+1j))*1j)
assert (abs(z2) - ln(z2))/exp(z1) == ComplexValue((abs(1+1j) - cmath.log(1+1j))/cmath.exp(1j))
def __init__(self, lvalue, rvalue):
"""
:arg lvalue: The coefficient to assign into.
:arg rvalue: The pointwise expression.
"""
if not isinstance(lvalue, ufl.Coefficient):
raise ValueError("lvalue for pointwise assignment must be a coefficient")
self.lvalue = lvalue
self.rvalue = ufl.as_ufl(rvalue)
n = len(self.lvalue.function_space())
if n > 1:
self.splitter = MemoizerArg(_split)
self.splitter.n = n
def compute_volume_large(self):
J_all = as_ufl(0)
for n in range(self.pb.pblarge.num_domains):
J_all += self.pb.pblarge.ki.J(
self.pb.pblarge.u) * self.pb.pblarge.dx_[n]
vol = assemble_scalar(J_all)
vol = self.pb.comm.allgather(vol)
volume_large = sum(vol)
if self.pb.comm.rank == 0:
print('Volume of myocardium: %.4e' % (volume_large))
sys.stdout.flush()
return volume_large
def set_homeostatic_threshold(self, t):
# time is absolute time (should only be set in first cycle)
eps = 1.0e-14
if t >= self.pb.t_gandr_setpoint - eps and t < self.pb.t_gandr_setpoint + self.pb.pbs.dt - eps:
if self.pb.comm.rank == 0:
print('Set homeostatic growth thresholds...')
sys.stdout.flush()
time.sleep(1)
growth_thresolds = []
for n in range(self.pb.pbs.num_domains):
if self.pb.pbs.mat_growth[n]:
growth_settrig = self.pb.pbs.constitutive_models[
'MAT' + str(n + 1) + '']['growth']['growth_settrig']
if growth_settrig == 'fibstretch':
growth_thresolds.append(self.pb.pbs.ma[n].fibstretch_e(
self.pb.pbs.ki.C(self.pb.pbs.u), self.pb.pbs.theta,
self.pb.pbs.fib_func[0]))
elif growth_settrig == 'volstress':
growth_thresolds.append(
tr(self.pb.pbs.ma[n].M_e(
self.pb.pbs.u,
self.pb.pbs.p,
self.pb.pbs.ki.C(self.pb.pbs.u),
ivar=self.pb.pbs.internalvars)))
else:
raise NameError(
"Unknown growth trigger to be set as homeostatic threshold!"
)
else:
growth_thresolds.append(as_ufl(0))
growth_thres_proj = project(growth_thresolds,
self.pb.pbs.Vd_scalar, self.pb.pbs.dx_)
self.pb.pbs.growth_thres.vector.ghostUpdate(
addv=PETSc.InsertMode.ADD, mode=PETSc.ScatterMode.REVERSE)
self.pb.pbs.growth_thres.interpolate(growth_thres_proj)
def dtheta_dp(self, u_, p_, ivar, theta_old_, thres, dt):
theta_ = ivar["theta"]
dFg_dtheta = self.F_g(theta_,tang=True)
ktheta = self.res_dtheta_growth(u_, p_, ivar, theta_old_, thres, dt, 'ktheta')
K_growth = self.res_dtheta_growth(u_, p_, ivar, theta_old_, thres, dt, 'tang')
if self.growth_trig == 'volstress':
tangdp = (ktheta*dt/K_growth) * ( diff(tr(self.M_e(u_,p_,self.kin.C(u_),ivar)),p_) )
elif self.growth_trig == 'fibstretch':
tangdp = as_ufl(0)
else:
raise NameError("Unkown growth_trig!")
return tangdp
def compute_solid_growth_rate(self, N, t):
dtheta_all = as_ufl(0)
for n in range(self.num_domains):
dtheta_all += (self.theta - self.theta_old) / (self.dt) * self.dx_[n]
gr = assemble_scalar(dtheta_all)
gr = self.comm.allgather(gr)
self.growth_rate = sum(gr)
if self.comm.rank == 0:
print('Solid growth rate: %.4e' % (self.growth_rate))
sys.stdout.flush()
if self.io.write_results_every > 0 and N % self.io.write_results_every == 0:
if np.isclose(t,self.dt): mode='wt'
else: mode='a'
fl = self.io.output_path+'/results_'+self.simname+'_growthrate.txt'
f = open(fl, mode)
f.write('%.16E %.16E\n' % (t,self.growth_rate))
f.close()
def test_comparison_checker(self):
cell = triangle
element = FiniteElement("Lagrange", cell, 1)
u = TrialFunction(element)
v = TestFunction(element)
a = conditional(ge(abs(u), imag(v)), u, v)
b = conditional(le(sqrt(abs(u)), imag(v)), as_ufl(1), as_ufl(1j))
c = conditional(gt(abs(u), pow(imag(v), 0.5)), sin(u), cos(v))
d = conditional(lt(as_ufl(-1), as_ufl(1)), u, v)
e = max_value(as_ufl(0), real(u))
f = min_value(sin(u), cos(v))
g = min_value(sin(pow(u, 3)), cos(abs(v)))
assert do_comparison_check(a) == conditional(ge(real(abs(u)), real(imag(v))), u, v)
with pytest.raises(ComplexComparisonError):
b = do_comparison_check(b)
with pytest.raises(ComplexComparisonError):
c = do_comparison_check(c)
assert do_comparison_check(d) == conditional(lt(real(as_ufl(-1)), real(as_ufl(1))), u, v)
assert do_comparison_check(e) == max_value(real(as_ufl(0)), real(real(u)))
assert do_comparison_check(f) == min_value(real(sin(u)), real(cos(v)))
assert do_comparison_check(g) == min_value(real(sin(pow(u, 3))), real(cos(abs(v))))
def set_variational_forms_and_jacobians(self):
# add constant Neumann terms for large scale problem (trigger pressures)
self.neumann_funcs = []
w_neumann = as_ufl(0)
for n in range(len(self.pbsmall.surface_p_ids)):
self.neumann_funcs.append(Function(self.pblarge.Vd_scalar))
for i in range(len(self.pbsmall.surface_p_ids[n])):
ds_ = ds(
subdomain_data=self.pblarge.io.mt_b1,
subdomain_id=self.pbsmall.surface_p_ids[n][i],
metadata={'quadrature_degree': self.pblarge.quad_degree})
# we apply the pressure onto a fixed configuration of the G&R trigger point, determined by the displacement field u_set
# in the last G&R cycle, we assure that growth falls below a tolerance and hence the current and the set configuration coincide
w_neumann += self.pblarge.vf.deltaW_ext_neumann_true(
self.pblarge.ki.J(self.pblarge.u_set),
self.pblarge.ki.F(self.pblarge.u_set),
self.neumann_funcs[-1], ds_)
self.pblarge.weakform_u -= w_neumann
def _simplify_abs_(o, self, in_abs):
if not in_abs:
return o
# Inline abs(constant)
return ufl.as_ufl(abs(o._value))
def set_variational_forms_and_jacobians(self):
self.cq, self.cq_old, self.dcq, self.dforce = [], [], [], []
self.coupfuncs, self.coupfuncs_old = [], []
# Lagrange multiplier stiffness matrix (most likely to be zero!)
self.K_lm = PETSc.Mat().createAIJ(size=(self.num_coupling_surf,
self.num_coupling_surf),
bsize=None,
nnz=None,
csr=None,
comm=self.comm)
self.K_lm.setUp()
# Lagrange multipliers
self.lm, self.lm_old = self.K_lm.createVecLeft(
), self.K_lm.createVecLeft()
# 3D constraint variable (volume or flux)
self.constr, self.constr_old = [], []
self.work_coupling, self.work_coupling_old, self.work_coupling_prestr = as_ufl(
0), as_ufl(0), as_ufl(0)
# coupling variational forms and Jacobian contributions
for n in range(self.num_coupling_surf):
self.pr0D = expression.template()
self.coupfuncs.append(Function(
self.pbs.Vd_scalar)), self.coupfuncs_old.append(
Function(self.pbs.Vd_scalar))
self.coupfuncs[-1].interpolate(
self.pr0D.evaluate), self.coupfuncs_old[-1].interpolate(
self.pr0D.evaluate)
cq_, cq_old_ = as_ufl(0), as_ufl(0)
for i in range(len(self.surface_c_ids[n])):
ds_vq = ds(
subdomain_data=self.pbs.io.mt_b1,
subdomain_id=self.surface_c_ids[n][i],
metadata={'quadrature_degree': self.pbs.quad_degree})
# currently, only volume or flux constraints are supported
if self.coupling_params['constraint_quantity'] == 'volume':
cq_ += self.pbs.vf.volume(self.pbs.u,
self.pbs.ki.J(self.pbs.u),
self.pbs.ki.F(self.pbs.u), ds_vq)
cq_old_ += self.pbs.vf.volume(
self.pbs.u_old, self.pbs.ki.J(self.pbs.u_old),
self.pbs.ki.F(self.pbs.u_old), ds_vq)
elif self.coupling_params['constraint_quantity'] == 'flux':
cq_ += self.pbs.vf.flux(self.pbs.vel,
self.pbs.ki.J(self.pbs.u),
self.pbs.ki.F(self.pbs.u), ds_vq)
cq_old_ += self.pbs.vf.flux(self.pbs.v_old,
self.pbs.ki.J(self.pbs.u_old),
self.pbs.ki.F(self.pbs.u_old),
ds_vq)
else:
raise NameError(
"Unknown constraint quantity! Choose either volume or flux!"
)
self.cq.append(cq_), self.cq_old.append(cq_old_)
self.dcq.append(derivative(self.cq[-1], self.pbs.u, self.pbs.du))
df_ = as_ufl(0)
for i in range(len(self.surface_p_ids[n])):
ds_p = ds(subdomain_data=self.pbs.io.mt_b1,
subdomain_id=self.surface_p_ids[n][i],
metadata={'quadrature_degree': self.pbs.quad_degree})
df_ += self.pbs.timefac * self.pbs.vf.surface(
self.pbs.ki.J(self.pbs.u), self.pbs.ki.F(self.pbs.u), ds_p)
# add to solid rhs contributions
self.work_coupling += self.pbs.vf.deltaW_ext_neumann_true(
self.pbs.ki.J(self.pbs.u), self.pbs.ki.F(self.pbs.u),
self.coupfuncs[-1], ds_p)
self.work_coupling_old += self.pbs.vf.deltaW_ext_neumann_true(
self.pbs.ki.J(self.pbs.u_old),
self.pbs.ki.F(self.pbs.u_old), self.coupfuncs_old[-1],
ds_p)
# for prestressing, true loads should act on the reference, not the current configuration
if self.pbs.prestress_initial:
self.work_coupling_prestr += self.pbs.vf.deltaW_ext_neumann_refnormal(
self.coupfuncs_old[-1], ds_p)
self.dforce.append(df_)
# minus sign, since contribution to external work!
self.pbs.weakform_u += -self.pbs.timefac * self.work_coupling - (
1. - self.pbs.timefac) * self.work_coupling_old
# add to solid Jacobian
self.pbs.jac_uu += -self.pbs.timefac * derivative(
self.work_coupling, self.pbs.u, self.pbs.du)
def _simplify_abs_(o, self, in_abs):
if not in_abs:
return o
# Inline abs(constant)
return ufl.as_ufl(abs(o._value))
def compute_argument_factorization(S, rank):
"""Factorizes a scalar expression graph w.r.t. scalar Argument components.
The result is a triplet (AV, FV, IM):
- The scalar argument component subgraph:
AV[ai] = v
with the property
SV[arg_indices] == AV[:]
- An expression graph vertex list with all non-argument factors:
FV[fi] = f
with the property that none of the expressions depend on Arguments.
- A dict representation of the final integrand of rank r:
IM = { (ai1_1, ..., ai1_r): fi1, (ai2_1, ..., ai2_r): fi2, }
This mapping represents the factorization of SV[-1] w.r.t. Arguments s.t.:
SV[-1] := sum(FV[fik] * product(AV[ai] for ai in aik) for aik, fik in IM.items())
where := means equivalence in the mathematical sense,
of course in a different technical representation.
"""
# Extract argument component subgraph
arg_indices = build_argument_indices(S)
AV = [S.nodes[i]['expression'] for i in arg_indices]
# Data structure for building non-argument factors
F = ExpressionGraph()
# Attach a quick lookup dict for expression to index
F.e2i = {}
# Insert arguments as first entries in factorisation graph
# They will not be connected to other nodes, but will be available
# and referred to by the factorisation indices of the 'target' nodes.
for v in AV:
graph_insert(F, v)
# Adding 1.0 as an expression allows avoiding special representation
# of arguments when first visited by representing "v" as "1*v"
one_index = graph_insert(F, as_ufl(1.0))
# Intermediate factorization for each vertex in SV on the format
# SV_factors[si] = None # if SV[si] does not depend on arguments
# SV_factors[si] = { argkey: fi } # if SV[si] does depend on arguments, where:
# FV[fi] is the expression SV[si] with arguments factored out
# argkey is a tuple with indices into SV for each of the argument components SV[si] depends on
# SV_factors[si] = { argkey1: fi1, argkey2: fi2, ... } # if SV[si]
# is a linear combination of multiple argkey configurations
# Factorize each subexpression in order:
for si, attr in S.nodes.items():
deps = S.out_edges[si]
v = attr['expression']
if si in arg_indices:
assert len(deps) == 0
# v is a modified Argument
factors = {(si, ): one_index}
else:
fac = [S.nodes[d]['factors'] for d in deps]
if not any(fac):
# Entirely scalar (i.e. no arg factors)
# Just add unchanged to F
graph_insert(F, v)
factors = noargs
else:
# Get scalar factors for dependencies
# which do not have arg factors
sf = []
for i, d in enumerate(deps):
if fac[i]:
sf.append(None)
else:
sf.append(S.nodes[d]['expression'])
# Use appropriate handler to deal with Sum, Product, etc.
factors = handler(v, fac, sf, F)
attr['factors'] = factors
assert len(F.nodes) == len(F.e2i)
# Find the (only) node in S that is marked as 'target'
# Should always be the last one.
S_targets = [i for i, v in S.nodes.items() if v.get('target', False)]
assert len(S_targets) == 1
S_target = S_targets[0]
# Get the factorizations of the target values
if S.nodes[S_target]['factors'] == {}:
if rank == 0:
# Functionals and expressions: store as no args * factor
factors = {(): F.e2i[S.nodes[S_target]['expression']]}
else:
# Zero form of arity 1 or higher: make factors empty
factors = {}
else:
# Forms of arity 1 or higher:
# Map argkeys from indices into SV to indices into AV,
# and resort keys for canonical representation
factors = {
tuple(sorted(arg_indices.index(si) for si in argkey)): fi
for argkey, fi in S.nodes[S_target]['factors'].items()
}
# Expecting all term keys to have length == rank
# (this assumption will eventually have to change if we
# implement joint bilinear+linear form factorization here)
assert all(len(k) == rank for k in factors)
# Indices into F that are needed for final result
for i in factors.values():
F.nodes[i]['target'] = []
for k in factors:
i = factors[k]
F.nodes[i]['target'] += [k]
# Compute dependencies in FV
for i, v in F.nodes.items():
expr = v['expression']
if not expr._ufl_is_terminal_ and not expr._ufl_is_terminal_modifier_:
for o in expr.ufl_operands:
F.add_edge(i, F.e2i[o])
return F
def neumann_bcs(self, V, V_real):
w, w_old = as_ufl(0), as_ufl(0)
for n in self.bc_dict['neumann']:
if n['dir'] == 'xyz':
func, func_old = Function(V), Function(V)
if 'curve' in n.keys():
load = expression.template_vector()
load.val_x, load.val_y, load.val_z = self.ti.timecurves(
n['curve'][0])(self.ti.t_init), self.ti.timecurves(
n['curve'][1])(self.ti.t_init), self.ti.timecurves(
n['curve'][2])(self.ti.t_init)
func.interpolate(load.evaluate), func_old.interpolate(
load.evaluate)
self.ti.funcs_to_update_vec.append({
func: [
self.ti.timecurves(n['curve'][0]),
self.ti.timecurves(n['curve'][1]),
self.ti.timecurves(n['curve'][2])
]
})
self.ti.funcs_to_update_vec_old.append({
func_old: [
self.ti.timecurves(n['curve'][0]),
self.ti.timecurves(n['curve'][1]),
self.ti.timecurves(n['curve'][2])
]
})
else:
func.vector.set(
n['val']
) # currently only one value for all directions - use constant load function otherwise!
for i in range(len(n['id'])):
ds_ = ds(subdomain_data=self.io.mt_b1,
subdomain_id=n['id'][i],
metadata={'quadrature_degree': self.quad_degree})
w += self.vf.deltaP_ext_neumann(func, ds_)
w_old += self.vf.deltaP_ext_neumann(func_old, ds_)
elif n['dir'] == 'normal': # reference normal
func, func_old = Function(V_real), Function(V_real)
if 'curve' in n.keys():
load = expression.template()
load.val = self.ti.timecurves(n['curve'])(self.ti.t_init)
func.interpolate(load.evaluate), func_old.interpolate(
load.evaluate)
self.ti.funcs_to_update.append(
{func: self.ti.timecurves(n['curve'])})
self.ti.funcs_to_update_old.append(
{func_old: self.ti.timecurves(n['curve'])})
else:
func.vector.set(n['val'])
for i in range(len(n['id'])):
ds_ = ds(subdomain_data=self.io.mt_b1,
subdomain_id=n['id'][i],
metadata={'quadrature_degree': self.quad_degree})
w += self.vf.deltaP_ext_neumann_normal(func, ds_)
w_old += self.vf.deltaP_ext_neumann_normal(func_old, ds_)
else:
raise NameError("Unknown dir option for Neumann BC!")
return w, w_old
def set_variational_forms_and_jacobians(self):
# set form for acceleration
self.acc = self.ti.set_acc(self.v, self.v_old, self.a_old)
# kinetic, internal, and pressure virtual power
self.deltaP_kin, self.deltaP_kin_old = as_ufl(0), as_ufl(0)
self.deltaP_int, self.deltaP_int_old = as_ufl(0), as_ufl(0)
self.deltaP_p, self.deltaP_p_old = as_ufl(0), as_ufl(0)
for n in range(self.num_domains):
if self.timint != 'static':
# kinetic virtual power
self.deltaP_kin += self.vf.deltaP_kin(self.acc, self.v,
self.rho[n], self.dx_[n])
self.deltaP_kin_old += self.vf.deltaP_kin(
self.a_old, self.v_old, self.rho[n], self.dx_[n])
# internal virtual power
self.deltaP_int += self.vf.deltaP_int(
self.ma[n].sigma(self.v, self.p), self.dx_[n])
self.deltaP_int_old += self.vf.deltaP_int(
self.ma[n].sigma(self.v_old, self.p_old), self.dx_[n])
# pressure virtual power
self.deltaP_p += self.vf.deltaP_int_pres(self.v, self.dx_[n])
self.deltaP_p_old += self.vf.deltaP_int_pres(
self.v_old, self.dx_[n])
# external virtual power (from Neumann or Robin boundary conditions, body forces, ...)
w_neumann, w_neumann_old, w_robin, w_robin_old = as_ufl(0), as_ufl(
0), as_ufl(0), as_ufl(0)
if 'neumann' in self.bc_dict.keys():
w_neumann, w_neumann_old = self.bc.neumann_bcs(
self.V_v, self.Vd_scalar)
if 'robin' in self.bc_dict.keys():
w_robin, w_robin_old = self.bc.robin_bcs(self.v, self.v_old)
# TODO: Body forces!
self.deltaP_ext = w_neumann + w_robin
self.deltaP_ext_old = w_neumann_old + w_robin_old
self.timefac_m, self.timefac = self.ti.timefactors()
### full weakforms
# kinetic plus internal minus external virtual power
self.weakform_u = self.timefac_m * self.deltaP_kin + (1.-self.timefac_m) * self.deltaP_kin_old + \
self.timefac * self.deltaP_int + (1.-self.timefac) * self.deltaP_int_old - \
self.timefac * self.deltaP_ext - (1.-self.timefac) * self.deltaP_ext_old
self.weakform_p = self.timefac * self.deltaP_p + (
1. - self.timefac) * self.deltaP_p_old
# Reynolds number: ratio of inertial to viscous forces
#self.Re = sqrt(dot(self.vf.f_inert(self.acc,self.v,self.rho), self.vf.f_inert(self.acc,self.v,self.rho))) / sqrt(dot(self.vf.f_viscous(self.ma[0].sigma(self.v, self.p)), self.vf.f_viscous(self.ma[0].sigma(self.v, self.p))))
if self.order_vel == self.order_pres:
raise ValueError(
"Equal order velocity and pressure interpolation requires stabilization! Not yet implemented! Use order_vel > order_pres."
)
#dx1_stab = dx(subdomain_data=self.io.mt_d, subdomain_id=1, metadata={'quadrature_degree': 2*3})
## stabilization stuff - TODO: FIXME and finish!!!
#res_v_strong = self.vf.residual_v_strong(self.acc, self.v, self.rho, self.ma[0].sigma(self.v, self.p))
#res_v_strong_old = self.vf.residual_v_strong(self.a_old, self.v_old, self.rho, self.ma[0].sigma(self.v_old, self.p_old))
#res_p_strong = self.vf.residual_p_strong(self.v)
#res_p_strong_old = self.vf.residual_p_strong(self.v_old)
#vnorm, vnorm_old = sqrt(dot(self.v, self.v)), sqrt(dot(self.v_old, self.v_old))
#nu = 0.004/self.rho
#Cinv = 16.*self.Re
#tau_SUPG = Min(self.io.h0**2./(Cinv*nu), self.io.h0/(2.*vnorm))
##tau = ( (2.*self.dt)**2. + (2.0*vnorm_/self.io.h0)**2 + (4.0*nu/self.io.h0**2.)**2. )**(-0.5)
##delta = conditional(ge(vnorm,1.0e-8), self.io.h0/(2.*vnorm), 0.)
##delta_old = conditional(ge(vnorm_old,1.0e-8), self.io.h0/(2.*vnorm_old), 0.)
#stab_v = tau_SUPG * dot(dot(self.v, grad(self.var_v)),res_v_strong)*dx1_stab
##stab_p = tau_PSPG * dot(dot(self.v, grad(self.var_v)),res_p_strong)*dx1_stab
#self.weakform_u += self.timefac * stab_v #+ (1.-self.timefac) * stab_old
##self.weakform_p += tau_SUPG*inner(grad(self.var_p), res_strong)*self.dx1
### Jacobians
self.jac_uu = derivative(self.weakform_u, self.v, self.dv)
self.jac_up = derivative(self.weakform_u, self.p, self.dp)
self.jac_pu = derivative(self.weakform_p, self.v, self.dv)
# for saddle-point block-diagonal preconditioner - TODO: Doesn't work very well...
self.a_p11 = as_ufl(0)
for n in range(self.num_domains):
self.a_p11 += inner(self.dp, self.var_p) * self.dx_[n]
def __init__(self,
io_params,
time_params,
fem_params,
constitutive_models,
bc_dict,
time_curves,
io,
comm=None):
problem_base.__init__(self, io_params, time_params, comm)
self.problem_physics = 'fluid'
self.simname = io_params['simname']
self.io = io
# number of distinct domains (each one has to be assigned a own material model)
self.num_domains = len(constitutive_models)
self.order_vel = fem_params['order_vel']
self.order_pres = fem_params['order_pres']
self.quad_degree = fem_params['quad_degree']
# collect domain data
self.dx_, self.rho = [], []
for n in range(self.num_domains):
# integration domains
self.dx_.append(
dx(subdomain_data=self.io.mt_d,
subdomain_id=n + 1,
metadata={'quadrature_degree': self.quad_degree}))
# data for inertial forces: density
self.rho.append(constitutive_models['MAT' + str(n + 1) +
'']['inertia']['rho'])
self.incompressible_2field = True # always true!
self.localsolve = False # no idea what might have to be solved locally...
self.prestress_initial = False # guess prestressing in fluid is somehow senseless...
self.p11 = as_ufl(
0
) # can't think of a fluid case with non-zero 11-block in system matrix...
# type of discontinuous function spaces
if str(self.io.mesh.ufl_cell()) == 'tetrahedron' or str(
self.io.mesh.ufl_cell()) == 'triangle3D':
dg_type = "DG"
if (self.order_vel > 1
or self.order_pres > 1) and self.quad_degree < 3:
raise ValueError(
"Use at least a quadrature degree of 3 or more for higher-order meshes!"
)
elif str(self.io.mesh.ufl_cell()) == 'hexahedron' or str(
self.io.mesh.ufl_cell()) == 'quadrilateral3D':
dg_type = "DQ"
if (self.order_vel > 1
or self.order_pres > 1) and self.quad_degree < 5:
raise ValueError(
"Use at least a quadrature degree of 5 or more for higher-order meshes!"
)
else:
raise NameError("Unknown cell/element type!")
# create finite element objects for v and p
self.P_v = VectorElement("CG", self.io.mesh.ufl_cell(), self.order_vel)
self.P_p = FiniteElement("CG", self.io.mesh.ufl_cell(),
self.order_pres)
# function spaces for v and p
self.V_v = FunctionSpace(self.io.mesh, self.P_v)
self.V_p = FunctionSpace(self.io.mesh, self.P_p)
# a discontinuous tensor, vector, and scalar function space
self.Vd_tensor = TensorFunctionSpace(self.io.mesh,
(dg_type, self.order_vel - 1))
self.Vd_vector = VectorFunctionSpace(self.io.mesh,
(dg_type, self.order_vel - 1))
self.Vd_scalar = FunctionSpace(self.io.mesh,
(dg_type, self.order_vel - 1))
# functions
self.dv = TrialFunction(self.V_v) # Incremental velocity
self.var_v = TestFunction(self.V_v) # Test function
self.dp = TrialFunction(self.V_p) # Incremental pressure
self.var_p = TestFunction(self.V_p) # Test function
self.v = Function(self.V_v, name="Velocity")
self.p = Function(self.V_p, name="Pressure")
# values of previous time step
self.v_old = Function(self.V_v)
self.a_old = Function(self.V_v)
self.p_old = Function(self.V_p)
self.ndof = self.v.vector.getSize() + self.p.vector.getSize()
# initialize fluid time-integration class
self.ti = timeintegration.timeintegration_fluid(
time_params, fem_params, time_curves, self.t_init, self.comm)
# initialize kinematics_constitutive class
self.ki = fluid_kinematics_constitutive.kinematics()
# initialize material/constitutive class
self.ma = []
for n in range(self.num_domains):
self.ma.append(
fluid_kinematics_constitutive.constitutive(
self.ki, constitutive_models['MAT' + str(n + 1) + '']))
# initialize fluid variational form class
self.vf = fluid_variationalform.variationalform(
self.var_v, self.dv, self.var_p, self.dp, self.io.n0)
# initialize boundary condition class
self.bc = boundaryconditions.boundary_cond_fluid(
bc_dict, fem_params, self.io, self.ki, self.vf, self.ti)
self.bc_dict = bc_dict
# Dirichlet boundary conditions
if 'dirichlet' in self.bc_dict.keys():
self.bc.dirichlet_bcs(self.V_v)
self.set_variational_forms_and_jacobians()
def build_uflacs_ir(cell, integral_type, entitytype,
integrands, tensor_shape,
coefficient_numbering,
quadrature_rules, parameters):
# The intermediate representation dict we're building and returning here
ir = {}
# Extract uflacs specific optimization and code generation parameters
p = parse_uflacs_optimization_parameters(parameters, integral_type)
# Pass on parameters for consumption in code generation
ir["params"] = p
# { ufl coefficient: count }
ir["coefficient_numbering"] = coefficient_numbering
# Shared unique tables for all quadrature loops
ir["unique_tables"] = {}
ir["unique_table_types"] = {}
# Shared piecewise expr_ir for all quadrature loops
ir["piecewise_ir"] = empty_expr_ir()
# { num_points: expr_ir for one integrand }
ir["varying_irs"] = {}
# Temporary data structures to build shared piecewise data
pe2i = {}
piecewise_modified_argument_indices = {}
# Whether we expect the quadrature weight to be applied or not
# (in some cases it's just set to 1 in ufl integral scaling)
tdim = cell.topological_dimension()
expect_weight = (
integral_type not in ("expression",) + point_integral_types
and (entitytype == "cell"
or (entitytype == "facet" and tdim > 1)
or (integral_type in custom_integral_types)
)
)
if integral_type == "expression":
# TODO: Figure out how to get non-integrand expressions in here, this is just a draft:
# Analyse all expressions in one list
assert isinstance(integrands, (tuple, list))
all_num_points = [None]
cases = [(None, integrands)]
else:
# Analyse each num_points/integrand separately
assert isinstance(integrands, dict)
all_num_points = sorted(integrands.keys())
cases = [(num_points, [integrands[num_points]])
for num_points in all_num_points]
ir["all_num_points"] = all_num_points
for num_points, expressions in cases:
# Rebalance order of nested terminal modifiers
expressions = [balance_modifiers(expr) for expr in expressions]
# Build initial scalar list-based graph representation
V, V_deps, V_targets = build_scalar_graph(expressions)
# Build terminal_data from V here before factorization.
# Then we can use it to derive table properties for all modified terminals,
# and then use that to rebuild the scalar graph more efficiently before
# argument factorization. We can build terminal_data again after factorization
# if that's necessary.
initial_terminal_indices = [i for i, v in enumerate(V)
if is_modified_terminal(v)]
initial_terminal_data = [analyse_modified_terminal(V[i])
for i in initial_terminal_indices]
unique_tables, unique_table_types, unique_table_num_dofs, mt_unique_table_reference = \
build_optimized_tables(num_points, quadrature_rules,
cell, integral_type, entitytype, initial_terminal_data,
ir["unique_tables"], p["enable_table_zero_compression"],
rtol=p["table_rtol"], atol=p["table_atol"])
# Replace some scalar modified terminals before reconstructing expressions
# (could possibly use replace() on target expressions instead)
z = as_ufl(0.0)
one = as_ufl(1.0)
for i, mt in zip(initial_terminal_indices, initial_terminal_data):
if isinstance(mt.terminal, QuadratureWeight):
# Replace quadrature weight with 1.0, will be added back later
V[i] = one
else:
# Set modified terminals with zero tables to zero
tr = mt_unique_table_reference.get(mt)
if tr is not None and tr.ttype == "zeros":
V[i] = z
# Propagate expression changes using dependency list
for i in range(len(V)):
deps = [V[j] for j in V_deps[i]]
if deps:
V[i] = V[i]._ufl_expr_reconstruct_(*deps)
# Rebuild scalar target expressions and graph
# (this may be overkill and possible to optimize
# away if it turns out to be costly)
expressions = [V[i] for i in V_targets]
# Rebuild scalar list-based graph representation
SV, SV_deps, SV_targets = build_scalar_graph(expressions)
assert all(i < len(SV) for i in SV_targets)
# Compute factorization of arguments
(argument_factorizations, modified_arguments,
FV, FV_deps, FV_targets) = \
compute_argument_factorization(SV, SV_deps, SV_targets, len(tensor_shape))
assert len(SV_targets) == len(argument_factorizations)
# TODO: Still expecting one target variable in code generation
assert len(argument_factorizations) == 1
argument_factorization, = argument_factorizations
# Store modified arguments in analysed form
for i in range(len(modified_arguments)):
modified_arguments[i] = analyse_modified_terminal(modified_arguments[i])
# Build set of modified_terminal indices into factorized_vertices
modified_terminal_indices = [i for i, v in enumerate(FV)
if is_modified_terminal(v)]
# Build set of modified terminal ufl expressions
modified_terminals = [analyse_modified_terminal(FV[i])
for i in modified_terminal_indices]
# Make it easy to get mt object from FV index
FV_mts = [None]*len(FV)
for i, mt in zip(modified_terminal_indices, modified_terminals):
FV_mts[i] = mt
# Mark active modified arguments
#active_modified_arguments = numpy.zeros(len(modified_arguments), dtype=int)
#for ma_indices in argument_factorization:
# for j in ma_indices:
# active_modified_arguments[j] = 1
# Dependency analysis
inv_FV_deps, FV_active, FV_piecewise, FV_varying = \
analyse_dependencies(FV, FV_deps, FV_targets,
modified_terminal_indices,
modified_terminals,
mt_unique_table_reference)
# Extend piecewise V with unique new FV_piecewise vertices
pir = ir["piecewise_ir"]
for i, v in enumerate(FV):
if FV_piecewise[i]:
j = pe2i.get(v)
if j is None:
j = len(pe2i)
pe2i[v] = j
pir["V"].append(v)
pir["V_active"].append(1)
mt = FV_mts[i]
if mt is not None:
pir["mt_tabledata"][mt] = mt_unique_table_reference.get(mt)
pir["V_mts"].append(mt)
# Extend piecewise modified_arguments list with unique new items
for mt in modified_arguments:
ma = piecewise_modified_argument_indices.get(mt)
if ma is None:
ma = len(pir["modified_arguments"])
pir["modified_arguments"].append(mt)
piecewise_modified_argument_indices[mt] = ma
# Loop over factorization terms
block_contributions = defaultdict(list)
for ma_indices, fi in sorted(argument_factorization.items()):
# Get a bunch of information about this term
rank = len(ma_indices)
trs = tuple(mt_unique_table_reference[modified_arguments[ai]] for ai in ma_indices)
unames = tuple(tr.name for tr in trs)
ttypes = tuple(tr.ttype for tr in trs)
assert not any(tt == "zeros" for tt in ttypes)
blockmap = tuple(tr.dofmap for tr in trs)
block_is_uniform = all(tr.is_uniform for tr in trs)
# Collect relevant restrictions to identify blocks
# correctly in interior facet integrals
block_restrictions = []
for i, ma in enumerate(ma_indices):
if trs[i].is_uniform:
r = None
else:
r = modified_arguments[ma].restriction
block_restrictions.append(r)
block_restrictions = tuple(block_restrictions)
# Store piecewise status for fi and translate
# index to piecewise scope if relevant
factor_is_piecewise = FV_piecewise[fi]
if factor_is_piecewise:
factor_index = pe2i[FV[fi]]
else:
factor_index = fi
# TODO: Add separate block modes for quadrature
# Both arguments in quadrature elements
"""
for iq
fw = f*w
#for i
# for j
# B[i,j] = fw*U[i]*V[j] = 0 if i != iq or j != iq
BQ[iq] = B[iq,iq] = fw
for (iq)
A[iq+offset0, iq+offset1] = BQ[iq]
"""
# One argument in quadrature element
"""
for iq
fw[iq] = f*w
#for i
# for j
# B[i,j] = fw*UQ[i]*V[j] = 0 if i != iq
for j
BQ[iq,j] = fw[iq]*V[iq,j]
for (iq) for (j)
A[iq+offset, j+offset] = BQ[iq,j]
"""
# Decide how to handle code generation for this block
if p["enable_preintegration"] and (factor_is_piecewise
and rank > 0 and "quadrature" not in ttypes):
# - Piecewise factor is an absolute prerequisite
# - Could work for rank 0 as well but currently doesn't
# - Haven't considered how quadrature elements work out
block_mode = "preintegrated"
elif p["enable_premultiplication"] and (rank > 0
and all(tt in piecewise_ttypes for tt in ttypes)):
# Integrate functional in quadloop, scale block after quadloop
block_mode = "premultiplied"
elif p["enable_sum_factorization"]:
if (rank == 2 and any(tt in piecewise_ttypes for tt in ttypes)):
# Partial computation in quadloop of f*u[i],
# compute (f*u[i])*v[i] outside quadloop,
# (or with u,v swapped)
block_mode = "partial"
else:
# Full runtime integration of f*u[i]*v[j],
# can still do partial computation in quadloop of f*u[i]
# but must compute (f*u[i])*v[i] as well inside quadloop.
# (or with u,v swapped)
block_mode = "full"
else:
# Use full runtime integration with nothing fancy going on
block_mode = "safe"
# Carry out decision
if block_mode == "preintegrated":
# Add to contributions:
# P = sum_q weight*u*v; preintegrated here
# B[...] = f * P[...]; generated after quadloop
# A[blockmap] += B[...]; generated after quadloop
cache = ir["piecewise_ir"]["preintegrated_blocks"]
block_is_transposed = False
pname = cache.get(unames)
# Reuse transpose to save memory
if p["enable_block_transpose_reuse"] and pname is None and len(unames) == 2:
pname = cache.get((unames[1], unames[0]))
if pname is not None:
# Cache hit on transpose
block_is_transposed = True
if pname is None:
# Cache miss, precompute block
weights = quadrature_rules[num_points][1]
if integral_type == "interior_facet":
ptable = integrate_block_interior_facets(weights, unames, ttypes,
unique_tables, unique_table_num_dofs)
else:
ptable = integrate_block(weights, unames, ttypes,
unique_tables, unique_table_num_dofs)
ptable = clamp_table_small_numbers(ptable, rtol=p["table_rtol"], atol=p["table_atol"])
pname = "PI%d" % (len(cache,))
cache[unames] = pname
unique_tables[pname] = ptable
unique_table_types[pname] = "preintegrated"
assert factor_is_piecewise
block_unames = (pname,)
blockdata = preintegrated_block_data_t(block_mode, ttypes,
factor_index, factor_is_piecewise,
block_unames, block_restrictions,
block_is_transposed, block_is_uniform,
pname)
block_is_piecewise = True
elif block_mode == "premultiplied":
# Add to contributions:
# P = u*v; computed here
# FI = sum_q weight * f; generated inside quadloop
# B[...] = FI * P[...]; generated after quadloop
# A[blockmap] += B[...]; generated after quadloop
cache = ir["piecewise_ir"]["premultiplied_blocks"]
block_is_transposed = False
pname = cache.get(unames)
# Reuse transpose to save memory
if p["enable_block_transpose_reuse"] and pname is None and len(unames) == 2:
pname = cache.get((unames[1], unames[0]))
if pname is not None:
# Cache hit on transpose
block_is_transposed = True
if pname is None:
# Cache miss, precompute block
if integral_type == "interior_facet":
ptable = multiply_block_interior_facets(0, unames, ttypes, unique_tables, unique_table_num_dofs)
else:
ptable = multiply_block(0, unames, ttypes, unique_tables, unique_table_num_dofs)
pname = "PM%d" % (len(cache,))
cache[unames] = pname
unique_tables[pname] = ptable
unique_table_types[pname] = "premultiplied"
block_unames = (pname,)
blockdata = premultiplied_block_data_t(block_mode, ttypes,
factor_index, factor_is_piecewise,
block_unames, block_restrictions,
block_is_transposed, block_is_uniform,
pname)
block_is_piecewise = False
elif block_mode == "scaled": # TODO: Add mode, block is piecewise but choose not to be premultiplied
# Add to contributions:
# FI = sum_q weight * f; generated inside quadloop
# B[...] = FI * u * v; generated after quadloop
# A[blockmap] += B[...]; generated after quadloop
raise NotImplementedError("scaled block mode not implemented.")
# (probably need mostly the same data as premultiplied, except no P table name or values)
block_is_piecewise = False
elif block_mode in ("partial", "full", "safe"):
# Translate indices to piecewise context if necessary
block_is_piecewise = factor_is_piecewise and not expect_weight
ma_data = []
for i, ma in enumerate(ma_indices):
if trs[i].is_piecewise:
ma_index = piecewise_modified_argument_indices[modified_arguments[ma]]
else:
block_is_piecewise = False
ma_index = ma
ma_data.append(ma_data_t(ma_index, trs[i]))
block_is_transposed = False # FIXME: Handle transposes for these block types
if block_mode == "partial":
# Add to contributions:
# P[i] = sum_q weight * f * u[i]; generated inside quadloop
# B[i,j] = P[i] * v[j]; generated after quadloop (where v is the piecewise ma)
# A[blockmap] += B[...]; generated after quadloop
# Find first piecewise index TODO: Is last better? just reverse range here
for i in range(rank):
if trs[i].is_piecewise:
piecewise_ma_index = i
break
assert rank == 2
not_piecewise_ma_index = 1 - piecewise_ma_index
block_unames = (unames[not_piecewise_ma_index],)
blockdata = partial_block_data_t(block_mode, ttypes,
factor_index, factor_is_piecewise,
block_unames, block_restrictions,
block_is_transposed,
tuple(ma_data), piecewise_ma_index)
elif block_mode in ("full", "safe"):
# Add to contributions:
# B[i] = sum_q weight * f * u[i] * v[j]; generated inside quadloop
# A[blockmap] += B[i]; generated after quadloop
block_unames = unames
blockdata = full_block_data_t(block_mode, ttypes,
factor_index, factor_is_piecewise,
block_unames, block_restrictions,
block_is_transposed,
tuple(ma_data))
else:
error("Invalid block_mode %s" % (block_mode,))
if block_is_piecewise:
# Insert in piecewise expr_ir
ir["piecewise_ir"]["block_contributions"][blockmap].append(blockdata)
else:
# Insert in varying expr_ir for this quadrature loop
block_contributions[blockmap].append(blockdata)
# Figure out which table names are referenced in unstructured partition
active_table_names = set()
for i, mt in zip(modified_terminal_indices, modified_terminals):
tr = mt_unique_table_reference.get(mt)
if tr is not None and FV_active[i]:
active_table_names.add(tr.name)
# Figure out which table names are referenced in blocks
for blockmap, contributions in chain(block_contributions.items(),
ir["piecewise_ir"]["block_contributions"].items()):
for blockdata in contributions:
if blockdata.block_mode in ("preintegrated", "premultiplied"):
active_table_names.add(blockdata.name)
elif blockdata.block_mode in ("partial", "full", "safe"):
for mad in blockdata.ma_data:
active_table_names.add(mad.tabledata.name)
# Record all table types before dropping tables
ir["unique_table_types"].update(unique_table_types)
# Drop tables not referenced from modified terminals
# and tables of zeros and ones
unused_ttypes = ("zeros", "ones", "quadrature")
keep_table_names = set()
for name in active_table_names:
ttype = ir["unique_table_types"][name]
if ttype not in unused_ttypes:
if name in unique_tables:
keep_table_names.add(name)
unique_tables = { name: unique_tables[name]
for name in keep_table_names }
# Add to global set of all tables
for name, table in unique_tables.items():
tbl = ir["unique_tables"].get(name)
if tbl is not None and not numpy.allclose(tbl, table, rtol=p["table_rtol"], atol=p["table_atol"]):
error("Table values mismatch with same name.")
ir["unique_tables"].update(unique_tables)
# Analyse active terminals to check what we'll need to generate code for
active_mts = []
for i, mt in zip(modified_terminal_indices, modified_terminals):
if FV_active[i]:
active_mts.append(mt)
# Figure out if we need to access CellCoordinate to
# avoid generating quadrature point table otherwise
if integral_type == "cell":
need_points = any(isinstance(mt.terminal, CellCoordinate)
for mt in active_mts)
elif integral_type in facet_integral_types:
need_points = any(isinstance(mt.terminal, FacetCoordinate)
for mt in active_mts)
elif integral_type in custom_integral_types:
need_points = True # TODO: Always?
else:
need_points = False
# Figure out if we need to access QuadratureWeight to
# avoid generating quadrature point table otherwise
#need_weights = any(isinstance(mt.terminal, QuadratureWeight)
# for mt in active_mts)
# Count blocks of each mode
block_modes = defaultdict(int)
for blockmap, contributions in block_contributions.items():
for blockdata in contributions:
block_modes[blockdata.block_mode] += 1
# Debug output
summary = "\n".join(" %d\t%s" % (count, mode)
for mode, count in sorted(block_modes.items()))
debug("Blocks of each mode: \n" + summary)
# If there are any blocks other than preintegrated we need weights
if expect_weight and any(mode != "preintegrated" for mode in block_modes):
need_weights = True
elif integral_type in custom_integral_types:
need_weights = True # TODO: Always?
else:
need_weights = False
# Build IR dict for the given expressions
expr_ir = {}
# (array) FV-index -> UFL subexpression
expr_ir["V"] = FV
# (array) V indices for each input expression component in flattened order
expr_ir["V_targets"] = FV_targets
### Result of factorization:
# (array) MA-index -> UFL expression of modified arguments
expr_ir["modified_arguments"] = modified_arguments
# (dict) tuple(MA-indices) -> FV-index of monomial factor
#expr_ir["argument_factorization"] = argument_factorization
expr_ir["block_contributions"] = block_contributions
### Modified terminals
# (array) list of FV-indices to modified terminals
#expr_ir["modified_terminal_indices"] = modified_terminal_indices
# Dependency structure of graph:
# (CRSArray) FV-index -> direct dependency FV-index list
#expr_ir["dependencies"] = FV_deps
# (CRSArray) FV-index -> direct dependee FV-index list
#expr_ir["inverse_dependencies"] = inv_FV_deps
# Metadata about each vertex
#expr_ir["active"] = FV_active # (array) FV-index -> bool
#expr_ir["V_piecewise"] = FV_piecewise # (array) FV-index -> bool
expr_ir["V_varying"] = FV_varying # (array) FV-index -> bool
expr_ir["V_mts"] = FV_mts
# Store mapping from modified terminal object to
# table data, this is used in integralgenerator
expr_ir["mt_tabledata"] = mt_unique_table_reference
# To emit quadrature rules only if needed
expr_ir["need_points"] = need_points
expr_ir["need_weights"] = need_weights
# Store final ir for this num_points
ir["varying_irs"][num_points] = expr_ir
return ir
def assertEqualValues(self, A, B):
B = as_ufl(B)
self.assertEqual(A.ufl_shape, B.ufl_shape)
self.assertEqual(inner(A-B, A-B)(None), 0)
def write_output(self, pb, writemesh=False, N=1, t=0):
if writemesh:
if self.write_results_every > 0:
self.resultsfiles = {}
for res in self.results_to_write:
outfile = XDMFFile(self.comm, self.output_path+'/results_'+pb.simname+'_'+res+'.xdmf', 'w')
outfile.write_mesh(self.mesh)
self.resultsfiles[res] = outfile
return
else:
# write results every write_results_every steps
if self.write_results_every > 0 and N % self.write_results_every == 0:
# save solution to XDMF format
for res in self.results_to_write:
if res=='displacement':
self.resultsfiles[res].write_function(pb.u, t)
elif res=='velocity': # passed in v is not a function but form, so we have to project
v_proj = project(pb.vel, pb.V_u, pb.dx_, nm="Velocity")
self.resultsfiles[res].write_function(v_proj, t)
elif res=='acceleration': # passed in a is not a function but form, so we have to project
a_proj = project(pb.acc, pb.V_u, pb.dx_, nm="Acceleration")
self.resultsfiles[res].write_function(a_proj, t)
elif res=='pressure':
self.resultsfiles[res].write_function(pb.p, t)
elif res=='cauchystress':
stressfuncs=[]
for n in range(pb.num_domains):
stressfuncs.append(pb.ma[n].sigma(pb.u,pb.p,ivar=pb.internalvars))
cauchystress = project(stressfuncs, pb.Vd_tensor, pb.dx_, nm="CauchyStress")
self.resultsfiles[res].write_function(cauchystress, t)
elif res=='trmandelstress':
stressfuncs=[]
for n in range(pb.num_domains):
stressfuncs.append(tr(pb.ma[n].M(pb.u,pb.p,ivar=pb.internalvars)))
trmandelstress = project(stressfuncs, pb.Vd_scalar, pb.dx_, nm="trMandelStress")
self.resultsfiles[res].write_function(trmandelstress, t)
elif res=='trmandelstress_e':
stressfuncs=[]
for n in range(pb.num_domains):
if pb.mat_growth[n]: stressfuncs.append(tr(pb.ma[n].M_e(pb.u,pb.p,pb.ki.C(pb.u),ivar=pb.internalvars)))
else: stressfuncs.append(as_ufl(0))
trmandelstress_e = project(stressfuncs, pb.Vd_scalar, pb.dx_, nm="trMandelStress_e")
self.resultsfiles[res].write_function(trmandelstress_e, t)
elif res=='vonmises_cauchystress':
stressfuncs=[]
for n in range(pb.num_domains):
stressfuncs.append(pb.ma[n].sigma_vonmises(pb.u,pb.p,ivar=pb.internalvars))
vonmises_cauchystress = project(stressfuncs, pb.Vd_scalar, pb.dx_, nm="vonMises_CauchyStress")
self.resultsfiles[res].write_function(vonmises_cauchystress, t)
elif res=='pk1stress':
stressfuncs=[]
for n in range(pb.num_domains):
stressfuncs.append(pb.ma[n].P(pb.u,pb.p,ivar=pb.internalvars))
pk1stress = project(stressfuncs, pb.Vd_tensor, pb.dx_, nm="PK1Stress")
self.resultsfiles[res].write_function(pk1stress, t)
elif res=='pk2stress':
stressfuncs=[]
for n in range(pb.num_domains):
stressfuncs.append(pb.ma[n].S(pb.u,pb.p,ivar=pb.internalvars))
pk2stress = project(stressfuncs, pb.Vd_tensor, pb.dx_, nm="PK2Stress")
self.resultsfiles[res].write_function(pk2stress, t)
elif res=='jacobian':
jacobian = project(pb.ki.J(pb.u), pb.Vd_scalar, pb.dx_, nm="Jacobian")
self.resultsfiles[res].write_function(jacobian, t)
elif res=='glstrain':
glstrain = project(pb.ki.E(pb.u), pb.Vd_tensor, pb.dx_, nm="GreenLagrangeStrain")
self.resultsfiles[res].write_function(glstrain, t)
elif res=='eastrain':
eastrain = project(pb.ki.e(pb.u), pb.Vd_tensor, pb.dx_, nm="EulerAlmansiStrain")
self.resultsfiles[res].write_function(eastrain, t)
elif res=='fiberstretch':
fiberstretch = project(pb.ki.fibstretch(pb.u,pb.fib_func[0]), pb.Vd_scalar, pb.dx_, nm="FiberStretch")
self.resultsfiles[res].write_function(fiberstretch, t)
elif res=='fiberstretch_e':
stretchfuncs=[]
for n in range(pb.num_domains):
if pb.mat_growth[n]: stretchfuncs.append(pb.ma[n].fibstretch_e(pb.ki.C(pb.u),pb.theta,pb.fib_func[0]))
else: stretchfuncs.append(as_ufl(0))
fiberstretch_e = project(stretchfuncs, pb.Vd_scalar, pb.dx_, nm="FiberStretch_e")
self.resultsfiles[res].write_function(fiberstretch_e, t)
elif res=='theta':
self.resultsfiles[res].write_function(pb.theta, t)
elif res=='phi_remod':
phifuncs=[]
for n in range(pb.num_domains):
if pb.mat_remodel[n]: phifuncs.append(pb.ma[n].phi_remod(pb.theta))
else: phifuncs.append(as_ufl(0))
phiremod = project(phifuncs, pb.Vd_scalar, pb.dx_, nm="phiRemodel")
self.resultsfiles[res].write_function(phiremod, t)
elif res=='tau_a':
self.resultsfiles[res].write_function(pb.tau_a, t)
elif res=='fiber1':
fiber1 = project(pb.fib_func[0], pb.Vd_vector, pb.dx_, nm="Fiber1")
self.resultsfiles[res].write_function(fiber1, t)
elif res=='fiber2':
fiber2 = project(pb.fib_func[1], pb.Vd_vector, pb.dx_, nm="Fiber2")
self.resultsfiles[res].write_function(fiber2, t)
else:
raise NameError("Unknown output to write for solid mechanics!")
if self.write_restart_every > 0 and N % self.write_restart_every == 0:
self.writecheckpoint(pb, N)
def __init__(self, io_params, time_params, fem_params, constitutive_models, bc_dict, time_curves, io, comm=None):
problem_base.__init__(self, io_params, time_params, comm)
self.problem_physics = 'solid'
self.simname = io_params['simname']
self.io = io
# number of distinct domains (each one has to be assigned a own material model)
self.num_domains = len(constitutive_models)
self.order_disp = fem_params['order_disp']
try: self.order_pres = fem_params['order_pres']
except: self.order_pres = 1
self.quad_degree = fem_params['quad_degree']
self.incompressible_2field = fem_params['incompressible_2field']
self.fem_params = fem_params
self.constitutive_models = constitutive_models
# collect domain data
self.dx_, self.rho0, self.rayleigh, self.eta_m, self.eta_k = [], [], [False]*self.num_domains, [], []
for n in range(self.num_domains):
# integration domains
self.dx_.append(dx(subdomain_data=self.io.mt_d, subdomain_id=n+1, metadata={'quadrature_degree': self.quad_degree}))
# data for inertial and viscous forces: density and damping
if self.timint != 'static':
self.rho0.append(constitutive_models['MAT'+str(n+1)+'']['inertia']['rho0'])
if 'rayleigh_damping' in constitutive_models['MAT'+str(n+1)+''].keys():
self.rayleigh[n] = True
self.eta_m.append(constitutive_models['MAT'+str(n+1)+'']['rayleigh_damping']['eta_m'])
self.eta_k.append(constitutive_models['MAT'+str(n+1)+'']['rayleigh_damping']['eta_k'])
try: self.prestress_initial = fem_params['prestress_initial']
except: self.prestress_initial = False
# type of discontinuous function spaces
if str(self.io.mesh.ufl_cell()) == 'tetrahedron' or str(self.io.mesh.ufl_cell()) == 'triangle3D':
dg_type = "DG"
if (self.order_disp > 1 or self.order_pres > 1) and self.quad_degree < 3:
raise ValueError("Use at least a quadrature degree of 3 or more for higher-order meshes!")
elif str(self.io.mesh.ufl_cell()) == 'hexahedron' or str(self.io.mesh.ufl_cell()) == 'quadrilateral3D':
dg_type = "DQ"
if (self.order_disp > 1 or self.order_pres > 1) and self.quad_degree < 5:
raise ValueError("Use at least a quadrature degree of 5 or more for higher-order meshes!")
else:
raise NameError("Unknown cell/element type!")
# create finite element objects for u and p
P_u = VectorElement("CG", self.io.mesh.ufl_cell(), self.order_disp)
P_p = FiniteElement("CG", self.io.mesh.ufl_cell(), self.order_pres)
# function spaces for u and p
self.V_u = FunctionSpace(self.io.mesh, P_u)
self.V_p = FunctionSpace(self.io.mesh, P_p)
# Quadrature tensor, vector, and scalar elements
Q_tensor = TensorElement("Quadrature", self.io.mesh.ufl_cell(), degree=1, quad_scheme="default")
Q_vector = VectorElement("Quadrature", self.io.mesh.ufl_cell(), degree=1, quad_scheme="default")
Q_scalar = FiniteElement("Quadrature", self.io.mesh.ufl_cell(), degree=1, quad_scheme="default")
# not yet working - we cannot interpolate into Quadrature elements with the current dolfinx version currently!
#self.Vd_tensor = FunctionSpace(self.io.mesh, Q_tensor)
#self.Vd_vector = FunctionSpace(self.io.mesh, Q_vector)
#self.Vd_scalar = FunctionSpace(self.io.mesh, Q_scalar)
# Quadrature function spaces (currently not properly functioning for higher-order meshes!!!)
self.Vd_tensor = TensorFunctionSpace(self.io.mesh, (dg_type, self.order_disp-1))
self.Vd_vector = VectorFunctionSpace(self.io.mesh, (dg_type, self.order_disp-1))
self.Vd_scalar = FunctionSpace(self.io.mesh, (dg_type, self.order_disp-1))
# functions
self.du = TrialFunction(self.V_u) # Incremental displacement
self.var_u = TestFunction(self.V_u) # Test function
self.dp = TrialFunction(self.V_p) # Incremental pressure
self.var_p = TestFunction(self.V_p) # Test function
self.u = Function(self.V_u, name="Displacement")
self.p = Function(self.V_p, name="Pressure")
# values of previous time step
self.u_old = Function(self.V_u)
self.v_old = Function(self.V_u)
self.a_old = Function(self.V_u)
self.p_old = Function(self.V_p)
# a setpoint displacement for multiscale analysis
self.u_set = Function(self.V_u)
self.p_set = Function(self.V_p)
self.tau_a_set = Function(self.Vd_scalar)
# initial (zero) functions for initial stiffness evaluation (e.g. for Rayleigh damping)
self.u_ini, self.p_ini, self.theta_ini, self.tau_a_ini = Function(self.V_u), Function(self.V_p), Function(self.Vd_scalar), Function(self.Vd_scalar)
self.theta_ini.vector.set(1.0)
self.theta_ini.vector.ghostUpdate(addv=PETSc.InsertMode.INSERT, mode=PETSc.ScatterMode.FORWARD)
# growth stretch
self.theta = Function(self.Vd_scalar, name="theta")
self.theta_old = Function(self.Vd_scalar)
self.growth_thres = Function(self.Vd_scalar)
# initialize to one (theta = 1 means no growth)
self.theta.vector.set(1.0), self.theta_old.vector.set(1.0)
self.theta.vector.ghostUpdate(addv=PETSc.InsertMode.INSERT, mode=PETSc.ScatterMode.FORWARD), self.theta_old.vector.ghostUpdate(addv=PETSc.InsertMode.INSERT, mode=PETSc.ScatterMode.FORWARD)
# active stress
self.tau_a = Function(self.Vd_scalar, name="tau_a")
self.tau_a_old = Function(self.Vd_scalar)
self.amp_old, self.amp_old_set = Function(self.Vd_scalar), Function(self.Vd_scalar)
self.amp_old.vector.set(1.0), self.amp_old_set.vector.set(1.0)
self.amp_old.vector.ghostUpdate(addv=PETSc.InsertMode.INSERT, mode=PETSc.ScatterMode.FORWARD), self.amp_old_set.vector.ghostUpdate(addv=PETSc.InsertMode.INSERT, mode=PETSc.ScatterMode.FORWARD)
# prestressing history defgrad and spring prestress
if self.prestress_initial:
self.F_hist = Function(self.Vd_tensor, name="Defgrad_hist")
self.u_pre = Function(self.V_u)
else:
self.F_hist = None
self.u_pre = None
self.internalvars = {"theta" : self.theta, "tau_a" : self.tau_a}
self.internalvars_old = {"theta" : self.theta_old, "tau_a" : self.tau_a_old}
# reference coordinates
self.x_ref = Function(self.V_u)
self.x_ref.interpolate(self.x_ref_expr)
if self.incompressible_2field:
self.ndof = self.u.vector.getSize() + self.p.vector.getSize()
else:
self.ndof = self.u.vector.getSize()
# initialize solid time-integration class
self.ti = timeintegration.timeintegration_solid(time_params, fem_params, time_curves, self.t_init, self.comm)
# check for materials that need extra treatment (anisotropic, active stress, growth, ...)
have_fiber1, have_fiber2 = False, False
self.have_active_stress, self.active_stress_trig, self.have_frank_starling, self.have_growth = False, 'ode', False, False
self.mat_active_stress, self.mat_growth, self.mat_remodel, self.mat_growth_dir, self.mat_growth_trig, self.mat_growth_thres = [False]*self.num_domains, [False]*self.num_domains, [False]*self.num_domains, [None]*self.num_domains, [None]*self.num_domains, []*self.num_domains
self.localsolve, growth_dir = False, None
self.actstress = []
for n in range(self.num_domains):
if 'holzapfelogden_dev' in self.constitutive_models['MAT'+str(n+1)+''].keys() or 'guccione_dev' in self.constitutive_models['MAT'+str(n+1)+''].keys():
have_fiber1, have_fiber2 = True, True
if 'active_fiber' in self.constitutive_models['MAT'+str(n+1)+''].keys():
have_fiber1 = True
self.mat_active_stress[n], self.have_active_stress = True, True
# if one mat has a prescribed active stress, all have to be!
if 'prescribed_curve' in self.constitutive_models['MAT'+str(n+1)+'']['active_fiber']:
self.active_stress_trig = 'prescribed'
if 'prescribed_multiscale' in self.constitutive_models['MAT'+str(n+1)+'']['active_fiber']:
self.active_stress_trig = 'prescribed_multiscale'
if self.active_stress_trig == 'ode':
act_curve = self.ti.timecurves(self.constitutive_models['MAT'+str(n+1)+'']['active_fiber']['activation_curve'])
self.actstress.append(activestress_activation(self.constitutive_models['MAT'+str(n+1)+'']['active_fiber'], act_curve))
if self.actstress[-1].frankstarling: self.have_frank_starling = True
if self.active_stress_trig == 'prescribed':
self.ti.funcs_to_update.append({self.tau_a : self.ti.timecurves(self.constitutive_models['MAT'+str(n+1)+'']['active_fiber']['prescribed_curve'])})
if 'active_iso' in self.constitutive_models['MAT'+str(n+1)+''].keys():
self.mat_active_stress[n], self.have_active_stress = True, True
# if one mat has a prescribed active stress, all have to be!
if 'prescribed_curve' in self.constitutive_models['MAT'+str(n+1)+'']['active_iso']:
self.active_stress_trig = 'prescribed'
if 'prescribed_multiscale' in self.constitutive_models['MAT'+str(n+1)+'']['active_iso']:
self.active_stress_trig = 'prescribed_multiscale'
if self.active_stress_trig == 'ode':
act_curve = self.ti.timecurves(self.constitutive_models['MAT'+str(n+1)+'']['active_iso']['activation_curve'])
self.actstress.append(activestress_activation(self.constitutive_models['MAT'+str(n+1)+'']['active_iso'], act_curve))
if self.active_stress_trig == 'prescribed':
self.ti.funcs_to_update.append({self.tau_a : self.ti.timecurves(self.constitutive_models['MAT'+str(n+1)+'']['active_iso']['prescribed_curve'])})
if 'growth' in self.constitutive_models['MAT'+str(n+1)+''].keys():
self.mat_growth[n], self.have_growth = True, True
self.mat_growth_dir[n] = self.constitutive_models['MAT'+str(n+1)+'']['growth']['growth_dir']
self.mat_growth_trig[n] = self.constitutive_models['MAT'+str(n+1)+'']['growth']['growth_trig']
# need to have fiber fields for the following growth options
if self.mat_growth_dir[n] == 'fiber' or self.mat_growth_trig[n] == 'fibstretch':
have_fiber1 = True
if self.mat_growth_dir[n] == 'radial':
have_fiber1, have_fiber2 = True, True
# in this case, we have a theta that is (nonlinearly) dependent on the deformation, theta = theta(C(u)),
# therefore we need a local Newton iteration to solve for equilibrium theta (return mapping) prior to entering
# the global Newton scheme - so flag localsolve to true
if self.mat_growth_trig[n] != 'prescribed' and self.mat_growth_trig[n] != 'prescribed_multiscale':
self.localsolve = True
self.mat_growth_thres.append(self.constitutive_models['MAT'+str(n+1)+'']['growth']['growth_thres'])
else:
self.mat_growth_thres.append(as_ufl(0))
# for the case that we have a prescribed growth stretch over time, append curve to functions that need time updates
# if one mat has a prescribed growth model, all have to be!
if self.mat_growth_trig[n] == 'prescribed':
self.ti.funcs_to_update.append({self.theta : self.ti.timecurves(self.constitutive_models['MAT'+str(n+1)+'']['growth']['prescribed_curve'])})
if 'remodeling_mat' in self.constitutive_models['MAT'+str(n+1)+'']['growth'].keys():
self.mat_remodel[n] = True
else:
self.mat_growth_thres.append(as_ufl(0))
# full linearization of our remodeling law can lead to excessive compiler times for ffcx... :-/
# let's try if we might can go without one of the critial terms (derivative of remodeling fraction w.r.t. C)
try: self.lin_remod_full = fem_params['lin_remodeling_full']
except: self.lin_remod_full = True
# growth threshold (as function, since in multiscale approach, it can vary element-wise)
if self.have_growth and self.localsolve:
growth_thres_proj = project(self.mat_growth_thres, self.Vd_scalar, self.dx_)
self.growth_thres.vector.ghostUpdate(addv=PETSc.InsertMode.ADD, mode=PETSc.ScatterMode.REVERSE)
self.growth_thres.interpolate(growth_thres_proj)
# read in fiber data
if have_fiber1:
fibarray = ['fiber']
if have_fiber2: fibarray.append('sheet')
# fiber function space - vector defined on quadrature points
V_fib = self.Vd_vector
self.fib_func = self.io.readin_fibers(fibarray, V_fib, self.dx_)
else:
self.fib_func = None
# for multiscale G&R analysis
self.tol_stop_large = 0
# initialize kinematics class
self.ki = solid_kinematics_constitutive.kinematics(fib_funcs=self.fib_func, F_hist=self.F_hist)
# initialize material/constitutive class
self.ma = []
for n in range(self.num_domains):
self.ma.append(solid_kinematics_constitutive.constitutive(self.ki, self.constitutive_models['MAT'+str(n+1)+''], self.incompressible_2field, mat_growth=self.mat_growth[n], mat_remodel=self.mat_remodel[n]))
# initialize solid variational form class
self.vf = solid_variationalform.variationalform(self.var_u, self.du, self.var_p, self.dp, self.io.n0, self.x_ref)
# initialize boundary condition class
self.bc = boundaryconditions.boundary_cond_solid(bc_dict, self.fem_params, self.io, self.ki, self.vf, self.ti)
if self.prestress_initial:
# initialize prestressing history deformation gradient
Id_proj = project(Identity(len(self.u)), self.Vd_tensor, self.dx_)
self.F_hist.interpolate(Id_proj)
self.bc_dict = bc_dict
# Dirichlet boundary conditions
if 'dirichlet' in self.bc_dict.keys():
self.bc.dirichlet_bcs(self.V_u)
self.set_variational_forms_and_jacobians()
def set_variational_forms_and_jacobians(self):
# set forms for acceleration and velocity
self.acc, self.vel = self.ti.set_acc_vel(self.u, self.u_old, self.v_old, self.a_old)
# kinetic, internal, and pressure virtual work
self.deltaW_kin, self.deltaW_kin_old = as_ufl(0), as_ufl(0)
self.deltaW_int, self.deltaW_int_old = as_ufl(0), as_ufl(0)
self.deltaW_damp, self.deltaW_damp_old = as_ufl(0), as_ufl(0)
self.deltaW_p, self.deltaW_p_old = as_ufl(0), as_ufl(0)
for n in range(self.num_domains):
if self.timint != 'static':
# kinetic virtual work
self.deltaW_kin += self.vf.deltaW_kin(self.acc, self.rho0[n], self.dx_[n])
self.deltaW_kin_old += self.vf.deltaW_kin(self.a_old, self.rho0[n], self.dx_[n])
# Rayleigh damping virtual work
if self.rayleigh[n]:
self.deltaW_damp += self.vf.deltaW_damp(self.eta_m[n], self.eta_k[n], self.rho0[n], self.ma[n].S(self.u_ini, self.p_ini, ivar={"theta" : self.theta_ini, "tau_a" : self.tau_a_ini}, tang=True), self.vel, self.dx_[n])
self.deltaW_damp_old += self.vf.deltaW_damp(self.eta_m[n], self.eta_k[n], self.rho0[n], self.ma[n].S(self.u_ini, self.p_ini, ivar={"theta" : self.theta_ini, "tau_a" : self.tau_a_ini}, tang=True), self.v_old, self.dx_[n])
# internal virtual work
self.deltaW_int += self.vf.deltaW_int(self.ma[n].S(self.u, self.p, ivar=self.internalvars), self.ki.F(self.u), self.dx_[n])
self.deltaW_int_old += self.vf.deltaW_int(self.ma[n].S(self.u_old, self.p_old, ivar=self.internalvars_old), self.ki.F(self.u_old), self.dx_[n])
# pressure virtual work (for incompressible formulation)
# this has to be treated like the evaluation of a volumetric material, hence with the elastic part of J
if self.mat_growth[n]: J, J_old = self.ma[n].J_e(self.u, self.theta), self.ma[n].J_e(self.u_old, self.theta_old)
else: J, J_old = self.ki.J(self.u), self.ki.J(self.u_old)
self.deltaW_p += self.vf.deltaW_int_pres(J, self.dx_[n])
self.deltaW_p_old += self.vf.deltaW_int_pres(J_old, self.dx_[n])
# external virtual work (from Neumann or Robin boundary conditions, body forces, ...)
w_neumann, w_neumann_old, w_robin, w_robin_old = as_ufl(0), as_ufl(0), as_ufl(0), as_ufl(0)
if 'neumann' in self.bc_dict.keys():
w_neumann, w_neumann_old = self.bc.neumann_bcs(self.V_u, self.Vd_scalar, self.u, self.u_old)
if 'robin' in self.bc_dict.keys():
w_robin, w_robin_old = self.bc.robin_bcs(self.u, self.vel, self.u_old, self.v_old, self.u_pre)
# for (quasi-static) prestressing, we need to eliminate dashpots and replace true with reference Neumann loads in our external virtual work
w_neumann_prestr, w_robin_prestr = as_ufl(0), as_ufl(0)
if self.prestress_initial:
bc_dict_prestr = copy.deepcopy(self.bc_dict)
# get rid of dashpots
if 'robin' in bc_dict_prestr.keys():
for r in bc_dict_prestr['robin']:
if r['type'] == 'dashpot': r['visc'] = 0.
# replace true Neumann loads by reference ones
if 'neumann' in bc_dict_prestr.keys():
for n in bc_dict_prestr['neumann']:
if n['type'] == 'true': n['type'] = 'pk1'
bc_prestr = boundaryconditions.boundary_cond_solid(bc_dict_prestr, self.fem_params, self.io, self.ki, self.vf, self.ti)
if 'neumann' in bc_dict_prestr.keys():
w_neumann_prestr, _ = bc_prestr.neumann_bcs(self.V_u, self.Vd_scalar, self.u, self.u_old)
if 'robin' in bc_dict_prestr.keys():
w_robin_prestr, _ = bc_prestr.robin_bcs(self.u, self.vel, self.u_old, self.v_old, self.u_pre)
self.deltaW_prestr_ext = w_neumann_prestr + w_robin_prestr
# TODO: Body forces!
self.deltaW_ext = w_neumann + w_robin
self.deltaW_ext_old = w_neumann_old + w_robin_old
self.timefac_m, self.timefac = self.ti.timefactors()
### full weakforms
# quasi-static weak form: internal minus external virtual work
if self.timint == 'static':
self.weakform_u = self.deltaW_int - self.deltaW_ext
if self.incompressible_2field:
self.weakform_p = self.deltaW_p
# full dynamic weak form: kinetic plus internal (plus damping) minus external virtual work
else:
self.weakform_u = self.timefac_m * self.deltaW_kin + (1.-self.timefac_m) * self.deltaW_kin_old + \
self.timefac * self.deltaW_damp + (1.-self.timefac) * self.deltaW_damp_old + \
self.timefac * self.deltaW_int + (1.-self.timefac) * self.deltaW_int_old - \
self.timefac * self.deltaW_ext - (1.-self.timefac) * self.deltaW_ext_old
if self.incompressible_2field:
self.weakform_p = self.timefac * self.deltaW_p + (1.-self.timefac) * self.deltaW_p_old
### local weak forms at Gauss points for inelastic materials
self.r_growth, self.del_theta = [], []
for n in range(self.num_domains):
if self.mat_growth[n] and self.mat_growth_trig[n] != 'prescribed' and self.mat_growth_trig[n] != 'prescribed_multiscale':
# growth residual and increment
a, b = self.ma[n].res_dtheta_growth(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt, 'res_del')
self.r_growth.append(a), self.del_theta.append(b)
else:
self.r_growth.append(as_ufl(0)), self.del_theta.append(as_ufl(0))
### Jacobians
# kinetic virtual work linearization (deltaW_kin already has contributions from all domains)
self.jac_uu = self.timefac_m * derivative(self.deltaW_kin, self.u, self.du)
# internal virtual work linearization treated differently: since we want to be able to account for nonlinear materials at Gauss
# point level with deformation-dependent internal variables (i.e. growth or plasticity), we make use of a more explicit formulation
# of the linearization which involves the fourth-order material tangent operator Ctang ("derivative" cannot take care of the
# dependence of the internal variables on the deformation if this dependence is nonlinear and cannot be expressed analytically)
for n in range(self.num_domains):
# material tangent operator
Cmat = self.ma[n].S(self.u, self.p, ivar=self.internalvars, tang=True)
if self.mat_growth[n] and self.mat_growth_trig[n] != 'prescribed' and self.mat_growth_trig[n] != 'prescribed_multiscale':
# growth tangent operator
Cgrowth = self.ma[n].Cgrowth(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
if self.mat_remodel[n] and self.lin_remod_full:
# remodeling tangent operator
Cremod = self.ma[n].Cremod(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
Ctang = Cmat + Cgrowth + Cremod
else:
Ctang = Cmat + Cgrowth
else:
Ctang = Cmat
self.jac_uu += self.timefac * self.vf.Lin_deltaW_int_du(self.ma[n].S(self.u, self.p, ivar=self.internalvars), self.ki.F(self.u), self.u, Ctang, self.dx_[n])
# Rayleigh damping virtual work contribution to stiffness
self.jac_uu += self.timefac * derivative(self.deltaW_damp, self.u, self.du)
# external virtual work contribution to stiffness (from nonlinear follower loads or Robin boundary tractions)
self.jac_uu += -self.timefac * derivative(self.deltaW_ext, self.u, self.du)
# pressure contributions
if self.incompressible_2field:
self.jac_up, self.jac_pu, self.a_p11, self.p11 = as_ufl(0), as_ufl(0), as_ufl(0), as_ufl(0)
for n in range(self.num_domains):
# this has to be treated like the evaluation of a volumetric material, hence with the elastic part of J
if self.mat_growth[n]:
J = self.ma[n].J_e(self.u, self.theta)
Jmat = self.ma[n].dJedC(self.u, self.theta)
else:
J = self.ki.J(self.u)
Jmat = self.ki.dJdC(self.u)
Cmat_p = diff(self.ma[n].S(self.u, self.p, ivar=self.internalvars), self.p)
if self.mat_growth[n] and self.mat_growth_trig[n] != 'prescribed' and self.mat_growth_trig[n] != 'prescribed_multiscale':
Cmat = self.ma[n].S(self.u, self.p, ivar=self.internalvars, tang=True)
# growth tangent operators - keep in mind that we have theta = theta(C(u),p) in general!
# for stress-mediated growth, we get a contribution to the pressure material tangent operator
Cgrowth_p = self.ma[n].Cgrowth_p(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
if self.mat_remodel[n] and self.lin_remod_full:
# remodeling tangent operator
Cremod_p = self.ma[n].Cremod_p(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
Ctang_p = Cmat_p + Cgrowth_p + Cremod_p
else:
Ctang_p = Cmat_p + Cgrowth_p
# for all types of deformation-dependent growth, we need to add the growth contributions to the Jacobian tangent operator
Jgrowth = diff(J,self.theta) * self.ma[n].dtheta_dC(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
Jtang = Jmat + Jgrowth
# ok... for stress-mediated growth, we actually get a non-zero right-bottom (11) block in our saddle-point system matrix,
# since Je = Je(C,theta(C,p)) ---> dJe/dp = dJe/dtheta * dtheta/dp
# TeX: D_{\Delta p}\!\int\limits_{\Omega_0} (J^{\mathrm{e}}-1)\delta p\,\mathrm{d}V = \int\limits_{\Omega_0} \frac{\partial J^{\mathrm{e}}}{\partial p}\Delta p \,\delta p\,\mathrm{d}V,
# with \frac{\partial J^{\mathrm{e}}}{\partial p} = \frac{\partial J^{\mathrm{e}}}{\partial \vartheta}\frac{\partial \vartheta}{\partial p}
dthetadp = self.ma[n].dtheta_dp(self.u, self.p, self.internalvars, self.theta_old, self.growth_thres, self.dt)
if not isinstance(dthetadp, constantvalue.Zero):
self.p11 += diff(J,self.theta) * dthetadp * self.dp * self.var_p * self.dx_[n]
else:
Ctang_p = Cmat_p
Jtang = Jmat
self.jac_up += self.timefac * self.vf.Lin_deltaW_int_dp(self.ki.F(self.u), Ctang_p, self.dx_[n])
self.jac_pu += self.timefac * self.vf.Lin_deltaW_int_pres_du(self.ki.F(self.u), Jtang, self.u, self.dx_[n])
# for saddle-point block-diagonal preconditioner
self.a_p11 += inner(self.dp, self.var_p) * self.dx_[n]
if self.prestress_initial:
# quasi-static weak forms (don't dare to use fancy growth laws or other inelastic stuff during prestressing...)
self.weakform_prestress_u = self.deltaW_int - self.deltaW_prestr_ext
self.jac_prestress_uu = derivative(self.weakform_prestress_u, self.u, self.du)
if self.incompressible_2field:
self.weakform_prestress_p = self.deltaW_p
self.jac_prestress_up = derivative(self.weakform_prestress_u, self.p, self.dp)
self.jac_prestress_pu = derivative(self.weakform_prestress_p, self.u, self.du)
