/** * Compute contribution from internal forces in elem_residual and contribution from * linearized internal forces (stiffness matrix) in elem_jacobian. */ bool SolidSystem::element_time_derivative(bool request_jacobian, DiffContext &context) { FEMContext &c = libmesh_cast_ref<FEMContext&>(context); // First we get some references to cell-specific data that // will be used to assemble the linear system. // Element Jacobian * quadrature weights for interior integration const std::vector<Real> &JxW = c.element_fe_var[var[0]]->get_JxW(); // The gradients of the shape functions at interior // quadrature points. const std::vector<std::vector<RealGradient> >& dphi = c.element_fe_var[var[0]]->get_dphi(); // Dimension of the mesh const unsigned int dim = this->get_mesh().mesh_dimension(); // The number of local degrees of freedom in each variable const unsigned int n_u_dofs = c.dof_indices_var[var[0]].size(); libmesh_assert(n_u_dofs == c.dof_indices_var[var[1]].size()); if (dim == 3) { libmesh_assert(n_u_dofs == c.dof_indices_var[var[2]].size()); } unsigned int n_qpoints = c.element_qrule->n_points(); // Some matrices and vectors for storing the results of the constitutive // law DenseMatrix<Real> stiff; DenseVector<Real> res; VectorValue<Gradient> grad_u; // Instantiate the constitutive law NonlinearNeoHookeCurrentConfig material(dphi, args); // Just calculate jacobian contribution when we need to material.calculate_linearized_stiffness = request_jacobian; // Get a reference to the auxiliary system TransientExplicitSystem& aux_system = this->get_equation_systems().get_system< TransientExplicitSystem>("auxiliary"); std::vector<dof_id_type> undefo_index; // Assume symmetry of local stiffness matrices bool use_symmetry = args("assembly/use_symmetry", false); // Now we will build the element Jacobian and residual. // This must be calculated at each quadrature point by // summing the solution degree-of-freedom values by // the appropriate weight functions. // This class just takes care of the assembly. The matrix of // the jacobian and the residual vector are provided by the // constitutive formulation. for (unsigned int qp = 0; qp != n_qpoints; qp++) { // Compute the displacement gradient grad_u(0) = grad_u(1) = grad_u(2) = 0; for (unsigned int d = 0; d < dim; ++d) { std::vector<Number> u_undefo; aux_system.get_dof_map().dof_indices(c.elem, undefo_index, undefo_var[d]); aux_system.current_local_solution->get(undefo_index, u_undefo); for (unsigned int l = 0; l != n_u_dofs; l++) grad_u(d).add_scaled(dphi[l][qp], u_undefo[l]); // u_current(l)); // - } // initialize the constitutive formulation with the current displacement // gradient material.init_for_qp(grad_u, qp); // Aquire, scale and assemble residual and stiffness for (unsigned int i = 0; i < n_u_dofs; i++) { res.resize(dim); material.get_residual(res, i); res.scale(JxW[qp]); for (unsigned int ii = 0; ii < dim; ++ii) { c.elem_subresiduals[ii]->operator ()(i) += res(ii); } if (request_jacobian && c.elem_solution_derivative) { libmesh_assert(c.elem_solution_derivative == 1.0); for (unsigned int j = (use_symmetry ? i : 0); j < n_u_dofs; j++) { material.get_linearized_stiffness(stiff, i, j); stiff.scale(JxW[qp]); for (unsigned int ii = 0; ii < dim; ++ii) { for (unsigned int jj = 0; jj < dim; ++jj) { c.elem_subjacobians[ii][jj]->operator ()(i, j) += stiff(ii, jj); if (use_symmetry && i != j) { c.elem_subjacobians[ii][jj]->operator ()(j, i) += stiff(jj, ii); } } } } } } } // end of the quadrature point qp-loop return request_jacobian; }
// The matrix assembly function to be called at each time step to // prepare for the linear solve. void assemble_solid (EquationSystems& es, const std::string& system_name) { //es.print_info(); #if LOG_ASSEMBLE_PERFORMANCE PerfLog perf_log("Assemble"); perf_log.push("assemble stiffness"); #endif // Get a reference to the auxiliary system //TransientExplicitSystem& aux_system = es.get_system<TransientExplicitSystem>("Newton-update"); // It is a good idea to make sure we are assembling // the proper system. libmesh_assert (system_name == "Newton-update"); // Get a constant reference to the mesh object. const MeshBase& mesh = es.get_mesh(); // The dimension that we are running const unsigned int dim = mesh.mesh_dimension(); // Get a reference to the Stokes system object. TransientLinearImplicitSystem & newton_update = es.get_system<TransientLinearImplicitSystem> ("Newton-update"); // New TransientLinearImplicitSystem & last_non_linear_soln = es.get_system<TransientLinearImplicitSystem> ("Last-non-linear-soln"); TransientLinearImplicitSystem & fluid_system_vel = es.get_system<TransientLinearImplicitSystem> ("fluid-system-vel"); #if VELOCITY TransientLinearImplicitSystem& velocity = es.get_system<TransientLinearImplicitSystem>("velocity-system"); #endif #if UN_MINUS_ONE TransientLinearImplicitSystem & unm1 = es.get_system<TransientLinearImplicitSystem> ("unm1-system"); #endif test(62); const System & ref_sys = es.get_system("Reference-Configuration"); test(63); // Numeric ids corresponding to each variable in the system const unsigned int u_var = last_non_linear_soln .variable_number ("u"); const unsigned int v_var = last_non_linear_soln .variable_number ("v"); const unsigned int w_var = last_non_linear_soln .variable_number ("w"); #if INCOMPRESSIBLE const unsigned int p_var = last_non_linear_soln .variable_number ("p"); #endif #if FLUID_P_CONST const unsigned int m_var = fluid_system_vel.variable_number ("fluid_M"); std::vector<unsigned int> dof_indices_m; #endif // Get the Finite Element type for "u". Note this will be // the same as the type for "v". FEType fe_vel_type = last_non_linear_soln.variable_type(u_var); test(64); #if INCOMPRESSIBLE // Get the Finite Element type for "p". FEType fe_pres_type = last_non_linear_soln .variable_type(p_var); #endif // Build a Finite Element object of the specified type for // the velocity variables. AutoPtr<FEBase> fe_vel (FEBase::build(dim, fe_vel_type)); #if INCOMPRESSIBLE // Build a Finite Element object of the specified type for // the pressure variables. AutoPtr<FEBase> fe_pres (FEBase::build(dim, fe_pres_type)); #endif // A Gauss quadrature rule for numerical integration. // Let the \p FEType object decide what order rule is appropriate. QGauss qrule (dim, fe_vel_type.default_quadrature_order()); // Tell the finite element objects to use our quadrature rule. fe_vel->attach_quadrature_rule (&qrule); test(65); // AutoPtr<QBase> qrule2(fe_vel_type.default_quadrature_rule(dim)); // fe_vel->attach_quadrature_rule (qrule2.get()); #if INCOMPRESSIBLE fe_pres->attach_quadrature_rule (&qrule); #endif // The element Jacobian * quadrature weight at each integration point. const std::vector<Real>& JxW = fe_vel->get_JxW(); // The element shape functions evaluated at the quadrature points. const std::vector<std::vector<Real> >& phi = fe_vel->get_phi(); // The element shape function gradients for the velocity // variables evaluated at the quadrature points. const std::vector<std::vector<RealGradient> >& dphi = fe_vel->get_dphi(); test(66); #if INCOMPRESSIBLE // The element shape functions for the pressure variable // evaluated at the quadrature points. const std::vector<std::vector<Real> >& psi = fe_pres->get_phi(); #endif const std::vector<Point>& coords = fe_vel->get_xyz(); // A reference to the \p DofMap object for this system. The \p DofMap // object handles the index translation from node and element numbers // to degree of freedom numbers. We will talk more about the \p DofMap // in future examples. const DofMap & dof_map = last_non_linear_soln .get_dof_map(); #if FLUID_P_CONST const DofMap & dof_map_fluid = fluid_system_vel .get_dof_map(); #endif // K will be the jacobian // F will be the Residual DenseMatrix<Number> Ke; DenseVector<Number> Fe; DenseSubMatrix<Number> Kuu(Ke), Kuv(Ke), Kuw(Ke), Kvu(Ke), Kvv(Ke), Kvw(Ke), Kwu(Ke), Kwv(Ke), Kww(Ke); #if INCOMPRESSIBLE DenseSubMatrix<Number> Kup(Ke),Kvp(Ke),Kwp(Ke), Kpu(Ke), Kpv(Ke), Kpw(Ke), Kpp(Ke); #endif; DenseSubVector<Number> Fu(Fe), Fv(Fe), Fw(Fe); #if INCOMPRESSIBLE DenseSubVector<Number> Fp(Fe); #endif // This vector will hold the degree of freedom indices for // the element. These define where in the global system // the element degrees of freedom get mapped. std::vector<unsigned int> dof_indices; std::vector<unsigned int> dof_indices_u; std::vector<unsigned int> dof_indices_v; std::vector<unsigned int> dof_indices_w; #if INCOMPRESSIBLE std::vector<unsigned int> dof_indices_p; #endif #if INERTIA test(67); const unsigned int a_var = last_non_linear_soln.variable_number ("a"); const unsigned int b_var = last_non_linear_soln.variable_number ("b"); const unsigned int c_var = last_non_linear_soln.variable_number ("c"); //B block DenseSubMatrix<Number> Kua(Ke), Kub(Ke), Kuc(Ke), Kva(Ke), Kvb(Ke), Kvc(Ke), Kwa(Ke), Kwb(Ke), Kwc(Ke); //C block DenseSubMatrix<Number> Kau(Ke), Kav(Ke), Kaw(Ke), Kbu(Ke), Kbv(Ke), Kbw(Ke), Kcu(Ke), Kcv(Ke), Kcw(Ke); //D block DenseSubMatrix<Number> Kaa(Ke), Kab(Ke), Kac(Ke), Kba(Ke), Kbb(Ke), Kbc(Ke), Kca(Ke), Kcb(Ke), Kcc(Ke); DenseSubVector<Number> Fa(Fe), Fb(Fe), Fc(Fe); std::vector<unsigned int> dof_indices_a; std::vector<unsigned int> dof_indices_b; std::vector<unsigned int> dof_indices_c; test(68); #endif DenseMatrix<Real> stiff; DenseVector<Real> res; VectorValue<Gradient> grad_u_mat; VectorValue<Gradient> grad_u_mat_old; const Real dt = es.parameters.get<Real>("dt"); const Real progress = es.parameters.get<Real>("progress"); #if PORO DenseVector<Real> p_stiff; DenseVector<Real> p_res; PoroelasticConfig material(dphi,psi); #endif // Just calculate jacobian contribution when we need to material.calculate_linearized_stiffness = true; MeshBase::const_element_iterator el = mesh.active_local_elements_begin(); const MeshBase::const_element_iterator end_el = mesh.active_local_elements_end(); for ( ; el != end_el; ++el) { test(69); const Elem* elem = *el; dof_map.dof_indices (elem, dof_indices); dof_map.dof_indices (elem, dof_indices_u, u_var); dof_map.dof_indices (elem, dof_indices_v, v_var); dof_map.dof_indices (elem, dof_indices_w, w_var); #if INCOMPRESSIBLE dof_map.dof_indices (elem, dof_indices_p, p_var); #endif const unsigned int n_dofs = dof_indices.size(); const unsigned int n_u_dofs = dof_indices_u.size(); const unsigned int n_v_dofs = dof_indices_v.size(); const unsigned int n_w_dofs = dof_indices_w.size(); #if INCOMPRESSIBLE const unsigned int n_p_dofs = dof_indices_p.size(); #endif #if FLUID_P_CONST dof_map_fluid.dof_indices (elem, dof_indices_m, m_var); #endif //elem->print_info(); fe_vel->reinit (elem); #if INCOMPRESSIBLE fe_pres->reinit (elem); #endif Ke.resize (n_dofs, n_dofs); Fe.resize (n_dofs); Kuu.reposition (u_var*n_u_dofs, u_var*n_u_dofs, n_u_dofs, n_u_dofs); Kuv.reposition (u_var*n_u_dofs, v_var*n_u_dofs, n_u_dofs, n_v_dofs); Kuw.reposition (u_var*n_u_dofs, w_var*n_u_dofs, n_u_dofs, n_w_dofs); #if INCOMPRESSIBLE Kup.reposition (u_var*n_u_dofs, p_var*n_u_dofs, n_u_dofs, n_p_dofs); #endif Kvu.reposition (v_var*n_v_dofs, u_var*n_v_dofs, n_v_dofs, n_u_dofs); Kvv.reposition (v_var*n_v_dofs, v_var*n_v_dofs, n_v_dofs, n_v_dofs); Kvw.reposition (v_var*n_v_dofs, w_var*n_v_dofs, n_v_dofs, n_w_dofs); #if INCOMPRESSIBLE Kvp.reposition (v_var*n_v_dofs, p_var*n_v_dofs, n_v_dofs, n_p_dofs); #endif Kwu.reposition (w_var*n_w_dofs, u_var*n_w_dofs, n_w_dofs, n_u_dofs); Kwv.reposition (w_var*n_w_dofs, v_var*n_w_dofs, n_w_dofs, n_v_dofs); Kww.reposition (w_var*n_w_dofs, w_var*n_w_dofs, n_w_dofs, n_w_dofs); #if INCOMPRESSIBLE Kwp.reposition (w_var*n_w_dofs, p_var*n_w_dofs, n_w_dofs, n_p_dofs); Kpu.reposition (p_var*n_u_dofs, u_var*n_u_dofs, n_p_dofs, n_u_dofs); Kpv.reposition (p_var*n_u_dofs, v_var*n_u_dofs, n_p_dofs, n_v_dofs); Kpw.reposition (p_var*n_u_dofs, w_var*n_u_dofs, n_p_dofs, n_w_dofs); Kpp.reposition (p_var*n_u_dofs, p_var*n_u_dofs, n_p_dofs, n_p_dofs); #endif Fu.reposition (u_var*n_u_dofs, n_u_dofs); Fv.reposition (v_var*n_u_dofs, n_v_dofs); Fw.reposition (w_var*n_u_dofs, n_w_dofs); #if INCOMPRESSIBLE Fp.reposition (p_var*n_u_dofs, n_p_dofs); #endif #if INERTIA //B block Kua.reposition (u_var*n_u_dofs, 3*n_u_dofs + n_p_dofs, n_u_dofs, n_u_dofs); Kub.reposition (u_var*n_u_dofs, 4*n_u_dofs + n_p_dofs, n_u_dofs, n_v_dofs); Kuc.reposition (u_var*n_u_dofs, 5*n_u_dofs + n_p_dofs, n_u_dofs, n_w_dofs); Kva.reposition (v_var*n_v_dofs, 3*n_u_dofs + n_p_dofs, n_v_dofs, n_u_dofs); Kvb.reposition (v_var*n_v_dofs, 4*n_u_dofs + n_p_dofs, n_v_dofs, n_v_dofs); Kvc.reposition (v_var*n_v_dofs, 5*n_u_dofs + n_p_dofs, n_v_dofs, n_w_dofs); Kwa.reposition (w_var*n_w_dofs, 3*n_u_dofs + n_p_dofs, n_w_dofs, n_u_dofs); Kwb.reposition (w_var*n_w_dofs, 4*n_u_dofs + n_p_dofs, n_w_dofs, n_v_dofs); Kwc.reposition (w_var*n_w_dofs, 5*n_u_dofs + n_p_dofs, n_w_dofs, n_w_dofs); test(701); //C block Kau.reposition (3*n_u_dofs + n_p_dofs, u_var*n_u_dofs, n_u_dofs, n_u_dofs); Kav.reposition (3*n_u_dofs + n_p_dofs, v_var*n_u_dofs, n_u_dofs, n_v_dofs); Kaw.reposition (3*n_u_dofs + n_p_dofs, w_var*n_u_dofs, n_u_dofs, n_w_dofs); Kbu.reposition (4*n_u_dofs + n_p_dofs, u_var*n_v_dofs, n_v_dofs, n_u_dofs); Kbv.reposition (4*n_u_dofs + n_p_dofs, v_var*n_v_dofs, n_v_dofs, n_v_dofs); Kbw.reposition (4*n_u_dofs + n_p_dofs, w_var*n_v_dofs, n_v_dofs, n_w_dofs); Kcu.reposition (5*n_u_dofs + n_p_dofs, u_var*n_w_dofs, n_w_dofs, n_u_dofs); Kcv.reposition (5*n_u_dofs + n_p_dofs, v_var*n_w_dofs, n_w_dofs, n_v_dofs); Kcw.reposition (5*n_u_dofs + n_p_dofs, w_var*n_w_dofs, n_w_dofs, n_w_dofs); //D block Kaa.reposition (3*n_u_dofs + n_p_dofs, 3*n_u_dofs + n_p_dofs, n_u_dofs, n_u_dofs); Kab.reposition (3*n_u_dofs + n_p_dofs, 4*n_u_dofs + n_p_dofs, n_u_dofs, n_v_dofs); Kac.reposition (3*n_u_dofs + n_p_dofs, 5*n_u_dofs + n_p_dofs, n_u_dofs, n_w_dofs); Kba.reposition (4*n_u_dofs + n_p_dofs, 3*n_u_dofs + n_p_dofs, n_v_dofs, n_u_dofs); Kbb.reposition (4*n_u_dofs + n_p_dofs, 4*n_u_dofs + n_p_dofs, n_v_dofs, n_v_dofs); Kbc.reposition (4*n_u_dofs + n_p_dofs, 5*n_u_dofs + n_p_dofs, n_v_dofs, n_w_dofs); Kca.reposition (5*n_u_dofs + n_p_dofs, 3*n_u_dofs + n_p_dofs, n_w_dofs, n_u_dofs); Kcb.reposition (5*n_u_dofs + n_p_dofs, 4*n_u_dofs + n_p_dofs, n_w_dofs, n_v_dofs); Kcc.reposition (5*n_u_dofs + n_p_dofs, 5*n_u_dofs + n_p_dofs, n_w_dofs, n_w_dofs); Fa.reposition (3*n_u_dofs + n_p_dofs, n_u_dofs); Fb.reposition (4*n_u_dofs + n_p_dofs, n_v_dofs); Fc.reposition (5*n_u_dofs + n_p_dofs, n_w_dofs); dof_map.dof_indices (elem, dof_indices_a, a_var); dof_map.dof_indices (elem, dof_indices_b, b_var); dof_map.dof_indices (elem, dof_indices_c, c_var); test(71); #endif System& aux_system = es.get_system<System>("Reference-Configuration"); std::vector<unsigned int> undefo_index; std::vector<unsigned int> vel_index; for (unsigned int qp=0; qp<qrule.n_points(); qp++) { Point rX; for (unsigned int l=0; l<n_u_dofs; l++) { rX(0) += phi[l][qp]*ref_sys.current_local_solution->el(dof_indices_u[l]); rX(1) += phi[l][qp]*ref_sys.current_local_solution->el(dof_indices_v[l]); rX(2) += phi[l][qp]*ref_sys.current_local_solution->el(dof_indices_w[l]); } #if INERTIA || DT test(72); Real rho_s=15; Point current_x; for (unsigned int l=0; l<n_u_dofs; l++) { current_x(0) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_u[l]); current_x(1) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_v[l]); current_x(2) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_w[l]); } Point old_x; for (unsigned int l=0; l<n_u_dofs; l++) { old_x(0) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_u[l]); old_x(1) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_v[l]); old_x(2) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_w[l]); } #if INERTIA Point old_vel; for (unsigned int l=0; l<n_u_dofs; l++) { old_vel(0) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_a[l]); old_vel(1) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_b[l]); old_vel(2) += phi[l][qp]*last_non_linear_soln.old_local_solution->el(dof_indices_c[l]); } Point current_vel; for (unsigned int l=0; l<n_u_dofs; l++) { current_vel(0) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_a[l]); current_vel(1) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_b[l]); current_vel(2) += phi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_c[l]); } #endif #if UN_MINUS_ONE Point unm1_x; for (unsigned int l=0; l<n_u_dofs; l++) { unm1_x(0) += phi[l][qp]*unm1.old_local_solution->el(dof_indices_u[l]); unm1_x(1) += phi[l][qp]*unm1.old_local_solution->el(dof_indices_v[l]); unm1_x(2) += phi[l][qp]*unm1.old_local_solution->el(dof_indices_w[l]); } #endif Point value_acc; Point value_acc_alt; #if DT for (unsigned int d = 0; d < dim; ++d) { value_acc_alt(d) = (rho_s)*( ((current_x(d)-rX(d))-(old_x(d)-rX(d)))-((old_x(d)-rX(d))- (unm1_x(d)-rX(d))) ); value_acc(d) = (rho_s)*((current_x(d))-2*(old_x(d))+ (unm1_x(d))); value_acc(d) = (rho_s)*((current_x(d))-(old_x(d))); } #endif Point res_1; Point res_2; #if INERTIA for (unsigned int d = 0; d < dim; ++d) { res_1(d) = (rho_s)*((current_vel(d))-(old_vel(d))); res_2(d) = current_x(d)-dt*current_vel(d)-old_x(d); } /* std::cout<<" current_vel "<<current_vel<<std::endl; std::cout<<" res_1 "<<res_1<<std::endl; std::cout<<" res_2 "<<res_2<<std::endl; */ #endif test(73); #endif Number p_solid = 0.; #if MOVING_MESH grad_u_mat(0) = grad_u_mat(1) = grad_u_mat(2) = 0; for (unsigned int d = 0; d < dim; ++d) { std::vector<Number> u_undefo; //Fills the vector di with the global degree of freedom indices for the element. :dof_indicies aux_system.get_dof_map().dof_indices(elem, undefo_index,d); aux_system.current_local_solution->get(undefo_index, u_undefo); for (unsigned int l = 0; l != n_u_dofs; l++) grad_u_mat(d).add_scaled(dphi[l][qp], u_undefo[l]); } #endif //#include "fixed_mesh_in_solid_assemble_code.txt" #if INCOMPRESSIBLE for (unsigned int l=0; l<n_p_dofs; l++) { p_solid += psi[l][qp]*last_non_linear_soln.current_local_solution->el(dof_indices_p[l]); } #endif #if INCOMPRESSIBLE Real m=0; Real p_fluid=0; #if FLUID_VEL for (unsigned int l=0; l<n_p_dofs; l++) { p_fluid += psi[l][qp]*fluid_system_vel.current_local_solution->el(dof_indices_p[l]); } //As outlined in Chappel p=(p_curr-p_old)/2 Real p_fluid_old=0; for (unsigned int l=0; l<n_p_dofs; l++) { p_fluid_old += psi[l][qp]*fluid_system_vel.old_local_solution->el(dof_indices_p[l]); } p_fluid=0.5*p_fluid+0.5*p_fluid_old; Real m_old=0; #if FLUID_P_CONST for (unsigned int l=0; l<n_p_dofs; l++) { m += psi[l][qp]*fluid_system_vel.current_local_solution->el(dof_indices_m[l]); } for (unsigned int l=0; l<n_p_dofs; l++) { m_old += psi[l][qp]*fluid_system_vel.old_local_solution->el(dof_indices_m[l]); } #endif material.init_for_qp(grad_u_mat, p_solid, qp, 1.0*m+0.0*m_old, p_fluid); #endif #endif //#if INCOMPRESSIBLE #if INCOMPRESSIBLE && ! PORO material.init_for_qp(grad_u_mat, p_solid, qp); #endif for (unsigned int i=0; i<n_u_dofs; i++) { res.resize(dim); material.get_residual(res, i); res.scale(JxW[qp]); #if INERTIA res.scale(dt); #endif #if DT res.scale(dt); #endif //std::cout<< "res "<<res<<std::endl; Fu(i) += res(0); Fv(i) += res(1) ; Fw(i) += res(2); #if GRAVITY Real grav=0.0; Fu(i) += progress*grav*JxW[qp]*phi[i][qp]; #endif #if INERTIA Fu(i) += JxW[qp]*phi[i][qp]*res_1(0); Fv(i) += JxW[qp]*phi[i][qp]*res_1(1); Fw(i) += JxW[qp]*phi[i][qp]*res_1(2); Fa(i) += JxW[qp]*phi[i][qp]*res_2(0); Fb(i) += JxW[qp]*phi[i][qp]*res_2(1); Fc(i) += JxW[qp]*phi[i][qp]*res_2(2); #endif // Matrix contributions for the uu and vv couplings. for (unsigned int j=0; j<n_u_dofs; j++) { material.get_linearized_stiffness(stiff, i, j); stiff.scale(JxW[qp]); #if DT res.scale(dt); #endif #if INERTIA stiff.scale(dt); Kua(i,j)+= rho_s*JxW[qp]*phi[i][qp]*phi[j][qp]; Kvb(i,j)+= rho_s*JxW[qp]*phi[i][qp]*phi[j][qp]; Kwc(i,j)+= rho_s*JxW[qp]*phi[i][qp]*phi[j][qp]; Kau(i,j)+= JxW[qp]*phi[i][qp]*phi[j][qp]; Kbv(i,j)+= JxW[qp]*phi[i][qp]*phi[j][qp]; Kcw(i,j)+= JxW[qp]*phi[i][qp]*phi[j][qp]; Kaa(i,j)+= -dt*JxW[qp]*phi[i][qp]*phi[j][qp]; Kbb(i,j)+= -dt*JxW[qp]*phi[i][qp]*phi[j][qp]; Kcc(i,j)+= -dt*JxW[qp]*phi[i][qp]*phi[j][qp]; #endif Kuu(i,j)+= stiff(u_var, u_var); Kuv(i,j)+= stiff(u_var, v_var); Kuw(i,j)+= stiff(u_var, w_var); Kvu(i,j)+= stiff(v_var, u_var); Kvv(i,j)+= stiff(v_var, v_var); Kvw(i,j)+= stiff(v_var, w_var); Kwu(i,j)+= stiff(w_var, u_var); Kwv(i,j)+= stiff(w_var, v_var); Kww(i,j)+= stiff(w_var, w_var); #if GRAVITY Kuu(i,j)+= 1*JxW[qp]*phi[i][qp]*phi[j][qp]; #endif } } #if INCOMPRESSIBLE && FLUID_P_CONST for (unsigned int i = 0; i < n_p_dofs; i++) { material.get_p_residual(p_res, i); p_res.scale(JxW[qp]); Fp(i) += p_res(0); } for (unsigned int i = 0; i < n_u_dofs; i++) { for (unsigned int j = 0; j < n_p_dofs; j++) { material.get_linearized_uvw_p_stiffness(p_stiff, i, j); p_stiff.scale(JxW[qp]); Kup(i, j) += p_stiff(0); Kvp(i, j) += p_stiff(1); Kwp(i, j) += p_stiff(2); } } for (unsigned int i = 0; i < n_p_dofs; i++) { for (unsigned int j = 0; j < n_u_dofs; j++) { material.get_linearized_p_uvw_stiffness(p_stiff, i, j); p_stiff.scale(JxW[qp]); Kpu(i, j) += p_stiff(0); Kpv(i, j) += p_stiff(1); Kpw(i, j) += p_stiff(2); } } #endif #if CHAP && ! FLUID_P_CONST for (unsigned int i = 0; i < n_p_dofs; i++) { Fp(i) += 0*JxW[qp]*psi[i][qp]; } for (unsigned int i = 0; i < n_p_dofs; i++) { for (unsigned int j = 0; j < n_p_dofs; j++) { Kpp(i, j) += 1*JxW[qp]*psi[i][qp]*psi[j][qp]; } } #endif }//end of qp loop newton_update.matrix->add_matrix (Ke, dof_indices); newton_update.rhs->add_vector (Fe, dof_indices); } // end of element loop // dof_map.constrain_element_matrix_and_vector (Ke, Fe, dof_indices); newton_update.rhs->close(); newton_update.matrix->close(); #if LOG_ASSEMBLE_PERFORMANCE perf_log.pop("assemble stiffness"); #endif #if LOG_ASSEMBLE_PERFORMANCE perf_log.push("assemble bcs"); #endif //Assemble the boundary conditions. assemble_bcs(es); #if LOG_ASSEMBLE_PERFORMANCE perf_log.pop("assemble bcs"); #endif std::ofstream lhs_out("lhsoutS3.dat"); Ke.print(lhs_out); lhs_out.close(); return; }