Hi Jean, Thanks a lot for your detailed answer. Using this idea, I was able to solve the standard eigenvalue problem.
Thanks again! Animesh On Saturday, September 26, 2020 at 1:49:10 AM UTC+5:30 Jean-Paul Pelteret wrote: > Hi Animesh, > > I’m glad to hear that you’ve gotten Arpack working, and that you’re > getting somewhere with your implementation. When you encounter a crash like > this, its best to run the program in debug mode (better yet, do all of your > development in debug mode). In debug mode, there are a lot of additional > checks that can help you find the source of issues. You can read more about > this here > <https://github.com/dealii/dealii/wiki/Frequently-Asked-Questions#how-can-dealii-help-me-find-bugs> > . > > I see that you invoke the identity matrix inline with the solve() step > like this: IdentityMatrix() . According to the documentation > <https://dealii.org/developer/doxygen/deal.II/classIdentityMatrix.html#a271487fb3b1eeec33ca734aa05601110>, > > this is incorrect because the matrix is zero sized (i.e. holds no entries). > You probably want to use this constructor > <https://dealii.org/developer/doxygen/deal.II/classIdentityMatrix.html#a849adfebd6f0068b9f49dd8e6a6c3746>, > > and because the matrix should have the same size as the tangent matrix you > would call it like this: IdentityMatrix(tangent_matrix > .block(u_dof,u_dof).m()). > > I am a bit confused of where actually to solve the standard eigenvalue > problem in the code for my purpose. > > > Well, from your original post it sounds to me like you want to investigate > some properties of the equilibrium state of your problem. That makes sense. > So really what you probably want to do is let the nonlinear solver iterate > until convergence, and then keep the system matrix and RHS vector around > (i.e. don’t clear them) and finally as a post-processing step solve the > eigenvalue problem. So I would suggest that the entire code block that you > added to the solve_linear_system() method simply be moved to the > output_results() method, or to some new method that is invoked between > solve_nonlinear_timestep() and time.increment() inside the main run() > function. > > Best, > Jean-Paul > > On 25 Sep 2020, at 08:36, Animesh Rastogi IIT Gandhinagar < > animesh...@alumni.iitgn.ac.in> wrote: > > Hi, > > I could figure out how to give Identity Matrix instead of the mass matrix > to solve the Standard Eigen value problem. However, on running, I am > getting the folllowing error. > > Timestep 1 @ 0.1s > > _______________________________________________________________________________________ > SOLVER STEP | LIN_IT LIN_RES RES_NORM RES_U > NU_NORM NU_U > > _______________________________________________________________________________________ > 0 CST ASM SLV CMakeFiles/run.dir/build.make:57: recipe for target > 'CMakeFiles/run' failed > make[3]: *** [CMakeFiles/run] Segmentation fault (core dumped) > CMakeFiles/Makefile2:131: recipe for target 'CMakeFiles/run.dir/all' failed > make[2]: *** [CMakeFiles/run.dir/all] Error 2 > CMakeFiles/Makefile2:138: recipe for target 'CMakeFiles/run.dir/rule' > failed > make[1]: *** [CMakeFiles/run.dir/rule] Error 2 > Makefile:144: recipe for target 'run' failed > make: *** [run] Error 2 > > > > I am copying here the solve_linear_system function and the part that I > updated is blue in color. Also, I wish to calculate the eigenvalues after > the newton method has converged. I am a bit confused of where actually to > solve the standard eigenvalue problem in the code for my purpose. Could > someone please help me with these questions? > > ------------------------------------------------ > template <int dim,typename NumberType> > std::pair<unsigned int, double> > Solid<dim,NumberType>::solve_linear_system(BlockVector<double> > &newton_update) > { > BlockVector<double> A(dofs_per_block); > BlockVector<double> B(dofs_per_block); > > unsigned int lin_it = 0; > double lin_res = 0.0; > { > timer.enter_subsection("Linear solver"); > std::cout << " SLV " << std::flush; > if (parameters.type_lin == "CG") > { > const int solver_its = static_cast<unsigned int>( > tangent_matrix.block(u_dof, u_dof).m() > * parameters.max_iterations_lin); > const double tol_sol = parameters.tol_lin > * system_rhs.block(u_dof).l2_norm(); > > SolverControl solver_control(solver_its, tol_sol); > > GrowingVectorMemory<Vector<double> > GVM; > SolverCG<Vector<double> > solver_CG(solver_control, GVM); > > PreconditionSelector<SparseMatrix<double>, Vector<double> > > preconditioner (parameters.preconditioner_type, > parameters.preconditioner_relaxation); > preconditioner.use_matrix(tangent_matrix.block(u_dof, u_dof)); > > solver_CG.solve(tangent_matrix.block(u_dof, u_dof), > newton_update.block(u_dof), > system_rhs.block(u_dof), > preconditioner); > > lin_it = solver_control.last_step(); > lin_res = solver_control.last_value(); > } > else if (parameters.type_lin == "Direct") > { > > SparseDirectUMFPACK A_direct; > A_direct.initialize(tangent_matrix.block(u_dof, u_dof)); > A_direct.vmult(newton_update.block(u_dof), > system_rhs.block(u_dof)); > > lin_it = 1; > lin_res = 0.0; > } > else > Assert (false, ExcMessage("Linear solver type not implemented")); > > std::vector<std::complex<double>> eigenvalues; > std::vector<Vector<double> > eigenvectors; > SparseDirectUMFPACK inverse; > inverse.initialize (tangent_matrix.block(u_dof,u_dof)); > SolverControl solver_control(1000, 1e-9); > > ArpackSolver eigensolver(solver_control, > ArpackSolver::AdditionalData(ArpackSolver::WhichEigenvalues::algebraically_smallest)); > eigensolver.solve(tangent_matrix.block(u_dof,u_dof), > IdentityMatrix(), inverse, eigenvalues, eigenvectors,eigenvalues.size()); > > timer.leave_subsection(); > } > --------------------------------------------- > On Friday, September 25, 2020 at 12:15:29 AM UTC+5:30 Animesh Rastogi IIT > Gandhinagar wrote: > >> I could finally configure dealii using ARPACK. I am facing issues in >> initialising the identity matrix that is to be given to the >> eigensolver.solve(). I am trying to replace the mass matrix with an >> identity matrix.. >> >> I am copying the code again in this thread.. >> >> std::vector<std::complex<double>> eigenvalues; >> std::vector<Vector<double> > eigenvectors; >> SparseMatrix<double> identity (IdentityMatrix(u_dof)); >> SparseDirectUMFPACK inverse; >> inverse.initialize (tangent_matrix.block(u_dof,u_dof)); >> const int eigensolver_its = static_cast<unsigned int>( >> tangent_matrix.block(u_dof, u_dof).m() >> * parameters.max_iterations_lin); >> SolverControl solver_control(eigensolver_its, 1e-9); >> ArpackSolver::ArpackSolver eigensolver(solver_control); >> eigensolver.solve(tangent_matrix.block(u_dof,u_dof), identity, >> inverse, eigenvalues, eigenvectors); >> >> I am getting the following error for this - >> >> error: request for member ‘vmult’ in ‘mass_matrix’, which is of non-class >> type ‘dealii::BlockSparseMatrix<double>(dealii::IdentityMatrix)’ >> mass_matrix.vmult(tmp, src); >> >> Could someone please help me with this? >> >> Thanks! >> >> Animesh >> On Thursday, September 24, 2020 at 4:40:47 PM UTC+5:30 Animesh Rastogi >> IIT Gandhinagar wrote: >> >>> Hi Jean, >>> >>> It turns out that I did not configure dealii with ARPACK. So I followed >>> the readme instructions and installed and compiled ARPACK as suggested. I >>> also tried to run the examples in the ARPACK directory and they are running >>> without any errors. However, when I tried to reconfigure deallii using the >>> selfcompiled version, it says the following error. >>> >>> Could not find the arpack library! >>> >>> Please ensure that a suitable arpack library is installed on your >>> computer. >>> >>> If the library is not at a default location, either provide some >>> hints for >>> autodetection, >>> >>> $ ARPACK_DIR="..." cmake <...> >>> $ cmake -DARPACK_DIR="..." <...> >>> >>> or set the relevant variables by hand in ccmake. >>> >>> I used the following cmake command to configure dealii again - >>> >>> cmake -DDEAL_II_WITH_PETSC=ON >>> -DPETSC_DIR=/home/animesh/Documents/petsc-3.13.5 >>> -DPETSC_ARCH=arch-linux-c-debug -DDEAL_II_WITH_METIS=ON >>> -DDEAL_II_WITH_MPI=ON -DDEAL_II_WITH_ARPACK=ON >>> -DARPACK_DIR=/home/animesh/Documents/ARPACK .. >>> >>> I checked that the path of the directory of ARPACK that I am giving here >>> is correct. >>> >>> I followed the instructions in this page - >>> https://www.dealii.org/current/external-libs/arpack.html. >>> Also, I couldn't figure out what it means when it says "For compilation >>> of ARPACK we emphasize adding the compiler flag -fPIC". What is -fPIC >>> and where should I use this flag? >>> >>> Could you please help me with this and the question about identity >>> matrix asked in the previous mail of this thread? >>> >>> Thanks a lot! >>> >>> Animesh >>> On Thursday, September 24, 2020 at 11:56:24 AM UTC+5:30 Animesh Rastogi >>> IIT Gandhinagar wrote: >>> >>>> Hi Jean, >>>> >>>> Thanks a lot for your response. I am trying as you suggested. I have >>>> added the following code inside the linear solver so that I can get the >>>> eigenvalues at every newton step. >>>> >>>> std::vector<std::complex<double>> eigenvalues; >>>> std::vector<Vector<double> > eigenvectors; >>>> SparseMatrix<double> identity (IdentityMatrix(u_dof)); >>>> SparseDirectUMFPACK inverse; >>>> inverse.initialize (tangent_matrix.block(u_dof,u_dof)); >>>> const int eigensolver_its = static_cast<unsigned int>( >>>> tangent_matrix.block(u_dof, u_dof).m() >>>> * parameters.max_iterations_lin); >>>> SolverControl solver_control(eigensolver_its, 1e-9); >>>> ArpackSolver::ArpackSolver eigensolver(solver_control); >>>> eigensolver.solve(tangent_matrix.block(u_dof,u_dof), identity, >>>> inverse, eigenvalues, eigenvectors); >>>> >>>> However, I am getting the follwing error while compiling. >>>> >>>> error: ‘ArpackSolver’ has not been declared >>>> >>>> I have included the header file #include <deal.II/lac/arpack_solver.h>. >>>> >>>> >>>> Also, could you please let me know if I have declared the identity >>>> matrix correctly? I am replacing the mass matrix B with the Identity >>>> matrix >>>> to compute the eigenvalues and eigenvectors. >>>> >>>> Thanks! >>>> >>>> Animesh >>>> On Thursday, September 24, 2020 at 1:10:35 AM UTC+5:30 Jean-Paul >>>> Pelteret wrote: >>>> >>>>> Hi Animesh, >>>>> >>>>> Although in that code-gallery example the system is assembled into a >>>>> BlockSparseMatrix<double>, the system actually has only one block. So you >>>>> can retrieve the underlying SparseMatrix<double> (and Vector<double> for >>>>> the RHS) via >>>>> tangent_matrix.block(u_dof, u_dof); >>>>> system_rhs.block(u_dof); >>>>> and use them with one of the standard eigensolvers (maybe ArpackSolver >>>>> <https://dealii.org/developer/doxygen/deal.II/classArpackSolver.html>, >>>>> since you want both the eigenvalues and eigenvectors). >>>>> >>>>> I hope that this helps you! >>>>> >>>>> Best, >>>>> Jean-Paul >>>>> >>>>> On 23 Sep 2020, at 20:53, Animesh Rastogi IIT Gandhinagar < >>>>> animesh...@alumni.iitgn.ac.in> wrote: >>>>> >>>>> Hi All, >>>>> >>>>> I am trying to play with the code of Quassi Static Finite Strain >>>>> Compressibility >>>>> <https://dealii.org/developer/doxygen/deal.II/code_gallery_Quasi_static_Finite_strain_Compressible_Elasticity.html>. >>>>> >>>>> I want to calculate the eigenvalues and eigenvectors of the System >>>>> Tangent >>>>> Matrix (BlockSparseMatrix<double> >>>>> <https://dealii.org/developer/doxygen/deal.II/classBlockSparseMatrix.html> >>>>> tangent_matrix) that we get at every time step after the Newton >>>>> method has converged. However, I could not find any function to calculate >>>>> the eigenvalues and eigenvectors of the BlockSparse Matrix. Could someone >>>>> please help me with this? >>>>> >>>>> Thanks! >>>>> >>>>> AR >>>>> >>>>> -- >>>>> The deal.II project is located at http://www.dealii.org/ >>>>> For mailing list/forum options, see >>>>> https://groups.google.com/d/forum/dealii?hl=en >>>>> --- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "deal.II User Group" group. >>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>> an email to dealii+un...@googlegroups.com. >>>>> To view this discussion on the web visit >>>>> https://groups.google.com/d/msgid/dealii/987eb63a-333e-43b3-a9e0-fbcbae2d567en%40googlegroups.com >>>>> >>>>> <https://groups.google.com/d/msgid/dealii/987eb63a-333e-43b3-a9e0-fbcbae2d567en%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>> . >>>>> >>>>> >>>>> > -- > The deal.II project is located at http://www.dealii.org/ > For mailing list/forum options, see > https://groups.google.com/d/forum/dealii?hl=en > --- > You received this message because you are subscribed to the Google Groups > "deal.II User Group" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to dealii+un...@googlegroups.com. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/dealii/d03b3524-83e6-473e-9d37-ff45f4bf6a6en%40googlegroups.com > > <https://groups.google.com/d/msgid/dealii/d03b3524-83e6-473e-9d37-ff45f4bf6a6en%40googlegroups.com?utm_medium=email&utm_source=footer> > . > > > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/542f4b09-47ba-482f-bf19-481ff32ff329n%40googlegroups.com.
