Abstract
Many applications in structural mechanics require the numerical solution of sequences of linear systems typically issued from a finite element discretization of the governing equations on fine meshes. The method of Lagrange multipliers is often used to take into account mechanical constraints. The resulting matrices then exhibit a saddle point structure and the iterative solution of such preconditioned linear systems is considered as challenging. A popular strategy is then to combine preconditioning and deflation to yield an efficient method. We propose an alternative that is applicable to the general case and not only to matrices with a saddle point structure. In this approach, we consider to update an existing algebraic or application-based preconditioner, using specific available information exploiting the knowledge of an approximate invariant subspace or of matrix-vector products. The resulting preconditioner has the form of a limited memory quasi-Newton matrix and requires a small number of linearly independent vectors. Numerical experiments performed on three large-scale applications in elasticity highlight the relevance of the new approach. We show that the proposed method outperforms the deflation method when considering sequences of linear systems with varying matrices.
Similar content being viewed by others
Notes
This is to be in agreement with all preconditioning strategies implemented in the host finite element software \(Code\_Aster\).
\({{P}}_\mathcal{{V}, \mathcal {W}}\) is the unique projection operator with range \(\mathcal {R}({{P}}_\mathcal{{V}, \mathcal {W}}) = \mathcal {V}\) and null space \(\mathcal {N}({{P}}_\mathcal{{V}, \mathcal {W}}) = \mathcal {W}\) [39].
We emphasize that this situation did not occur in all our numerical experiments.
The CPU time for the Ritz-LMP strategy does include the cost of forming X, Y as required in (20).
This estimate has been obtained as a by-product of the application of MUMPS (here in double-precision arithmetic) to the first matrix \({\mathcal {K}}_1\) in the sequence.
References
Amestoy PR, Duff IS, L’Excellent J-Y (2000) Multifrontal parallel distributed symmetric and unsymmetric solvers. Comput Methods Appl Mech Eng 184:501–520
Amestoy PR, Duff IS, L’Excellent J-Y, Koster J (2001) A fully asynchronous multifrontal solver using distributed dynamic scheduling. SIAM J Matrix Anal Appl 23(1):15–41
Amestoy PR, Guermouche A, L’Excellent JY, Pralet S (2006) Hybrid scheduling for the parallel solution of linear systems. Parallel Comput 32(2):136–156
Baglama J, Calvetti D, Golub G, Reichel L (1998) Adaptively preconditioned GMRES algorithms. SIAM J Sci Comput 20:243–269
Bai Z, Demmel J, Dongarra J, Ruhe A, van der Vorst HA (2000) Templates for the solution of algebraic eigenvalue problems: a practical guide, vol 11. SIAM, Philadelphia
Bargellini R (2011) Elasto-visco-plastic modeling with isotropic hardening in large deformations. Technical Report, EDF R&D, AMA R5.03.21
Bellavia S, Bertaccini D, Morini B (2011) Nonsymmetric preconditioner updates in Newton–Krylov methods for nonlinear systems. SIAM J Sci Comput 33(5):2595–2619
Bellavia S, Morini B, Porcelli M (2014) New updates of incomplete LU factorizations and applications to large nonlinear systems. Optim Methods Softw 29:321–340
Benzi M (2002) Preconditioning techniques for large linear systems: a survey. J Comput Phys 182(2):418–477
Benzi M, Golub GH, Liesen J (2005) Numerical solution of saddle point problems. Acta Numer 14(1):1–137
Bonet J, Wood RD (1997) Nonlinear continuum mechanics for finite element analysis. Cambridge University Press, Cambridge
Broyden CG (1969) A new method of solving nonlinear simultaneous equations. Comput J 12(1):94–99
Carpentieri B, Duff IS, Giraud L (2003) A class of spectral two-level preconditioners. SIAM J Sci Comput 25(2):749–765
Chapman A, Saad Y (1997) Deflated and augmented Krylov subspace techniques. Numer Linear Algebra Appl 4(1):43–66
Dostál Z (1988) Conjugate gradient method with preconditioning by projector. Int J Comput Math 23(3):315–323
Duintjer Tebbens J, Tuma M (2007) Efficient preconditioning of sequences of nonsymmetric linear systems. SIAM J Sci Comput 29(5):1918–1941
Duintjer Tebbens J, Tuma M (2010) Preconditioner updates for solving sequences of linear systems in matrix-free environment. Numer Linear Algebra Appl 17:997–1019
Eirola T, Nevanlinna O (1989) Accelerating with rank-one updates. Linear Algebra Appl 121:511–520
Elman HC, Silvester DJ, Wathen AJ (2014) Finite elements and fast iterative solvers: with applications in incompressible fluid dynamics, 2nd edn. Oxford University Press, Oxford
Erlangga Y, Nabben R (2008) Deflation and balancing preconditioners for Krylov subspace methods applied to nonsymmetric matrices. SIAM J Matrix Anal Appl 30(2):684–699
Gaul A (2014) Recycling Krylov subspace methods for sequences of linear systems : analysis and applications. Ph.D. thesis, TU Berlin, Germany
Gaul A, Gutknecht MH, Liesen J, Nabben R (2013) A framework for deflated and augmented Krylov subspace methods. SIAM J Matrix Anal Appl 34(2):495–518
Gaul A, Schlömer N (2015) Preconditioned recycling Krylov subspace methods for self-adjoint problems. Electron Trans Numer Anal 44:522–547
Giraud L, Gratton S, Martin E (2007) Incremental spectral preconditioners for sequences of linear systems. Appl Numer Math 57:1164–1180
Golub GH, Greif C (2003) On solving block-structured indefinite linear systems. SIAM J Sci Comput 24(6):2076–2092
Gosselet P, Rey C, Pebrel J (2013) Total and selective reuse of Krylov subspaces for the resolution of sequences of nonlinear structural problems. Int J Numer Methods Eng 94(1):60–83
Gratton S, Mercier S, Tardieu N, Vasseur X (2016) Limited memory preconditioners for symmetric indefinite problems with application to structural mechanics. Numer Linear Algebra Appl 23(5):865–887
Gratton S, Sartenaer A, Tshimanga J (2011) On a class of limited memory preconditioners for large scale linear systems with multiple right-hand sides. SIAM J Opt 21(3):912–935
Gutknecht MH (2012) Spectral deflation in Krylov solvers: a theory of coordinate space based methods. Electron Trans Numer Anal 39:156–185
Gutknecht MH (2014) Deflated and augmented Krylov subspace methods: a framework for deflated BiCG and related solvers. SIAM J Matrix Anal Appl 35:1444–1466
Ipsen ICF (2001) A note on preconditioning nonsymmetric matrices. SIAM J Sci Comput 23(3):10501–1051
Jönsthövel TB, Van Gijzen MB, MacLachlan S, Vuik C, Scarpas A (2012) Comparison of the deflated conjugate gradient method and algebraic multigrid for composite materials. Comput Mech 50:321–333
Jönsthövel TB, Van Gijzen MB, Vuik C, Scarpas A (2013) On the use of rigid body modes in the deflated preconditioned conjugate gradient method. SIAM J Sci Comput 35:B207–B225
Kelley CT (2003) Solving nonlinear equations with Newton’s method, vol 1. SIAM, Philadelphia
Kharchenko S, Yeremin A (1995) Eigenvalue translation based preconditioners for the GMRES(k) method. Numer Linear Algebra Appl 2:51–77
Kilmer M, de Sturler E (2006) Recycling subspace information for diffuse optical tomography. SIAM J Sci Comput 27(6):2140–2166
Liesen J, Strakoš Z (2013) Krylov subspace methods principles and analysis. Oxford University Press, Oxford
Mercier S (2015) Fast nonlinear solvers in solid mechanics. Ph.D. thesis, University of Toulouse, Toulouse, France, http://thesesups.ups-tlse.fr/3000/1/2015TOU30305.pdf
Meyer CD (2000) Matrix analysis and applied linear algebra. SIAM, Philadelphia
Michel-Ponnelle S (2013) Modeling of the pretension cables. Technical report, EDF R&D, AMA R7.01.02
Morales JL, Nocedal J (2000) Automatic preconditioning by limited memory quasi-Newton updating. SIAM J Opt 10(4):1079–1096
Nash S (1984) Newton-type minimization via the Lanczos method. SIAM J Numer Anal 21:770–778
Nečas J, Hlavácek I (1980) Mathematical theory of elastic and elastoplastic bodies. Elsevier, Amsterdam
Nicolaides RA (1987) Deflation of conjugate gradients with applications to boundary value problems. SIAM J Numer Anal 24(2):355–365
Nocedal J, Wright SJ (1999) Numerical optimization. Springer series in operations research and financial engineering. Springer, Berlin
Notay Y (2014) A new analysis of block preconditioners for saddle point problems. SIAM J Matrix Anal Appl 35(1):143–173
O’Leary DP, Yeremin A (1994) The linear algebra of block quasi-Newton algorithms. Linear Algebra Appl 212(213):153–168
Olshanskii MA, Tyrtyshnikov EE (2014) Iterative methods for linear systems: theory and applications. SIAM, Philadelphia
Paige CC, Parlett BN, van der Vorst HA (1995) Approximate solutions and eigenvalue bounds from Krylov subspaces. Numer Linear Algebra Appl 2:115–134
Parks M, de Sturler E, Mackey G, Johnson D, Maiti S (2006) Recycling Krylov subspaces for sequences of linear systems. SIAM J Sci Comput 28(5):1651–1674
Pestana J, Wathen AJ (2015) Natural preconditioning and iterative methods for saddle point systems. SIAM Rev 57(1):71–91
Proix JM (2013) Elasto-visco-plastic modeling of Chaboche-type. Technical Report, EDF R&D, AMA R5.03.04
Rey C, Risler F (1998) A Rayleigh–Ritz preconditioner for the iterative solution to large scale nonlinear problems. Numer Algorithms 17:279–311
Risler F, Rey C (1998) On the reuse of Ritz vectors for the solution to nonlinear elasticity problems by domain decomposition methods. Contemp Math 218:334–340
Saad Y (2003) Iterative methods for sparse linear systems. SIAM, Philadelphia
Saad Y, Schultz MH (1986) GMRES: a generalized minimal residual algorithm for solving nonsymmetric linear systems. SIAM J Sci Stat Comput 7(3):856–869
Saad Y, Yeung M, Erhel J, Guyomarc’h F (2000) A deflated version of the conjugate gradient algorithm. SIAM J Sci Comput 21(5):1909–1926
Schnabel RB (1983) Quasi-Newton methods using multiple secant equations. Technical report, Technical report CU-CS-247-83, Department of Computer Sciences, University of Colorado, Boulder, CO
Shewchuk J (2002) What is a good linear finite element ? Interpolation, conditioning, anisotropy, and quality measures, University of California at Berkeley, unpublished preprint
Simoncini V, Szyld DB (2007) Recent computational developments in Krylov subspace methods for linear systems. Numer Linear Algebra Appl 14(1):1–59
Soodhalter KM, Szyld DB, Xue F (2014) Krylov subspace recycling for sequences of shifted linear systems. Appl Numer Math 81:105–118
Tang JM, MacLachlan SP, Nabben R, Vuik C (2010) A comparison of two-level preconditioners based on multigrid and deflation. SIAM J Matrix Anal Appl 31(4):1715–1739
Tardieu N, Cheignon E (2012) A Newton–Krylov method for solid mechanics. Eur J Comput Mech 21(3–6):374–384
Toselli A, Widlund O (2005) Domain decomposition methods: algorithms and theory. Springer, Berlin
van der Vorst HA (2003) Iterative Krylov methods for large linear systems. Cambridge University Press, Cambridge
Vassilevski PS (2008) Multilevel block factorization preconditioners: matrix-based analysis and algorithms for solving finite element equations. Springer, New-York
Wang S, de Sturler E, Paulino G (2007) Large-scale topology optimization using preconditioned Krylov subspace methods with recycling. Int J Numer Methods Eng 69(12):2441–2468
Wathen AJ (2015) Preconditioning. Acta Numer 24:329–376
Yang UM (1995) A family of preconditioned iterative solvers for sparse linear systems. Ph.D. thesis, University of Illinois at Urbana-Champaign
Yang UM, Gallivan KA (1995) A new family of preconditioned iterative solvers for nonsymmetric linear systems. Appl Num Math 19(3):287–317
Acknowledgements
The authors thank the two anonymous reviewers for their comments. The authors are indebted to Jed Brown and Barry Smith (PETSc Team) and Thomas de Soza (EDF) for their comments and suggestions related to the implementation of the limited memory preconditioners into PETSc and \(Code\_Aster\). The authors would like to acknowledge ANRT (Association Nationale de la Recherche et de la Technologie) for the CIFRE Grant dotation (2012/0687) supporting Sylvain Mercier.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
We give in this section the two theorems cited in Sect. 4.1.
Theorem 1
Let \(A \in \mathbb {R}^{N \times N}\) be a nonsingular matrix and assume that \(S \in \mathbb {R}^{N \times k}\) is of full column rank k, with \(k \le N\) and denote \({\mathscr {S}}={\mathcal {R}}(S)\). The preconditioner H given in Definition 1 is nonsingular if and only if
Proof
Using relation (7), the preconditioner given in Definition 1 can be written as
Hence, the preconditioned operator HA simply reads
Since \(\mathcal{{R}}(P_{(A{\mathscr {S}})^{\perp }}A) = (A {\mathscr {S}})^{\perp }\), \(\mathcal{{N}}(P_{(A{\mathscr {S}})^{\perp }}A) = {\mathscr {S}}\), \(\mathcal{{N}}(I_N-Q) = (A^T A {\mathscr {S}})^{\perp }\) and \(\mathcal{{R}}(I_N-Q) = {\mathscr {S}}\), necessary and sufficient conditions for HA to be invertible are given by
We note that the condition \(\mathbb {R}^{N} = (A^T A {\mathscr {S}})^{\perp } \oplus {\mathscr {S}}\) holds since \(A^TA\) is symmetric positive definite. This completes the proof, since A is supposed to be nonsingular. \(\square \)
Theorem 2
Let \(A \in \mathbb {R}^{N \times N}\) be a nonsingular matrix and H be given by (15) in Definition 1. Assume that the columns of \(W \in \mathbb {R}^{N \times k}\) and \(W_{\perp } \in \mathbb {R}^{N \times (N-k)}\) form an orthonormal basis for \(A{\mathscr {S}}\) and \((A{\mathscr {S}})^{\perp }\), respectively. The spectrum of the preconditioned operator AH is then given by
Proof
Using relations (7) and (23) leads to
Since the columns of \([W, W_{\perp }]\) form an orthonormal basis of \(\mathbb {R}^{N}\), we have
\(\varLambda (AH)=\varLambda ([W,W_{\perp }]^TAH[W,W_{\perp }])\) and
Using relation (25) and basic properties of the orthogonal projection \(P_{(A{\mathscr {S}})^{\perp }}\), we finally obtain
which completes the proof. \(\square \)
Rights and permissions
About this article
Cite this article
Mercier, S., Gratton, S., Tardieu, N. et al. A new preconditioner update strategy for the solution of sequences of linear systems in structural mechanics: application to saddle point problems in elasticity. Comput Mech 60, 969–982 (2017). https://doi.org/10.1007/s00466-017-1450-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-017-1450-z