Advancements in Eigenvalue Technology
The standard eigenvalue problem for Implicit Mechanics in LS-DYNA is ΚΦ = ΜΦΛ , where Κ and Μ , the stiffness and mass matrices, are real and symmetric positive semi-definite in most applications. LS-DYNA offers three main algorithms for this problem, chosen with EIGMTH (field 7 of line 1 of *CONTROL_IMPLICIT_EIGENVALUE). The three algorithms currently available are • Lanczos (EIGMTH=2); • LOGPCG (EIGMTH=102); • Fast Lanczos (EIGMTH=103). This paper will give an overview of these methods, guidelines on when each should be used, performance comparisons, and recent enhancements. For non-symmetric problems, LS-DYNA also has an SMP solver, which relies on ARPACK [1,2]. An MPP solver relying on a different algorithm is being actively developed and is briefly mentioned in the last section.
https://www.dynalook.com/conferences/17th-international-ls-dyna-conference-2024/nvh-implicit/rouet_ansys.pdf/view
https://www.dynalook.com/@@site-logo/DYNAlook-Logo480x80.png
Advancements in Eigenvalue Technology
The standard eigenvalue problem for Implicit Mechanics in LS-DYNA is ΚΦ = ΜΦΛ , where Κ and Μ , the stiffness and mass matrices, are real and symmetric positive semi-definite in most applications. LS-DYNA offers three main algorithms for this problem, chosen with EIGMTH (field 7 of line 1 of *CONTROL_IMPLICIT_EIGENVALUE). The three algorithms currently available are • Lanczos (EIGMTH=2); • LOGPCG (EIGMTH=102); • Fast Lanczos (EIGMTH=103). This paper will give an overview of these methods, guidelines on when each should be used, performance comparisons, and recent enhancements. For non-symmetric problems, LS-DYNA also has an SMP solver, which relies on ARPACK [1,2]. An MPP solver relying on a different algorithm is being actively developed and is briefly mentioned in the last section.
Rouet_Ansys.pdf
— 467.6 KB