A Massively Parallel Sparse Eigensolver for Structural Dynamics Finite Element Analysis
- Sandia National Laboratories
Eigenanalysis is a critical component of structural dynamics which is essential for determinating the vibrational response of systems. This effort addresses the development of numerical algorithms associated with scalable eigensolver techniques suitable for use on massively parallel, distributed memory computers that are capable of solving large scale structural dynamics problems. An iterative Lanczos method was determined to be the best choice for the application. Scalability of the eigenproblem depends on scalability of the underlying linear solver. A multi-level solver (FETI) was selected as most promising for this component. Issues relating to heterogeneous materials, mechanisms and multipoint constraints have been examined, and the linear solver algorithm has been developed to incorporate features that result in a scalable, robust algorithm for practical structural dynamics applications. The resulting tools have been demonstrated on large problems representative of a weapon's system.
- Research Organization:
- Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 7019
- Report Number(s):
- SAND99-1158; ON: DE00007019
- Country of Publication:
- United States
- Language:
- English
Similar Records
Experiences with FETI-DP in a Production Level Finite Element Application
A comparison of eigensolvers for large-scale 3D modal analysis using AMG-preconditioned iterative methods.