Object-oriented concurrent solution algorithms for nonlinear structural dynamics
Abstract
This research addresses the time history analysis of structures subjected to dynamic loads using high performance computing environments. In this work, structural mechanics, parallel computing, and object-oriented programming methodologies are integrated. Sequential and parallel solution algorithms for nonlinear structural dynamics are investigated and implemented. Specific contributions are as follows: (1) A generalized numerical implicit solution algorithm for structural dynamics equations is developed which contains most of the existing solution algorithms as special cases. This algorithm is optimized for desirable criteria, such as second order accuracy, unconditional stability, overshoot control, and controllable dissipation for higher modes. (2) A new iteration strategy is developed for solving nonlinear structural dynamics problems. This iterative procedure can be more efficient than the full or modified Newton-Raphson iterative methods. Because of the explicit nature of its individual iterations, this algorithm is highly scalable for parallel implementation. (3) An adaptive time-step control strategy for structural dynamics algorithms is developed based on consideration of local truncation error, convergence, and stability. (4) A concurrent solution algorithm, termed as Iterative-Group-Implicit algorithm, has been developed for the analysis of structural dynamics problems. Finally, the efficiency and accuracy of the parallel solution algorithm for linear structural dynamics problems is demonstrated by subjecting a structural system to earthquake loading. An object-oriented approach is employed in the design and implementation of the aforementioned solution procedures in order to facilitate extensibility, reusability, maintainability and simplicity. A framework for parallel and sequential transient finite element analysis is designed and implemented. In particular, the Iterative-Group-Implicit algorithm is tested using this framework. The proposed research should help in the structural analysis of large structural systems in research and practice, resulting in more realistic and rational design of structures.
Degree
Ph.D.
Advisors
Sotelino, Purdue University.
Subject Area
Civil engineering|Mechanics|Mechanical engineering|Systems design|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.