Characterizing Errors and Evaluating Performance of Transient Simulations Using Multi-Time-Step Integration

Abstract

AbstractThe multi-time-step method of time integration for problems in structural dynamics allows one to decompose the problem domain into small subdomains and use different time steps within each subdomain to reduce the computational cost of solving such problems. However, the number of possible decompositions and their associated time steps for a given model is huge and grows exponentially with the number of elements. To find an optimal decomposition that minimizes error in the solution while maintaining a bound on the computational cost is challenging. In this work, existing multi-time-step methods are used and, for the first time, a systematic approach for traversing the space of possible decompositions to characterize the nature of how solution errors and computational costs vary for different decompositions is devised. Through numerical examples for three different types of structures, trusses, frames, and continuum solid bodies, it is shown that the characteristics of these error and cost functions...