Different Techniques of Model Simplifications Applied for Crashworthiness Analysis - Neural Network Approach

This paper on "Different Techniques of Model Simplifications Applied for Crashworthiness Analysis - Neural Network Approach" was presented at the NAFEMS World Congress on Effective Engineering Analysis - 25-28 April 1999, Newport, Rhode Island, USA.

Abstract

In the aim to automate the production of accurate and minimal complexity models destined for crashworthiness analysis of automobiles, model reduction algorithm based on neural network have been coupled with finite element simulation system.
Preliminary numerical simulations of the crash of the sophisticated (non-simplified) model provide a data to be used as an input sequence for artificial neural network. The ability of neural network to predict the response of nonlinear system is invoked to simulate desired parts of the whole model to be simplified. The expense of computing such reduced model is highly reduced while its crucial properties are preserved. The simplified model can be used in the subsequent iterations of the design cycle. Thus a designer can dramatically cut down the computation time associated with the whole design cycle and is able to perform more iterations within given quantity of time.