Abstract
Optimization of composite materials is essential considering their high costs and complex manufacturing processes. This paper includes optimization of composite material square tube for crashworthiness requirements using Artificial Neural Networks and Genetic Algorithms. Objectives of optimization were peak crushing force and maximizing specific energy absorbed by the composite square tube. Training data for the artificial neural network was obtained by crash analysis of a square tube against a rigid wall for various samples using LS Dyna. Latin Hypercube sampling technique was used to generate the required sampling plan. Neural Network Training was performed to obtain the highest regression value and minimum mean square error. The trained network was further combined with Multi-Objective Genetic Algorithms (MOGA) to find optimized values of the variables to fulfill the crashworthiness requirements. The fitness function used in MOGA was obtained by ANN with a 5% allowable error in the objective values. Optimized solutions obtained were presented by the Pareto frontier curve. The methodology was developed for square tubes with four and eight plies.
