An optimization framework to tune a lattice model for improving crashworthiness using surrogates

The goal of this work is to develop a framework for optimization problems involving complex expensive functions for which traditional optimization techniques are not practical. Important concerns in such problems are high simulation cost and unreliable or unavailable derivative information. It's a common practice to work on a surrogate model (also called metamodel) of the original problem. We have used a 'kriging metamodel' to build the surrogate of the original problem and pattern search methods to solve the surrogate optimization problem. In turn the surrogate model is used to predict the optimum of original problem. We investigate the use of a lattice structure as a possible surrogate model for a more complex structure. We are concerned with design for crashworthiness of structures and an attempt has been made to 'tune' the lattice such that the response of lattice can be used to predict the response of the structure. The advantage of using such lattice model is low computational cost and less modeling time which, can help reduce the design cycle time and improved accuracy.