Crashworthiness Optimization in LS-OPT: Case Studies in Metamodeling and Random Search Techniques

This crashworthiness optimization study compares the use of three metamodeling techniques while using a sequential random search method as a control procedure. The three methods applied are (i) the original Successive Linear Response Surface Method, (ii) the Neural Network method and (iii) the Kriging method. It is shown that, although NN and Kriging seem to require a larger number of initial points, the three metamodeling methods have comparable efficiency. The random search method is surprisingly efficient in some instances, but by nature much less predictable.