MULTIDISCIPLINARY DESIGN OPTIMIZATION OF AUTOMOTIVE CRASHWORTHINESS AND NVH USING LS-OPT

This paper describes the multidisciplinary design optimization of a full vehicle to minimize mass while complying with crashworthiness and Noise, Vibration and Harshness (NVH) constraints. A full frontal impact is used for the crashworthiness simulation in the nonlinear dynamics code, LS-DYNA. The NVH constraints are evaluated from an implicit modal analysis of a body-in- white vehicle model using LS-DYNA. Seven design variables describe the structural components of which the thickness can be varied. The crashworthiness constraints relate to crush energy and displacement, while the torsional frequency characteristics are obtained from the modal analysis. The Multidisciplinary Feasible (Fully Integrated) formulation, in which full sharing of the variable sets is employed, is used as the reference case. In an attempt to investigate global optimality, three starting designs are used. Based on a Design of Experiments analysis of variance of the fully-shared variable results for each starting design, discipline-specific variables are selected from the full set using the sensitivity of the disciplinary responses. The optimizer used in all cases is the Successive Response Surface Method as implemented in LS-OPT. It is shown that partial sharing of the variables not only reduces the computational cost in finding an optimum due to fewer, more sensitive variables, but also leads to a better result. The mass of the vehicle is reduced by 4.7% when starting from an existing baseline design, and by 2.5% and 1.1% when starting from a lightest and heaviest starting design respectively.