Improved Structural Impact Performance Using Sequential Multiquadric Regression with Output Space Mapping

Cheng-Ho Tho,Bo Ping Wang

DOI: https://doi.org/10.2514/6.2007-1899

2007-01-01

Abstract:Crashworthiness design optimization problems are continuously challenged by inherent complexity of long simulation time, numerical instability, and unavailability of sensitivity analysis. In recent years, surrogate-model-based design optimization using design of experiments and response surface methodology (DOE/RSM) has proven to be a promising way to achieve the improved design for crashworthiness problems. However, the method is still hampered by a large number of function evaluations for a large number of design variables or low accuracy for small sample size. This paper presents an effective simulation-based optimization algorithm for optimal design of large-scale, computationally expensive crashworthiness problems. The proposed optimization algorithm is called the sequential regularized multiquadric regression with output space mapping (SRMQ/OSM). The proposed SRMQ/OSM method is demonstrated on two real-world crashworthiness design optimization problems to enhance the structural impact performance with prescribed design constraints: 1. level landing of a helicopter skid landing gear, 2. frontal impact of a full pick-up truck. The results show that the proposed approach converged quickly to a feasible design with no constraint violation in both problems.