Moving Least Square Method for Reliability-Based Design Optimization

1. Abstract A new reliability-based design optimization (RBDO) method is presented, using a moving lease square (MLS) method and performance measure approach (PMA). This method will provide so-called 6-sigma optimum design. Based on the least square method, the MLS method is introduced to better approximate implicit responses by imposing a variable weight over a compact support. In addition, design sensitivity data is incorporated in the MLS method to demonstrate substantial improvement in accuracy of the approximate sensitivity for RBDO. By taking advantage of the inverse PMA problem, a design of experiment (DOE) framework, that is suitable for reliability analysis, is proposed to properly reproduce the main effects and interactions of design parameters by combining the axial star (AS) and selective interaction (SI) sampling in a single DOE building block. PMA is shown to be much more effective than the reliability index approach (RIA) when response surface methodology (RSM) is used for RBDO. The proposed RSM is integrated with the hybrid mean value (HMV) method of PMA to develop a robust and efficient RBDO process. A large-scale vehicle side impact model example problem is employed to demonstrate the new RBDO method.