Abstract:
This paper introduces and evaluates the Successive Response Surface Method (SRSM), a robust and adaptive domain reduction scheme designed for simulation-based optimization, especially in the context of nonlinear structural dynamics and crashworthiness. Using linear response surfaces constructed through D-optimal experimental designs, SRSM iteratively reduces the design space based on oscillation and move distance criteria, enabling effective navigation through noisy and non-smooth simulation responses. A key feature is its compatibility with parametric modeling, allowing for flexible and efficient shape optimization with minimal user input. The method is benchmarked against traditional optimization approaches using analytical test problems and real-world engineering applications, including shape optimization of a vehicle front structure, material parameter identification, and a head impact scenario. Results demonstrate SRSM’s robustness in handling noise and design variability, making it especially suitable for shape optimization tasks that involve complex parametric models.
