A multi-objective optimization approach for lightweight design of automotive rear seats based on hybrid surrogate model

In the multi-objective optimization design of automotive seats, traditional single surrogate models often fail to achieve the desired fitting accuracy when handling complex nonlinear responses. Moreover, complex engineering multi-objective optimization problems employing high-fidelity simulation models frequently face challenges associated with high computational costs. To overcome these challenges, this study proposes an integrated optimization framework comprising optimal Latin hypercube experimental design (OLHD), a Pointer algorithm-based RF–RBFNN hybrid surrogate modeling approach, the Multi-objective gray wolf optimization algorithm (MOGWO), and the MPSI-based decision-making method. Furthermore, for the rear seat of a passenger car, the headrest displacement ( Dis 1 ), seatback displacement ( Dis 2 ), Cost , and Mass were selected as the primary performance indicators. The proposed optimization framework was further validated through the multi-objective lightweight optimization problem of the rear seat, demonstrating its feasibility and effectiveness in practical engineering applications. The optimization results indicate that, compared with the original design, the seatback mass was significantly reduced by 14.71%, successfully achieving the lightweighting objective. In terms of performance, the headrest displacement ( Dis 1 ) decreased by 3.34%, while the seatback displacement ( Dis 2 ) showed only a slight increase of 0.22%. Meanwhile, the cost dropped from 44.83 (¥) to 33.428 (¥), representing a reduction of 25.43%. Simulation results confirmed that all prediction errors were within acceptable limits, and all performance indicators met the predefined development targets of the rear seat. These findings demonstrate that the proposed approach effectively balances lightweight design, cost, and safety performance, meeting the multi-objective optimization requirements of passenger car seat frames while highlighting the robustness and engineering practicality of the proposed method.