Abstract:In the current engineering and scientific research, finite element models for large-scale structural optimization face limitations due to high computational costs and complexity. The integration of response surface models has emerged as an effective approach to overcome these challenges, enabling researchers to significantly reduce computational costs while maintaining acceptable accuracy. Howev-er, when fitting response surfaces for complex models, conventional parameter screening methods often lead to reduced accuracy and efficiency, particularly when considering individual variations and the high costs of sensitivity analysis. Focusing on the finite element model of a 26-story frame-shear wall structure, this study integrated two preprocessing steps—single-factor experiments and hill-climbing tests—during response surface construction. These steps aimed to narrow the search space, screen key factors, and provide gradient information,making the construction of the response surface more accurate and operable, and providing a reliable foundation for subsequent model processing. By integrating multiple intelligent algorithms, this study completed the model updating and optimization operations for the response surface. The research results showed that the response surface constructed using parameters screened through preprocessing steps maintained consistently low error rates with identification results when multiple algorithm types were applied. This study provides valuable guidance for future engineering practices and research on related fields, offering a more flexible and universal optimization solution for enhancing the accuracy and efficiency of finite element model updating in largescale structural optimization.
