A surrogate-driven optimization framework for thermal parameter inversion in hydraulic concrete structures

Abstract

Thermal parameters are key inputs for numerical thermal analysis of hydraulic concrete structures. Under long-term service conditions, however, these parameters may progressively deviate from their design values because of environmental exposure, material deterioration, and complex operational loads, thereby reducing the accuracy of temperature field simulation and structural safety assessment. Therefore, efficient and accurate inversion of thermal parameters is essential for reliable numerical evaluation of in-service hydraulic concrete structures. To address the low computational efficiency and limited inversion accuracy of conventional methods for transient heat transfer problems, this study develops an improved surrogate model-based inversion and optimization method. In the proposed framework, multi-output least squares support vector regression (MLSSVR) is used to construct a surrogate model that describes the nonlinear mapping between thermal parameters and spatiotemporal temperature responses. An improved grey wolf optimizer (IGWO), incorporating a nonlinear convergence factor and a self-adaptive dynamic weighting strategy, is then coupled with the surrogate model to enhance global search capability and reduce the risk of premature convergence. Engineering applications are conducted to evaluate the performance of the proposed method. The results show that the proposed method achieves higher inversion accuracy while reducing computational cost by more than 80% compared with conventional approaches. These findings demonstrate that the proposed surrogate model-based framework provides an efficient and practical tool for thermal parameter inversion and can support more accurate numerical analysis and safety evaluation of hydraulic concrete structures.