Artificial intelligence‐based prediction of cooling pipe system parameters for thermal field control and crack prevention in mass concrete structures

Abstract The pile cap of a bridge pier is susceptible to thermal cracking caused by heat generated during cement hydration at early ages. Although numerical and machine‐learning methods have been used to analyze the thermal behavior of mass concrete with cooling pipe systems (CPSs), an integrated approach for rapid prediction and inverse optimization of CPS parameters, together with explicit equations for engineering applications, remains limited. This study develops an integrated artificial neural network–genetic algorithm (ANN–GA) framework for rapid thermal prediction and optimization of CPS parameters for mass concrete pile caps under construction and climatic conditions representative of Vietnam. A database of 675 numerical cases was generated using MIDAS Civil 2022 for ANN training, validation, and testing. The proposed framework achieved a prediction accuracy of 98.7% relative to the numerical simulations. Quadratic regression models derived from the optimized ANN–GA results yielded R 2 values of 0.938 and 0.908, with RMSE values of 1.94 and 1.52°C for the maximum temperature and maximum temperature difference, respectively. All optimized solutions satisfied the prescribed thermal‐cracking control criteria. The results demonstrate that the proposed ANN–GA framework provides a rapid and practical approach for thermal prediction and CPS optimization in mass concrete pile caps.