Artificial Intelligence-Based Surrogate Model for the Analysis and Prediction of Blast-Induced Behavior of RC Wall Structures in Small Modular Reactors

Small modular reactor (SMR) structures can be constructed near power-demand sites, thereby allowing for flexible adaptation to changes in power demand and relatively short construction periods. However, owing to limited construction sites, high levels of performance are required in terms of facility protection and security. To analyze the behavior of reinforced-concrete (RC) structures under blast loads and effects of design parameters, dynamic and nonlinear analyses are required, which can be time-consuming. Accordingly, studies have been conducted utilizing artificial neural network-based structural response prediction models. In this study, a multi-layer perception (MLP)-based surrogate model was developed to effectively analyze the behavior of RC wall structures in terms of influencing parameters in SMRs under blast loads. To prevent overfitting of the output data on a limited dataset, K-fold cross-validation was applied, resulting in satisfactory learning rates and output data accuracy.