Machine learning prediction of structural response of steel fiber-reinforced concrete beams subjected to far-field blast loading

Abstract

Owing to its superior mechanical properties, enhanced energy absorption, and improved blast resistance, steel fiber-reinforced cementitious composites are a prime choice for the development of resilient structures. Sizable advancements have recently been made in assessing the blast performance of steel fiber-reinforced concrete (SFRC), high-strength steel fiber-reinforced concrete (HSFRC), and ultra-high-performance steel fiber-reinforced concrete (UHPFRC). However, there is still need for concerted research to explore the effects of fiber parameters under blast loading and to develop simple and robust predictive models. Thus, the present study develops a machine learning model to predict the maximum displacement of SFRC, HSFRC, and UHPFRC beams subjected to far field blast loading. Using Gaussian Process (GP) regression and Conditional Tabular Generative Adversarial Networks, a novel model was developed considering 117 experimental and 300 synthetic datasets. The proposed model was appraised through several statistical performance metrics and compared to existing analytical and numerical methods. The model imparts noteworthy reduction in modeling complexity in terms of static and dynamic material properties. A parametric analysis was conducted using the developed model to investigate the effects of fiber properties for diverse beam configurations. The proposed simplified model achieved highly favorable predictive performance with MAE of 2.75, R2 of 90.32%, and MAPE of 13.8%.