A machine learning approach for predicting a full load-deflection behaviour of strengthened beams using fabric-reinforced cementitious matrix (FRCM)

Machine learning (ML) methods in structural engineering are typically applied to predict single values, such as failure load. This study presents an advanced application of ML by predicting five critical points on the load-deflection curve of fabric-reinforced cementitious matrix (FRCM)-strengthened rectangular beams. Three separate ML models were developed, each tailored to predict specific key points on the curve. The ML models consider variables such as mechanical and geometric properties of a beam, steel reinforcement, and the FRCM properties such as nominal thickness, number of layers, type of fabric and presence of anchorage. Based on these models, a user-friendly application was developed. In the second part of this study, validated Finite Element (FE) models examine the robustness of ML models on unseen data. The accuracy of the load-deflection curves is evaluated using three parameters: load capacity, stiffness, and absorbed energy. Results indicate that the proposed ML-based models effectively capture the entire response of FRCM-strengthened beams, achieving RMSE of 15 kN, 6 kN/mm, and 2kN-m for load capacity, stiffness, and absorbed energy, respectively. • Three ML models accurately predict the load-deflection behavior of FRCM-strengthened beams. • ML-based predictions are faster than FE methods with comparable accuracy. • Increasing FRCM layers improves strength until debonding dominates at higher layers. • Engineers can quickly assess strengthening designs without extensive testing or FEM. • Load capacity, stiffness, and absorbed energy predictions align well with experiments.