Gene Expression Programming to High-Cycle and Low-Cycle Fatigue Life Prediction of Polymer-Reinforced Concrete in the Vicinity of Room, Water, and Seawater Considering Different Immersion Times

Nowadays, the use of machine learning techniques to solve industrial problems is very popular, although the prediction accuracy of these techniques depends on various factors such as input data, neural network architecture, and its settings. But by combining several destructive phenomena at the same time, the complexity of using machine learning increases. The current research presents a gene expression programming (GEP)-based mathematical equations to predict both high-cycle fatigue (HCF) and low-cycle fatigue (LCF) behaviors of polymer-reinforced concrete (PC) exposed to various environmental degradation. To reach this goal, 54 cylindrical samples of PC were prepared in the same conditions. Also, two degradation environments with different pH values were used representing water and seawater, and the samples were immersed in them for zero, one, and 6 months. Next, axial fatigue test was performed at three different load levels. Furthermore, pH value, immersion time, and maximum cyclic force were considered as input parameters to the GEP. Also, the average number of cycles to failure corresponding to each load level was considered as output. At first, according to the range of one thousand cycles, the separation of LCF and HCF regions was done. Furthermore, a unique formula was presented for each fatigue zone. Finally, the model accuracy was assessed for all cases. © 2025 Elsevier B.V., All rights reserved.