An efficient random forest-based subset simulation method for reliability analysis of the marine structure piles subject to scour

This study proposes a hybrid random forest-based subset simulation (RFSS) method for probabilistic assessment of the scour around pile groups under waves. In the RFSS, random forest (RF) is employed to replace the true limit state function, and it is updated based on the design samples in the first and last levels of subset simulation (SS) method. In this regard, 127 laboratory datasets collected from the literature were used to modeling. First, an existing equation for predicting the scour depth around piles was modified by using a metaheuristic approach. The performance of the modified equation was compared with four equations and two artificial intelligence (AI) models. The comparisons demonstrated that modified equation more accuracy than existing formulas and previous AI-based models. Then, a probabilistic model based on the RFSS was developed by considering the modified formula as the limit state function of scour depth. Solving two numerical, one hydraulic engineering, and scour of piles group problems, validate the robustness and accuracy of the developed structural reliability method. Results showed that the novel proposed RFSS is a robust and efficient method for solving high-dimensional real-world problems. Furthermore, compared to the Monte Carlo Simulation (MCS), the RFSS enable to estimate the reliability index with fewer computational cost and same accuracy. The function call number of RFSS was obtained 160 at the first example, 100 at the second example, 100 at the hydraulic example, and 150 at the scour of piles example.