Research on determining the piles bearing capacity using a random forest model considering the randomness of the soil data

Abstract

Bearing capacity is one of the most important parameters when designing piles. However, determining the exact bearing capacity of piles is a difficult job due to the influence of many parameters. The traditional methods of calculating the axial load capacity of piles all use a predefined problem, that is, determining only a single load capacity value, which is not entirely consistent with the actual working of the piles, where the input parameters affecting the bearing capacity of the piles are random. In this study, an advanced machine learning model based on artificial intelligence, the Random Forest, was developed and applied to predict the bearing capacity of piles. This model is used as a predefined model applied in the Monte-Carlo simulation method to determine the reliability of the pile-bearing capacity. The results show that the Random Forest model very well predicts the bearing capacity of piles on both training and testing data. In addition, the Monte-Carlo simulation results with random soil data show that there is still the possibility of unsafe pile operation even when the pile top load is lower than the expected average bearing capacity of the pile. Furthermore, the maximum load to the top of the pile should not exceed 99.2% of the mean load value, to achieve a high probability of safe working, on this data set.