Estimation of pile bearing capacity using hybrid models based on modified radial base function
摘要
Pile bearing capacity (Pu) is a vital aspect of geotechnical engineering and foundation design, representing the maximum load a single pile or a group of piles can support without excessive settlement or failure. Accurate determination of this capacity is essential for structural stability and involves a combination of analytical and empirical methods, considering factors like soil properties, pile materials, and construction techniques. A main emphasis of several research projects has been the prediction of Pu utilizing machine learning (ML) techniques. It is crucial to remember that certain ML techniques have inherent limits. These constraints, which are mostly related to the approach's sluggish convergence rates and difficulties in discovering global minima, should be noted. The goal of the present study was to improve a prediction model based on ML by including optimization methods. The development of models included the use of a radial basis function (RBF) in combination with two separate optimization strategies: Runge Kutta optimization (RKO) and Alibaba and the forty thieves (AFT). It provides adaptation to changing circumstances, and RBF provides a flexible framework for modeling intricate interactions among input variables, such as pile and soil parameters. Precision is increased by efficiently adjusting model parameters with AFT and RKO optimizers. After being trained using input variables, the models were validated, and statistical measures were used to assess the testing outcomes. The study's findings showed a strong connection between the actual observed bearing capacity and the Pu predictions produced by the RBAF (RBF optimized with AFT). High R2 and low RMSE values, notably at 0.991 and 33.414, respectively, demonstrated this significant association. These findings highlight the usefulness and efficiency of using enhanced RBF models as reliable resources for Pu prediction in the area of civil engineering; they also provide several noteworthy benefits.