Estimating the bearing capacity of driven piles in sandy clay soils using radial neural network and multilayer perceptron

Abstract

In this research, neural network models have been used to estimate the bearing capacity of percussion piles in different types of clay and pile soils. Input data include soil cohesion in kilopascals, soil friction angle in degrees, soil friction angle and pile in degrees, soil specific gravity in kilowatts per cubic meter, number of impacts per unit, pile cross section in square meters And the length of the candle is defined in meters and the output of the bearing capacity of the candles is in kilotons, which has been obtained in a laboratory. To increase the accuracy and find the least amount of error, the combination of multilayer perceptron neural network and gray wolf optimization algorithm and the combination of fuzzy neural network and gray wolf optimization algorithm have been used to determine the most accurate network and the least amount of error. In general, based on the obtained results, both models can be used to estimate the bearing capacity of piles. However, the results obtained from the multilayer perceptron model and the gray wolf optimization algorithm with the proposed structure show better performance. Also, the final selected model was approved and new data sets were obtained.