Quantcast

Prediction of ultimate bearing capacity through various novel evolutionary and neural network models

Research paper by Hossein Moayedi, Arash Moatamediyan, Hoang Nguyen, Xuan-Nam Bui, Dieu Tien Bui, Ahmad Safuan A. Rashid

Indexed on: 26 Mar '19Published on: 11 Mar '19Published in: Engineering with Computers



Abstract

In the current study, various evolutionary artificial intelligence and machine learning models namely, optimized artificial neural network (ANN), genetic algorithm optimized with ANN (GA-ANN) and particle swarm optimization optimized with ANN (PSO-ANN), differential evolution algorithm (DEA), adaptive neuro-fuzzy inference system (ANFIS), general regression neural network (GRNN), and feedforward neural network (FFNN) were optimized and applied to predict the ultimate bearing capacity (Fult) of shallow footing on two-layered soil condition. Due to a lot of input variables such as (upper layer thickness/foundation width (h/B) ratio, footing width (B), top and bottom soil layer properties) finding a reliable solution for such a complex engineering problem is difficult. Most of the available solutions are based on very limited experimental works. To assess the capability of proposed methods a new ranking system called CER (color intensity rating) based on their result of above indices was developed. As a result, although all provided methods, after being optimized, could successfully predict the bearing capacity of shallow footing in the two-layer subsoil and PSO-ANN could perform better compared to other techniques. Based on RMSE, R2 and VAF, values of (0.01, 0.99, and 99.90) and (0.01, 0.99, and 99.90) were found, respectively, for the training and testing datasets of PSO-ANN model. In this regard, the accuracy of other hybrid algorithm of GA-ANN model with RMSE, R2 and VAF of (0.05, 0.99, and 97.80) and (0.06, 0.99, and 97.57), respectively, for the training and testing datasets was slightly lower than the PSO-ANN model. This shows the superiority of the PSO-ANN model in the prediction of a highly complex real-world engineering problem.