2015 ISFAS
The 3^rd International Symposium on Fundamental and Applied Sciences
March 22-24, 2015, Osaka, Japan

Keynote Speech

Topic:
Applications of hybrid-artificial neural network in geotechnical engineering

Abstract:
Geotechnical engineers face many uncertainties in relation with their design works. This is due to the non- homogeneity of the soil materials that they are dealing with. Recently several studies have proved that feasibility of artificial neural network (ANN) in solving different geotechnical problems. It is well established that ANN as a function approximation tool that tries to find a nonlinear relationships between parameters can be implemented for estimating several soil properties, including the compression index, shear strength, permeability, soil compaction, lateral earth pressure, and others. However, over the last few years, some researchers addressed ANN deficiency in finding global minima and slow rate of learning. Alternatively, to improve the ANN prediction power, they suggested the use of hybrid ANN which combines global search algorithms such as particle swarm optimization (PSO) and genetic algorithm (GA) with conventional ANN. The main objective of this paper is to provide a general view of hybrid ANN application in geotechnical engineering. The paper is not intended to provide details of ANN modelling procedure, but rather, for brevity purpose, it tries to cover a number of hybrid-ANN application in predicting the bearing capacity and settlement of both deep and shallow foundations. The paper suggest the prediction performance of hybrid-ANN is better than that of conventional ANN. It recommends the hybrid ANN implementation, as a more reliable, economic and quick tool, in solving geotechnical problems more specifically when the contact nature between parameters is unknown and complex. Results base from the author’s works indicate that the pile bearing capacities predicted by GA-based ANN are in close agreement with measured bearing capacities. Coefficient of determination as well as mean square error equal to 0.990 and 0.002 for testing datasets respectively. The model performance using PSO-based ANN model also shows a close agreement between the measured bearing capacities for shallow foundation based on the coefficient of determination value. The determination values equal to 0.997 and 0.991 for training and testing datasets respectively.While using a Back Propagation techniques shows that ANN is a powerful and accurate tool in predicting the bearing capacity of spread foundations. The coefficient of determination equals to 0.99 suggests that the predicted bearing capacity by ANN model is in close agreement to the measured bearing capacity. The results conclude that the Hybrid ANN model serves as a reliable and simple predictive tool to appropriately consider various essential parameters for predicting the bearing capacity of either shallow or deep foundations.
(Keywords: Artificial Neural Network, bearing capacity, shallow foundation, deep foundation, hybrid.)