PREDICTION OF SOIL BEARING CAPACITY USING SOFT COMPUTING TECHNIQUES

Abstract

The nature and manner in which structures are collapsing in Nigeria are alarming. It creates a room in which structural engineers, the building industry, government, estate developers, building consultants and other relevant stakeholders in the department building industry ask many questions about how and what is behind the sudden collapse of structures. Therefore, this research aimed to predict the ultimate bearing capacity of the square, strip and circular footing from shearing strength parameters using ANN and ANFIS. This paper, 200 data sets were used to develop the model; 75% were used for training and 25% for testing the model. ANN and ANFIS learning algorithms were employed in developing the models under various foundation types. Eventually, Various error measures, such as coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R), were employed to compare the efficiency of the models. The performance comparison findings indicated that the soft-computing system is an efficient instrument for risk reduction in soil engineering projects. The models were validated using external data and the correlation prediction capacity of the models where ANN-STRIP (89%), ANN-SQUARE (83%), ANN-CIRCULAR (89%), ANFIS-STRIP (86%), ANFIS-SQUARE (79%) and ANFIS-CIRCULAR (96%). All the models have shown a quite good and reliable prediction capacity, with ANFIS-CIRCULAR having 96% prediction accuracy of soil bearing capacity.

Keywords

• neural network
• neuro-fuzzy
• cohesion
• angle of internal friction