Determination of load-bearing capacity in deep foundations using artificial neural networks

Abstract

Knowing the geotechnical characteristics of the construction site is of utmost importance for the engineer when designing a project, regardless it is large, medium, or small in scale. By using SPT-N values, the designer will be capable of determining the load-bearing capacity of the soil. Semi-empirical models have been proposed to calculate the load-bearing capacity in deep foundations. However, these models are developed based mathematical frameworks and approximations, where the calculation is not always accurate. Artificial Neural Networks (ANNs) are considered a mechanism capable of accurately calculating these values because they use more reliable variables. For this reason, the present study aims to propose the use of Multi-Layer Perceptron (MLP) artificial neural networks to determine the load-bearing capacity of deep foundations based on SPT-N values and depth. The database compiled from a series of 68 load tests carried out in the city of Balsas-MA, for which SPT-N values and their respective depths were available. To evaluate the obtained results, the roots of the mean squared errors (RMSE) and Pearson correlation coefficients were calculated. The results indicated that the proposed model can determine the load-bearing capacity with a small error rate (RMSE) and high correlation with the values calculated using the classical formulas from the literature.