CPT-based method using hybrid artificial neural network and mathematical model to predict the load-settlement behaviour of shallow foundations

ABSTRACT

This paper proposes a new hybrid artificial neural network and mathematical (ANN-MATH) model to improve the prediction capability of the load-settlement behaviour of shallow foundations in sandy soils. The hybridisation process is performed by replacing the conventional activation function of the ANN output layer by a new mathematical model. Thereafter, 110 full-scale loading tests of shallow foundations, carried out in sand with the cone penetration test (CPT) results, are used to build and validate the proposed model. In terms of accuracy, the proposed model shows a high correlation between the predicted results and the measured data. In addition, the proposed method was compared to the available methods in the literature. It was found that the proposed model is superior to classical and artificial intelligence-based methods by more than 29% and 35%, respectively, in terms of root mean square error (RMSE). Furthermore, a parametric study was undertaken to assess the robustness of the proposed model and to derive the bearing capacity factor based on the CPT test. The bearing capacity factor was found close to those recommended by the French standard NF P 94–261.

KEYWORDS:

• Shallow foundation
• sand
• load
• settlement
• cone penetration test
• hybrid model
• artificial neural network

Disclosure statement

We declare that there is no conflict of interest or financial support associated with this work.