Use of Neural Networks Within Constitution Models of Soils

Abstract

This paper focuses on the innovative use of machine learning and neural networks in constitutive modelling of soils, a material with complex and nonlinear behaviour. Traditional constitutive models, based on Hooke’s law or the Mohr-Coulomb model, often show significant discrepancies from the real-world behaviour of soils, leading to high costs and uncertainties in construction projects. The aim of this work is to lay the groundwork for a neural network capable of learning and reproducing results that are closer to the real behaviour of soils than current constitutive models. This approach could bring about a revolutionary change in the fields of geotechnics and construction by providing more accurate and efficient models for analysis and design of structures. The results could lead to the optimization of materials, cost reduction, and increased safety and sustainability of construction projects. This interdisciplinary approach opens up new possibilities for research and applications, with the potential to significantly contribute to innovations in geotechnics and construction.