Recent Advances in Soil Constitutive Modeling Using Machine Learning

In recent years, many computer-aided pattern recognitions have been developed for constitutive modeling of soil behavior. The main idea behind a pattern recognition system is that it learns from experience and can provide predictions for new cases. These systems have been increasingly employed for constitutive modeling, especially in cases where the behavior is too complex. Machine learning (ML) can directly learn from raw data to develop constitutive models for soils by using pure mathematic skills without limitations of constitutive formulations. However, current studies on the ML-based constitutive modeling of soils are still very limited. Therefore, there is significant potential for ML to achieve great advances in the constitutive modeling of soils. However, a common challenge and criticism of ML is a lack of interpretability and generalization ability. This paper provides a detailed review of the recent advances in ML-based studies of soil constitutive modeling.

کلیدواژه ها:

Soil ، Constitutive Modeling ، Machine Learning ، Artificial Intelligence