Revealing the nature of concrete materials using soft computing models

Abstract

The durability of materials has long been a topic undergoing intense study in civil engineering. This paper used a novel approach to foster current research in two aspects: identifying dominant concrete properties and simulating chloride transport in concrete. Machine learning methods have proven to have remarkable abilities to approximate complex functions, which provides a potential tool for realizing the above ideas. This paper used deep neural networks to identify dominant concrete parameters and simulate the transport process by integrating physical constraints into models. Results showed that deep neural networks combined with physical constraints could effectively identify material properties with high accuracy (relative error <1%) and simulate the aggression of chloride. This study also demonstrated that machine learning models could be interpretable. In addition, this paper also explored the influence of time intervals of detection datasets, the size of datasets, and the noise in datasets (−5%–5%) on the convergence of identification. Results showed that short time intervals and big datasets could accelerate convergence while the noise in datasets had negligible influences on this method. In addition, with exact physical constraints, this method could predict the performance of materials without the assistance of any dataset.