Machine learning applications in designing cementitious materials

This review explores the development and application of machine learning (ML) algorithms in cementitious materials, and some highlighting and potential ML-related applications are emphasized. This review takes the commonly employed ML algorithms and training strategies as clues, and it covers commonly used ML models, including Neural Networks based (NN-based) algorithms and Classification and Regression Trees based (CART-based) algorithms, along with transfer learning concepts. Then, the impact of ML on material mechanics is analyzed, emphasizing improved reliability in phenomenal analysis, composite design, and predictive modeling of material properties. The role of ML algorithms in visual material identification and physics-informed modeling is discussed, along with applications in model interpretability, physical constraints, in-situ damage identification. The integration of Large Language Models (LLMs) is also introduced as an emerging research avenue. By providing an overview of ML's role in material mechanics, this review offers insights for researchers and engineers in the field.