A Comprehensive Review of Machine Learning and Traditional Models for Predicting the Physical and Mechanical Properties of Cement and Cement Mortar

Traditional methods of testing the physical and mechanical properties of cement and cement mortar are reliable, but they are time-consuming and labor-intensive. Therefore, it is necessary to develop robust and consistent models that can reduce dependency on prolonged experimental testing. In this review, different types of models are critically examined, including both machine learning and traditional approaches. The complexity of cement hydration, chemical properties, and microstructural development presents challenges for predictive modeling, but several models have yielded satisfactory results. This review includes various Machine Learning (ML) models such as Back-Propagation Neural Network (BPNN), Support Vector Machine (SVM), Decision Tree, Random Forest (RF), AdaBoost, k-Nearest Neighbors (k-NN), a hydration reaction analysis model, hybrid SVM-Genetic Algorithm (GA), Deep Convolutional Neural Network (DCNN) with Particle Swarm Optimization (PSO), contact mechanics approach, Extreme Gradient Boosting (XGBoost), Gradient Boosting Regression, fuzzy logic-based model, Physics-Informed Neural Network (PINN-CHK), Deep Neural Network (DNN) with Generative Adversarial Networks (GANs), and Group Method of Data Handling (GMDH). In addition, mathematical and numerical models have been studied to predict cement hydration kinetics, compressive strength, microstructure evolution, slurry behavior, and hydration heat. These include MATHCAD7, Scheffe’s simplex technique, continuum-based hydration models, constitutive equations, integrated numerical models, and thermodynamic and finite element models. Such predictive models can significantly accelerate decision-making in cement technology. The insights presented in this review will serve as a valuable reference for researchers and engineers. Major Findings: This study highlights that traditional testing methods used for cement and mortar are reliable, but they are time-consuming. Several ML and mathematical models have demonstrated strong forecasting capabilities. A few models, such as SVM-GA, DNN-GANs, and PINN-CHK, have shown excellent accuracy. These models can save significant time and enable faster decision-making in the construction sector.