Cement-based grouting material development and prediction of material properties using PSO-RBF machine learning

Grouting technique is one of the main methods to improve mechanical properties of fractured rock. The study of grouting material is essential for improving the effect of grouting. To develop a new low water-cement ratio cement-based grouting material, the influence of additives on cement strength and Pearson correlation analysis method was adopted to obtain main material compositions. Then, the coupled particle swarm optimization algorithm and radial basis function (PSO-RBF) model was established for material properties prediction with proportion as input. The prediction results show that the proposed PSO-RBF model has a higher accuracy compared to the RF, BP, and RBF models. Furthermore, combined the results of PSO-RBF with entropy weight method, the optimal proportion of grouting material was developed. The results of mechanical properties indicated that this proposed cement-based material has the characteristics of reducing water-cement ratio and porosity and increasing strength, fluidity, and contact angle. The proposed material proportion intelligent optimization approach and grouting material can provide a reference for material design and engineering application.