Development, optimisation and performance prediction of a novel cement-based materials for borehole sealing

Abstract

Achieving high-quality borehole sealing is critical for effective gas extraction, yet conventional cement-based materials often suffer from shrinkage-induced cracking, poor fluidity, and insufficient adaptability to fractured formations. This study presents the development and multi-level optimization of a novel high-performance cement-based sealing material, using Portland cement as the matrix and incorporating polycarboxylate superplasticizer, calcium sulphoaluminate expansion agent, and sodium gluconate retarder. A comprehensive methodology was employed that single-factor experiments initially identified the influence of individual components on fluidity, expansibility, and compressive strength, while orthogonal design combined with response surface analysis (RSA) enabled multivariable optimization of admixture dosages. In addition, a backpropagation (BP) neural network based on the Levenberg-Marquardt algorithm was constructed to predict material performance across varying formulations. The integrated experimental–computational framework led to the identification of optimal parameters with the water-cement ratio of 0.8, superplasticizer at 0.7 %, expansion agent at 3 %, and retarder at 0.07 %. The BP neural network accurately predicted fluidity, expansion, and strength with average errors of 6.396 %, 3.794 %, and 4.042 %, respectively. This innovative approach not only enhances material performance but also establishes a predictive foundation for designing application-specific sealing materials, offering a practical and adaptable solution for improving borehole sealing reliability in complex geological conditions.