Constitutive Modeling of Cemented Tailings Backfill With Different Saturation States Based on Particle Swarm Optimization and Support Vector Machine

• Main Authors: Zhuoqun Yu, Yongyan Wang, Hao Wang
• Format: Article
• Language: English
• Published: IEEE 2021-01-01
• Series: IEEE Access
• Subjects: Machine learning; materials testing; mechanical variables measurement
• Online Access: https://ieeexplore.ieee.org/document/9317773/

Mine tailings disposal has been a serious environmental issue for decades. The wide application of cemented tailings backfill (CTB) technology could indirectly abate tailings pollution by recycling the tailings for backfilling. CTB constitutive modeling helps with design by improving the understanding of its compressive behavior. This study focused on CTB intelligent constitutive modeling considering the coupled effects of the cement content and saturation state. An artificial intelligence model was established and utilized based on particle swarm optimization (PSO) and the support vector machine (SVM). CTB samples with different cement contents and water saturation states were prepared, and unconfined compression tests were conducted to obtain the dataset. We verified the feasibility of using integrated PSO and SVM (P-S) in the CTB constitutive model using experimental data. We analyzed model errors. The results showed that the CTB stress strain curve was complex and nonlinear and could be significantly affected by the saturation states. PSO was feasible and efficient for tuning the SVM hyperparameters. The lowest minimum MSE value of 0.0108 was achieved in the eighth iteration. The PSO and SVM modeling was indicated to be accurate in the CTB constitutive model (a high R-square value of 0.9935 and a low mean squared error value of 0.001664 were achieved on the testing set). This model may accelerate the CTB structure design process.