Comparative prediction performance of the strength of a new type of Ti tailings cemented backfilling body using PSO-RF, SSA-RF, and WOA-RF models

Abstract

The uniaxial compressive strength (UCS) of the backfilling body plays a crucial role in backfilling safety. To study the feasibility of preparing new cementitious materials by Ti tailings, 90 sets of backfilling ratio tests and UCS tests with different Ti tailings replacement ratios were carried out. The test results showed that the UCS tended to decrease as the ratio of Ti tailings replacing cement increased, and the UCS decreased sharply after the ratio of Ti tailings replacing cement exceeded 80%; however, the UCS increased at 28 days when the ratio of Ti tailings replacing cement did not exceed 10%. Meanwhile, to accurately predict the UCS, the whale optimization algorithm and random forest (WOA-RF), particle swarm optimization and random forest (PSO-RF), and sparrow search algorithm and random forest (SSA-RF) hybrid models were constructed. Moreover, the input parameters of the models include the ratio of Ti tailings, concentration, curing age, ratio of P.O42.5 cement and cement-sand ratio. In addition, a single RF model was constructed for comparative analysis, the four models (RF, WOA-RF, PSO-RF, and SSA-RF) were trained and tested, and the root mean square error (RMSE), coefficient of determination (R2), and mean absolute error (MAE) of the four models were obtained: (0.532, 0.922, 0.416), (0.254, 0.975, 0.209), (0.367, 0.961, 0.300) and (0.318, 0.965, 0.258), respectively. The research showed that the WOA-RF model achieves better performance (R2 = 0.975), which proves that it is a better prediction model for predicting the USC of new cementitious filler with Ti tailings replacement.

Keywords

• New backfilling material
• Uniaxial compressive strength (UCS)
• Ti tailings
• Machine learning (ML)
• Hybrid model