Uniaxial compressive strength prediction and ratio parameter optimization of titanium tailings composite backfill materials using intelligent hybrid models

This study analyzed the feasibility of using titanium (Ti) tailings as a backfilling material in an underground mine. The backfilling ratio tests were performed using Ti tailings as the backfilling aggregate and flotation tailings, fly ash, and phosphogypsum as auxiliary materials. The uniaxial compressive strength (UCS) of the backfill body under different curing times was measured. Ninety datasets, including UCS, ratio parameters, and curing time, were collected. According to these datasets, extremely randomized trees (ERT) and four optimization algorithms were used to build four hybrid models to predict the UCS of the backfill body. A five‐fold cross‐validation was adopted to verify the validity of the model. The testing R 2 values of the four hybrid models were greater than 0.950. The slime mold algorithm‐ERT (SMA‐ERT) model achieved the best performance ( R 2 = 0.970). To analyze the superiority of these hybrid models, eight machine learning (ML) models (random forest, support vector machine, etc.) were also developed. Four hybrid models outperformed these eight ML models on the testing datasets. In addition, an intelligent optimization design system was developed to ensure strength compliance in mix proportioning. This indicated that the proposed hybrid models were effective tools for predicting the UCS of the backfill body. Finally, the SMA‐ERT model was used to determine the optimal backfilling ratio parameters for achieving a backfill body with a UCS of 2.0 MPa after 28 days of curing. The rationality of the ratio parameters was verified by tests. The testing results showed that the backfill body prepared using the ratio parameters provided by the SMA‐ERT model can achieve a UCS of 2.02 MPa. This result suggested that the SMA‐ERT model is the most effective tool for UCS prediction, and can reliably guide the design of the backfill ratio parameters.