Abstract: In order to solve the problem of strength instability of cemented tailings backfill (CTB) under low temperature environment ( 20 ◦C), the strength optimization and prediction of CTB under the influence of multiple factors were carried out. The response surface method (RSM) was used to design the experiment to analyze the development law of backfill strength under the coupling effect of curing temperature, sand-cement ratio and slurry mass fraction, and to optimize the mix proportion; the artificial neural network algorithm (ANN) and particle swarm optimization algorithm (PSO) were used to build the prediction model of backfill strength. According to the experimental results of RSM, the optimal mix proportion under different curing temperatures was obtained. When the curing temperature is 10—15 ◦C, the best mix proportion of sand-cement ratio is 9, and the slurry mass fraction is 71%; when the curing temperature is 15—20 ◦C, the best mix proportion of sandcement ratio is 8, and the slurry mass fraction is 69%. The ANN-PSO intelligent model can accurately predict the strength of CTB, its mean relative estimation error value and correlation coefficient value are only 1.95% and 0.992, and the strength of CTB under different mix proportion can be predicted quickly and accurately by using this model.

Key words: cemented tailings backfill (CTB), response surface method (RSM), multi-factor coupling, strength optimization, intelligent prediction model

摘要： 为解决低温环境下（≤20℃）尾砂胶结充填体强度不稳定问题，开展了多因素影响下尾砂胶结充填体强度优化与预测。采用响应面法进行试验设计，分析了固化温度、砂灰比和料浆质量分数耦合作用下充填体强度的发展规律，并对配合比进行了优化；采用人工神经网络算法（ANN）和粒子群优化算法（PSO），建立了充填体强度预测模型。根据响应面法试验结果，得到了不同固化温度下的最佳配比。固化温度为10 ~ 15℃时，砂灰比最佳配合比为9，料浆质量分数为71%；固化温度为15 ~ 20℃时，砂灰比最佳配合比为8，料浆质量分数为69%。ANN-PSO智能模型能够准确地预测尾砂胶结充填体强度，其平均相对估计误差值和相关系数值仅为1.95%和0.992，可以快速准确地预测不同混合比例下的尾砂胶结充填体强度。

关键词: 尾砂胶结充填体，响应面法，多因素耦合，强度优化，智能预测模型