基于CSSA-BPNN模型的胶结充填体动态抗压强度预测

充填采矿法二步骤回采时胶结充填体稳定性受爆破扰动而降低。为快速准确地获得充填体动态抗压强度，利用分离式霍普金森压杆(SHPB)进行了40组不同应变率的单轴冲击实验，以灰砂比、充填体密度、养护龄期和平均应变率作为输入参数，充填体动态抗压强度作为输出参数，建立了一种基于Logistic混沌麻雀搜索算法(CSSA)优化BP神经网络(BPNN)的预测模型，并与传统BPNN和麻雀搜索算法优化的BPNN进行了对比分析。研究结果表明：CSSA-BPNN模型的平均相对误差为4.11％，预测值与实测值之间拟合的相关系数均在0.96以上，模型预测精度高。CSSA-BPNN模型的均方根误差为0.3950 MPa，平均绝对误差为0.3592 MPa，决定系数为0.9952，均优于另外两种预测模型。实现了对充填体动态抗压强度的准确预测，可大幅减小物理实验量，为矿山胶结充填体的强度设计提供了一种新方法。

Prediction of Dynamic Compressive Strength of Cemented Backfill Based on CSSA-BPNN Model

The stability of cemented backfill is compromised by blasting disturbances in the two-step stoping method. To obtain the dynamic compressive strength of the backfill quickly and accurately,40 sets of uniaxial impact experiments with different strain rates were conducted using the separated Hopkinson pressure bar (SHPB). Input parameters included the lime-sand ratio, backfill density, curing age, and average strain rate,while the dynamic compressive strength of the backfill served as the output parameter. A prediction model, optimized using the backpropagation neural network (BPNN) based on the logistic chaotic sparrow search algorithm (CSSA), was established and compared with the BPNN optimized by the traditional BPNN and sparrow search algorithm. The results demonstrate that the average relative error of the CSSA-BPNN model is 4.11%, with fitting correlation coefficients between the predicted and measured values exceeding 0.96, indicating high prediction accuracy. The root-mean-square error of the CSSA-BPNN model is 0.3950 MPa, the average absolute error is 0.3592 MPa, and the coefficient of determination is 0.9952, all of which outperform the other two prediction models. This enables accurate prediction of the dynamic compressive strength of the backfill, greatly reducing the need for physical experiments and providing a novel approach to the strength design of mine cemented backfill.