基于粒子群优化算法的充填体单轴抗压强度预测研究

Research on prediction of backfill strength based on particle swarmoptimization algorithm

为快速有效确定充填体单轴抗压强度,以灰砂比、固体含量和养护龄期作为输入因子,充填体单轴抗压强度作为输出因子,建立一种粒子群优化算法(PSO),对支持向量机(SVM)参数进行全局优化的预测模型。结果表明,该模型预测性能较好,相关系数高(训练集为0.996,测试集为0.993),均方误差值低(训练集为0.000393,测试集为0.00072613);通过室内试验对采集的216个试样进行预测与对比,证明模型可以准确地预测充填体单轴抗压强度,大幅度减少物理试验量及缩短试验周期,为矿山充填提供一种新思路。

In order to quickly and effectively determine the strength of the backfill,a particle swarm optimization algorithm(PSO)was established for the global optimization of the support vector machine parameters by taking the ratio of lime to sand,solid content and curing age as input factors,and the output factor as the uniaxial compressive strength of the backfill.The research results show that the model had good prediction performance,achieving high correlation coefficients(training set 0.996,test set 0.993),and low mean square error value(training set is 0.000393,test set is 0.00072613).The comparison and prediction of 216 samples collected through indoor experiments proved that the model could accurately predict the strength of the filling body,greatly reduce the amount of physical test and the test period,and provide a new idea for mine filling.