基于PSO-BPNN的充填体围岩界面抗剪强度预测

Prediction on the Interface Shear Strength of Backfill and Surrounding Rock Based on PSO-BPNN Algorithm

在地下采场胶结充填体强度设计过程中,充填体围岩界面峰值抗剪强度是充填体安全系数计算的一个重要参数。为给出一定充填体侧压力下,充填体围岩界面峰值抗剪强度预估方法,使用非线性拟合能力强大的神经网络进行预估。基于试验所得考虑胶结剂含量、养护温度、养护时间和剪切面法向压力的试验数据,对胶结充填体围岩界面峰值抗剪强度进行预计。使用粒子群优化算法初始化反向传播神经网络权值和偏置参数,训练后的神经网络可以有效地对给定法向压力下胶结充填体围岩界面的峰值抗剪强度进行预估,预估结果可作为胶结充填体强度设计时的参考。

During the strength design of cemented backfill,the peak interface shear strength of backfill and surrounding rock is a key parameter for estimating the safety factor of backfill.In order to obtain the prediction method of the peak interface shear strength of backfill and surrounding rock under a certain side pressure of the backfill,the neural network with strong nonlinear fitting ability was used to predict.Taking the cement content,curing temperature,curing time and normal pressure of shear plane as test data,the peak interface shear strength of backfill and surrounding rock was predicted.The weights and bias parameters of back propagation neural network were initialized by particle swarm optimization algorithm.The trained neural network can effectively predict the peak shear strength of cemented backfill under a given normal pressure.The predicted results can be used as a reference for the strength design of cemented backfill.