基于改进BP神经网络的煤矿冲击地压预测方法研究

摘要

In order to effectively predict and prevent the mine pressure bump occurred in coal mine,in combination with the acoustic emission technology and the neural network,taking the characteristic parameters of the acoustic emission activity as the basic data,according to the slow convergent speed of the BP neural network,easy in a local extremum and other problems,the BP neural predicted network was improved.The particle swarm optimization algorithm was applied to optimize the BP neural network and the particle swarm algorithm was applied to train the weight value and threshold value of the BP neural network.The results showed that under the condition of the training errors all to be 0.001,in comparison with the not optimized conventional BP neural network,the convergent speed of the particle swarm optimization algorithm neural network would be 4 ～ 5 times fast than not optimized conventional BP neural network.The high convergent speed,high predicted accuracy and other features of the provided prediction method were proved.In the application of the mine pressure bump,the BP neural network was feasible and effective and could provide the theoretical support to the prediction of the mine disaster.%为有效预测与防治煤矿冲击地压灾害的发生,将声发射技术与神经网络结合,把声发射活动的特征参数作为基础数据,针对BP神经网络收敛速度慢且易陷入局部极值等问题,改进BP神经预测网络.采用粒子群优化算法对BP神经网络进行优化,利用粒子群算法训练BP神经网络的权值和阈值.结果表明:在训练误差均要求达到0.001的情况下,与未经优化的传统BP神经网络相比,粒子群优化过的BP神经网络的收敛速度要较其加快了4～5倍,证明该预测方法具有收敛速度快,预测精度高等特点,在煤矿冲击地压预测的应用中具有可行性与有效性,为煤矿灾害的预测提供了理论支持.