基于改进果蝇优化GRNN的冲击地压预测研究

Research on the Prediction of Rock Burst Based on Generalized Regression Neural Network Optimized by Improved Drosophila Algorithm

针对煤矿开采过程中非线性,强耦合性等特点所致的动力灾害难以预测的问题,选用一种新的线性生成机制(LGMS)改进果蝇算法(FOA)优化广义回归神经网络(GRNN)的方法,建立了冲击地压煤岩灾害预测模型。采用LGMS改进FOA,避免FOA优化GRNN时陷入局部最优,增强了其搜索全局最优解的能力,提高了GRNN的收敛性与和预测精度。选取冲击地压前的电磁辐射、声发射、红外辐射3个主要指标,根据3种指标的单项危险指数求得综合危险指数,构建冲击地压动力灾害预测的LGMSFOAGRNN模型。研究表明,所构建的LGMSFOA-GRNN模型具有很好的预测能力和泛化能力。

Aimed at the problems of unpredictable dynamic disasters due to the non-linearity and strong coupling in the coal mining process,a new linear generation mechanism(LGMS)was used to improve Drosophila algorithm(FOA)to optimize the generalized regression neural network(GRNN).And a prediction model of rock burst was established.The FOA was improved by LGMS to avoid local optimization when FOA optimized GRNN,increasing its ability for searching the optimal solution from the total results,and enhancing the convergence and prediction accuracy of GRNN.Selecting three main indexes before the rock burst,including electromagnetic radiation,acoustic emission and infrared radiation,the comprehensive risk index was obtained on the basis of the single risk index of these three indexes.And the dynamic disaster prediction model of rock burst LGMSFOA-GRNN was established.The research showed that the constructed LGMSFOA-GRNN model had good prediction and generalization ability.