摘要： 针对尾矿库事故具有随机波动性和非线性的特点,提出采用修正型果蝇优化算法优化广义回归神经网络的尾矿库安全评价模型(MFOA-GRNN)。该方法利用修正型果蝇优化算法的全局寻优特性对广义回归神经网络进行参数优化,同时应用去相关性分析选取尾矿库安全评价指标,实现尾矿库的安全预测。以辽宁本溪南芬尾矿库为研究实例进行拟合预测,实验结果表明,将MFOA 方法与GRNN 网络有机结合,有利于平滑因子σ 的选择,相 较于FOA-GRNN 模型70% 的预测准确度,采用修正型果蝇算法优化的GRNN 模型预测准确度高达100% ,预测精度更高,适用性更强。

关键词: 尾矿库, 果蝇优化算法, 广义回归神经网络, 平滑因子, 参数优化, 安全预测

Abstract: At the mine tailings’ characteristics of stochastic fluctuation and nonlinear,and its safety prediction can be affected by many factors,a prediction model for mine tailings is put forward by adopting Modified fruit Fly Optimization Algorithm of the Generalized Regression Neural Network ( MFOA-GRNN ). The method introduces the global optimization characteristics of MFOA to optimize the parameter of GRNN,while using correlation analysis to select the mine tailings safety evaluation to achieve forecast. Taking Liaoning Benxi Nanfen mine tailing as research instance to fit forecast,it shows that combining MFOA with GRNN is beneficial to select the smoothing factor and compared with prediction accuracy 70% of the FOA-GRNN model,MFOA-GRNN model prediction accuracy is as high as 100% and has higher prediction precision and stronger applicability.

Key words: mine tailings facilities, Fly Optimization Algorithm ( FOA ), Generalized Regression Neural Network (GRNN), smoothing factor, parameter optimization, safety prediction

中图分类号: 

• TP183