煤尘爆炸火焰传播特性因子分析与BP网络组合预测研究

Study on combined prediction of flame propagation characteristics in coal dust explosion based on factors analysis and BP network

为研究煤尘爆炸火焰传播规律,利用因子分析和BP网络定量预测煤尘爆炸火焰传播特性,探究工业分析指标、煤尘着火特性、环境因素和物理因素对煤尘爆炸火焰传播特性的影响。选取40个煤样,通过改造试验激波管体,实现25 m内煤尘爆炸火焰传播特性测试。将前30个煤样因子得分作为输入变量、火焰传播特性数据作为输出变量,同时添加附加动量进行BP网络训练。结果表明,煤尘爆炸火焰传播特性预测误差小于5%,网络训练预测成功,基于因子分析的网络预测精度得到提高。

In order to study the coal dust explosion flame propagation law,factor analysis and BP network were used for predicting explosion flame propagation characteristics of coal dust. Selected influencing factors are industry analysis indexes,coal dust ignition characteristics,environmental factors and physical factors. Coal dust explosion flame propagation characteristics experiments were carried out in a 25 meterslong transformed shock tube. Factor scores of top thirty kinds of coal samples were considered as input variables,flame propagation characteristic data were considered as output variable. Meanwhile an additional momentum item was added to the BP network to train. Results show that coal dust explosion flame propagation characteristics prediction error is less than 5%,network training is successful,and that by using factor analysis,the BP network prediction accuracy has been improved greatly.