群空区危险度的结构尺寸效应灰关联神经网络模型

Risk factor forecast based on neural network and grey relational analysis on physical dimension effect of mine goafs

在力学分析的基础上，耦合灰关联分析及神经网络方法，建立空区危险度的结构尺寸效应灰关联神经网络模型，实现基于空区结构尺寸参数的五级空区危险度评价，分析空区危险度与结构尺寸参数间的影响程度及响应关系。研究结果表明：建立的基于空区结构尺寸的危险度灰关联神经网络预测模型，能够实现群空区下的危险度的精确预测，其预测误差范围为0．08230A-6．7483％，平均误差为3．0546％；空区危险度的影响因素中，矿柱宽度对空区危险度的影响最大，埋深对危险度的影响最小；在其他影响因素不变的情况下，空区的危险度与暴露面积、埋深呈正相关性，与矿柱宽度呈负相关性，而与高宽比的关系式随着高宽比增大先增大后减小并最终趋于定值。

On the basis of mechanics, coupled grey relational analysis method and neural network method, the neural network forecasting model based on grey relational analysis of physical dimension effect on risk factor of mine goals was created, five-level risk evaluation based on physical parameter of mine goafs was realized, and relation ＆ response between physical dimension and risk factor of skip area were analyzed. The results show that neural network forecasting model created based on grey relational analysis is an accurate realization method of forecast on risk factor of mine goaf with the error range between 0.082 3%-6.748 3% and average error 3.054 6%. Among influential factors, width of pillar and buried depth are respectively the most and the least influential factor for risk factor of mine goals. Besides, with other influential factors unchanged, risk factor of mine goals positively correlates with exposed area and buried depth, attd negatively correlates with width of pillar. As the height-width ratio increases, risk factor firstly increases and then decreases and finally tends to be a fixed value.