煤层底板破坏深度预测的GRA-FOA-SVR模型

摘要

针对煤层底板破坏深度影响因素冗余、预测方法种类繁多且参数优化较为困难的问题,选取支持向量回归机来预测煤层底板破坏深度,引入灰色关联度分析法和近年来较为流行的果蝇优化算法以用于影响煤层底板破坏深度的主控因素的提取和对常规支持向量回归机的优化,提出了煤层底板破坏深度预测的GRA-FOA-SVR模型;选取开采深度、煤层倾角、开采厚度、工作面斜长、煤层底板损伤变量和煤层切穿型断层或破碎带数等6个影响煤层底板破坏深度的因素,采用灰色关联度分析法选出关联度在80% 以上的因素组成煤层底板破坏深度的主控因素;用果蝇优化算法对支持向量回归机参数进行迭代寻优,将最优参数代入支持向量回归机模型中.将测试样本的模型预测值与实测值、规程公式计算值、常规支持向量回归机计算值对比分析,结果表明:GRA-FOA-SVR模型比规程公式和常规支持向量回归机的预测误差更小,更能有效地预测煤层底板破坏深度.利用排序加权平均算子法将GRA-FOA-SVR模型同"下四带"理论公式、岩石力学试验和规程公式融合,提出了预测煤层底板破坏深度的多源信息融合的方法,并以良庄井田51302工作面为例,表明了该融合方式的实用性,最后据此方法预测出了肥城煤田6个工作面的煤层底板破坏深度.%Aiming at the problems of redundant influence factors of failure depth of coal seam floor,various prediction methods and difficult parameter optimization,support vector regression was selected to predict the failure depth of coal seam floor,and grey relational analysis and the recently popular fruit fly optimization algorithm were introduced to extract the main control factors affecting the failure depth of coal seam floor and optimize the conventional support vector regression machine.The GRA-FOA-SVR model was proposed to predict the failure depth of coal seam floor.The study selects six factors such as mining depth,coal seam inclination,mining thickness,length of working face,coal floor damage variable and the number of faults cut through the coal seam or broken belts.Then,the controlling factors of the failure depth of coal seam floor are selected by the correlation de-gree of more than 80% using the grey relational analysis.The optimization algorithm of Drosophila is used to iterate the parame-ters of the support vector regression machine,and the optimal parameters are replaced in the support vector regression model. The model predictive values of the tested sample are compared with the measured values,the calculated values of the regulation formula and the conventional support vector regression machine.The result shows that the prediction error of the GRA-FOA-SVR model is smaller than that of regulatory formula and conventional support vector regression machine,so it can effectively predict the failure depth of coal seam floor.Based on the weighted average operator method,the GRA-FOA-SVR model is mer-ged with the"down four zones"theoretical formula,rock mechanics test and the regulatory formula,and a multi-source informa-tion fusion method for predicting failure depth of coal seam floor is put forward.Taking the 51302 working face of Liangzhuang Minefield as an example,the practicability of the fusion method is demonstrated.Finally,the failure depth of coal seam floor of 6 working faces in Feicheng coalfield is predicted.