1. Strategic Planning Ｒesearch Institute，China Coal Ｒesearch Institute ，Beijing ，China;
2. School of Ｒesource and Safety Engineering，China University of Mining and Technology( Beijing) ，Beijing ，China

为了更好地预测深部煤层底板突水危险性,将深部煤层底板破坏深度多元非线性数学模型给出的预测值,与底板含水层水压力、隔水层厚度、工作面斜长、埋深4个因素相结合,共同作为深部煤层底板突水危险性预测的输入向量;运用PSO获取SVM的最优惩罚因子和核函数参数,建立了深部煤层底板突水危险性预测模型,并与突水系数法、马氏距离判别法、贝叶斯判别法的预测精度进行对比。结果表明:深部煤层底板破坏深度数学模型预测的准确度较高;考虑煤层底板采动破坏影响的深部煤层底板突水危险性预测模型,较好地预测了煤层底板突水危险性,其准确度相比于突水系数法、马氏距离判别法和贝叶斯判别法均较高。

In order to better predict water inrush risk from deep coal seam floor，the mathematic model for failure depth predicted value of deep coal seam floor was obtained by using multiple linear regression theory．Based on this，the influences of confined aquifer pressure， thickness of aquifuge，length of panel and buried depth were considered as input vectors of the prediction model for evaluating deep water inrush risk． The prediction model for evaluating water inrush risk from deep coal seam was built，which optimal penalty factor and optimal kernel function parameter of SVM was obtained by PSO． The predictive accuracy of the model was compared with water inrush coefficient method，Mahalanobis distance discriminant method and Bayes discriminant method． The results show that the correlation coefficient be- tween fitting results calculated by mathematic model for failure depth of deep coal seam floor and measured results had a high fitting de- gree，which means the mathematic model had a higher accuracy． Taking the failure depth of deep coal seam floor into consideration，the model for water inrush risk from deep coal seam floor provided better results，which prediction accuracy was higher than the water inrush coefficient method，Mahalanobis distance discriminant method and Bayes discriminant method．

国家自然科学基金青年科学基金资助项目(51704158)；中国煤炭科工集团科技创新青年基金资助项目(2014QN005)；