基于SVM的采区地表沉陷预测与分析

Subsidence Prediction and Analysis of Mining Area Surface Based on SVM

针对现有地表沉陷预测方法在处理岩体运移过程中的非线性与不连续性时的不足,考虑地下开采引起的地表沉陷的众多影响因素,基于支持向量机对地表沉陷的最大值进行了预测,以地表最大沉陷值为因变量,以采厚、煤层倾角、采深、开采长度、开采宽度和覆岩特性等为自变量,得出了地表最大沉陷值的预测模型。并采用平均影响值(MIV)方法对各影响因素的重要度进行了排序。结果表明:校正模型的相关系数R2=0.99166,校正误差均方根MSE=0.00188,模型的预测精度满足实际工作所需;MIV法分析结果表明,采厚、倾角、覆岩岩性三者的影响作用最为明显。

Aiming at the existing surface subsidence prediction method＇s deficiency in dealing with nonlinearity and discontinuity of rockmass movement and considering numerous factors of surface subsidence caused by un‐derground mining ,the paper predicted the maximum of surface subsidence based on support vector machine ,ob‐tained the predictive models of surface subsidence maximum by using surface subsidence maximum as dependent variable and the mining thick ,coal seam dip ,mining depth ,mining length ,width ,and mining overburden charac‐teristics as independent variables ,and finally sorted the importance of each factor by using the average impact value （MIV） method .Results showed that ：the correlation coefficient of calibration model R2 =0 .99166 ,the root mean square of correction error MSE=0 .00188 ,the prediction accuracy of model met the job requirements;MIV analysis results showed that the most important factors were the mining thickness ,inclination and the rock property of the overburden rock .