中文摘要：

为了减少矿井深部巷道顶板事故，对矿井深部开采中煤层巷道的动压规律，采用一种基于EEMD—SVM—DS—ARIMA的多模态软测量的来压预测方法进行了预测研究。首先利用聚合经验模态分解（EEMD）方法对非线性、非平稳来压监测信号进行模态分解得到多个模态函数序列（IFM）；第二运用支持向量机（SVM）方法对各IFM分量进行训练，重构得到各样本输出函数；第三用证据理论（DS）合成规则得到多个证据概率分配函数，将其作为权值因子对子函数的输出进行融合得到多函数的输出；最后应用单整自回归移动平均（ARIMA）模型对合成序列进行动态校正。实际应用表明，多模态软测量的来压预测模型能提高顶板压力的的预测能力，反映动压大变形规律的变化，捕捉预板灾害的预兆信息，满足安全生产的需求。

英文摘要：

To reduce the roof accidents of deep roadway in mine, the dynamic pressure law of coal seam roadway in deep mining was predicted and studied by a multi-model and soft-sensing prediction method based on EEMD-SVM- DS-ARIMA. Firstly, the mode decompositions of non-linear and non-stationary monitoring signals were carried out by EEMD, and several mode function sequences （IFM） were obtained. Secondly, each component of IFM was trained by SVM, and each sample output function was received after reconstitution. Thirdly, several distribution functions of evidence probability were obtained by the synthetic rules of DS, which were viewed as weight factors to merge the sub-function output. And then the output of multiple functions was acquired. Finally, the composite sequence was dynamically checked by ARIMA models. The practical application showed that this method can improve the predictive ability on roof pressure, reflect the large deformation law of dynamic pressure, capture the omen information of roof disaster and meet the needs of production safety.