基于神经网络和Dempster-Shafter信息融合的煤岩界面预测

针对煤岩界面识别精度无法满足采煤机自动调度的情况，提出采用神经网络融合工作面的三边信息，使用D－S证据理论再将此信息和不断获得的煤岩界面识别信息进行二次融合，从而实现在线融合和在线预测，不断提高预测精度，仿真结果显示：该方法不仅对地质条件好的工作面有效，而且对断层也有一定的适应性；同时，具有优良的容错性。

Forecast of coal-rock interface based on neural network and dempster-shafter theory

A method of improving the accuracy of forecast coal-rock interface is presented,the method is that fusion the information of gateway of coal seam based on neural network firstly, the result of fusion and that of identification are fused based on Dempster-Shafter in a sample period, then, the coal-rock seam is forecast based on the second fusion in the same sample period. The accuracy of forecast is enough for horizon control of shear. The result of simulation shows that the method is not only fit for continue interface but also interface with fault and the method is robust to error of data. This is confirmed by the result of simulation and experiment.