CAJ | 학술논문

　　在总结全国各典型煤矿断层防水煤柱相关资料的基础上，以水头压力、煤层厚度、安全系数、煤的抗张强度为主要影响因子，选择有代表性的样本数据，通过Matlab软件构建了BP和RBF神经网络模型，对各煤矿断层防水煤柱的留设宽度进行了预测，并与规程经验公式计算的结果进行了对比。结果显示，在煤矿断层防水煤柱留设宽度预测中，RBF 神经网络比 BP 神经网络的训练速度更快，效率更高，具有更加广阔的应用前景。
　　재총결전국각전형매광단층방수매주상관자료적기출상，이수두압력、매층후도、안전계수、매적항장강도위주요영향인자，선택유대표성적양본수거，통과Matlab연건구건료BP화RBF신경망락모형，대각매광단층방수매주적류설관도진행료예측，병여규정경험공식계산적결과진행료대비。결과현시，재매광단층방수매주류설관도예측중，RBF 신경망락비 BP 신경망락적훈련속도경쾌，효솔경고，구유경가엄활적응용전경。
On the basis of summing up the relevant information of the safety pillar against water-inrush from fault in typical coal mines throughout the country, taking head pressure, coal seam thickness, safe coefficient, coal ten-sile strength as the main influence factors, choosing the representative sample data, we built the BP and RBF neural network model through the Matlab software, with the model we forecasted the designed width of the water-proof coal pillar in each coal mine. Compared with the calculated results of procedure empirical formula, on the research of the design about water-proof pillar from coalfield, we found that it was faster for RBF neural networks in train-ing speed than the BP neural networks, furthermore, RBF is much more efficient than BP and can get broader ap-plication prospects.