模糊神经网络在开采沉陷预测中的应用

Application of FNN to Mining Subsidence Prediction

用模糊数学与神经网络结合的方法研究开采沉陷的预测。与单一神经网络的预计结果相比 ,模糊神经网络方法效果更佳 ,克服了单一神经网络易陷入局部最优和收敛速度慢等缺点 ,不用对原始数据和地质条件作相应假设 ,预测结果完全建立在反映客观原型的基础之上。研究结果仅仅针对静态的开采沉陷的预测 ,没有考虑地表水平移动和变形 ,也没有考虑工作面的推进速度等动态因素对开采沉陷的影响 ,这是今后应继续研究的问题。

The mining subsidence prediction with combination of fuzzy inference and neural networks is investigated. Compared to the artificial neural works, fuzzy neural networks is more effective method, the disadvantages of the artificial neural works such as limitation of local optimal and slow convergence rate are overcome, and the subsidence data predicted is completely based on the objective prototype without hypothesis on primary data and geological condition. The results are the prediction only for the static mining subsidence, the horizontal displacement and distortion of the ground are not included in the prediction, and the effect of the dynamic factors such as rate of stoping face on the prediction are not considered.