BP神经网络在煤级划分中的应用

The Application of BP Neural Network in the Coalification Stages Division

煤化过程的复杂性和煤级指标的多元性,导致不同煤化阶段中的各煤级指标显著不同,煤级划分比较困难。利用BP神经网络的非线性映射特性,以典型的煤样主要煤级指标做训练样本,对网络进行训练,对其中各煤化阶段指标的权值和神经元内部的阈值沿误差下降的方向不断修正,最终达到了精度要求。经检验,训练后的网络对煤化阶段划分效果较好,对实际工作中具有一定参考价值和指导意义。

Because of the complexity of the coalification process and the diversity of the coal - level indicators, it was induced that re- sulting in different coalification stages the coal - level indicators were significantly different and coal - division was rather difficult. In this paper, based on Non - linear mapping of the BP neural network, with the coal - level indicators of typical coal samples as the training samples, the network was trained, in which the weights of all the coal - level indicators and the thresholds in the neurons were modified all the time in the direction of the error decrease, until the required precision was reached. The conclusion was that after tested, the trained network did better in classifying the coalification stages and this method of classifying the coalification stages has some reference value and guiding significance in the practical work.