PCA-BP模型在判别基于LIF技术煤矿突水水源的应用

防治煤矿突水时需迅速精准地判别突水水源，激光诱导荧光(LIF)光谱技术具有灵敏度高、快速准确监测特点，为检测突水水源提供了一种新的方法。该研究引入该技术以获取突水荧光光谱数据。采用卷积(SG)平滑和多元散射校正(MSC)方法对光谱图进行预处理，以消除光谱采集过程中噪声干扰。采用主成分分析(PCA)方法提取特征信息，针对SG预处理后的数据，当主成分个数为3时，累积贡献率可达到99.76%，已基本保留原数据的全信息。选择3层结构BP神经网络建立分类判别模型，通过不同方式构造训练集和测试集，SG预处理数据构建的分类模型可以达到精准判别，而对于MSC预处理和原始数据出现很少的误判。实验结果表明SG预处理结果要优于MSC预处理。研究结果表明，将PCA和BP神经网络结合建立分类模型，能有效判别煤矿突水水源，且具有较强的自组织、自学习能力。

Identification of Coalmine Water Inrush Source with PCA-BP Model Based on Laser-Induced Fluorescence Technology

The water inrush should been rapidly and accurately identified during preventing coalmine water inrush .The laser induced fluorescent (LIF) spectrum technology provides a new method to identify water inrush with the characteristics of high sensitivity,quick and accurate monitoring .In order to identify water inrush,t his paper introduces the spectrum technology of LIF to obtain water inrush fluor escence spectra data .The spectral preprocessing methods of Savitzky-Golay(SG) and Multiplicative Scatter Correction (MSC) have been used to eliminate noise spectra in collecting process .Principal component analysis (PCA) extracts feature information,for SG reprocessing data,when the number of principal component is 3,the cumulative contribution rate can reach 9976 percent .This method has largely retained the information of original data .This paper chooses the classification model with 3 layers BP neural network,constructing by different training and testing sets .The classification model with SG preprocessing has achieved accurate identification,however,appeared few false identification for MSC and original data .The result shows that SG preprocessing is better than MSC .Research results show that the classification model with PCA and BP neural network ca n effectively identify coalmine water inrush,and have the strong self-organizi ng,self-learning ability .