本文基于SIMCA和PLS-DA这两类方法, 对激光荧光技术用于山西某煤矿突水水源识别的应用进行了研究, 分别采集老窑水、冲积层水、砂岩水三种水样荧光光谱数据, 对数据进行了压缩, 采用Moving-Average、Craussiaa-Filter算法进行光谱预处理。对建立的PLS-DA模型、SIMCA模型可行性进行验证。在SIMCA模型中, 三种水样模型对水样样本识别正确率均为100％, 三种水样模型的验证集水样样本识别正确率均为100％。在PLS-DA模型中, 各模型建模集识别正确率均为100％, 相关系数r分别为0.998、0.992、0.988, RMSF-LW分别为0.038、0.066、0.070; 验证集识别正确率均为100％, RMSEP分别为0.063、0.094、0.152, 三种水样模型的验证集样品预测值均在1附近。在不同光谱预处理方法分析下, SIMCA、PLS-DA建模均对水源进行快速识别。

In this paper, based on the two kinds of methods of SIMCA and PLS-DA, the application of laser fluorescence technology in Water Inrush Source Identification of a coal mine in Shanxi Province is studied. The fluorescence spectrum data of three kinds of water samples, i.e. old kiln water, alluvium water and sandstone water, are collected respectively, and the data are compressed. The moving average and craussiaa filter algorithms are used for spectrum preprocessing. Verify the feasibility of PLS-DA model and SIMCA model. In the SIMCA model, the recognition accuracy of three water sample models is 100%, and the recognition accuracy of three water sample models is 100%. In PLS-DA model, the recognition accuracy of each model set is 100%, the correlation coefficient r is 0.998, 0.992, 0.988, RMSF-LW is 0.038, 0.066, 0.070, respectively; the recognition accuracy of the validation set is 100%, the RMSEP is 0.063, 0.094, 0.152, respectively, and the prediction values of the validation set samples of the three water sample models are all around 1. Under the analysis of different spectral pretreatment methods, SIMCA and PLS-DA models can identify the water source quickly.