【摘要】 针对煤矿井下近红外煤岩识别中所存在的粉尘问题,采用无烟煤与抑爆剂9∶1混合的混合物模拟煤矿井下粉尘环境,构建了粉尘环境煤岩光谱识别实验装置。为了研究粉尘环境对典型煤岩近红外光谱的影响,从全国各地收集了页岩、砂岩、灰岩3类岩石样本及无烟煤、烟煤、褐煤3类煤类样本的原位典型煤岩试样23个,采集无粉尘情况下的23个煤岩样本表面近红外波段1000-2500nm的反射光谱作为实验标准数据库,分别从实验标准样本库中3类典型煤样本与3类典型岩样本中随机选择1个样本作为实验样本,分别采集测试样本在600, 1 000, 1 500和3 000 mg·m^-3粉尘浓度下的近红外波段的反射光谱数据,结果显示:粉尘的加入导致1 000～1 200 nm波段与2 400～2 500 nm波段的光谱图像信噪比降低;随着粉尘浓度的增加,粉尘中的无烟煤的不透明物质使得实验样本中的特征吸收谷减弱;采用光谱角度匹配SAM以及皮尔逊相关系数对试样和标准样本库进行相关性分析,无烟煤类样本、烟煤类样本、砂岩类样本、灰岩类样本在光谱角度匹配SAM匹配模型下有着较高的匹配度,匹配度在各个粉尘浓度下均处于0.9以上;相关系数匹配模型匹配度受粉尘的影响剧烈,平均相关系数为0.73;实验标准数据库及实验样本经SG卷积和SNV标准正态预处理后,预处理后的样本数据库与实验样本光谱角度匹配SAM匹配模型匹配度无明显变化,相关系数匹配模型匹配度显著提升,平均相关系数为0.78;除褐煤2号外,所有的样本光谱相关系数平均提升0.13,无烟煤2号样本各个浓度平均相关系数提升76.3%,而样本12褐煤2号的光谱相关系数经光谱预处理降低。建立光谱角度匹配SAM以及皮尔逊相关系数煤岩识别模型,二值化煤岩样本,煤为"0"岩为"1",通过两种识别模型对不同浓度下的6个实验样本进行煤岩识别,光谱角度匹配SAM的识别准确率P为100%,识别时间为8 ms,皮尔逊相关系数的识别准确率P为87.5%,识别时间为852 ms。更多还原

【Abstract】 In order to study the dust problem existing in the identification of near-infrared coal and rock in the underground coal mine, the mixture of anthracite and the anti-explosive agent was used to simulate the dust environment of coal mine underground so that the experimental device for spectral identification of coal and rock in dust environment was constructed. 23 samples of in-situ typical coal samples of shale, sandstone and limestone samples and anthracite, bituminous coal and lignite coal samples were collected from all over the country. 23 coal and stone samples reflection spectrum collecting without the dust of the near-infrared band（1 000～2 500 nm） was used as an experimental standard database. 1 sample was randomly selected from the three typical coal samples and three typical rock samples in the experimental standard the sample library as experimental samples, and the test samples’ reflectance spectra of the near-infrared bands at 600, 1 000, 1 500 and 3 000 mg·m^-3 dust concentrations were collected. The results showed that the addition of dust led to a decrease in the signal-to-noise ratio of the spectral image between 1 000～1 200 and 2 400～2 500 nm; With the increase of dust concentration, The opaque substance of anthracite in the dust made the characteristic absorption valley in the experimental sample weaken; Correlation analysis between sample and standard sample library was carried out by spectral angle matching SAM and Pearson correlation coefficient. Anthracite samples, bituminous coal samples, sandstone samples and limestone samples had a high matching degree under SAM matching model. The cosine angle was above 0.9 at each dust concentration; Correlation coefficient matching model matching degree was strongly affected by dust, and the average correlation coefficient was 0.73; After the experimental standard database and special envoy samples were normal preprocessed by SG convolution and SNV standard, the matching degree of SAM matching model did not change significantly. Correlation coefficient matching model matching degree was significantly improved, the average correlation coefficient was 0. 78; The correlation coefficient matching model excepted for lignite No.2, the spectral correlation coefficient of all samples increased by 0.13, anthracite No.2 The sample correlation coefficient increased by 76.3%, while the spectral correlation coefficient of sample 12 lignite No.2 was reduced by spectral pretreatment. The spectral angle matching SAM and Pearson correlation coefficient coal and rock identification model was established. The two models were used to identify coal samples under different concentrations and binarized coal rock sample, the coals were "0", and the rocks were "1". Coal rock identification was performed on 6 experimental samples at different concentrations by two recognition models. The identification accuracy of SAM was 100%, and the recognition time was 8 ms. Pearson correlation recognition accuracy of the coefficient P was 87.5%, and the recognition time was 852 ms.更多还原