基于近红外光谱的煤粉样品定量检测研究

针对煤质快速在线检测的需求,采用傅里叶变换近红外光谱结合不同的光谱预处理方法,即平滑处理方法、微分方法、多元散射校正方法、标准归一化处理方法分别建立了煤粉样品的水分、灰分和挥发分的偏最小二乘模型,并对模型的检测结果进行了十字交叉验证。结果表明,基于25点平滑处理方法建立的水分偏最小二乘模型较优,基于标准归一化处理方法建立的灰分偏最小二乘模型最佳,基于5点平滑处理方法建立的挥发分偏最小二乘模型精度最高,验证了应用傅里叶变换近红外光谱技术定量分析煤粉指标的可行性。

Research of quantitative detection of coal samples based on near infrared spectrum

In view of requirement of quick and online detection for coal quality,the paper uses Fourier transform near infrared spectrum to separately build partial least square models of water,ash and volatile combining with different spectrum preprocessing methods,namely smooth processing method,differential method,multiplicative signal correction method and standard normal variate method,and makes decussation verification for detecting result of the models.The result shows that the partial least square model of water built by 25 points smooth processing method is better,the partial least square model of ash built by standard normal variate method is the best,precision of the partial least square model of volatile built by 5 points smooth processing method is the highest,which validates feasibility of applying Fourier transform near infrared spectrum technology to analyze coal indexes.