基于正交信号校正的喷气燃料近红外光谱去噪研究  

Study on denoising near infrared spectra of jet fuel on orthogonal signal correction

作　　者：李华[1] 王菊香[2] 邢志娜[2] 申刚[2] 

机构地区：[1]海军航空工程学院飞行器工程系研究生管理大队,山东烟台264001 [2]海军航空工程学院飞行器工程系,山东烟台264001

出　　处：《计算机与应用化学》2011年第1期99-102,共4页Computers and Applied Chemistry

摘　　要：为了提高喷气燃料近红外光谱模型的预测精度和稳健性,结合近红外光谱的特点,将正交信号校正法(OSC)用于喷气燃料近红外光谱的预处理。在正交信号校正过程中,通过K矩阵法建模选择正交信号校正的最佳主成分数。将正交信号校正后的光谱分别与K矩阵法、主成分回归(PCR)和偏最小二乘法(PLS)结合建立校正模型,对喷气燃料的密度和20℃运动粘度的进行预测分析。结果表明使用正交信号校正对光谱进行预处理后,简单K矩阵法校正模型可达到与PCA、PLS模型一致的预测效果。本文还比较了正交信号校正和平滑-导数预处理方法对校正模型的影响,比较结果表明,正交信号校正后的光谱提高了模型的精度,降低了PLS模型的主成分数。In order to improve the prediction and the robustness of the near infrared spectroscopy（NIR） calibration models of jet fuel, the orthogonal signal correction（OSC） was used to preprocess the near infrared spectroseopy（NIR） of jet fuel. in this study, the numbers of principal component of OSC was chose by K-Matrix. The preprocessed spectrum were subjected to the multivariate calibrations of K-Matrix, principle component regression（PCR） and partial least squares（PLS） to predict the density and the kinematic viscosity（20℃） of jet fuel, the results show that the precision of K-Matrix model was the same as the PCR model and PLS model. The OSC-PLS model was compared with what combined with smooth-derivative and the OSC was proved to have the ability of improving the prediction and reducing the number of principal component for building model.

关 键 词：近红外 正交信号校正 K矩阵法 航空煤油 

分 类 号：O657.33[理学—分析化学]