Near Infrared Spectroscopy Study Of Trace Moisture In Oil

Near Infrared Reflectance Spectroscopy as a analytical techniques of rapid,non-destructive, safe and environmental protection, has a wide range ofapplications and development prospects in the field of petrochemical. Thedetection of moisture in the oil is an important indicator of the oil the basis ofperformance detections, the presence of water pollutants in the oil will speed upall the performance deterioration of the oil, reduce the useful life of equipments,pose a serious threat to the normal operation of equipments. Due to complicatedto operate, time-consuming, seriously subject to environmental impact, sideeffects, reagent toxic, such as reasons, conventional oil moisture detectionmethod can not meet most user requirements the rapid detection of oil waterdemand. This topic is based on this situation, a quantitative analysis method byusing of modern near-infrared spectroscopy, combining with the stoichiometryand computer technology to detect moisture content in the oil, aims to oilmoisture detection for online detection of a rapid non-destructive quantitativeanalysis method.This article first describes the characteristics of the near-infrared spectroscopy andtheoretical foundation, including the near-infrared spectroscopy technology works, thePrincipia Mathematica, and quantitative analysis of the basic laws. It also describescommonly used pretreatment methods and the optimization of the spectral informationof Near Infrared Spectroscopy, using configuration modeling required to injectablemicrospheres of water in the diesel oil, by using ultrasound to make oil and waterquickly to achieve a homogeneous, then the use of nearly infrared Spectroscopy nearInfrared Spectroscopy. Pretreatment of the raw spectral data, the main use of amicro-order derivative algorithm to remove the baseline drift of the spectral data, Thenuse the Ensemble Empirical Mode Decomposition (EEMD) to remove the noisespectrum, the original spectral data are decomposed into nine intrinsic mode function(IMF) by the Ensemble Empirical Mode Decomposition. After the study concludedthat: in the case of removing the first layer and the eighth layer of intrinsic modefunctions, the noise reduction processing of the raw spectral data is the best. Making usethe software of Matlab, using the method of Least Squares Support Vector Machine(LS-SVM), establish a prediction model for the trace water in diesel Studies have shown that the original spectrum after decomposition by EEMD can achieve very goodprediction results; Making use the software of MATLAB, using the method ofRelevance Vector Machines (RVM), establish a prediction model for the trace water indiesel Studies have shown that. The modeling effect of RVM is better than theLS-SVM, the model be built by RVM and EEMD decomposition will be obtain thebest prediction results.