The Lubricant Quality Near-infrared And Raman Spectroscopy Testing Methods

Lubricant oil is well known as the ‘blood’ of automobile engine, it relates to the healthstatus of the engine. The suitable lubricants not only be able to guarantee the normal operatingstate and can also extend its life and performance. Nowadays, there are many brands oflubricants with different qualities. It is difficulty to distinguish between good and inferiorfrom the outside. The quality of lubricant is a reference index of whether to replace the oil onthe care and maintenance. The traditional way of test the brand and quality of the lubricant oilis rely on experience or chemical instruments to chemical composition analysis, this processis not only cost time but also may cause pollution. So it is necessary to study a fast,non-destructive, economy method to detecting the quality of the lubricant oil.In this paper, near infrared spectroscopy and Raman spectroscopy are applied to studythe feasibility of lubricant brand and the quality indicators （viscosity, water content）, analyzethe relationship between them, and establishment the quantitative qualitative analysis of themodel, in order to achieve a accurate, rapid and non-destructive way to detection thelubricating oil by spectroscopy. In this paper, the main research content and results wereshown as follows:（1） Study the quantitative determination of lubricant water content and viscosity by NIR.Using interval partial least-squares （iPLS） algorithm to model input variables to band selection. The resultswere that the wave bands7432-6321cm^-1and5583-4846cm^-1were considered thecharacteristic band of measure the water content of lubricants by NIR. The correlationcoefficient（R） achieve0.993and low RMSECV6.4×10^-4. The wave bands8528cm^-1-8094cm^-1、7371cm^-1-6937cm^-1、6069cm^-1-5346cm^-1and5201cm^-1-4478cm^-1were considered the characteristicband of measure the oil viscosity of lubricants by NIR. The correlation coefficient（R）achieves0.994and low RMSECV4.252and achieved better forecast results.（2） Study the qualitative identification of lubricant brand by NIR. Combined with partialleast squares discriminant analysis （PLS-DA） method to identify the seven kinds oflubricating oil （ACDelco, HELIX, Dazhong, JinBaoLi, Kunlun TianGe, Mobil and ShanghaiTongyong）. The results showed that seven kinds lubricant oil were identified with PLS-DAmodels by NIR spectra, the identification rates were all100%. Screening for variables with genetic algorithm, the identification rates of the model for the unknown samples were100%,Greatly reduce the amount of calculation for modeling.（3） Study the adulteration of lubricant brand by NIR. Screening the characteristic variableof original spectra by two methods iPLS and GA. The wave bands8466cm^-1-8285cm^-1、7378cm^-1-6653cm^-1、6110cm^-1-5747cm^-1、5566cm^-1-4478cm^-1and69variable points wereselected through the iPLS and GA algorithm respectively. Through the PLS modeling resultscontrast, we can find the characteristics of iPLS selected band can replace the full spectrum ofmodeling and were batter than GA.（4） Study the qualitative identification of lubricant brand by Raman. Using the iPLS bandscreening method to the original spectrum characteristic variable screening, and the1344cm^-1-1457.5cm^-1and the1571cm^-1-1684.5cm^-1were selected as the characteristic band ofthe Raman qualitative identification oil brand. Two characteristic bands were used as the inputvariables of the PLS-DA discrimination method for the qualitative analysis of six lubricants.In addition to the correct rate were99.17%of Mobil lubricants, and the remaining five（ACDelco, HELIX, JinBaoLi, Kunlun TianGe, Mobil and Shanghai Tongyong） are up100%.（5） Study the adulteration of lubricant brand by Raman. The iPLS and GA method wereused to reduce the input variables of calibration model. The correlation coefficient（R）achieves over0.972and low RMSECV0.0277and achieved better forecast results. Screeningfor variables with iPLS and GA, selecting the1148-1484cm^-1wave-band and the185variable points separately. The Mobil lubricants adulterated Kunlun Tiange lubricants wereused as validation samples to detect the selected wave-band and achieve good results.（6） Study the method of kinematic viscosity of lubricant by Raman. Screening thecharacteristic variable of original spectra by two methods iPLS and GA. The model that theinput variables were screened by iPLS is batter than GA. Model correlation coefficient Rp0.963, and the root mean square error of prediction set RMSEP for10.1078.This paper shows that NIR and Raman combined with PLS modeling algorithm to detectthe quality of lubrication oil （moistutre content, viscosity）, adulteration and brandidentification is feasible. Using the spectral band screening methods can greatly reduce theinput variables of modeling and the computational process; improve the robustness andaccuracy of the model.