| Lubricant oil is also known as engine oil, is vividly called as the "blood" of automobile engine. Lubricant oil with good quality is important to extend the work life of engine and improve eninge's efficience. Nowadays, there are numerous brands of lubricants with different qualities. It is not easy to tell that they are good or inferior. The quality of lubricant oil determines the time to change the lubricant oil for automible engine and the period time for maintenance and repairmen of engine. However, common used methods to distinguish the brands and to determine the qualities of lubricant oil are usually based on artificial experience which is not very accurate or chemical analysis which needs numerous manpower and materials.In this paper, visible and near infrared (Vis-NIR) spectroscopy was applied to non-invasively distinguish the brands of lubricant oil and determine the internal qualities which include viscosity, water content and acid value, against disadvantages of common used lubricant oil quality determination techniques. Relationships were analyzed between visible and near infrared spectra and brands and qualities of lubricant oil, and were used to establish the quantitative prediction model. The final purpose of this study was to achieve the fast, accurate and non-invasive discrimination of the brands and determination of the qualities of lubricant oil.This paper focused on the development of new method for the fast brand discrimination and quality determination of lubricant oil based on Vis-NIR spectroscopy. The main research contents and results were shown as follow:(1) Typical and common brands of lubricant oil were chose as the experiment samples. The relationship between the spectra and brands was analyzed. The most efficient spectral wavelengths were chose to establish the quantitative model based on the relationship of spectra and brands. The results showed the feasibility of accurate, fast and non-invasive discrimination of different brands of lubricant oil which gave a new method for the brand discrimination of lubricant oil. (2) Based on the brand discrimination analysis, the relationship between the oil qualities of viscosity, water content and acid value and the oil spectra was analyzed. The effective wavelengths were obtaind to improve the performances of the quantitative quality models. Finally a new fast and accurate lubricant oil quality determination method was proposed.(3) Vis-NIR spectroscopy technique was applied for the fast brand disrcrimination of lubricant oil in two experiments. Firstly, principal component analysis (PCA) and multiclass discrimination analysis were combined to establish lubricant oild brand discrimination model. The model obtained a good result of 100% discrimination rate on the prediction samples. It was shown that Vis-NIR spectroscopy technique could fast and accurate distinguish different brands of lubricant oils. Secondly, a fast method based on successive projections algorithm (SPA) for the brands of lubricants discrimination was developed using Vis-NIR spectroscopy. SPA was used to select the effective Vis-NIR spectra data, and combined with partial least squares (PLS) to build models for brands discrimination of lubricants.360 lubricants samples were selected randomly as 240 (40 for each brands) for calibration set, and the left 120 (20 for each brands) for perdition set. The correlation coefficient (r), root mean square error of prediction (RMSEP) and bias in prediction set were 0.9721,0.4055 and -0.0145, and the discrimination rate was reached at 91.7%. It indicated that the result for discrimination brands of machine oil was very good based on SPA, and it offers a new approach to the fast discrimination of lubricants brands.(4) Vis-NIR spectroscopy was investigated for the fast and nondestructive determination of viscosity of lubricating oil. A total of 150 oil samples were scanned, and different calibration models were developed with the pretreatment of smoothing and standard normal variate. The input variables of calibration were the principal component selected by PCA and characteristic wavelengths selected by SPA. The the calibration model were developed using PLS, multiple linear regression (MLR) and back propagation neural networks (BPNN). The results indicated that PCA-BPNN and SPA-BPNN models were better than the linear models (PCA-PLS, PCA-MLR, SPA-PLS and SPA-MLR). The correlation coefficients were 0.971 for PCA-BPNN and 0.964 for SPA-BPNN. This demonstrated that BPNN could make good use of the nonlinear information in spectral data, and SPA was a powerful way for the selection of characteristic wavelengths. The selected wavelengths were helpful for the development of portable lubricating oil viscosity detection instrument.(5) Vis-NIR spectroscopy was applied to non-invasively measure water content in engine lubricant. Based on measured spectra, several spectral calibration algorithms were adopted to improve accuracy and simply calculation. PCA and SPA were separately used to reduce variables of spectral model. Nine effective variables,476, 483,544,925,933,938,952,970 and 974 nm, were selected by SPA, and were inputted into PLS and MLR models. Both two models obtained better results than full-spectra-PLS model and PCA-PLS model. It shows that SPA does not select uninformative but effective variables from full-spectrum. Least-square support vector machine (LS-SVM) was operated to improve Vis-NIR spectroscopy's ability based on full-spectrum and SPA, separately. High coefficients of determination for prediction set (rp2) up to 0.9 were obtained by both full-spectrum-LS-SVM and SPA-LS-SVM models. SPA-LS-SVM is better than full-spectrum-LS-SVM. rp2 of SPA-LS-SVM is 0.983 and residual predictive deviation (RPD) is 6.963. It is concluded that Vis-NIR spectroscopy can be used in the non-invasive measurement of water content in engine lubricant, and SPA is a feasible and efficient algorithm for the spectral variable selection.(6) Vis-NIR spectroscopy was applied to non-invasively measure acid value in lubricant. The spectra of 575-975 nm were measured. LS-SVM was used to establish the Vis-NIR spectral model. Uninformative variable elimination (UVE) and successive projections algorithm were combined to select wavelength from Vis-NIR spectroscopy. Eight wavelength variables, namely 489,553,591,874,893,910,935 and 951nm, were selected. The UVE-SPA-LS-SVM model was established based on these eight wavelength variables, and obtained the results of the coefficient of determination for prediction set of 0.9546, the root mean square error for prediction of 0.0081, and residual predictive deviation of 4.5663. It was concluded that Vis-NIR spectroscopy could be used to non-destructively measure the acid value of lubricant and UVE-SPA was a feasible and efficient algorithm for the spectral variable selection. |
PDF Full Text Request