A novel method which is a combination of wavelet packet transform (WPT),\r\nuninformative variable elimination by partial least squares (UVE-PLS) and simulated\r\nannealing (SA) to extract best variance information among different varieties of lubricants\r\nis presented. A total of 180 samples (60 for each variety) were characterized on the basis of\r\nvisible and short-wave infrared spectroscopy (VIS-SWNIR), and 90 samples (30 for each\r\nvariety) were randomly selected for the calibration set, whereas, the remaining 90 samples\r\n(30 for each variety) were used for the validation set. The spectral data was split into\r\ndifferent frequency bands by WPT, and different frequency bands were obtained. SA was\r\nemployed to look for the best variance band (BVB) among different varieties of lubricants.\r\nIn order to improve prediction precision further, BVB was processed by UVE-PLS and the\r\noptimal cutoff threshold of UVE was found by SA. Finally, five variables were mined, and\r\nwere set as inputs for a least square-support vector machine (LS-SVM) to build the\r\nrecognition model. An optimal model with a correlation coefficient (R) of 0.9850 and root\r\nmean square error of prediction (RMSEP) of 0.0827 was obtained. The overall results\r\nindicated that the method of combining WPT, UVE-PLS and SA was a powerful way to\r\nselect diagnostic information for discrimination among different varieties of lubricating oil,\r\nfurthermore, a more parsimonious and efficient LS-SVM model could be obtained
Loading....