Data fusion of near infrared, Fourier Transform infrared and Raman spectroscopy for quantifying the conversion of Poly Alpha Oil (PAO)

The conversion of PAO is a key parameter in the production of lubricating oils, which can be analyzed by NIR (near infrared), FT-IR (Fourier Transform infrared) and Raman spectroscopy respectively, in combination with chemometrics methods. In order to improve the prediction accuracy, NPLS (N-way partial least squares) fusion strategy was proposed in this paper and utilized in this study. In addition, traditional data fusion strategies such as low-level, mid-level, high-level data fusion method as well as SO-PLS (sequential orthogonalized partial least squares) fusion that has recently been proposed were also carried out in this study. Comparisons were conducted between models established based on fusion methods and individual spectroscopy. The results indicated that NPLS fusion method was more efficient than other fusion strategies which can be used to improve the model performance and robustness significantly.