Soft sensor modeling method for freezing point of diesel fuel based on PCA and LS-SVM

To solve the problems of real-time on-line measurements of freezing point of diesel fuel, a novel method of soft sensor with near-infrared (NIR) spectrometry was proposed based on the integration of both principal component analysis (PCA) and least squares support vector machines (LS-SVM). In this method, the PCA was incorporated into the model, which not only solved the linear correlation of the input, but also simplified the LS-SVM structure and improved the training speed. Then, the soft sensor model for freezing point was established using LS-SVM regression algorithm. The model performance has been tested and the results show that the propose method is superior to the soft sensor model based on BP neural network or PCA+SVM. So it can satisfy the demand of the on-lines measurement of freezing point.