The improvement of MPO and LS-SVM algorithm in soft measurement model of NIS

摘要

Aiming at the shortage of accuracy and authenticity on data corrections in soft measurement Near Infrared Spectrometer, in order to overcome the traditional''s disadvantages of “over-fitting” and “owe-fitting” at modelling and predicting, improved the algorithm of MPO and LS-SVM separately in the paper, adopt a complex correction algorithm based on MPSO and VWLS-SVM. When Near Infrared Spectrometer multiple correction model is built, MPSO algorithm is adopted to search the optimal sample weight. In the other, VWLS-SVM algorithm can using more wavelength variables. Finally, the model is built. The analysis shows that the algorithm has a good astringency and ergodicity. In the other, the algorithm is helpful to keep multichannel advantage and stability, and to obtain more precise and truly correction model.