Treatment methods of abnormality in FIR model identification

This article mainly focuses on two treatment methods in FIR model identification for the abnormality in measured data set. One is called linear interpolation method(LIM), whose essence is to rebuild the data set according to linear interpolation after indicating the abnormal data. The other is the method of identification based on segments of data(ISDM). The idea is to remove the abnormal data and divide the original data set into two or more inconsecutive data sets, then perform model identification using those data sets respectively, finally merge the results with different weighted means. The guidelines of the proposed methods are enumerated. The two methods are illustrated with FIR model identification, and simulations with the Shell heavy oil fractionator model verify the feasibility and effectiveness.