Abstract: During the research and application processes of the soft sensing technology for pulp Kappa number‚the pre-dicting precision decrease of the soft sensing method under complex conditions based directly on experiential models was observed．To overcome this problem‚a new soft sensing method was introduced．In this method‚the characteristic in-formation of temperature-rising process extracted by Daubechies wavelet transform was used to classify the producing con-ditions‚and a classified prediction model of Kappa number was then proposed．It is proved by the analysis of130sample data from a certain pulping workshop that‚under complex working conditions and comparing with the results by the meth-ods based directly on experiential models‚better prediction results would be obtained by the proposed method．

Key words: soft sensing, Kappa number, wavelet transform, producing condition classification