Application Of Kalman Filter In Soft Sensor Modeling

Soft-sensing technology is an important method to solve the real-time estimation ofunmeasured variables in the field of process control and process detects. With the increasinglycomplex of the industrial process, the soft-sensor model based on data driven is often difficultto meet the accuracy of the complex system which is nonlinear, multivariable andtime-varying. Mechanism modeling method based on the analysis of reaction process candescribe the process properties of the object better, while the simplified mechanism model isstill difficult to obtain satisfactory results. A practical industrial process is taken as abackground in this paper, and mechanism modeling combined with hybrid modeling correctedby field data correction can significantly improve the soft measurement model estimationaccuracy and generalization performance.The cation exchange resin is used as the catalyst in the Bisphenol A production. Onlinedetection of the catalytic activity has an important impact on the BPA production. On the basisof studying the mechanistic of catalyst deactivation, this paper builds a mechanism model todescribe the catalytic activity, and the state equation and observation equation of catalystdeactivation is derived through simplification and derivation of the model, and then usesUnscented Kalman Filtering to estimate the catalyst activity. Simulation results confirm theaccuracy of the model.Simplifications and assumptions in modeling process unavoidably result in somedisadvantages. First, a certain error between the model and the actual process is generated.Second, it will aggravate the estimation error of this model. In order to improve the accuracyand generalization ability of the model, a hybrid model of catalyst deactivation is constructedand the observation equation of catalyst deactivation is reconstructed.The reactor outlet Bisphenol A is a direct response to the production quality index, andonline soft measurement of Bisphenol A content is very necessary. Based on the mechanismof Bisphenol A condensation process, the mechanism equations describing the dynamicbehavior of reactor are established, and the state equation and observation equation ofBisphenol A derived through simplification and derivation of the soft measurement model,and then uses Unscented Kalman Filtering to estimate the Bisphenol A content. Simulationresults confirm that the method is feasible and effective.