Bayesian dynamic linear model-based modeling method for an aniline stripping column

Abstract

The step test method is a common modeling approach for industrial process and has the disadvantages of time consuming and production disruption. One of the major features of the Bayesian methodology is the ability of integrating the subjective intervention information into existing Bayesian dynamic linear model (DLM) and sometimes the subjective information is also important for a MPC controller design. This paper takes the bottom temperature control problem of an aniline stripping column as the research object, and the model calculated by the Bayesian DLM and linear fitness is adopted to design a MPC controller. Finally, the practical application result shows that the model based on the Bayesian DLM is useful and effective, and by using this modeling method, the disruptive and expensive step test may be reduced.