A Data-Driven Modeling and Control Scheme Design Methodology for a Class of SISO Industrial Processes

In order to solve the problem of difficult mathematical modeling of complex industrial processes, this thesis proposes a data-driven modeling method that performs dynamic linearization near the working point. Since all real industrial processes have a certain working range, they can be regarded as linear objects containing unmodeled dynamics within this range. Therefore, a dynamic modeling algorithm based on incremental least squares is designed in this thesis. This algorithm is able to compute a linearized model of the dynamics in the vicinity of the operating point based on actual production data. Considering the controller implementation problems in the industrial field, the model is computed up to the 2nd order, i.e., the higher-order characteristics are also regarded as part of the unmodeled dynamics. Finally, on the basis of the dynamic model, Smith predictor is used to simulate the control. The simulation results show that control based on the model derived from this algorithm results in better dynamic and steady state characteristics.