Improved Robust Predictive Control For Uncertain Multi-phase Batch Process

Batch production process,as an indispensable production mode in modern industrial process,is favored because of its characteristics of multiple varieties,small batch and high added value,so it enjoys the significant position in modern industrial production.With the abundance of contemporary developed material,people's requirements for products are increasingly improved.The complexity of technology and processes for batch processes is also increasing.The factors like uncertainties and external disturbances in the system operation have become obstacles in the development of batch process.Addressing such issues to improve system performance becomes critical now.Therefore,it is imperative and of great significance to study effective advanced control methods to deal with issues like uncertainties.With the model predictive control presented in the end of 20 th century,it was found that rolling optimization and feedback correction in predictive control could predict and optimize the future behavior,and the control effect was better than the traditional global optimal control.How to apply its advantages to batch processes and optimize the performance of production process has become the focus of research in this field.In this paper,the uncertainty problem in multi-phase batch process is studied by using the predictive control method under the one-dimensional theoretical framework.The main research contents are as follows.1.An improved min-max linear quadratic tracking control method is proposed aiming at uncertain multi-phase batch processes.Firstly,the multi-stage extended non-minimum state space model was established to introduce the state error,output tracking error and extended information.The quadratic performance index function with external interference is constructed,using Pontryagin optimization theory to design control law and to analyze the stability and robustness of the system.Then the average dwell time is used in the switching system theory to obtain the minimum running time for stable operation of a subsystem.Finally,the feasibility and advantages of this method are verified by classical injection molding simulation.2.An improved optimization design of constrained model predictive tracking control is proposed for multi-phase batch process with time-varying uncertainties.Firstly,according to the multi-model characteristics of batch process,the input-output process data discrete system model is further processed and transformed into a comprehensive state space model.An improved extended state space model is established by combining state variables and tracking errors and introducing an extended dynamic model.Considering the uncertainty in the system,the extended model is transformed into an observable discrete model with polyhedron uncertainty.By designing MPC controller based on min-max optimization,the issue of controller gain being not adjustable is solved while improving performance.The system stability condition based on LMI constraints is given and then obtained the control law.At the same time,combining with the classical switching system theory,the average resident time method is used to calculate the minimum running time of the subsystem.Finally,the effectiveness and superiority of this method are demonstrated by comparing the cases of single mode and multi-mode.