Data-Driven Q-Learning-Based Fault-Tolerant Tracking Control For Industrial Processes

With the changes in the production environment,actuator faults,external interference and the two-dimensional characteristics of the batch process are key factors that affect whether the industrial process can achieve high quality production.In the current environment,traditional model-based fault-tolerant control methods have a certain effect on performance improvement,but there are limitations such as excessive reliance on system models.Therefore,it is urgent to explore data-driven control approaches that can get rid of system model information.The research work of this paper is as follows:For the actuator faults problem that arises in industrial processes,a data-driven Q-learning fault-tolerant tracking control method is proposed.The main work is to put forward performance indicators based on a new state space model,and give the traditional model-based controller design based on the Lyapunov stability theory.By building the V function and the Q function,giving the design of a controller based on data-driven Q-learning that does not require system model information.Finally,taking the three-container water tank as the simulation object and by comparing it with traditional model-based fault-tolerant tracking control method,the proposed error control method is verified as more effective.For industrial processes with external interference and actuator faults,a H∞fault-tolerant tracking control is proposed based on data-driven Q-learning,the main work is to establish an expanded model equivalent to the original system model and build performance indicator functions,combined with the zero-sum game minmax theory to convert the optimal control strategy and the worst external disturbance design problem to solve the problem of the game Riccati equation,and therefore give the data-driven Q-learning controller design solution.Finally,the excellence of the proposed method is demonstrated by the injection molding process as an example.For batch processes with actuator faults,a data-driven Q-learning-based 2D fault-tolerant tracking control problem is proposed.Based on the 2D-Roesser model and the Lyapunov stability theory,the traditional model-based error tolerance control methods are introduced under the conditions of system stability being met,the Q-learning concept in the reinforced learning is based on the design of data-driven fault tolerant control method that is irrelevant to the model parameters and the two-dimensional Q-learning algorithm is proposed.Finally,by using the preservation stage of the injection molding process as an example,the proposed data-driven control method has been verified to have a better control effect.