Research On Neural Network Iterative Learning Control Of Tobacco Fermentation

As a result of tobacco fermentation process with non-linear, time-varying, uncertainty, characterized by strong coupling, making use of traditional methods of control it is very difficult to control the production process to meet the performance requirements, a direct impact on tobacco production. With this situation on the urgent need for an advanced control methods to resolve this problem.Iterative learning control is a new control algorithm, it does not rely on precise dynamic mathematical model of the system, through the introduction of learning mechanisms, continue to accumulate knowledge of the object, on-line to complete the design and improvement of the controller by means of online learning, on-line control to improve system performance and control functions are integrated in an algorithm, the adoption of the industrial process to be repeated to achieve. It exists in the settlement of non-linear object or as a result of bad modeling uncertainty caused by problems with the unique advantages in the learning process continued to make up for the lack of prior knowledge, so that system performance has been gradually improving.With the based on the analysis of the characteristics of tobacco fermentation process based on neural network control with a combination of iterative learning control, the neural network optimization of iterative learning control with PD feedback controller combination. Through the introduction of feedback control systems to reduce the tracking error; iterative learning control of feed-forward compensation, and optimization through neural network iterative learning controller fitting parameters, construction of new learning control law, and then with a combination of feedback control, common role in the charged object. The fixed-gain learning law will make the learning speed algorithm reduced the number of iterations increase. In this paper, using neural networks to optimize the controller parameters used in tobacco fermentation to overcome the least square method for calculating the existence of large and slow convergence shortcomings, to achieve better control of tobacco fermentation temperature control system. In order to enhance robustness of control system in the learning controller based on PD feedback compensation by adding controllers.In this paper, neural network iterative learning control method has been applied to tobacco fermentation process in the fermentation room temperature control and fixed gain of the iterative process and the use of neural networks to optimize the temperature of the iterative process output are compared. Simulation results show that the method of complex temperature control tobacco fermentation process is effective to ensure that the less number of iterations, the temperature of the output trajectory can converge to the expectations of the track.