Applications Of Iterative Learning Control

ABSTRACT:As the injection molding machine has an irreplaceable position in the production of plastic products, the research on its clamping mechanism localization is also highly anticipated. The existing clamping mechanism control system is an open loop control system, and users control the clamping action by setting the hydraulic flow, location and other information of various stages in the interface. However, the existing control system is defective, and the final stop position of the core plate can’t be precisely controlled, which may lead to mechanical shock, and even short the life of the machine. For this problem, a PD-type ILC algorithm is researched in this paper to improve the control performance of the existing control system.Firstly, in this paper, the significance of research on clamping mechanism is explained, and the status of research in this field is analyzed. Then some appropriate solutions for the problems are proposed.Secondly, various model-free control algorithms from the sight of Systematic Theory are analyzed and compared.Thirdly, an improved PD-type iterative learning control algorithm for multiple-input systems with delay is proposed. The sufficient condition of the iteration output converging to the desired output is also given when delay is unknown but its scope is determined.Then, an optimize algorithm is researched to find the minimum output error and the corresponding PD coefficients of each control inputs by using neural networks with genetic and PSO algorithms. On this basis, the neural network with genetic PD iterative learning algorithm and PSO PD iterative learning algorithm are proposed. By analyzing the result of the simulation, the both control algorithms are proved to be effective.In order to verify the feasibility of the above algorithm, the experiment of clamping mechanism in the actual environment of factory is done.The iterative learning algorithm is embedded into injection molding machine clamping mechanism control process in the experiment. The final positioning error is controlled by changing the hydraulic, the flow and the stop position.During the experiment, molds of different dimensions and weight are mounted, and the positioning error curve is drawn by recording the final positioning data of core plate. Then the control effects of the traditional control method, the PD iterative learning algorithm and the neural network with genetic PD iterative learning algorithm are compared.At last, the conclusions and the prospects of the paper are proposed.