Computer Modeling for Injection Molding: Simulation, Optimization, and Control

Chapter 13: Learning Control

Yi Yang

Deptartment of Control Science and Engineering, Zhejiang University, Hangw, Zhejiang, China

Furong Gao

Department of Chemical and Biomolecular Engineering, Hong Kong University of Science and Technology, Kowloon, Hong Kong, China

As mentioned in Chapter 2, injection molding is a batch process with a repetitive nature, information of past controls may be explored for improvement of the current and future cycles. This kind of cycle-to-cycle improvement is similar to that of the human learning process, therefore referred to as learning control.

13.1 Learning Control

Generally, there are two ways to improve the learning control performance along cycle direction, one is controller parameter learning and the other is control signal learning control.

The control parameter improvement learning control strategy can be described as Fig. 13.1. The system has two feedback loops, one is timewise feedback control that uses convention feedback control strategy maintaining timewise control performance within a cycle and the other one is a batchwise controller adjustment loop that updates the controller parameters from cycle to cycle according to certain batch control performance index. Yang and Gao¹ proposed cycle-to-cycle improvement for the nozzle pressure control in 1999. Li et al.² proposed to adjust the fuzzy controller parameters based on the information of the previous cycle. Yang³ further proposed to describe the dynamics of injection filling process ...