No description available.

No description available.

The thermal control of a die is crucial for the development of high efficiency injection moulds. For successful thermal management, this research provides an effective control strategy to find sensor locations, identify thermal dynamic models, and design controllers. By applying a clustering method and sensitivity analysis, sensor locations are identified. The neural network and finite element analysis techniques enable the modeling to deal with various cycle-times for the moulding process and uncertain dynamics of a die. A combination of off-line training through finite element analysis and training using on-line learning algorithms and experimental data is used for the system identification. Based on the system identification which is experimentally validated using a real system, controllers are designed using fuzzy-logic and self-adaptive PID methods with backpropagation (BP) and radial basis function (RBF) neural networks to tune control parameters. Direct adaptive inverse control and additive feedforward control by adding direct adaptive inverse control to self-adaptive PID controllers are also provided. Through a comparative study, each controller’s performance is verified in terms of response time and tracking accuracy under different moulding processes with multiple cycle-times. Additionally, the improved cooling effectiveness of the conformal cooling channel designed in this study is presented by comparing with a conventional straight channel.

The thermal control of a die is crucial for the development of high efficiency injection moulds. For successful thermal management, this research provides an effective control strategy to find sensor locations, identify thermal dynamic models, and design controllers. By applying a clustering method and sensitivity analysis, sensor locations are identified. The neural network and finite element analysis techniques enable the modeling to deal with various cycle-times for the moulding process and uncertain dynamics of a die. A combination of off-line training through finite element analysis and training using on-line learning algorithms and experimental data is used for the system identification. Based on the system identification which is experimentally validated using a real system, controllers are designed using fuzzy-logic and self-adaptive PID methods with backpropagation (BP) and radial basis function (RBF) neural networks to tune control parameters. Direct adaptive inverse control and additive feedforward control by adding direct adaptive inverse control to self-adaptive PID controllers are also provided. Through a comparative study, each controller’s performance is verified in terms of response time and tracking accuracy under different moulding processes with multiple cycle-times. Additionally, the improved cooling effectiveness of the conformal cooling channel designed in this study is presented by comparing with a conventional straight channel.