Modelling and predictive control for a RTM mold

Abstract

The objective of this work is the control of a Resin Transfer Moulding (RTM) mold trying to obtain a temperature in the whole cavity as much homogeneous as possible. This is a complex process, mainly because of two factors: temperature distribution in a complex geometry and variability of the system parameters. In order to solve these difficulties, a design procedure based on models is carried out. Firstly, the detailed modelling of the thermal system, discretized and solved by means of finite elements (FEM) and validated by thermal tests in the physical mold. Then, a simplified reduced-order model (ROM) is obtained by means of ARX approach based on the former model. Finally, the design of a  robust MPC (Model Predictive Control) which is able to take the system variability, the geometry complexity and the limitations of the actuation systems (resistances) into account. In order to afford the nonlinearities intrinsically present in the system, in addition to the controller, a perturbation estimator is implemented, which evaluates the changes produced in the mold with regard to the expected behaviour according to the ROM. In this way, the control system is able to detect the variations and compensate them in real time. These algorithms are validated both by simulation and experimentally. From this verification, it follows that the application of these modelling, estimation and control techniques allows to control the mold with temperature variations much tighter than for PID standard controllers.

Full document