Injection Pultrusion (IP) is a novel process that combines the best features of traditional Resin Transfer Molding (RTM) and wet bath pultrusion for manufacturing polymer composites. Few researchers have dealt with the quality control aspects of this process which is extremely crucial to the plant economics. The main contribution of this paper is the study of the IP process from the point of view of product quality control while maximizing production rates. A mathematical model that includes resin flow and cure, and heat transfer in the part and the die is used to determine sensitivity of the quality variables to important processing variables and parameters. Similarities and differences with traditional pultrusion process are pointed out. Control-relevant features of the process are identified and the requirements necessary for a control system are presented. A cascaded control strategy which uses an on-line process model is proposed.

In this paper, we address the quality control issues in manufacturing of fiber-resin composites through a prototype process called injection pultrusion (IP). The objective is to maximize production rates while maintaining quality. It is demonstrated that a cascaded control strategy which uses an online process model is suitable in meeting this objective. The strategy uses steady-state cure and pressure models, in conjunction with an optimizer. A first principles model is exercised to generate input data for the feature selection procedures. Based on statistical significance tests, streamlined regression models are generated by identifying processing variables and parameters having a crucial bearing on the part quality and eliminating superfluous variables. Infrequent quality measurements are used to correct for modeling errors. Closed loop control results are presented to demonstrate the successful working of the strategy.

In this paper, we present a model predictive inferential control (MPIC) strategy to address the problem of controlling unmeasured output variables (such as quality) using readily available secondary measurements. First we establish the relationship between inferential control and other classical control strategies such as cascade and internal model control. Next we present a framework for incorporating the inferential control strategy within the framework of the often used model predictive control (MPC). This framework, termed model predictive inferential control (MPIC), is general enough to accommodate multiple secondary measurements as well as nonlinear estimators and controllers. The advantages of inferential control are established using two case studies. One is the Shell challenge problem which employs linear transfer function models. The second is a nonlinear, multivariable problem on the control of product composition using secondary measurements on a simulated injection pultrusion process. Problems of collinearity are addressed using principal component analysis (PCA) during the construction of the dynamic estimator. These simulations demonstrate the advantages of the proposed model predictive inferential control strategy.

No abstract