Model based control of polymer composite manufacturing processes

The objective of this research is to develop tools that help process engineers design, analyze and control polymeric composite manufacturing processes to achieve higher productivity and cost reduction. Current techniques for process design and control of composite manufacturing suffer from the paucity of good process models that can accurately represent these non-linear systems. Existing models developed by researchers in the past are designed to be process and operation specific, hence generating new simulation models is time consuming and requires significant effort. To address this issue, an Object Oriented Design (OOD) approach is used to develop a component-based model building framework. Process models for two commonly used industrial processes (Injected Pultrusion and Autoclave Curing) are developed using this framework to demonstrate the flexibility. Steady state and dynamic validation of this simulator is performed using a bench scale injected pultrusion process. This simulator could not be implemented online for control due to computational constraints. Models that are fast enough for online implementation, with nearly the same degree of accuracy are developed using a two-tier scheme. First, lower dimensional models that captures essential resin flow, heat transfer and cure kinetics important from a process monitoring and control standpoint are formulated. The second step is to reduce these low dimensional models to Reduced Order Models (ROM) suited for online model based estimation, control and optimization. Model reduction is carried out using Proper Orthogonal Decomposition (POD) technique in conjunction with a Galerkin formulation procedure. Subsequently, a nonlinear model-based estimation and inferential control scheme based on the ROM is implemented. In particular, this research work contributes in the following general areas: (1) Design and implementation of versatile frameworks for modeling and simulation of manufacturing processes using object-oriented concepts. (2) Reduced order modeling of Nonlinear Distributed Parameter Systems (NDPS) by combining mathematical models with the available experimental/simulation data. (3) Implementation of robust nonlinear model based estimation and inferential control strategies for NDPS.