Application of reduced order process models for nonlinear inferential control

Most of the industrial scale polymeric composites manufacturing processes are nonlinear systems with spatially varying outputs, states, controls and parameters. In practice, the spatially distributed nature of these processes is generally overlooked or ignored and conventional control and design techniques are applied using lumped models identified with input/output data from numerical simulations or real life experiments. These models suffer from strong interactions, apparent time delays and poor predictive capabilities due to the inability to capture the convective and diffusive phenomena. To resolve this issue, we develop fast and accurate online models using a proper orthogonal decomposition (POD) technique in conjunction with a Galerkin formulation procedure. The utility of these reduced order models for online estimation and control of an experimental composite manufacturing unit is evaluated