Reduced-Order Kalman Filter for Surface Temperature Monitoring of parameter-variant thermal Systems

A real-time model-based surface temperature monitoring approach for large scale systems is presented in this manuscript. An insulated heating plate, which is part of a curing mold, is used to exemplify the method. The heat equation is formulated and boundary conditions such as thermal radiation and convection are included, to solve for temperature distributions over time. Complex geometry and different components require fine finite element meshes for spatial discretization, which leads to a large system scale. Furthermore, the heating plate can operate in a wide temperature range that impedes linearization of temperature dependent parameters. Model order reduction is performed to reduce computation time to real-time capabilities and enable iteratively solved identification of boundary condition parameters. A special remark is given to preservation of identifiable physical parameters and application of model order reduction, despite parameter-variant system description. Eventually, real-time monitoring of surface temperature distribution on a programmable logic controller is performed, using validated thermal models in a Kalman filter. A deviation below 1K is reached throughout most of the monitoring, during different temperature set-point trajectories.