Real-Time Prediction of Curing Processes using Model Order Reduction

Abstract

Manifold engineering applications are directly affected by temperature. For rubber or composite curing processes, temperature distributions over time inside the compounds are crucial for chemical cross-linking reactions. Most of these reactions occur subsequently to a heating process during product cool down. Online prediction of cooling phases is performed during the actual heating process and hence, final cure status can be estimated before the actual process finishes. Therefore, mold temperatures and heating duration can be adapted in regard to current ambient conditions, and hence product quality is increased. In order to achieve longterm thermal predictions for complex product geometries, simulating nonlinear thermal finite element models is unfeasible, due to high computational effort. Hence, a prediction-model is derived from finite element analysis using matrix export, linearization, model order reduction algorithms such as rational Krylov or iterative rational Krylov and correction of operating point deviation. A special remark is given to temperature dependent boundary conditions, choice of time discretization and choice of solving algorithm, to address arising conflicting goals between execution time and simulation accuracy. Eventually, a complete process simulation is performed during the task-cycle time on a PLC control with a sufficiently high accuracy.