Multiphysics Process Modeling of Single Magnetron Microwave Cured Carbon Fiber Reinforced Composites

Abstract

Abstract

Abstract Carbon fiber reinforced polymer (CFRP) composites are in increasing demand in critical applications, yet the standard manufacturing methodology, autoclave curing, remains costly and energy intensive. Microwave curing offers an out-of-autoclave (OOA) solution to this limitation, but many existing systems rely on similarly expensive and large multi-mode systems. In this work, 16-layer unidirectional Cycom 5320-1 prepreg laminates, with 2.5 mm thickness, are analyzed to study their behavior as they are heated and cured in a domestic microwave modified with a custom mode stirring mechanism. A Multiphysics finite element (FE) model is created to predict the complex coupled electromagnetic, thermal, and curing physics using COMSOL Multiphysics software. The model successfully reproduced the fiber orientation dependent temperature differentials observed experimentally by loading the composites one at a time into the microwave with the fibers oriented parallel and perpendicular to the dominant field direction, confirming that fiber-field coupling persists in single magnetron systems despite the presence of mode stirring. The differences in power required to cure the two panels are quantified along with the maximum temperature differences that the panels experience through their thickness and across the surface. Variation in the resulting degree of cure across the part is also reported using a modified Arrhenius equation. A parametric study on the location of the mode stirrer within the microwave cavity is performed to optimize its placement for curing uniformity. Deposited power is shown to depend on fiber orientation relative to the cavity field polarization. The predicted ratio of deposited power between orientations agrees with measured magnetron power requirements to within 6.4%. Temperature predictions by the multiphysics model exhibited agreement with experiment with a normalized root mean square error below 3% at all six sensor locations across the two panels.