Abstract
Establishing a relationship between processing conditions during the manufacturing process and end-part properties is essential for the efficient fabrication of advanced composites. For fiber-reinforced polymer composites, this manufacturing process typically includes multiple complex stages such as lay-up, curing, high-temperature pyrolysis, several resin backfill steps followed by pyrolytic process repeats, and graphitization. In particular, the high-temperature pyrolysis step is critical, because the process temperature cycle significantly affects the degradation reactions and phase transformations undergone by the polymer resin. Therefore, it is crucial to identify the ideal parameters for optimal performance. Characterizing pyrolysis kinetics to achieve this commonly requires an exhaustive testing campaign to study the effect of different temperature cycles on final properties. This involves thousands of possible combinations of multi-heating ramps and holds, resulting in an intractable search space. Additionally, established literature succeeds in characterizing the pyrolysis kinetics for simple dynamic temperature cycles, but fails to account for the effect of more complex cycles. This paper aims to address the above challenges by introducing a novel probabilistic machine learning-based framework for accelerated characterization of pyrolysis kinetics using theory-based transformations of limited and noisy experimental data. The degradation reactions and fraction of volatile components of a polymer-based composite are studied using thermogravimetric analysis techniques, and results are analyzed using theory-guided Gaussian Processes (GPs) to accurately characterize the pyrolysis kinetics of complex temperature cycles. The framework is then used to identify optimal parameters for achieving the desired yield. Using this approach, the experimental effort is considerably reduced while developing an accurate pyrolysis kinetics model for the material and establishing optimal processing parameters.
