Abstract

Process-induced deformations result from internal residual stress caused by the anisotropic properties of thermoset composite parts. The study’s focus is diagnosing the polymerization process, or curing, and considering how uncertainties in boundary conditions affect cure kinetics. This is achieved through a Particle Filter approach, utilizing a Bayesian framework. This framework recursively estimates the evolving cure state’s posterior distribution based on observed measurements from Differential Calorimetry Scanning tests and thermocouples. The algorithm simultaneously estimates the cure state parameters and predicts part temperature and process-induced deformations, which are closely tied to cure behaviour. This is accomplished using diffusion cure kinetics and analytical deformation models. Furthermore, it introduces an augmented cure state formulation to address uncertainties in cure boundary conditions, which conventional models overlook. The developed stochastic approach adeptly captures uncertainties related to cure evolution while providing comparable deformation predictions with minimal computational costs and memory usage. Experimental measurements of process-induced deformations in C-shaped thermosetting parts, made of unidirectional 8552/AS4 fibres and cured following the Manufacturing Recommended Curing Cycle, are validated using the developed algorithm. After validation, the proposed model is employed to predict outcomes, which are then utilized to determine the optimal curing cycle using a Genetic Algorithm.