A theory-guided probabilistic machine learning framework for accelerated prediction of process-induced deformations in advanced composites

This paper introduces an innovative framework for efficient analysis of composites manufacturing processes and phenomena. The method combines sparse probabilistic characterizations, multi-fidelity simulation schemes, and limited experiments to train surrogate machine learning (ML) models. Guided by a probabilistic technique, Spatially Weighted Gaussian Process Regression (SWGPR), predictive models are constructed from multi-fidelity data to perform rapid and accurate manufacturing assessments. This study demonstrates the effectiveness of the framework in accurately predicting process-induced deformations (PIDs) for L-shaped composite parts using minimal experimental efforts. The method introduced in this work aims to offer a cost-efficient and broadly applicable framework for potentially mitigating PIDs and solving other composites manufacturing problems.