Simulation-in-the-loop additive manufacturing for real-time structural validation and digital twin development

Ensuring end-use quality is essential for batch-produced parts, particularly for load-bearing components, where defects can significantly compromise structural integrity. Traditionally, finite element analysis (FEA) has been employed either in pre-process design or as a post-process troubleshooting tool. This paper introduces a novel, in-process, simulation-in-the-loop FEA system for real-time validation of the structural quality of additively manufactured components as they are being produced. We present a case study using a consumer-grade 3D material extrusion printer to validate the proposed system. Defect information is segmented from the layer image using a U-net architecture and fed into a finite element solver to predict the potential structural failure of the specimen in real-time. The proposed vision-based damage detection system achieved a segmentation accuracy of 92.79% on the test data, while the FEA model showed final errors of 4.92% and 3.36% in terms of tensile strengths when compared to the measured specimens with and without impactful defects, respectively. The real-time FEA validation process varies depending on the computer system and the complexity of detected defects. Overall, the framework introduced in this work progresses the state-of-the-art towards ensuring real-time validation and timely decision-making during printing. The proposed algorithm is effective for automatic real-time product structural quality validation and decision-making, as demonstrated in three case studies. Result show that for the three different test cases with different levels of defects, the model predicted the failure strength of the specimen within 5%. The contributions of this paper are threefold: First, a simulation-in-the-loop framework was developed for in-process real-time structural validation of additively manufactured components. Second, advanced image segmentation was integrated for adaptive defect detection, enabling precise localization of defects without prior training on each defect size. Third, a flexible decision-making system was created to evaluate product quality using tailored structural metrics, allowing timely responses to maintain integrity. Together, these innovations form a comprehensive real-time FEA validation system, enhancing reliability in structural assessment for additive manufacturing. © 2025 Elsevier B.V.