A Hybrid Digital Twin Framework for Modeling Machine Process Parameters for Automated Fiber Placement-based Manufacturing of Composite Structures

The aerospace and defense sectors require lightweight, high-performance composite airframes, necessitating advancements in Automated Fiber Placement (AFP) for high-rate, cost-effective manufacturing with minimal defects. In-situ consolidation of thermoplastic composites eliminates autoclave post-processing but demands precise control of temperature, compaction pressure, strip tension, and layup speed to ensure defect-free parts. Temperature governs resin melting and interlayer bonding; insufficient heat causes voids and delamination, while excessive heat degrades polymers, reducing stiffness and fatigue resistance. Compaction affects adhesion and residual stress, while improper tension or speed induces wrinkles, gaps, or slippage. Uncontrollable variables like tape and process variations complicate control, requiring extensive testing. Current AFP workflows rely on kinematics-focused CAM tools without physics-based models to predict thermal/stress evolution, thus leading to defects, rework, and reduced efficiency.This effort proposes a physics-based, AI-enhanced Digital Twin (DT) framework to optimize AFP for complex composite structures, addressing process variations and enabling defect-minimized manufacturing. In Phase I, a multi-axis thermal model will simulate dynamic heat source, roller compaction, and strip tension, predicting transient temperature distributions for single-curvature geometry. In-situ data from infrared cameras and sensors will capture process variations, processed with AI to classify defects (e.g., folds, twists) and deviations in heating/compaction. A data-driven thermal correction scheme will update thermal analysis inputs based on process variation data, creating an enhanced multi-physics model. Phase I will develop this hybrid DT framework, ensuring feasibility for integrating physics-based models with in-situ data.The Phase I Option will map AI-processed defects onto thermo-structural models, analyzing their impact on part performance. A phenomenological model will predict crystalline evolution and interlayer adhesion based on thermal history and compaction, optimizing parameters to mitigate defects like delamination. The defects will be ranked by their impact on thermal buildup and mechanical performance, with the goal of providing guidance to operators on which defects require rework, reducing the need for extensive operator intervention in corrective actions.Phase II will develop a closed-loop control framework, integrating multi-physics simulations and AI for predictive planning, defect mapping, and real-time optimization of layup sequences on complex geometries. Once developed and validated on a representative component, this approach will deliver a machine-agnostic software tool, addressing the Navy’s requirements for a smart AFP process with enhanced in-process monitoring, reduced operator intervention, and improved part performance, ultimately enabling high-rate, cost-effective composite manufacturing.