Adaptive Knowledge-Informed Error Compensation for Process Planning in Extrusion-Based Additive Manufacturing

Abstract Achieving high-throughput production without sacrificing quality remains a critical challenge in extrusion-based additive manufacturing. Traditional approaches rely on real-time sensing for feedback control of process parameters that indirectly reflect printing quality. Many extrusion printers lack online monitoring systems, making such closed-loop solutions to direct quality control impractical. This article proposes a framework which enables adjustments of kinematic parameters such as printing speed and acceleration along different regions of printing paths to compensate for defects. To make this planning process computationally feasible, this article discovers and leverages process knowledge that links fast printing-induced defects to infill slicing patterns. This insight enables a targeted, localized search for compensation strategies, drastically reducing the design space for planning. A two-scale adjustment for compensation is developed to reduce the printing time. At a global scale, the method identifies how the base setting affects the spatial distribution of defects. At a local scale, it adjusts the base setting in regions prone to defects. The cloud framework enables the sharing of Bayesian models to achieve compensation using just very limited user-provided samples. It rapidly updates printing plans and estimates defect regions, optimal compensation, and production time. Our case studies show that local-scale compensation through speed adjustment in different part designs and commercial printers reduces the printing time by an average of 54% without introducing infill defects. Moreover, the proposed two-scale compensation through cloud-based model adaptation achieves reductions of up to 65%.