2022-069 Real-Time ML Surrogate for Spatter Mapping and Build Optimization in L-PBF
Benefit
This technology is a fast-response machine learning surrogate model for predicting spatter transport and contamination in laser powder bed fusion (L-PBF) additive manufacturing. The innovation combines gradient-boosted random forest modeling with custom loss functions, trained on computational fluid dynamics (CFD-DPM) simulations, to accurately and rapidly forecast where spatter particles will land during the printing process. Integrated into a build planning application, the system allows users to optimize machine settings (gas flow, material, laser parameters) and part placements to minimize contamination risk, maximize quality yield, and avoid costly defects. A spatter heatmap and digital twin visualization empower manufacturers with real-time, interactive layout optimization enabling scalable, contamination-aware production planning for L-PBF systems, with future potential to support diverse machines and materials.
Market Application
● Aerospace and defense: Qualification-friendly planning to minimize lack-of-fusion and foreign object debris, improving first-pass yield for flight-critical parts.
● Medical devices: Clean build planning for high-integrity implants and instruments, reducing contamination hotspots and ensuring regulatory-grade consistency.
● Energy and turbomachinery: Contamination-aware layouts for nickel superalloys (e.g., Inconel 718), improving internal passage quality and fatigue life.
● Automotive and motorsports: Fast iteration on multi-part trays to boost throughput while controlling spatter-driven defects on dense builds.
● Industrial tooling and molds: Optimized placement to protect fine features and surfaces from spatter deposition, reducing finishing and scrap.