Machine Vision to Provide Quantitative Analysis of Meltpool Stability for a Coaxial Wire Directed Energy Deposition...

Area of Science:

• Materials Science
• Manufacturing Engineering
• Computer Vision

Background:

• Wire-based additive manufacturing (AM), specifically directed energy deposition (DED), offers scalability and high deposition rates for complex metal parts.
• Maintaining process stability during wire DED is crucial, as insufficient energy leads to wire stubbing and meltpool turbulence, while excessive energy causes overheating and liquid drips.

Purpose of the Study:

• To propose and evaluate a computer vision technique for real-time state and event detection of wire stability in DED.
• To develop a non-machine learning based model for process monitoring that allows for user interpretation and adjustment.

Main Methods:

• A computer vision model was developed using intensity variations and frame-to-frame difference calculations to assess wire stability.
• The model's performance was validated through a 1D laser power experiment, generating diverse stability conditions.
• Accuracy was assessed by comparing model predictions with 3D geometry data of deposited beads.

Main Results:

• The proposed computer vision model effectively detected various states of process stability during wire DED.
• The model demonstrated high accuracy in identifying stable and unstable deposition conditions.
• The system proved capable of differentiating between optimal, insufficient, and excessive energy input based on visual cues.

Conclusions:

• The developed computer vision technique is a capable and accurate system for monitoring wire stability in DED.
• The non-machine learning approach allows for interpretability and potential manual adjustments by operators.
• This method shows significant potential for implementation as a real-time control system in wire-based AM.