Determining optimal bead central angle by applying machine learning to wire arc additive manufacturing (WAAM)

Area of Science:

• Materials Science
• Manufacturing Engineering
• Machine Learning Applications

Background:

• Wire arc additive manufacturing (WAAM) is widely adopted across industries.
• Uncontrolled bead deposition in WAAM can lead to shape collapse in multi-layer structures.
• Optimizing deposition parameters is crucial for achieving desired bead geometry and structural integrity.

Purpose of the Study:

• To develop a novel method for optimizing bead geometry in WAAM using a support vector machine (SVM) classifier.
• To determine and verify optimal deposition condition ranges for stable bead formation.
• To enable precise control over bead shape in WAAM processes, including re-manufacturing applications.

Main Methods:

• Utilized a support vector machine (SVM) binary classifier to predict optimal deposition conditions.
• Extracted geometric data of deposited beads using a laser profiler.
• Conducted single-layer experiments to gather data and multi-layer experiments for verification, classifying based on a 4° central angle.

Main Results:

• The SVM classifier effectively identified optimal ranges for deposition conditions, preventing bead shape collapse.
• Verification experiments confirmed accurate classification of deposition results based on the SVM boundary.
• Demonstrated the SVM classifier's efficiency and accuracy even with limited input data.

Conclusions:

• The developed SVM-based method successfully optimizes bead geometry in WAAM.
• This approach facilitates the creation of desired bead shapes, crucial for applications like re-manufacturing.
• The method shows potential for real-world industrial implementation with further multi-layer deposition research.