A multi-source melt pool compilation for vision-based analytics applications in additive manufacturing

Area of Science:

• Materials Science
• Manufacturing Engineering
• Data Science

Background:

• Metallic additive manufacturing (AM) relies on melt pool dynamics for part quality.
• Current melt pool monitoring lacks standardization due to diverse materials, parameters, and sensors.
• Standardized data is crucial for developing robust machine learning models for AM.

Purpose of the Study:

• To introduce Melt-Pool-Kinetics, a comprehensive and standardized dataset for metallic AM melt pool analysis.
• To facilitate machine learning applications for in-situ monitoring, process optimization, and control.
• To provide a foundation for future research and development in AM data analytics.

Main Methods:

• Compiled 1.9 TB of raw data from 32 datasets across 23 sources into a 48.6 GB HDF5 collection.
• Processed images using standard techniques: cropping, centering, resizing, grayscaling, denoising, and debayering.
• Structured the dataset into raw, processed, and diverse subsets for flexible use.

Main Results:

• Created Melt-Pool-Kinetics, a large-scale, curated dataset for metallic AM melt pool signatures.
• The dataset supports machine learning tasks including defect detection and microstructure prediction.
• Standardized data processing enables consistent analysis across different AM systems.

Conclusions:

• Melt-Pool-Kinetics addresses the critical need for standardized melt pool data in metallic AM.
• The dataset empowers advancements in real-time monitoring, optimization, and control of AM processes.
• This resource is designed for scalability and future expansion with new data.