The Object Segmentation from the Microstructure of a FSW Dissimilar Weld.

Area of Science:

• Materials Science
• Metallurgy
• Computer Vision

Background:

• Friction stir welding (FSW) is an advanced, eco-friendly solid-state joining process.
• Analyzing the microstructure of FSW welds, especially dissimilar alloys, is crucial for understanding mechanical properties.
• Traditional microstructure analysis can be challenging due to noise and complex feature distribution.

Purpose of the Study:

• To develop and present a two-stage image processing framework for analyzing FSW weld microstructures.
• To segment and characterize particle distributions within different layers of dissimilar aluminum alloy welds.
• To demonstrate the effectiveness of computer vision techniques in overcoming limitations of manual analysis.

Main Methods:

• A dual-speed tool was used to prepare the dissimilar aluminum alloy weld.
• A two-stage image processing framework involving segmentation and object detection was applied.
• Digital analysis utilized basic segmentation, morphological operations, point transformations, and domain knowledge on microscopy images.

Main Results:

• The developed segmentation method successfully identified areas with particles forming bands in the weld microstructure.
• The image analysis effectively separated microstructural features, even in noisy images.
• Analysis of particle distribution, shape, and size variability across different weld layers was achieved.

Conclusions:

• The proposed computer vision-based framework is effective for analyzing complex microstructures in FSW welds.
• This approach enables detailed characterization of microstructural features, essential for material performance prediction.
• The method overcomes previous limitations, allowing for quantitative analysis of dissimilar alloy FSW joints.