MObjects--a novel method for the visualization and interactive exploration of defects in industrial XCT data

Area of Science:

• Materials Science
• Computer Science
• Non-Destructive Testing

Background:

• Industrial 3D X-Ray Computed Tomography (XCT) generates large datasets with numerous defects.
• Analyzing individual defects like pores, inclusions, and cracks is challenging.
• Existing methods lack efficient ways to explore and classify large numbers of defects.

Purpose of the Study:

• To develop an advanced visualization method for analyzing defects in 3D XCT data.
• To enable interactive exploration and classification of individual objects (defects) within datasets.
• To demonstrate the method's utility in porosity determination for carbon fiber reinforced polymers (CFRP).

Main Methods:

• Calculating object properties (volume, dimensions, shape factors) for individual defects.
• Clustering objects into representative Mean Objects (MObjects).
• Visualizing MObject parameter spaces using radial and parallel arrangements for interactive exploration and sub-classing.

Main Results:

• Developed and demonstrated an interactive MObject visualization technique.
• Showcased iterative classification and visualization along selected analysis paths.
• Exported representative MObjects as volumetric datasets for simulations (e.g., ultrasonic calibration, thermography).

Conclusions:

• The novel MObject visualization method enhances defect analysis in 3D XCT data.
• The approach is practical for domain specialists, particularly in CFRP porosity determination.
• Visual linking and local MObject visualization with homogeneity mapping improve defect property understanding.