The School of Computer Science is pleased to present…
Multi-Angle Virtual Flaw Augmentation Toward the Automation of PAUT Data Analysis
MSc Thesis Proposal by: Abdul Rafey Khan
Date: Wednesday, June 18, 2025
Time: 4:00 pm
Location: Essex Hall, Room 122
Automation of the detection of flaws in Phased Array Ultrasonic Testing data is critical for reliable non-destructive evaluation. However, the availability of diverse and well-annotated defect data is limited, making it difficult to train and validate analysis tools. This work presents a method for Multi-Angle Virtual Flaw Augmentation, where real flaws extracted from previously recorded defective scans are carefully transferred onto clean, defect-free PAUT scans. The placement of these flaws is guided by known flaw characteristics and occurs in spatially appropriate regions of the scan data.
A key focus of this work is maintaining geometric consistency and amplitude integrity of the flaws across different probe angles. This is important because PAUT inspections are conducted from multiple views, and any augmented flaws must appear naturally aligned across these perspectives. The method ensures that the flawed morphology and signal response remain realistic after augmentation, both in individual views and across the full set of inspection angles.
This approach is tested using real-world nuclear scanned data from Ontario Power Generation, and results show that it can significantly enhance the variety and realism of flaw data available for analysis. Ultimately, this augmentation method supports the development of more reliable and consistent automated tools for PAUT data interpretation.
Internal Reader: Dr. Hamidreza Koohi
External Reader: Dr. Caniggia Castro Diniz Viana
Advisor: Dr. Ziad Kobti
Co-Advisor: Dr. Roman Maev
