Mastering the Match: A Comprehensive Validation of Run Comparison Software Using Synthetic Data
This study rigorously validates run comparison (RC) software, essential for accurate corrosion growth rate assessments in pipelines, using an extensive synthetic dataset and an experimental K-nearest neighbours-based algorithm across 2,000 diverse spools. Detailed within the paper are the synthetic data generation, in-line inspection (ILI) run simulations, and algorithm validation processes, facilitating a nuanced understanding of the algorithm’s performance under varying conditions.
Results highlight the significant impact of anomaly distribution on RC accuracy, with noticeable performance declines as anomaly counts increase, especially in scenarios with circumferentially concentrated anomalies. However, the algorithm maintains commendable accuracy and robust performance across various tests.
This research not only sheds light on critical factors affecting RC performance but also sets a robust framework for future evaluations and underscores the need for representativeness in synthetic datasets, guiding enhancements in RC algorithms. Future directions include improving algorithmic resilience in high-density anomaly conditions, refining validation frameworks, and diversifying matching scenarios.