Stress Concentration Tomography (SCT) has become a vital tool for inspecting pipelines across the globe, covering thousands of kilometers. This study presents an extensive analysis of SCT’s performance through a binary classification system, utilizing direct assessment excavations as a benchmark. The focus lies on two crucial metrics: Probability of Detection (PoD) and Probability of False Alarm (PoFA). Results demonstrate SCT's proficiency in detecting stress concentration zones (SCZs) within pipelines.

Furthermore, this study introduces a novel machine learning approach in conjunction with SCT for defect severity classification. Machine learning algorithms are employed to enhance classification accuracy. The best-performing model achieves an accuracy rate of 80% when utilizing both excavation and Inline inspection (ILI) data for verification.

In addition, the study highlights the significance of considering defect variety and dataset size when refining machine learning models for defect severity prediction. This approach ensures the practicality of these models in real-world industrial applications.

In summary, this research underscores SCT's effectiveness in identifying SCZs in pipelines and introduces a promising machine learning-based approach for defect severity classification. These findings have practical implications for the inspection and categorization of pipeline defects, improving the overall quality control in pipeline-related industries.