Pitting corrosion inversion of buried pipelines based on weak magnetic detection and interpretable SACCO-ELM

Abstract

Buried pipelines play an important role in long-distance energy transportation. However, buried pipelines are prone to pitting corrosion due to harsh soil environment, making quantitative pitting corrosion inversion critical for safety. This paper proposes an interpretable inversion method based on weak magnetic detection (WMD) to simultaneously predict defect depth, diameter, and angle. A dedicated WMD experiment of buried pipeline is conducted to record magnetic signals under varying defect sizes and angles. Key features are extracted to construct the dataset, and correlation analysis is performed using Pearson correlation coefficient. A hybrid inversion model combining extreme learning machine (ELM) with a simulated annealing‑improved cuckoo catfish optimiser (SACCO) is then developed, followed by SHapley Additive exPlanations (SHAP) for model interpretability. The results show that the proposed SACCO‑ELM model achieves average relative errors of 4.72% for defect depth, 1.27% for defect diameter, and 7.15% for defect angle, outperforming comparative methods. This study provides a data‑driven and interpretable framework for pitting corrosion assessment, offering a theoretical basis for ensuring the transportation safety of buried pipelines.