Objective Buried pipelines play a vital role in the transportation of resources such as petroleum and natural gas. However, pitting defects frequently occur in these pipelines exposed to surrounding soil environments, seriously compromising their reliability. As a novel non-destructive testing technique, weak magnetic testing has been widely adopted in the detection of buried oil and gas pipelines due to its advantages, including simplicity of operation, high sensitivity, and strong adaptability. Currently, the dispersion of weak magnetic testing data and a lack of relevant sample data hinder the accuracy of defect inversion. Detecting pitting defects in engineering practices presents significant challenges. Additionally, previous research has overlooked the prediction of defect angles.

Methods This study proposes a methodology for inverting pitting defects in buried pipelines based on weak magnetic testing, utilizing a combination of the Chaotic Improved Whale Optimization Algorithm (CIWOA) and the BP neural network, with a focus on the size and angle of the defects. First, the impact of various pitting defects on the characteristics of weak magnetic signals was analyzed through experimental weak magnetic testing on pipelines. The results were then used to select inversion parameters and establish data sets, which were enhanced using the Synthetic Minority Oversampling Technique (SMOTE) algorithm. Next, Kernel Principal Component Analysis (KPCA) was applied to reduce the dimensionality of the data and identify the principal components contributing to the sizes and angles of pitting defects. Finally, a CIWOA-BP model was developed to predict the size and angle of pitting defects in buried pipelines. A comparison of the prediction results from the proposed model was conducted against those from other existing models.

Results The results demonstrate the accuracy of the CIWOA-BP model in predicting the size and angle of pitting defects in buried pipelines. With smaller values of mean absolute percentage error (MAPE), mean squared error (MSE), and mean absolute error (MAE), along with the coefficient of determination R² closer to 1 compared to WOA-BP (Whale Optimization Algorithm-Back Propagation), BP, RF (Random Forest), and SVM (Support Vector Machine), the CIWOA-BP model shows superior prediction accuracy for both the sizes and angles of pitting defects.

Conclusion Therefore, the proposed method enables accurate inversion of pitting defects in buried pipelines using weak magnetic testing, presenting promising application prospects for ensuring the safe transportation of these pipelines.