What type of study is this?

This is a Quantitative Study study.

October 9, 2025

A Novel Identification Method of Thaw Settlement Induced Bending Region of Buried Pipeline Based on Multi-Round IMU Inspection Data and Artificial Neural Network

Key Points

The intelligent method achieves over 90% accuracy in identifying thaw risk sections of pipelines.
Using IMU detection data, the method efficiently evaluates bending strain changes and improves safety assessments.
Variational mode decomposition combined with permutation entropy enhances data accuracy by reducing interference noise.
The neural network model rapidly classifies risk sections, offering a significant advancement over traditional manual analysis.

Abstract

Abstract The frequent occurrence of frost heave and thaw settlement of crude oil pipelines in the permafrost zone poses a serious threat to the structural safety of buried crude oil pipelines. The pipeline internal detector based on Inertial Measurement Unit (IMU) can realize the detection of pipeline bending strain. By analyzing the detection data of multi-wheel IMU, the change of bending strain can be evaluated, and then whether the pipeline needs maintenance can be judged, which provides a scientific basis for pipeline operation safety. At present, the identification of the strain change risk section of the pipeline in the thaw section mainly depends on the traditional manual analysis method, which leads to the problem of low identification efficiency and high misjudgment rate. Therefore, based on multiround IMU strain data, this paper proposes an intelligent identification method of melting risk section based on IMU strain data. This method uses the variational mode decomposition combined with the permutation entropy (VMD-PE) method to remove the interference noise in the IMU strain data, uses the geometric and magnetic flux leakage detection data to locate the geometric features in the IMU strain detection data, and establishes the pipeline strain change section database. Ten typical data features are extracted and feature selection is completed by Shapley Additive exPlanations (SHAP) analysis. Based on the neural network deep learning model, the classification is completed to realize the intelligent identification of the thaw risk section. The results show that the neural network model based on the database has a recognition accuracy of more than 90% for the pipeline thaw risk section, and the recognition efficiency is fast, which provides an effective technical means for the identification of the bending strain change greater than 0.02% in the pipeline integrity evaluation.

KI fragen

Bookmark