Alignment method of internal and external pipeline inspection data based on machine learning algorithm

Abstract

As an important part of pipeline integrity management, the alignment of the external and internal pipeline inspection data is to align the external inspection points to the internal inspection centerline, so as to fully mine the value of internal and external inspection data. Herein, from the perspective of the relationship between the surface marking points and the mileage piles of the pipelines, a relation model of internal inspection points and external inspection mileages was constructed using the machine learning algorithm based on the external and internal inspection data from the pipeline integrity management system of a long-distance pipeline company, and the mileage information of the internal inspection points in the external inspection was predicted to increase the mapping between the internal inspection points and the external inspection mileage, further realizing the data enhancement. In addition, the pipelines were segmented by the surface marking points and mileage piles, and the external inspection points were aligned to the internal inspection centerline with the linear stretching algorithm, so as to realize the alignment of the external and internal inspection data. The results show that the average absolute percent error is less than 0.10% and the determination coefficient is 99.99% for the internal inspection point based external inspection mileage prediction model established with the machine learning algorithm. Moreover, the model could capture the relationship between the internal inspection points and the external inspection mileages, so as to support the automatic alignment of the external and internal inspection data of the pipelines.

Keywords

• pipelines
• integrity
• internal inspection data
• external inspection data
• data alignment
• machine learning