Session: 03-04-02 Inline Inspection Performance III

Paper Number: 133959

133959 - Data Fusion of Complementary Axial and Circumferential Magnetic Flux Leakage Inline Inspections and Effects on Safe Remaining Life 

Abstract: 

 

Data Fusion of Complementary Axial and Circumferential Magnetic Flux Leakage Inline Inspections and Effects on Safe Remaining Life

The inspection capability of magnetic flux leakage (MFL) is subject to the angle between its magnetic field and the metal loss defect. To get a comprehensive assessment of defects, more and more pipelines are inspected with two MFL techniques with perpendicular magnetic fields (i.e., axial MFL [MFL-A] and circumferential MFL [MFL-C]). Currently, the inspection data from each MFL tool is analyzed separately, and two inspection reports are generated, respectively. Combining the two inspection reports, it becomes obvious that some conflicting results of the two technologies may in fact increase the uncertainty of the measurement due to the ambiguity of the magnetic signals. Fusing the original complementary data sets together, however, overcomes the ambiguity and enables one unique outcome.

 

To take advantage of the complementary information of two MFL inspections, a machine learning-based fusion model is proposed in this paper. The process to fuse the data from both MFL tools starts with a preprocessing step on the signal data to remove any background noise. The two signals are then automatically aligned to each other with a registration technique that provides a high level of alignment accuracy. The neural network fusion model, which has been trained on historical MFL and laser scan data, is then applied to the aligned data to generate a high-resolution 3D metal loss profile. Provided there are additional laser scans that can be supplied for an individual line, an additional fine-tuning process may be carried out in which the model is further trained for that specific pipeline, leading to an improved 3D metal loss profile.

 

The validity of the proposed model is verified using the field data from six operational pipelines with different diameters and wall thicknesses. The performance regarding depth prediction and burst pressure estimation is studied, respectively. The depth comparison of the derived 3D metal loss profiles versus laser scan profiles shows a very strong correlation in both morphology and depth. The 3D metal loss profiles are then used as inputs to RSTRENG as well as P² methodologies. The detail of the fusion-derived profiles, compared to the box-derived profiles, leads to a significantly more accurate estimation of the pipeline burst pressure. The consistent performance on all six pipelines proves the generalization of the proposed fusion model. Furthermore, three of the lines are also validated using the finetuning process and show additional small improvements of the metal loss profiles.

Presenting Author: Kevin Siggers ROSEN Group

Presenting Author Biography: Kevin Siggers graduated from the University of British Columbia with a Masters of Applied Science and has been a Pipeline Integrity Engineer for 15 years. Kevin began his career with Enbridge Liquid Pipelines and FortisBC before his move to ROSEN. Over the last 8 years at ROSEN Kevin has focused on new ways of handling the flow of information with clients and using novel techniques to improve the quality of information provided.

Authors:

Xiang Peng ROSEN Group
Kevin Siggers ROSEN Group
Mark Wright ROSEN USA
Johannes Palmer ROSEN Group