FREE ENBIS Webinar by Kieran Peel: "Gaussian Process Regression Modelling of Pipe Defects using Magnetic Flux Leakage Technology"

2 July 2020; 12:30 – 13:15

Kieran Peel hints at how to predict pipeline defects out in the field. The webinar will be moderated by Shirley Coleman

Gaussian Process Regression Modelling of Pipe Defects using Magnetic Flux Leakage Technology

Kieran Peel, Shirley Coleman, Jian Shi

Metallic pipelines are widely used to transport water, gas and sewage and need to be regularly checked for defects that may lead to leakage. Bespoke equipment developed by a small to medium enterprise (SME) records changes in magnetic flux through hall sensors passed along the pipes. Statistical models of these signals can be trained using sample pipes with known defects. The models can then be used to give information about the location and size of defects in pipes in the field. The relationship between signals and defect size and shape is complex and is affected by properties of the pipe such as wall thickness. Gaussian Process Regression models have been shown to fit the data and have the potential to be flexible enough to meet future business needs. The webinar describes the models and how they have been fitted, presents the outcomes for a sample of steel pipes, and discusses the implementation in practice.