Machine learning based oil pipeline diagnostics

The magnetic flux leakage (MFL) method is the most common approach for non-destructive testing of oil and gas pipelines. As a result of MFL detection, magnetograms are obtained, often analyzed by semi-automated methods, which leads to a decrease in accuracy and an increase in analysis time. The paper proposes a new CNN architecture for automatic image classification based on magnetograms for oil pipeline diagnostics. As a result of testing the developed algorithms on a deferred sample, the high accuracy and efficiency of the developed solution were proved.