IEEE Transactions on Magnetics · 2006 · 94 citations · 4 references
ComputingComputer ScienceElectrical EngineeringComputer VisionFeature EngineeringMachine LearningDeep LearningDefect DetectionFault DetectionLeak DetectionSignal ProcessingSinkhole DetectionDepth Profile AnalysisLeakage FluxMachine Learning-based Defect DetectionMagnetic Flux Leakage InspectionMagnetic Leakage AssessmentPipeline Condition MonitoringRadial Basis Function NetworksWavelet Analysis
Magnetic flux leakage (MFL) is routinely used to detect cracks and corrosion in natural gas pipelines, with defect dimensions (length, width, depth) extracted to estimate maximum safe operating pressure (MAOP). The study aims to improve MAOP estimation by inverting MFL signals to recover the full 3‑D depth profile of defects. An iterative inversion method employing adaptive wavelets and a radial basis function neural network reduces data dimensionality and predicts the 3‑D depth profile. Simulated data demonstrate the method’s ability to recover 3‑D defect profiles, indicating promising accuracy for MAOP estimation.
Natural gas transmission pipelines are commonly inspected using magnetic flux leakage (MFL) method for detecting cracks and corrosion in the pipewall. Traditionally the MFL data obtained is processed to estimate an equivalent length (L), width (W), and depth (D) of defects. This information is then used to predict the maximum safe operating pressure (MAOP). In order to obtain a more accurate estimate for the MAOP, it is necessary to invert the MFL signal in terms of the full three-dimensional (3-D) depth profile of defects. This paper proposes a novel iterative method of inversion using adaptive wavelets and radial basis function neural network (RBFNN) that can efficiently reduce the data dimensionality and predict the full 3-D depth profile. Initials results obtained using simulated data are presented
4