After to be preprocessed, such as elimination of noise, image enhancement and so on, pipeline MFL (Magnetic Flux Leakage) image will be segmented with max entropy algorithm and dilation algorithm, the position, width, etc. of defect can be calculated according to defect image. Quantitative recognition of the length, the depth of defect can be realized by use the BP neural network, the recognition result error is smaller than 10 %. After use two methods ,which is the weighted average method and the autoadapted weighted average method ,to fuse the radial and axial MFL signals, the BP neural network recognition results precision and reliability received improvement.
Study on intelligent quantitative recognition of defect in pipeline magnetic flux leakage inspection
Untersuchung einer intelligenten quantitativen Fehlererkennung in Pipelines mit Magnetpulverprüfung
2006
5 Seiten, 8 Bilder, 2 Tabellen, 5 Quellen
Conference paper
English
Magnetic flux leakage device for offshore oil pipeline defect inspection
Tema Archive | 2005
|Pipeline Inspection by Intelligent High Resolution and Conventional Magnetic Flux Leakage Pigs
British Library Conference Proceedings | 1992
|Defect separation considerations in magnetic flux leakage inspection
Tema Archive | 2005
|Flux-leakage vehicles pass tests for pipeline inspection
Tema Archive | 1988
|