The Defect Parameter Predicted And Quantitative Analysis In MFL Inspection

There are many ways in industry no damage inspection, such as eddy inspection, magnetic powder inspection, magnetic flux leakage (MFL) inspection, ultrasonic inspection and so on. Among them, MFL inspection has many advantages, such as quickly speed, high detection rate, high sensitivity, operation easily and so on. MFL inspection can test in varieties complex environment, and has a characteristic of realizing quantitative inspection of the defect parameter easily. With the development of modern industry, people have higher quality requirements for pipeline inspection. When inspect, not only find the defect, but also quantitative analysis to the defect especially. To realize the defect parameter accurately prediction and quantitative analysis, Not only understand the relation between defect parameter and MFL field, but also study the problem anti-push defect parameter from MFL signal.In this paper, it firstly use magnetic dipole to study the relation between defect and MFL field with one kind of cylinder magnetic dipole model to describe the corrosion pits , cracks , holes, grooves and other types of common defect by changing the long -axis, branchy-axis and height of cylinder. Not only get the change rule of defect and MFL field, but also achieve a kind of magnetic charge model to describe a wide range of defects.Secondly, mention one kind of the system analysis model of MFL inspection to analyze the forward problem (calculate MFL field from given impel source and defect which is known parameter) and an-reverse problem (estimate defect parameter or contour from given MFL field) in MFL inspection. Furthermore, study several kinds of common methods and analyze its characteristic about defect parameters prediction.And then, prospers iterative reverse solution algorithm based on the wavelet neural network to solve the problem of defect quantitative detection in the MFL inspection, Among them, treat wavelet neural networks as solution forward process, using MFL signal which obtained by test piece as training samples to train wavelet neural network, to take formation the reflection from defect parameters to the test sample signal; use genetic simulated annealed algorithm to solve optimizing the process.Finally, it can verify the result of defect quantitative prediction and prove that the accuracy and reliability of method by the experiment.