|
In recent years , magnetic flux leakage detection method is widely used , especially in the detection of corrosion defects research and extension applications, magnetic flux leakage testing has become a relatively mature detection technology. During their service in tanks and pipes , in addition to the presence of corrosion defects , may also be a greater risk of the formation of many cracks . Because crack MFL results were affected by many factors , which gives crack MFL as well as evaluating add a great deal of difficulty. In this paper, MFL detected as a research object, the integrated use of theoretical analysis, finite element analysis , experimental research methods MFL detection and analysis , and application of neural network technology MFL detection signal identification . This simplifies the tank bottom surface cracks into the shape of several relatively simple analytical model that is V-shaped face cracks, as well as combinations of rectangular -shaped end face crack crack . Equivalent in theory apply the crack band dipole magnetic flux leakage were analyzed ; and Maxwell's equations as the theoretical basis , using ANSYS finite element software crack leakage field analysis of finite element simulation analysis reveals that the crack depth, width , aspect ratio , parallel to the crack spacing parameters on crack crack leakage magnetic field , and analyze external testing conditions, such as sensor lift-off , the pole piece gap height of the crack leakage magnetic field ; according MFL inspection project the actual situation , the formation of cracks MFL experimental system through experiments on dipole model analysis and finite element analysis conclusions into the form validation. The crack leakage field using ANSYS finite element simulation and analysis of the data , the leakage magnetic field characteristic quantity extracting crack , crack shape obtained by analyzing the parameters of the relationship between the amount of magnetic flux leakage characteristics . According to the characteristics of magnetic flux leakage signal , combined with neural network theory to construct the crack geometry parameters predicted BP neural network model, the integrated use of finite element analysis and experimental data obtained in the analysis of the data obtained , as BP network training samples , training BP neural network, crack depth and width characteristic parameters to identify and verify that the network on the crack defect prediction reliability.
|