Multi-parameter Simulation Neural Network Algorithm for Wire Rope MFL Detection

Magnetic flux leakage (MFL) detection is the most effective method for nondestructive testing of wire ropes. Neural network is the main processing method of MFL signals. However, it is difficult to use a lot of measured samples to make the calculation accuracy of neural network meet the requirements, and it cannot be universally applied to a variety of wire ropes. It is difficult to promote in actual testing. This paper proposes a neural network quantitative calculation method based on multi-parameter simulation. Only a small number of measured samples can achieve high accuracy in detection, and it can detect a variety of wire ropes without retraining the neural network. The magnetic dipole model and three-dimensional model simulation are used to calculate the leakage magnetic field distribution of the wire rope, analyze the law of the defect leakage magnetic field signals, prove that the finite element simulation can accurately analyze, establish the connection between the actual defect and the simulated defect. A lot of samples are trained through multi-parameter simulation, and only a few samples are needed to achieve quantitative detection of MFL detection. It is verified by the experimental results of the wire rope that only two measured samples are needed, the error of the defect width is less than 10 mm, and the error of the section loss rate is less than 1%.