English 
日本語 
한국어 
Deutsch 
Español 
italiano 
русский 
português 
العربية 
tiếng việt 
ไทย 
中文
NEWS
In equipment maintenance and management, corrosion detection is a crucial step. Although traditional detection techniques are effective, with technological advancements, pulse eddy current testing has...
Using the principle of electromagnetic induction, the ECT NDT method can assess certain properties of the conductive material and its components through the change of induced eddy current in the tested workpiece, or finding defects. ECT NDT is one of the main methods to control the quality of various metal materials and a few non-metals (such as graphite, carbon fiber composite materials) and their products. Compared with other non-destructive testing methods, ECT NDT is easier to achieve automation, especially for pipes, rods and wires and other profiles.
Eddy current is when a conductor is placed in a changing magnetic field, because there is a vortex induced electric field around the changing magnetic field, the induced electric field acts on the free charge in the conductor to move the charge and form an eddy current.
Eddy current non-destructive testing is based on Faraday's law of electromagnetic induction, connecting alternating current to the detection coil to generate an alternating magnetic field perpendicular to the workpiece. When the detection coil is close to the workpiece to be inspected, the surface of the workpiece induces eddy currents and simultaneously generates a magnetic field opposite to the original magnetic field, which partially offsets the original magnetic field, resulting in changes in resistance and inductance of the detection coil. If there is a defect in the metal workpiece, the intensity and distribution of the eddy current field will be changed, and the impedance of the coil will change. By detecting the change, it can be judged whether there is a defect.
With the development of microelectronics and computer technology and the adoption of various signal processing technologies, eddy current testing transducers, eddy current testing signal processing technology and eddy current testing equipment have made great progress.
ECT NDT needs to improve the signal-to-noise ratio and anti-interference ability of the signal, realize signal identification, analysis and diagnosis, in order to obtain the best signal characteristics and detection results.
1. Signal feature extraction
Feature extraction methods commonly used in ECT NDT include Fourier description method, principal component analysis method and wavelet transform method.
Fourier description method is a common method to extract feature values. The advantage is that it is not affected by the probe speed, and the impedance map can be reconstructed by this description method. The more the number of sampling points, the closer the reconstructed curve is to the original curve. But this method is only sensitive to the shape of the curve, it is not sensitive to the zero point and gain of the eddy current detector, and does not change with the curve rotation, translation, size transformation and starting point selection.
The method of using the eigenvalues and eigenvectors of the autocorrelation matrix of the test signal to describe the signal characteristics is called the principal component analysis method, which has strong resolution for similar defects.
Wavelet transform is an advanced signal time-frequency analysis method. The multi-resolution analysis in wavelet transform is applied to the eddy current detection signal analysis, and the different wavelet coefficients are processed before reconstruction. The signal-to-noise ratio of this signal processed by wavelet transform will be greatly improved.
2. Signal analysis
Artificial neural network
The input vector of the artificial neural network is the characteristic parameter of the signal. The correct selection and extraction of the characteristic parameter of the signal is the key to the success of the intelligent judgment of the neural network. The combined neural network model adopts the hierarchical discrimination method to reduce the dimensionality of the network input variables from N2 to N. The network structure is greatly simplified, the training speed is fast, and it has a high defect recognition rate and practical value.
Neural network can realize defect classification, has the advantage of high recognition accuracy, and is also effective for incomplete and insufficiently clear data.
Information fusion technology
Information fusion is the multi-level processing of detection, correlation, correlation, estimation and synthesis from different information sources to obtain a unified best estimate of the measured object.
The fusion of eddy current scanning images decomposes the image into multiple sub-band images, and uses a fusion algorithm in the conversion area to achieve image fusion. Ka Bartels et al. used the best method of signal-to-noise ratio to combine the eddy current signals, and used the spatial frequency compensation method to make the high-frequency signal blur and the low-frequency signal clear. Z Liu et al. used the maximum value criterion to select the discrete wavelet transform coefficients of different signals, and selected the maximum absolute value of the coefficient to be fused as the combined conversion coefficient. Therefore, the fusion signal can be reconstructed based on these coefficients using inverse wavelet transform. Wavelet transform can effectively extract salient features at different scales. In the process of fusing the signals, all useful features of the signals are preserved, so the internal and surface defect information is enhanced.
(3) Solving the inverse problem of ECT NDT
The signals detected by the eddy current testing equipment contain information such as the position, shape, size, and properties of the medium, and the known signal is used to inversely infer the medium parameters (conductivity) or shape (defects), which is an inverse problem in the electromagnetic field theory.
In order to solve the eddy current inverse problem, a mathematical model for defect identification must be established first. There are models for artificial defects with regular shapes, natural defects with complex boundaries, single defects and multiple defects. In terms of medium types, there are models for composite materials and surface permeability changes of the tested part.
With the development of computer technology, various numerical solutions of defect models have also made progress. There are finite element method, moment method and boundary element method and so on.
