Research On Methods Of Interference Elimination And Defect Detection In Multi-frequency Eddy Current Nondestructive Testing

As one new eddy current nondestructive testing (NDT) technology, multi-frequency eddy current testing (MFECT) has the merits of interference elimination in testing procedure and multi-parameter detection of unit under test (UUT). This paper lucubrated signal parameter estimation and mixing, crest factor optimization of synchronously synthetical excitation and spectrum analysis of testing signals, linear frequency modulation excitation and zoom spectrum analysis of testing signals, multi-frequency analysis of pulsed excitation, et al in MFECT technology, and these are applied in interference elimination, defect detection and so on. The brief of these researches and the novel approaches are as follows:Based on time harmonic electromagnetic field theory, physical model and calculation method of MFECT are discussed. Induced eddy current distribution and impedance characteristic of the probe coil are simulated and researched with finite element method. According to Maxwell law, time harmonic electromagnetic field problem based on sine signal excitation is expressed mathematically with electromagnetic potential. The typical eddy current testing problem with an aircore coil is simplified and analyzed in cylindrical coordinate, and then the coil's equivalent impedance affected by eddy current effect is analyzed. The arbitrary eddy current testing problem is expressed and calculated with finite element method. The common data processing methods of MFECT including linear algebra method, phase rotation and subtraction, spectrum analysis method are introduced. The eddy current testing problem is modeled with Ansoft Maxwell finite element software, and then eddy current distribution and coil's equivalent impedance characteristic influenced by excitation frequency, lift-off height, UUT's thickness, defects and so on are simulated and researched.MFECT technology based on parameter mixing is researched. Testing signal's parameter estimation method based on orthogonal lock-in amplifiers is introduced and a new method based on linear least square method is proposed. Support interference elimination in tube testing and lift-off effect elimination in dual-frequency eddy current testing are researched. When the amount of frequency components is relative small in MFECT, two parameters of each frequency component are extracted and mixed to characterize UUT. Parameter estimation method based on band pass filters and orthogonal lock-in amplifiers is discussed but it takes a high hardware cost. Sine signal linearzed, a new method for MFECT signal's parameter estimation based on linear least square method is proposed, and it reduces the hardware cost. Support interference elimination method is researched. The support interference is removed based on phase rotation and subtraction, so the defect signal is enhanced. Lift-off effect elimination in dual-frequency eddy current testing is researched. The component variation difference between lift-off and defects is analyzed and the defect zone is proposed to distinguish lift-off signals and defect signals and to realize lift-off effect elimination.MFECT technology based on synchronously synthetical excitation and spectrum analysis method is researched. Crest factor optimization of the multi-frequency excitation signal based on genetic algorithm is proposed. The eddy current sensor with a rectangular excitation coil is adopted. The length of defects is quantitatively detected based on spectrum energy variation, and the classification of defects is realized based on principal component analysis (PCA). The eddy current sensor with a rectangular excitation coil is designed which can convert eddy current field in the surface of UUT to a uniform field. Defect characteristic can be identified by detecting the disturbance field which arises from defects. When the multi-frequency excitation signal is generated in synchronously synthetical mode, the crest factor of excitation signals will increase with the amount of frequency components rapidly. Therefore, that broadens the work voltage range of eddy current sensor's driving circuit. An optimization method of excitation signal's crest factor based on genetic algorithm is proposed. A fitness function is defined and the optimized initial phase of each frequency component is got. Compared with direct search algorithm, this method improves the search speed obviously. When the eddy current sensor is drived with a multi-frequency signal, the length of defects can be quantitatively detected based on spectrum energy variation. The spectrum energy reaches maximum when the eddy current sensor comes into defects. The spectrum energy reaches minimum when the eddy current sensor departs from defects. The length can be calculated with the time span between maximum and minimum and the moving speed of the eddy current sensor. Each frequency component is affected differently and the trend of spectrum curve varies differently when different types of defects are detected. The signal dimension is reduced based on PCA and defect classification is realized.MFECT technology based on frequency modulation excitation and zoom spectrum analysis method is researched. Modulated MFECT is proposed, and five features including spectrum energy, spectrum barycenter, spectrum kurtosis, spectrum skewness and spectrum difference's peak frequency point are defined to realize identification and classification of defects and depth detection of inner defects. Modulated MFECT technology proposed in this paper adopts frequency modulation signals as the excitation signal of eddy current sensors. After a multi-level amplification, testing signals are sampled and analyzed in frequency domain to realize interference elimination or multi-parameter detection. Compared with conventional MFECT technology, the crest factor of excitation signals is reduced because of frequency modulation excitation, and the testing time is reduced because that the frequency modulation signal is a continuous excitation of many frequency components and the multiplex module is not needed. As a result, the discrete spectrum of testing signals becomes continuous and it will be easily characterized and analyzed. But this technology can't satisfy the need of component parameter mixing. The linear frequency modulation is adopted to generate excitation signals and testing signals are amplified, sampled and analyzed with zoom spectrum method (Chirp-Z Transform). Tukey window function is added to testing signals to eliminate ripples in spectrum curves. According to spectrum characteristic, five features such as spectrum energy, spectrum barycenter, spectrum kurtosis, spectrum skewness and spectrum difference's peak frequency point are defined to realize multi-parameter detection of defects. The eddy current testing experiment shows that spectrum energy can be used to identify defects, spectrum barycenter, spectrum kurtosis, spectrum skewness can be used to classify defects, and spectrum difference's peak frequency point can be used to detect inner defect's depth position.Pulsed excitation is considered as a multi-frequency signal, and multi-frequency analysis method of pulsed eddy current testing (PECT) is researched. Spectrum relative variation (SRV) in amplitude frequency spectrum is proposed to classify defects, and phase cross zero point (PCZP) in phase frequency spectrum is proposed to eliminate lift-off effect and classify defects. Generally speaking, PECT can be considered as a kind of MFECT technology, and its time domain analysis method has been investigated widely. The duty ratio and time shift position of pulsed excitation signals is compared and selected, then the excitation signal is expressed as a series of unequal amplitude harmonics and testing signal's spectrum characteristic is analyzed. Because the energy of each harmonic component decreases with frequency rapidly, relative variation in testing signal's amplitude frequency spectrum can be used to reflect the variation characteristic of each harmonic component influenced by different types of defects effectively. SRV in amplitude frequency spectrum is proposed to classify different types of defects. This feature with clearly physical meanings varies obviously. High frequency harmonic component's SRV of surface defects keeps constant or has a little upward trend, but that of inner defects has an obvious downward trend. The reason is that the skin depth of high frequency harmonic components becomes smaller. In phase frequency spectrum, the phase value of low frequency harmonic components are positive and it decreases with frequency gradually to negative. PCZP is proposed from that. When the lift-off of the eddy current sensor varies, PCZP keeps constant. PCZP of surface defect's testing signals become bigger than that of no defect. PCZP of inner defect's testing signals become smaller than that of no defect. When the defects are too tiny, this feature varies a little and the classification function is not evident.MFECT systems for parameter mixing and spectrum analysis method are designed and implemented. MFECT systems are more complicated than other eddy current testing systems. DDS technology is adopted to generate sine signals in MFECT system for parameter mixing, and these sine signals are added to synthesize the multi-frequency excitation signal. Then the parameters of weak eddy current testing signals are estimated based on band pass filters and orthogonal lock-in amplifiers. MFECT system for spectrum analysis method is realized based on virtual instrument (VI) technology. Arbitrary multi-frequency excitation signals are generated by a D/A module. The weak eddy current testing signal is amplified enough and sampled, and then it is converted to frequency domain and analyzed.