| Operational safely of ferromagnetism hardware is directly related to the regular development of national economy, people's lives and property security. The traditional NDT (ultrasonic testing, magnetic powder testing, X -ray testing) only can test macroscopic flaws existed, but cannot prevent accidental fatigue damage of equipment, this is the main reason to generate equipment damages and accidents, which is solved by metal magnetic memory inspection technology. However, its technological development process is relatively short; there still has a lot of work to do on theoretical study and application. For example:â‘ Study of basic theory of magnetic memory phenomenon and microscopic mechanism;â‘¡The metal magnetic memory signal analysis processes and confirmative experiments;â‘¢Qualitative and quantitative analysis method of magnetic memory inspection signal and damage of examination objectives;â‘£Specific study of equipment inspection method and inspection standard making and so on.Under the circumstance, wavelet analysis method developed fast recently is adopted to analyze and process the gathered metal magnetic memory testing signals and characteristic quantity extraction research.This paper mainly focuses on:1,Based on metal magnetic memory testing mechanism, the following conclusions are got: Magnetic memory testing signal mainly uses the normal vector signal to recognize and judge faults. Original environment had better not been separated during testing; The sample tested and sensor level had better been laid horizontally; Small effect of inspection speed to signals, and non-steady spatial domain of magnetic memory signals. The noises in signals mainly result fromâ‘ Survey noise;â‘¡Probe head vibration;â‘¢Non-flaw factor of measured articles.2,Digital smooth technology is adopted. Data are pretreated by eliminating wild value points, after gathered multi-channel data are average weighted; Spectral analysis is finished by using Fourier transformation, which confirms that the main energy of magnetic memory signal concentrates on the low frequency part, based on the magnetostriction effect.3,On the basis of non-steady characteristic of metal magnetic memory signal, various ingredients which has good time - frequency characteristic, can analyze and process signals with different scales, make the singular point discontinuous points of get bigger, and enhances the resolution of signals and signal-to-noise ratio. It is new and significant to introduce wavelet transformation into data analysis in magnetic memory domain. Through studying spatial frequency of metal magnetic memory inspection signals by means of wavelet transformation theory, it is indicated that in different spatial frequency sectors, the survey noise mainly is the high frequency ingredient, which corresponds small scales, separate from noises and superficial deposit; signals generated by the carriage are mainly low frequency components, which corresponds big scales. The signal wavelet transformation is strongly related and its peak value will not be reduced with the increase of scale in many small scales, so that the truth signals can be found out, and reconstruct signals and effectively eliminate noises.4,By the application of various wavelet methods as compulsory denoising, experimental denoising, wavelet index denoising, self-adapted threshold value denoising, second denoising, and small wave packet denoising to analyze and process metal magnetic memory signal, the signal-to-noise ratio can be increased and noises can be removed, and data are quite smooth after filter processing. In light of wavelet multi-resolution analysis, it is better for wavelet function to choose Db and Sym series from the perspective of signal-to-noise ratio and evenness. Based on qualitative analysis, the former analysis methods can meet the requirements. However, refinement lamination method must be used to precisely position accidents points based on quantitative analysis. In addition, wavelet decomposition layers, valve value determination and the selection of wavelet generating function, all of which affect the final result of analysis processing.5,It is proposed that the peak-to-peak value,the difference ultra limiting value,the gradient and the zero crossing spot should be combined together to measure the stress concentration degree. Singularity of wavelet transformation is the index which is used to measure the concentration degree. Various characteristic quantities are used to judge the overall characteristic to magnetic memory signals, which enhances the accurate qualitative and quantitative judgment for misfunction.6,Based on MATLAB , many kinds of wavelet analysis algorithm programs had been made, which has good effect through a great deal of data tests. But its usability and versatility still had to be studied further. |
