The Study Of Feature Criterion Of Metal Magnetic Memory Signals Of The Welding Crack

In morden Industry, with the products developed towards high speed, hightemperature and high load, working conditions for equipments are becoming harsherthus performance requirements are higher. Sometimes equipments have already beendamaged in less than a normal inpecting cycle, causing serious consequences andlosses. Numerous ferromagnetic structures, especially pressure containers like boilers,pipelines, bridges, railways, steam turbine laminas, rotors and important welding parts,working as bearing structures and enduring fatigure loads to different degrees, mayengender fatigure cracks and damages under high stress concentration.Metal magnetic memory (MMM for short) inspecting method is a brand newdiscipline of non-destructive testing. It is the only technique which can be used forearly damnification assessment for ferromagnetic structures, and an NDT methodutilizing self-emitted information. Utilizing self-emitted information in geomagneticfield, it works by recording leakage field in stress-concentrating areas in equipmentsand metal structures under service load.Because MMM signals of welding cracks are complicated and identification ofsignals largely depend on experiences, popularization and application of MMMmethod become difficult. To promote quantified application of MMM method inwelding quality inspecting, MMM features for welding cracks are studied in thispaper.Based on a self-designed tensile test, adopting several signal processing methodssuch as power spectral density (PSD for short) evaluation, MMM signals areprocessesd and their features studied. Five informative features of MMM signals arefound out —— difference of maximum and minimum value of original signal,maximum value of the detail component, difference of maximum and minimum valueof detail component, maximum Burg PSD value of detail component, maximum peakvalue of FFT amplitude of 1st scale detail signal with original signal decomposed into4 scales.Finally, synthesizing all feature analyses, adopting principal component analysis,a model of MMM signal features is established. Two principal constituents denotingorigin of essence of MMM signals and degree of stress concentration respectively, arefound out. Results of model testing indicate that, abstracted features can be used ascriterions of existence of welding cracks.