Research On The Magnetic Memory Detection And Feature Extraction Of Ferromagnetic Component

The large ferromagnetic components (such as heavy-duty gearbox, gantry cranes,steel converter, etc.) often work in low speed, heavy load, high temperature, highpressure, corrosion and other harsh conditions, in the event of failure will bring hugebusiness economic losses. If the equipment is in working condition, the vibrationsignals and acoustic emission signal can be seen as the information carrier ofcondition monitoring and fault diagnosis, but when the equipment stops running orrunning in latent fault condition, abnormal vibration signals and acoustic emissionsignals is weak, even valid abnormality information can not be detected. while theheavy load operation of the equipment can generate local stress concentration andsubtle deformation and irreversibly retained in the form of magnetic memory,thereby it laid a theoretical foundation for assessment of the state and early faultdiagnosis. Aiming at large ferromagnetic components, this dissertation focuses onpractical techniques of magnetic memory fault diagnosis. The main contents are asfollows:(1) The basic principle of the metal magnetic memory testing technology isdescribed. The tangential component magnetic memory signal has a peak and thenormal component has zero-crossing feature and a peak gradient. Both of themincreases when the degree of stress concentration increases, so it can effectively detectthe ferromagnetic component failure, especially the stress concentration areas can notbe observed with the naked eye. The detection of gear and gantry cranes failure candetermine the stress concentration area, it verify the effectiveness of early fault featureextraction with metal magnetic memory testing technology.(2) Aiming at the non-stationary characteristics of the magnetic memory signal, anew method based on the Intrinsic Time-Scale Decomposition (ITD) is proposed toachieve the extraction of magnetic memory signal. Firstly, the magnetic memorysignals are decomposed into several proper rotation components (PRC) and a drabtendency item by ITD. Then reconstruct the first four order PRCs to eliminate the bigcycle composition of magnetic memory signal and magnetic noise to improve thesignal-to-noise ratio.(3) The SVM (Support Vector Machine SVM) method is used to extract thedifferent states of the gear status multidimensional characteristics vector value, thenestablish training samples to assess the gear stress state which provide a strong basisfor early warning mechanisms of gear differences. It makes a difference warning decision method to solve the lack of gear latent failures case with magnetic memorytesting.(4) Aiming at strong noise characteristics of magnetic memory signal, theMRSVD (Multi-resolution Singular Value Decomposition) method is used to processmagnetic memory signal. The detail signal the noise component of the magneticmemory signals, the approximate signal corresponding to the effective magneticmemory signals after denoising. The experiment results of a certain yard cranemagnetic memory signal processing shows that the method can improvesignal-to-noise ratio of the reconstructed signal effectively, and judge the stressconcentration area accurately, it will lay the foundation for the stress state evaluationand early fault diagnosis.(5) Aiming at dependence on imported and high-cost of the metal magneticmemory equipment, combined with knowledge of digital circuitry and analogcircuitry, a three-dimensional magnetic memory testing instrument is developed basedon high-resolution weak magnetic sensor HMC1043(three-dimensional magneticfield sensor), microcontroller, LCD hardware etc.Satisfactory results have been achieved when using above method to extract thefeature from magnetic memory signal, it has broad application prospects innon-destructive testing and early fault diagnosis.