非侵入性的电动机的转子条的故障诊断问题属于难问题[借鉴].pdfVIP

精品资料 1 Introduction The problem of non-invasive diagnosis of the faults of rotor bars in the induction motor belongs to difficult problems [5, 7]. Motor with this type of damage is still functioning but such work causes increased currents at both sides of the broken bar. This causes the damage of the next bars. The result of this avalanche process is the damage of all bars, and finally the machine stops functioning. Thus,the most important problem is to catch the moment when only one or at most few bars are broken (the beginning of the avalanche process). There has been a lot of research reported over the past years devoted to the development of various steady state condition monitoring techniques. Most of them use Fourier transformation of the stator current in a steady state [4, 14, 19]. Some apply more sophisticated method of wavelet analysis of stator current in transient state [18]. There are also solutions relying on the analysis of the magnetic field space vector orientation [13]. All these methods of current preprocessing are combined with different tools of analysis of the results of this preprocessing stage, forming the final classification stage. Among them, we can mention the statistical approach to classification [4, 6, 8], the artificial neural networks [5, 9, 17] or support vector machine [1, 3, 10]. They are responsible for the automatic recognition of fault. The observation of motors working in the normal state or at faulty conditions of the bars allows to point out some typical symptoms 可编辑修改 精品资料 indicating the bar faults [4