期刊名称：Leonardo Electronic Journal of Practices and Technologies
印刷版ISSN：1583-1078
出版年度：2009
卷号：8
期号：15
页码：1-14
出版社：Academic Direct Publishing House
摘要：Self-Organizing Maps (SOM) is an excellent method of analyzingmultidimensional data. The SOM based classification is attractive, due to itsunsupervised learning and topology preserving properties. In this paper, theperformance of the self-organizing methods is investigated in induction motorrotor fault detection and severity evaluation. The SOM is based on motor currentsignature analysis (MCSA). The agglomerative hierarchical algorithms using theWard’s method is applied to automatically dividing the map into interestinginterpretable groups of map units that correspond to clusters in the input data. Theresults obtained with this approach make it possible to detect a rotor bar fault justdirectly from the visualization results. The system is also able to estimate theextent of rotor faults.
