A new computational method for stator faults recognition in induction machines based on hyper-volumes

To remain competitive, the manufacturing industry is always innovating and developing new cost-efficient ways to produce goods. That is why today, extensive automation is applied in nearly every type of manufacturing and assembly processes. Automation improves productivity, quality and robustness of products. It also increases the predictability of production lines mainly constituted of asynchronous machines. These machines, however, need regular maintenance. Time-based maintenance is labor-intensive, ineffective in identifying problems that develop between scheduled inspections, and is not cost-effective. For these reasons, researchers and companies are now investigating new methods to develop what is called preventive maintenance. It involves the use of sensors (vibrations, load cells, electrical, etc.) placed on the machine to monitor its actual state in order to detect engine failures. For some years, works presenting interesting methods and results [1-35] have been published, but few of these investigated effective preventive maintenance capable to clearly characterize the type and the importance of failures. In this paper, we propose a new computational approach for detection and characterization of stator faults of asynchronous machines based on electrical signal analysis. Our method is able to detect, locate, and quantify the severity of a failure. To do so, we use the frequency characteristics [6, 7] for simple detection, the currents [6] and the performance speed of the induction machine for localization and quantification of the failures. Moreover, we exploit hyper-volumes in the model of defective asynchronous machines. We present an experiment conducted on a model which shows promising results.