A generic diagnosis protocol for the monitoring of induction motors based on multiple statistical references in the torque-speed plane

This article introduces an original and generic monitoring protocol suitable to build robust references for the diagnosis of mechanical drives. In a general way, an efficient diagnosis scheme is related to the ability of properly separate healthy and faulty operations. However, during healthy working, the level of a fault indicator may significantly vary with the load characteristics or the system's operating point. Consequently, a threshold-based distinction between healthy and faulty operation may be blurred by operating point changes in the torque-speed plane. This issue is however little discussed in literature and most diagnosis schemes are designed for a system whose behaviour is well known and for specific operating conditions. An innovative diagnosis strategy is therefore presented in this article to overcome these limitations. It is based on two principles : an adequate segmentation of the torque-speed plane and the design of multiple statistical references. They are built during healthy conditions by means of statistical tools and enable an efficient diagnosis, independent of load and speed conditions. After a detailed description of this method, experimental results demonstrate its performance for the detection of a mechanical unbalance occurring in an electro-mechanical system, regardless of the operating conditions.