Comprehension and Limitations of Different Types of Data-Driven Condition Monitoring Systems for Electrical Machines in the Context of Bearing Condition Monitoring

In the field of bearing condition monitoring, the utilisation of disparate data-driven condition monitoring systems (DD-CMSs) presents a multitude of advantages and challenges. This paper assesses the efficacy and constraints of three principal types of condition monitoring systems: fault classification, fault detection, and anomaly detection. The study employs data sets such as the Case Western Reserve University (CWRU) data set and from a published research in 2024 set to evaluate system performance across a spectrum of training scenarios and data conditions. The findings were also evaluated a second time by using a synthetic data set of commutator surface faults. The necessity for extensive data sets to enhance the reliability of condition monitoring systems is highlighted.