An anomaly detection method for rotating machinery monitoring based on the most representative data

摘要: With the development of concepts industry 4.0, condition monitoring techniques are changing. Large amounts generated data require diagnostic procedures to be automated, which drives need for new and better methods autonomous interpretations vibration data. However, if operational, they verified under real industrial conditions compared with well-established expert-based techniques. This article introduces novel algorithm preprocessing nearest-neighbor-based anomaly detection. approach is validated on machinery in a series case studies. The population over-hung centrifugal fans, employed same process, were monitored continuously according proposed methodology an extended time period. Piezoceramic accelerometers used register time-domain processed extract several signal features serve as inputs detection algorithm. solution approach. Presented include not only intact state but also machine breakdown serious deterioration condition. influence maintenance work presented article. Authors show data-driven monitoring, can one many predictive