Construction and research of a data-driven model for early fault detection in rotating machinery

The advent of the industry 4.0 era has precipitated unparalleled transformations within the manufacturing sector, with the condition monitoring of energy equipment in smart factories being of particular significance. Conventional monitoring systems, which utilise fixed threshold alarm methods categorised by equipment type, frequently result in false alarms or missed detections within intelligent manufacturing environments. The present study introduces a data-driven model for early fault detection in rotating machinery. The model utilises advanced signal processing and machine learning techniques to analyse extensive operational data, facilitating real-time monitoring of equipment operating states. The model’s foundation lies in the integration of time-domain and frequency-domain features, in conjunction with an enhanced Gini coefficient, Principal Component Analysis (PCA), T²statistical analysis, SPE statistical analysis, and time-delayed stochastic resonance feedback technology. This results in the construction of an early impact-type fault detection model for rotating machinery, based on vibration signal analysis. The model’s validation was conducted using accelerated life test data for rolling bearings from the Joint Laboratory for Mechanical Equipment Health Monitoring, Xi’an Jiaotong University. The outcomes demonstrate that, in comparison with conventional vibration intensity-based alarm methods, this early impact-type fault detection model can detect faults earlier and more accurately, effectively highlighting fault characteristics.