Abstract: The degradation feature of components based on vibration signals is easily affected by noise and therefore cannot accurately reflect their operation state. Considering both degradation features and fault features, we propose a condition-based maintenance model. A progressive fault feature identification method based on adaptive singular spectrum decomposition is first created, followed by the description of component reliability evolution by the proportional hazard model and the establishment of a joint maintenance model where the cumulative number of fault features is identified as the spare parts ordering threshold and the maintenance cost rate is used as the optimization objective. Finally, the accelerated life test data of a rolling bearing is used as an example to verify the effectiveness of the proposed method. The results show high diagnostic accuracy of adaptive singular spectrum decomposition for different signals, effectively reducing the number of manual diagnoses. The joint decision making method with the cumulative number of fault feature identification as the spare parts ordering threshold can compensate for the lack of uniqueness of the degradation feature thresholds and further obtain the optimal maintenance cost rate.

Key words: condition-based maintenance, spare parts ordering, fault feature, vibration monitoring, bearing, reliability analysis