A Degradation Identification Method Combining Cost Matrix and Prediction Window Width for Mechanical Equipment

To improve the efficiency of preventive maintenance management for complex mechanical equipment or components, it is necessary to identify the degradation state from both historical and online data. It may be more important to recognize the alarming state due to the high downtime cost. However, uncertainties of monitoring data may reduce the prediction accuracy. Therefore, a new self-organizing map-based method combined with misclassification cost matrix and window width of prediction (SOM+COST+WWP) is proposed to improve the prediction accuracy of the alarming state. Case studies show that SOM+COST+WWP has advantages for higher recognition accuracy for the degradation state compared with other SOM-based methods.