A Degradation Analysis and Prediction Method Oriented to the Whole-Life-Cycle Measurement Signal of Wind Turbine Gearboxes

Given the complexity of whole-life monitoring data processing for wind turbine gearboxes, this paper proposes a degradation analysis and prediction method oriented to the whole-life-cycle signal. First, based on historical measurement data and combined with fault category and condition assessments, the whole-life-cycle measurement signal of wind turbine gearboxes (WL-WTG) were compiled, thereby providing support for subsequent degradation analysis and prediction. Second, the Mean-Gray Relation Index (MGR) degradation analysis method was used to optimize the amount of complex data and visualize the process of degradation. Then, the segmented superposition training mechanism (ST) was designed, andPlandPitraining methods were introduced to construct the Long short-term memory-Segmented superposition training mechanism (LSTM-ST) prediction model, and degradation prediction was performed via repeated cyclic testing. Finally, the feasibility analysis of deterioration and prediction of IMS 2-1 signal from the Intelligent Maintenance System Center (IMS) is extended to validate the effectiveness of WL-WTG signal. Tests show that the method enables deterioration analysis and prediction, and provides a reference for the study of whole life cycle signals in the field of wind power engineering.