Lifetime prediction of wind turbine blades using degradation model considering variable load profiles

Abstract

Assessment of fatigue damage, classification of system state, prediction of a remaining life as well as the extension of maintenance intervals become a challenge in structural health monitoring of wind turbines. Almost all wind turbine parts are subjected to different load combinations due to the variation of wind profiles. The dynamic complexity of wind profiles causes limited ability to define failure thresholds, as well as estimating current health status of the system. Due to numerous factors affecting system operation it is hardly possibly to define solutions calculating degradation and related remaining useful lifetime. Various possible degradation scenarios could be arisen due to a variety of circumstances. Instead of using analytical models, in this contribution numerical (data driven) models with the capability to handle such scenarios and providing more effective degradation prediction are used. This allows an increased accuracy for  estimating lifetime. A newly introduced state machine-based prognostic model is used to enable flexibility in deterioration modeling while concerning
the relation between load and system lifetime. In addition to the previous development, here a suitable collection of different loads and wear-dependent basic degradation processes are defined to identify the load-lifetime relation. Using core wear units, the real load time series is composed of this units allowing to learn about the effective load-lifetime relation, which is used for training of the state machine lifetime model. To observe model applicability, wind turbine blade
moments time series data are used. Then, damage degradation data for various power-dynamic relations generated using a reference software to train this model and work as reference datasets. Results demonstrate the strong potential of the proposed approach for wind turbine lifetime prediction.