Damage detection of wind turbine blades by Bayesian multivariate cointegration

Changes in the vibration responses of wind turbine blades (WTBs) can be used to detect the presence of damages. However, constantly changing environmental circumstances might cause vibration responses to change, complicating the damage detection procedure. This study introduces an approach of time series analysis termed Bayesian cointegration for damage detection. This approach is extended from a univariate to a multivariate case, allowing for the simultaneous inclusion of more than two damage-sensitive features (DSFs) for analysis. The common trend induced by environmental fluctuations in the DSFs series is purged using the Bayesian cointegration approach. As a result, structural degradation can be detected when the underlying long-run equilibrium relationship among the time series suffers noticeable changes. The proposed approach has clear advantages over the frequently used Johansen test for cointegration: (1) Artificial parameter selection is unnecessary because all involved parameters can be inferred probabilistically; (2) It has a greater capacity to capture possible cointegrating relationships. Benchmark data from a numerical WTB are used to demonstrate the performance of the proposed method. Results show that this method can effectively eliminate the influence of environmental variations and detect the progressive damage of the WTB.

Automated summary

This study introduces a Bayesian cointegration approach for damage detection in wind turbine blades, which can effectively eliminate the influence of environmental variations and accurately detect structural degradation.