Abstract Summary
The challenge that is at the forefront of data-driven vibration-based structural health monitoring (VSHM) is the detrimental effect caused by environmental and operational variations (EOVs). Therefore, action must be taken in order to mitigate the effects of the EOVs without affecting the influence of damage. A number of regression-based approaches have been applied in VSHM, using measured environmental and operational parameters to model damage sensitive features. However, action is infrequently taken to address the problem of the inherent variabilities in the data, as well as the effect of biases in the creation of the regression models. This work aims to combat these issues with a demonstration on a multivariate nonlinear regression model, using data taken from an operational wind turbine blade. A number of metrics were used to assess the stability of the regression coefficient matrices. This included assessing the effect that was observed in the distribution of the outlier analysis, distribution of main group and number of outliers, as well as the variance of the covariance of the input variables. These metrics were tested by using a number of different sets of observations for training the regression models. The results of the analysis showed that by addressing the aforementioned issues, the damage detection actually decreased. However, the parasitic nuances cannot be left untreated. Ultimately, whilst the damage detection does decrease, the robustness of the system increases and so too does the confidence that can be given to each observation.
