Garcia Cava, David

Lu, Yong

Roberts, Callum

The Carnegie Trust for the Universities of Scotland

2023-10-25T12:49:08Z

2023-10-25T12:49:08Z

2023-10-25

Whilst wind power is a promising alternative to wasteful and polluting fossil fuels, there are a number of issues that must be addressed. Many difficulties lie in the maintenance of the ever increasing size of the blades, especially in offshore environments. The current industry standard of visual inspection is outdated and needs to be replaced with real-time and online monitoring. Vibration-based Structural Health Monitoring (VSHM) has been proposed as a potential solution to this problem. However, the presence of Environmental and Operational Variations (EOVs) causes VSHM methods to struggle to differentiate between damaged and undamaged observations. The Damage Sensitive Features (DSFs) measured from the wind turbine blades are heavily influenced by the EOVs and effort has to be made to mitigate their effects to ensure the damage detection is reliable. Through regression analysis, relationships can be established between the DSFs and measured Environmental and Operational Parameters (EOPs). Subsequently, EOP-normalised DSFs are created by the difference between the original DSFs and those predicted by the regression models. The reliability in the predictions, beyond what they were trained with, is an extremely important but often overlooked aspect of regression design. Uncertainty can easily be introduced by overfitting model orders, including non-influential EOPs and by benign trends present in the training data. Through considered design, this work aims to address such issues through the application of a comprehensive nonlinear forward stepwise regression method for the purpose of monitoring an operational wind turbine blade. The proposed methodology employs methods to remove collinear variables, identify the most influential EOPs, reduce model orders and determine which DSFs should be regressed. The combination of these methods facilitates a compact regression basis, purged of as much uncertainty as possible. Lasso regression is used for comparison, as it is a similar and established type of stepwise regression. Ultimately, reducing biases and overfitting through considered design will increase the robustness of the system, as well as increasing confidence in the decision making process.

https://hdl.handle.net/1842/41101

http://dx.doi.org/10.7488/era/3840

en

The University of Edinburgh

Roberts, C., Garcia Cava, D., Avendãno-Valencia L.D. (2023). Addressing practicalities in multivariate nonlinear regression for mitigating environmental and operational variations. In: Structural Health Monitoring, vol 22. Sage. Available at: https://doi.org/ 10.1177/14759217221091907

Roberts, C., Avenda˜no-Valencia, L.D., Garcia Cava, D. Robust Mitigation of EOVs Using Multivariate Nonlinear Regression within a Vibration-Based SHM Methodology. In: Mechanical Systems and Signal Processing. Elsevier. Available at SSRN: https://ssrn.com/abstract=4387099

Garcia Cava, D., Avenda˜no-Valencia, L.D., Movsessian, A., Roberts, C., Tcherniak, D. (2022). On Explicit and Implicit Procedures to Mitigate Environmental and Operational Variabilities in Data-Driven Structural Health Monitoring. In: Cury, A., Ribeiro, D., Ubertini, F., Todd, M.D. (eds) Structural Health Monitoring Based on Data Science Techniques. Structural Integrity, vol 21. Springer, Cham. https://doi.org/10.1007/978-3-030-81716-9_15

Roberts, C., Cava, D.G., Avenda˜no-Valencia, L.D. (2021). Understanding the Influence of Environmental and Operational Variability on Wind Turbine Blade Monitoring. In: Rizzo, P., Milazzo, A. (eds) European Workshop on Structural Health Monitoring. EWSHM 2020. Lecture Notes in Civil Engineering, vol 127. Springer, Cham. Available at: https://doi.org/10.1007/978-3-030-64594-6_12

Roberts, C., Garcia, D., Tcherniak, D. (2020). A Comparative Study on Data Manipulation in PCA-Based Structural Health Monitoring Systems for Removing Environmental and Operational Variations. In: Wahab, M. (eds) Proceedings of the 13th International Conference on Damage Assessment of Structures. Lecture Notes in Mechanical Engineering. Springer, Singapore. Available at: https: //doi.org/10.1007/978-981-13-8331-1_13

Roberts, C., Isbister, S., Murphy, C., Nisbet, C., Sweeney, P., Tcherniak, D., Garcia Cava, D. (2018). Strain estimation using modal expansion approach via virtual sensing for structural asset management. In: 1st International Conference on Structural Integrity for offshore energy industry. Available at: https://strathprints. strath.ac.uk/65677

structural health monitoring

environmental and operational variabilities

multivariate nonlinear regression

forward stepwise regression

lasso regression

mutual information

f-statistic

Mahalanobis distance

Addressing regression architecture for the robust mitigation of environmental and operational variations in wind turbine blade monitoring

Thesis or Dissertation

Doctoral

PhD Doctor of Philosophy