A physics-based leading edge erosion growth prediction model utilizing Bayesian updating and drone-inspection data

Abstract

Leading edge (LE) erosion causes reduced power output and a reduction in efficiency by impairing the aerodynamics of the blade. In severe cases, it can reduce the structural integrity of the blade. It is preferable and cheaper to repair LE erosion before it reaches the sub-surface. LE erosion is commonly widespread across a wind turbine blade and is monitored using drone inspection with manual review of images, which is time-consuming due to the large number of images generated. In addition, it is hard to forecast the evolution of a LE erosion defect, and to make prioritization of damages across a fleet of wind turbines. The approach proposed in this paper combines Bayesian updating and physics model, with the aid of drone inspection failure data to predict the future evolution of LE erosion. The method, based on the Bayesian updating method, can capture complex interactions in the degradation process. The physics-based approach can reflect physical degradation mechanisms. Fusion of knowledge from physics-based predictive models with information mined from failure databases using Bayesian updating can combine advantages of the two methods, and also make full use of available knowledge. A case study is presented which predicts the evolution of LE erosion using the proposed method. After the application of Bayesian updating method, the uncertainty of material property distribution decreased by 43.43%. The method predicts the future evolution of LE erosion damages across a wind turbine blade, providing more information to an engineer to prioritize which defect to repair first.