Uncertainty quantification and global sensitivity analysis of composite wind turbine blades

In this paper, a framework for uncertainty quantification (UQ) and global sensitivity analysis (GSA) of composite wind turbine blades is presented. Because of the presence of uncertainties, the performance and reliability of wind turbine blades are adversely affected. Uncertainties must therefore be accounted for during the design phase. However, performing UQ of composite blades while considering a large number of random parameters is computationally intensive. To make the process tractable, this work is based on an approach referred to as polynomial chaos expansion (PCE) with l(1)-minimization. PCE also enables one to perform GSA to assess the relative importance of random parameters using Sobol Indices. This article also introduces an anisotropic formulation of PCE for dimension adaptive basis expansion. In addition, the UQ framework can handle random inputs with arbitrary distributions as well as spatial variations of material and geometric properties using Karhunen-Loeve expansion. The presented framework was applied to three composite wind turbine blade problems - modal analysis, failure analysis, and buckling analysis - by considering the randomness in material and geometric parameters as well as loading conditions. The test case selected in this study is a blade from the National Renewable Energy Laboratory 5 Megawatt wind turbine. Results obtained with PCE were compared to Monte Carlo simulations. In addition, the influential random parameters were identified using Sobol Indices, obtained as an inexpensive sub-product of the PCE approximation.

Automated summary

This paper presents a framework for uncertainty quantification (UQ) and global sensitivity analysis (GSA) of composite wind turbine blades using polynomial chaos expansion (PCE) with l(1)-minimization. The framework is capable of handling a large number of random parameters and can assess the relative importance of these parameters using Sobol Indices. It also allows for arbitrary distributions of random inputs and spatial variations of material and geometric properties. The presented framework is applied to three composite wind turbine blade problems, and results are compared to Monte Carlo simulations.