Structural design methodology for composite wind turbine blades using data mining techniques

Abstract

Modern wind turbine blades are large composite structures with complex geometry, subjected to many different loading cases and structural constraints. As a consequence, design and analysis of wind turbine blades is laborious and computational resource requirements are high. In this thesis, an improved design methodology for composite wind turbine blades has been developed that allows a larger scope and greater flexibility than the traditional design process, whilst maintaining moderate computational resource requirements. The methodology incorporates a highly flexible parametric model into a structural optimisation procedure, and uses data mining techniques to reduce the computational resource requirements without sacrificing flexibility or accuracy. The method is demonstrated by application to two wind turbine blade design problems: a standard blade for use in high wind speed areas and a specialised blade designed for use in areas with low wind speeds. The specific requirements of each design problem were determined with a quantitative comparison of high and low wind speed blade structures. Common simplifications that are used in wind turbine blade design were assessed, namely material placement and internal structural geometry along with critical load assumptions and the loading approximation method. The results indicated that the problem formulation should include additional materials placement variables, a wider range of constraints, and additional loading cases using accurate loading approximation methods. The consequence of incorporating the additional variables, loading cases and constraints is that the complexity of the optimisation problem is increased. In order to remedy this, data mining techniques were used to systematically simplify the problem formulations. Correlations between pairs of structural responses were identified, which allowed the number of loading case evaluations and constraints to be reduced, and the most significant variables were identified, and the remainder eliminated. The simplified problem formulations were solved using a genetic algorithm to determine optimum high and low wind speed blade designs. The resulting designs have significantly reduced mass and materials cost compared to the traditionally-designed baseline blades, particularly for the low wind speed blade design, which features a spar cap that is predominately CFRP, and a greater proportion of material in the aerodynamic shell than the traditional design.