Accelerated Industrial Blade Design Based on Multi-Objective Optimization Using Surrogate Model Methodology
The intention of this paper is to provide an advanced aerodynamic blade design approach for industrial purposes which are basically characterized by limited development time, human and computing resources. From the industrial point of view, the demand for process acceleration and design optimization cannot be sufficiently satisfied with traditional human-based design methods. Recent investigations on blade optimization have shown some potential in performance improvements, however, this is typically obtained by high computational efforts in particular when using multi-objective optimization methods. In order to combine the benefits of numerical optimization with the requirements of industrial needs for design acceleration, a new automated blade optimization strategy is required. The accelerated industrial blade design process in this paper is based on a three-dimensional parameterization approach using non-dimensional parameter distributions which always guarantee desired blade geometry smoothness. In order to approximate the design objectives and constraints, a response surface methodology is applied where the design parameter variation is obtained by the quasi-random SOBOL sequence. Based on that, a highly sophisticated multi-objective genetic algorithm is used with reasonable numbers for individuals and generations for solving the contradicting design goals of an aerodynamic blade design problem by considering multiple aerodynamic and geometric constraints. This approach offers a set of non-dominated solutions on the Pareto-front which are subsequently evaluated with the exact flow analysis. In case of objective function value discrepancies between model and exact evaluations, an update of the surrogate model is performed including these additional solutions until the approximation response is equivalent to the exact analysis within a predefined tolerance. This new methodology shows a significant overall design time reduction particularly a decrease of required function evaluations without loosing the benefit of multi-objective optimization in providing Pareto-optimal solutions. Based on a typical industrial compressor test case, an aerodynamic performance improvement and process acceleration by factor greater than 10 could be achieved.