Integrating a genetic algorithm code with a response surface methodology code based upon the artificial neural network model, this paper develops an optimization system. By introducing a quasi-three dimensional through-flow design code and a design code of axial compressor airfoils with camber lines of arbitrary shape, and involving a three-dimensional computational fluid dynamics solver, this paper establishes a numerical aerodynamic optimization platform for the three-dimensional blades of axial compressors. The optimization in this paper mainly has four features. First, it applies the conventional inverse design method instead of the common computer aided design parameterization method to generate a three-dimensional blade. Second, it chooses aerodynamic parameters with physical meaning as optimization design variables instead of purely geometrical parameters. Third, it presents a stage-by-stage optimization strategy about the multistage turbomachinery optimization. Fourth, it introduces the visual sensitivity analysis method into optimization, which can adjust variation ranges of variables by analysing how great the variables influence the objective function. The above techniques were applied to the redesign of a single rotor row and two double-stage axial fans separately. The departure angles and work distributions in the inverse design were taken as design variables separately in optimizations of the single rotor and double-stage fans, and they were parametrically represented by means of Be´zier curves, whose parameters were used as the optimization variables in the practical operation. The three investigated examples elucidate that not only the techniques mentioned above are appropriate and effective in engineering, but also the design guidance for similar inverse design problems can be obtained from the optimization results.

This content is only available via PDF.
You do not currently have access to this content.