Abstract: Employing the strategy of "experimental design+ response surface model + genetic algorithm",a response surface model code based on the artificial neural network and a genetic algorithm code were developed;and this paper established a numerical aero-optimization platform for three-dimensional(3-D) axial-flow compressor blade.This integral platform contains also the commonly used arbitrary camber line airfoil and blade formatting code and 3-D computational fluid dynamics solver.A fulfilled inverse-design of a zero-stage compressor rotor of a turbojet engine was optimized by this platform.The optimization object is to achieve the maximum adiabatic efficiency at the design point,while the mass flow rate and total pressure ratio are kept unchanged as the constraints.The relative flow-angle and the inner-blade deviation-angle are chosen separately as the independent-variables in two optimization cases.In order to directly couple the optimization method with the modern design system,the aerodynamic variables with physical meanings were chosen instead of the commonly used geometric variables as the optimization independent-variables.As compared with the original rotor,the adiabatic efficiencies of two optimal rotors increase by 0.82 and 0.73 percentage points,respectively.It shows that 3-D optimization method based on the aerodynamic variables has a good performance.