Optimal design of an automotive fan using the Turb’Opty meta-model

Abstract

Optimal design techniques recently gained a wide popularity in the industry as relatively powerful computers have become broadly available and as attractive tools such as surrogate models and evolutionary optimization went through maturation. However, when dealing with complex geometries and difficult physical phenomena to be modeled, computing costs still remain high, due to the large number of required numerical simulations feeding the traditional surrogate models. Turb’Opty© is a meta-model which only requires a single CFD simulation at a reference configuration point, based on automatic differentiation of the discretized Reynolds-Averaged Navier–Stokes equations and high-order Taylor-series expansions. A flow database containing the derivatives of the physical variables with respect to the design variables is produced by this parameterization tool and thoroughly explored, in the post-processing step, by a multi-parameter and multi-objective genetic algorithm coupled to the associated extrapolation tool. In this paper, post-processing of the derivative database will be depicted through a 3D study of an automotive shrouded fan with casing treatment.