ABSTRACT
Multidisciplinary optimization methods are routinely applied in product development processes in the automotive industry. The wide popularity gained by gradient-free methods – such as Evolutionary Algorithms – is due to their generality and relative simplicity in implementation. Nevertheless, the new performance goals demanded to the turbocharging business require more detailed geometrical investigations and comprehensive optimizations capable of capturing the interactions of the different involved physical phenomena. Such holistic view is not easily achievable by gradient-free methods which suffer from the so-called “curse of dimensionality” and high computational cost. A more feasible approach to richer design spaces is therefore offered by gradient-based methods, and in particular by the adjoint methods, that prove to have the highest potential in this field.
This paper discusses the development of a multidisciplinary gradient-based methodology, which includes the CAD model for the shape parametrization, hence allowing the involvement of manufacturing considerations within the optimization. In addition to the aerodynamic performance, mechanical and vibrational predictions are invoked as constraints in order to prevent the search beyond mechanically feasible shapes.
The method is applied to the design optimization of a turbocharger radial turbine rotor for automotive applications. Multiple operative conditions within the engine map are included in the definition of the objective function, allowing for optimal performance over a large range of regimes. The results demonstrate that with limited computational resources a competitive rotor design can be efficiently achieved through a holistic design approach.
