Robust Design Optimization Of A Ford Turbocharger Compressor

Combining standard numerical optimization with sensitivity analysis find designs that are less sensitive to manufacturing and operating uncertainties.


What is robust design optimization (RDO), and is it better than standard optimization? In this customer case, we describe a multi-disciplinary optimization of a turbocharger compressor from Ford Motor Company, in which we demonstrate it is.

RDO combines standard numerical optimization with sensitivity analysis to take into account the influence of manufacturing variations and operating uncertainties on product performance early in the design process. This type of CFD optimization enables engineers to find optimal designs that are less sensitive to these uncertainties, giving more stable performance results.

This Ford turbocharger optimization case aimed to extend the compressor’s flow capacity and improve surge margins. We used an active self-recirculation casing treatment.

Simulation set-up

Given their varying geometries, we created two separate meshes for the points near the choke and the stall with Fidelity Automesh. The simulation domain consisted of the impeller, a vaneless diffuser, the self-recirculation casing and the volute for fluid flow simulation (CFD), and the solid blade with a back plate for structural simulations (CS). Fidelity Optimization was used to run the optimization on the impeller shape.

The design space consisted of 19 design variables, performed as multi-point optimizations, and three operating points on two different speed lines. The robust optimization accounted for the following uncertainties:

  • Tip gap height ±25%
  • Blade thickness ±1%
  • ±1% of the total pressure at the inlet or static pressure at the outlet, respectively
  • Different boundary conditions for choke and stall

Comparison of standard deterministic versus robust design optimization

Comparing the performances of Deterministic Design 1 and Robust Design 1, we learned that the mean values of choke mass flow and efficiency varied in a similar range. However, the standard deviations of these objectives were drastically reduced by the robust design only.

Mean values of choke mass flow and efficiency vary in a similar range.

Standard deviations of choke mass flow and efficiency were drastically reduced by the robust designs only.

The resulting blade designs were also notably different:


Accounting for uncertainties in the optimization process resulted in designs that are less sensitive to uncertainties. The standard deviation of the choke mass flow was reduced by 33% and the efficiency by nearly 48% in the robust design, leading to significantly more performance stability.

Leave a Reply

(Note: This name will be displayed publicly)