The development of robust design strategies coupled with 
detailed simulation models requires the introduction of 
advanced algorithms and computing resource management tools. 
On the algorithmic side, we explore the use of simplex-based 
stochastic collocation methods to characterize 
uncertainties, and a Probabilistic Non-dominated Sorting 
Genetic Algorithm (P-NSGA) for multi-objective optimization 
under uncertainty. The goal of this research is to optimize 
a large-scale, three-dimensional geometry using a very large 
number (extreme ensemble) of CFD (computational fluid 
dynamics) simulations on high performance computing 
clusters. 

The problem of interest is the optimization under 
uncertainty of a Formula 1 tire brake intake to maximize 
cooling efficiency and minimize aerodynamic resistance. 
Sculptor is used to introduce uncertainties into the problem 
by locally deforming the rubber tire. Depending on racing 
conditions the tire geometry can vary significantly at 
different turns and tracks. The impact of geometric 
uncertainty on the aerodynamics needs to be quantified by 
Formula 1 engineers.

Since we are using an iterative approach to quantifying the 
uncertainty, Sculptor is run in batch mode on our cluster. 
The design and uncertain parameters in the Sculptor input 
files are modified autonomously depending on the 
optimization process. A simulations environment (Leland) has 
been developed to dynamically schedule, monitor and stir 
the calculation ensemble and extract run-time information as 
well as simulation results and statistics.