14 Best Practices for Surrogate Based Uncertainty Quantification in. . .
439
Fig. 14.1 Comparison of Kriging and Gradient-Enhanced Kriging for a given set of sampling
points
14.3.4 Computing Statistics on Surrogate Models
Despite the reduced computational cost to evaluate the statistic of interest using
surrogate models, the use of Monte Carlo for the evaluation of statistics may be
impractical due to its low convergence rate.
A possible alternative is the use of Quasi Monte Carlo techniques [13], as they
have a faster convergence rate. The sampling is predefined (derandomized) through
Sobol Sequences [14] and the stochastic space is evaluated more efficiently.
14.4 Optimization of RAE2822 Airfoil Under Uncertainty
This section focuses on the application of the concepts previously introduced to
the aerodynamic shape optimization under uncertainty of the RAE2822 airfoil at
transonic conditions.
439
Fig. 14.1 Comparison of Kriging and Gradient-Enhanced Kriging for a given set of sampling
points
14.3.4 Computing Statistics on Surrogate Models
Despite the reduced computational cost to evaluate the statistic of interest using
surrogate models, the use of Monte Carlo for the evaluation of statistics may be
impractical due to its low convergence rate.
A possible alternative is the use of Quasi Monte Carlo techniques [13], as they
have a faster convergence rate. The sampling is predefined (derandomized) through
Sobol Sequences [14] and the stochastic space is evaluated more efficiently.
14.4 Optimization of RAE2822 Airfoil Under Uncertainty
This section focuses on the application of the concepts previously introduced to
the aerodynamic shape optimization under uncertainty of the RAE2822 airfoil at
transonic conditions.