WITHDRAWN: Physics-Based Uncertainty Quantification for Subsonic Wind Tunnel Testing Using a Validated CFD Model Part II: Holistic UQ Analysis

As modeling and simulation (M&S) is becoming a more pervasive tool in the design and analysis cycles of complex systems, the need for the assessment of its validity and uncertainty is also becoming more critical. For example, CFD models used for the simulation of the external aerodynamics of aircraft and vehicles are invariably validated against wind tunnel experimental data. However, both the numerical results and the test measurements feature a certain degree of uncertainty, which needs to be estimated and taken into account in order to increase the confidence in and the usability of the simulation data. To this avail it is necessary to develop a framework which captures the uncertainties that originate from the measured test conditions and manufacturing tolerances, as well as the uncertainties that arise from the M&S approach. In the present work, a practical uncertainty quantification framework is demonstrated on the wind tunnel test of the AGARD-B vehicle. This framework includes the effects generated by uncertainties i) in the vehicle manufacturing process, ii) in the wind tunnel operational uncertainties as well as iii) the errors introduced by the simulation model. The method is purposedly based on off-the-shelf commercial software since one key aspect of this effort is to democratize the application of uncertainty quantification to M&S of wind tunnel models. The baseline for this study is a validated CFD model of the flow around the AGARD-B vehicle presented in Part I of this work. Starting from this CFD model, a Reduced Order Model (ROM) was created for the entire set of input parameters including geometrical tolerances, surface finish, operating wind tunnel conditions and simulation settings. This ROM was then utilized for a time-efficient variational study that provided insight into which input parameters were the most critical and how they impacted the variations of the output parameters, nominally lift, drag, pressures and temperatures at specific locations on the vehicle. The main benefits of this approach are that it considers a wide base of the influencing factors that impact the accuracy of the simulation, it identifies the most important factors and it quantifies their impact on the most significant output variations driven by these factors, all in a computationally efficient manner.