Aeroelastic Uncertainty Quantification Studies Using the S4T Wind Tunnel Model

Abstract

This paper originates from the joint efforts of an aeroelastic study team in the Applied Vehicle Technology Panel from NATO Science and Technology Organization, with the Task Group number AVT-191, titled "Application of Sensitivity Analysis and Uncertainty Quantification to Military Vehicle Design." We present aeroelastic uncertainty quantification studies using the SemiSpan Supersonic Transport wind tunnel model at the NASA Langley Research Center. The aeroelastic study team decided treat both structural and aerodynamic input parameters as uncertain and represent them as samples drawn from statistical distributions, propagating them through aeroelastic analysis frameworks. Uncertainty quantification processes require many function evaluations to asses the impact of variations in numerous parameters on the vehicle characteristics, rapidly increasing the computational time requirement relative to that required to assess a system deterministically. The increased computational time is particularly prohibitive if high-fidelity analyses are employed. As a remedy, the Istanbul Technical University team employed an Euler solver in an aeroelastic analysis framework, and implemented reduced order modeling with Polynomial Chaos Expansion and Proper Orthogonal Decomposition to perform the uncertainty propagation. The NASA team chose to reduce the prohibitive computational time by employing linear solution processes. The NASA team also focused on determining input sample distributions.

Summary Findings from the AVT-191 Project to Assess Sensitivity Analysis and Uncertainty Quantification Methods for Military Vehicle Design

“…The Aeroelasticity Team's experience on the supersonic transport model 10 (See, Figure 6.) were reported in these AIAA special sessions by Cunningham and Holman 11 , Nikbay and Heeg 12 , and Tartaruga et al 13 , and their methods and findings are summarized in Table 3. All the team members used or developed surrogate models to which they applied the uncertainty methods because they found that such models represented the features of interest in the analysis with sufficient accuracy and they significantly reduced the computational resources required for the analysis.…”