Publications

Uncertainty quantification for a turbulent round jet using multifidelity karhunen-loève expansions

Jivani, Aniket; Huan, Xun; Safta, Cosmin S.; Zhou, Beckett Y.; Gauger, Nicolas R.

Understanding the behavior of turbulent jets under variable environment and uncertain conditions is critical for predicting and mitigating aircraft jet noise. However, uncertainty quantification (UQ) for jet noise, which requires repeated expensive eddy-resolving simulations, is often computationally prohibitive. We thus build surrogate models, in particular Karhunen-Loève expansions (KLEs) for field quantities of interest in three-dimensional turbulent round jets. We build them in a multifidelity manner by combining simulation data from high-fidelity enhanced delayed detached-eddy simulation (EDDES) and low-fidelity Reynolds-averaged Navier-Stokes (RANS), generated under uncertain nozzle exit stagnation pressure and inlet eddy viscosity ratio. Furthermore, we form the KLEs in conjunction with polynomial chaos expansions in order to explicitly associate their randomness to each physical source of uncertainty, and so justifying the combining procedure in the multifidelity construct. We illustrate advantages of the new multifidelity KLE against single-fidelity KLEs, with the former achieving more accurate predictions at locations away from existing high-fidelity training data. With the KLE surrogate, we conduct UQ inexpensively.