GPU-Accelerated Numerical Study of Temperature Effects in Choked Under-expanded Supersonic Jets

Temperature effects on the nearfield noise radiated by an under-expanded round supersonic jet operating at a nozzle pressure ratio of 2.97 are studied by performing Large Eddy Simulations using a new high-resolution gradient-based Navier-Stokes algorithm proposed by Chamarthi [1]. The algorithm employs 4^th-order (for linear cases) monotonicity preserving reconstruction for computing inviscid fluxes and the 4^th -order α-damping approach to evaluate the viscous fluxes. The preliminary cold jet results are in good quantitative agreement with experimental measurements reported in the literature. The screech frequency, amplitude, and jet oscillation mode are captured accurately. The effect of jet temperature is studied by varying the ratio of the chamber to atmospheric temperature. Three temperature ratios of values equal to 1 (cold), 2 (mid-hot), and 3 (hot) are considered in this study. The increase in the temperature ratio has triggered the jet to undergo a mode staging behavior and radiate Mach waves downstream with increased amplitude. Proper Orthogonal Decomposition (POD) analysis of the temporal density field suggests that the jet shear layer gets increasingly dominated by vorticial structures and turbulent mixing as the temperature ratio increases. The Mach wave radiation phenomenon is only observed in the ‘mid-hot’ and ‘hot’ jet cases with higher Overall Sound Pressure Levels (OASPL) and Mach angle for the ‘hot’ jet case. The GPU acceleration has resulted in a speed-up of about two hundred times (compared to single-core CPU), thus enabling each simulation to be run in a reasonable turnaround time of ≈ 35 hours. Performance statistics corresponding to single and multi-GPU parallelization models employed on the latest generation of data center GPUs are presented to conclude this paper.