Lyle N. Long
Penn State Univ., 233 Hammond Bldg., University Park, PA 16802
There is renewed interest in predicting aeroacoustic noise, especially for jets, rotating blades, and shock waves. These flows involve nonlinear, three-dimensional, turbulent phenomena, and nonuniform free streams. Simulating these flows requires algorithms quite different than those traditionally used in computational fluid dynamics (CFD). The time-dependent nature of aeroacoustic problems requires the algorithm to correctly simulate the dispersion and dissipation features of the flow. Good CFD algorithms usually rapidly damp out all but the steady-state portion of the flow, and are inappropriate for aeroacoustics. Computational aeroacoustics schemes have more in common with large eddy simulation (LES) algorithms than those used in CFD. Recent progress in higher-order algorithms for supersonic jets [T. S. Chyczewski and L. N. Long, 16th AIAA Aeroacoustics Conference, Paper 95-011 (1995)] and fan noise [Y. Ozyoruk and L. N. Long, 16th AIAA Aeroacoustics Conference, Paper 95-063 (1995)] illustrates that quite complicated aeroacoustic problems can be simulated. These algorithms require roughly 5-10 grid points per wavelength. The large demand on computer memory and speed requires that one use modern parallel computers, such as the IBM SP2 and the TMC CM-5. One must be careful to properly load balance the scheme and to minimize interprocessor communication. Kirchhoff surfaces are very effective in predicting the far-field solution.