Advanced science.  Applied technology.


Noise Emitted from Turbulent Gas Flows, 14-R8916

Principal Investigators
Matthew Hoffmeyer
Inclusive Dates 
03/04/19 to 07/05/19


In this internal research project, we investigated the efficacy of using computational fluid dynamics simulations to assist in the design of quiet throttling elements in high-pressure gas systems. High-pressure gas systems require such throttling elements to reduce pressures to operate valves and provide low flow rates. It is not easy to reduce pressure without generating very high gas velocity and turbulence, which creates noise. Design of throttling elements is usually done by gradual evolution of old proven designs and the use of scaling laws to change flow rates. In this work, we attempt to use simulations of the flow on a computer to predict the details (frequency spectrum or the directionality) of the sound emitted. We compared the results with the results of an experimental realization of the same scenario.


We designed a simple orifice plate to simulate the aeroacoustic noise generated by gas flows during throttling. The orifice plate was modeled and incorporated into an ANSYS Fluent® simulation of gas flows that were also generated in an experimental setup. The noise from the experiments was measured using a set of five microphones at various angles to the flow, stationed 1 meter away. The power spectrum of the noise was compared to the noise predicted by Fluent and the differences were reconciled by adding in a post-simulation processing step where the noise from the gas was propagated to the far field.


We started out relatively inexperienced at simulating this kind of noise generation. After the work was over, we gained proficiency in setting up the mesh, choosing the method of solution, and analyzing the simulation results in a way that corresponded to what real observations would measure. We also performed research on simulation techniques that could get detailed information about the spectrum content of the noise from simulations. Our overall noise predictions from Fluent were accurate within 4 dB, but details of the power spectrum were incorrect. Future internal research is planned to try to get the details right.