Short answer. CFD predicts how the gas or air flows: velocities, pressures and pressure drop. Acoustic simulation predicts how sound waves travel through and out of a system: transmission loss, resonances and sound levels at chosen positions. A steady CFD result does not contain the sound itself, though its velocities can be used to estimate the noise a flow generates. On engine and generator work the two are usually run on the same geometry, because a silencer or a ventilation path has to satisfy both.
What each one solves
| Aspect | CFD (the flow) | Acoustic simulation (the sound) |
|---|---|---|
| Physics | The mean motion of the gas or air: conservation of mass, momentum and energy, usually time-averaged | Small pressure waves at each frequency, travelling at the speed of sound |
| Typical outputs | Pressure drop, flow distribution, velocity and temperature fields | Transmission loss, resonances, and sound levels and spectra at agreed positions |
| What drives the answer | Flow rate, temperature and geometry | Geometry, materials and the strength of the sound source, from measured or manufacturer data |
| Typical methods | Steady RANS CFD, with sub-models for perforated tubes and packing | Plane-wave (transfer-matrix) models at low frequency; 3-D finite- and boundary-element models above the first cross-mode frequency |
| Judged against | The engine manufacturer’s back-pressure or restriction limit; the allowable cooling-air restriction | An attenuation target or a sound-level limit |
Can CFD predict generator noise?
Not directly. A steady, time-averaged CFD solution has no fluctuations in it, so it contains no sound. What it does contain is the velocity and pressure field, which points to regions where the flow is likely to make noise: high velocity, separation, strong turbulence. Even the noise-source models that work from steady CFD results are documented as showing which parts of the flow are responsible for noise generation, not the sound that reaches a listener.
For noise the flow generates inside a system, at silencer baffles, louvers, grilles, dampers and bends, the CFD velocities can go further. Published prediction methods for this regenerated noise relate the sound power an in-duct element generates to the pressure loss across it and the flow velocity through empirical relationships. SH Consulting uses the CFD velocities with empirical correlations of this kind to estimate flow-generated noise, for example to check whether a silencer or louver will add noise of its own at the design flow. It is an engineering estimate for comparing options and spotting risk, not a simulation of the sound field.
Simulating the sound a flow generates directly is a separate and much more expensive kind of simulation: time-accurate, resolving the turbulence, and coupled to an acoustic method that carries the sound to the listener. Because acoustic energy is tiny compared with the other fluctuations in a flow, it demands very accurate numerical schemes. SH Consulting doesn’t offer that kind of simulation.
For most engine and generator questions it isn’t needed. Exhaust and intake noise is usually strongest at the engine’s firing frequency and its harmonics, and the exhaust outlet and the enclosure’s cooling-air inlet and outlet are often the dominant noise paths. Those are acoustic problems: given the engine and fan sound data, an acoustic model predicts how much a silencer removes and what level an enclosure lets out (generator and engine noise reduction).
Where the two meet
- Silencers. The chambers, baffles, perforations and lining that attenuate noise are the same features that add back pressure. An acoustic model predicts the transmission loss and CFD predicts the pressure drop of the same geometry, so each variant is judged on both (side-by-side comparison).
- Hot gas and mean flow. The flow changes the acoustics: hot exhaust has a higher speed of sound, which shifts a silencer’s tuned frequencies upward, and mean flow changes performance, so acoustic models use the actual gas temperature and flow.
- Ventilation paths. Splitter silencers and louvers that attenuate fan and engine noise also restrict cooling air. The acoustic model sizes the attenuation; an airflow model checks the radiator still gets enough air (enclosure ventilation and cooling airflow).
Which one does your question need?
| Your question | Usually answered by |
|---|---|
| Will this silencer stay within the engine’s back-pressure limit? | CFD, or a hand calculation for a simple layout with published data |
| How much will this silencer attenuate at the firing frequency and its harmonics? | Acoustic simulation |
| Will the radiator still get enough air through the louvers and splitters we need for noise? | CFD of the cooling-air path |
| Given the engine and fan sound data, is this enclosure likely to meet the site’s sound limit? | Acoustic simulation |
| Where is the flow separating, recirculating or unevenly distributed? | CFD |
| Will this silencer or louver add flow-generated noise of its own at the design flow? | An estimate from the CFD velocities with empirical correlations |
| Which noise path should we treat first? | Source and path ranking (noise control) |
| Is this silencer quiet enough and within back pressure? | Both, on the same geometry |
When neither is the first step
If hardware already exists and the question is one measured number, a test may be quicker. If a published loss coefficient or a supplier’s attenuation data already answers the question, a hand calculation may be enough. Part of scoping a study is saying so.
Related services
Read next
- Generator & engine noise reductionBoth models applied to exhaust silencers and generator-set enclosures.
- Silencer design: attenuation and back pressureAttenuation and back pressure, side by side.
- What Is CFD? A Practical Guide to Computational Fluid DynamicsHow a CFD model is built and checked.
Sources
- Ansys, Broadband noise source models (Fluent Theory Guide §11.1.4)
- COMSOL, Aeroacoustics (Multiphysics Cyclopedia)
- COMSOL, Automotive muffler (application model documentation)
- C. M. Mak, CIBSE, Towards generalised prediction techniques for regenerated noise in ventilation systems (BSERT 16(2), 1995)
- NASA Glenn Research Center, Navier-Stokes Equations