Gas flow is not exactly my field - my areas of expertise are electrical engineering and acoustics - but here are some suggestions that may help.

Noise can enter a gas flow monitoring system in several ways:

- from turbulence in the gas flow

- from other mechanical vibrations in the system, since pressure sensors will have some sensitivity to acceleration as well as pressure

- from electromagnetic interference in the cabling from the sensor to the preamplifier or digitiser

- from electrical thermal noise in the preamplifier

The appropriate method for removing the noise depends on the source of the noise.

Filtering to remove the unwanted component of the signal should be applied with caution, as it will change the wanted signal as well as the unwanted signal. It is better to remove as much of the noise as possible at the source, and leave the filtering as a last resort.

For example, turbulence in the gas flow can be reduced by passing the gas through a wire mesh screen or some steel wool, as long as this does not interfere with the steady gas flow too much. Ensuring that the pipe is smooth and free of sharp bends, sharp transitions and interior protrusions and obstacles will also help.

To reduce mechanical vibrations at the sensor, it can be isolated from the rest of the system by inserting a flexible section of pipe upstream and downstream and bracing the measurement section against something stationary and rigid.

Electromagnetic interference can be reduced by keeping the cable run as short as possible, or even using a sensor with a digital output if this is available. If this is not an option, make sure the cable has adequate electrical shielding and, if possible, use a differential amplifier rather than an unbalanced amplifier.

Thermal noise in the preamplifier may be reducible by careful choice of amplifier hardware, keeping the amplifier cool and selecting an appropriate bandwidth for the task - thermal noise increases with bandwidth.

If all of these other options have been tried and the noise is still too high, then apply some low-pass filtering but be aware that the rise time of the system will be affected, so your transient information may reflect the filter characteristic more than the actual system behaviour.

From a quick search of the literature, it appears that gas flow transients are sometimes modelled, rather than measured, which, depending on the complexity of the model, will probably give the smooth results you are seeing in prior publications on this topic. But the publication should clearly state that the data are either measured or modelled. You shouldn't need to guess this.

I hope this helps.

I almost forgot the most important tool in the noise reduction toolbox - repeating and averaging. If you conduct the same experiment n times and average the result, the noise will be reduced by square root of n. If you do the experiment 9 times and average the results, the noise will be reduced by 2/3, as long as it is truly random and not some coherent interfering signal like mains hum.