Open for pre-order at Crowd Supply: Acoustic Leak Detector

[microphonon]

Anyone interested in the recent YouTube teardown of the ultrasonic gas leak detector may find my recently launched Crowd Supply project worth a look. It is a small sensor board that combines the principles of glass breakage and smoke detection to identify the presence of very small water leaks emanating from pressurized plumbing. Development and testing has focused on water leaks, but it should be effective with gases, refrigerants, steam, etc. It is designed for battery-powered IoT applications and cannot be used to listen to conversations, i.e. eavesdropping is physically impossible. Link to the Crowd Supply page is here:

https://www.crowdsupply.com/microphonon/aquaping

This was a challenging project that tried to balance sensitivity with reliability and high immunity to false alarms. This involved mechanical, electronic, and digital filtering along with a statistical analysis of the acoustic environment. The design is completely open-source and I welcome any questions or comments.

[Tigerwoods]

Hi,

Just interested how you balance and handle:

- the sensor-leakage distance (manually input required?),
- sound attenuation (the sound strength you measured),
- orientation of the microphone to the leakage position (different angle, different sound strength),
- orifice shape model you use (different shape may lead to different sound strength and frequency),
- flow velocity in the pipe (may impact on frequency and sound strength)
- leakage position (such as the difference between "close to the wall" or "in the center of the room" )

Honestly, I can imagine such tools can somehow work but it must be poor in accuracy.

[microphonon]

The sensor is best understood as a binary, yes/no leak detector. It is not attempting to determine flow rate, the nature of the leak, or even pinpoint its location. It is trying to assess, with reasonable certainty, whether or not a leak is present somewhere in the vicinity. To do this, it assumes that a leak signal is steady and persistent. Not all leaks will present this way, but the very weak ones are likely to.

Hi,

Just interested how you balance and handle:

- the sensor-leakage distance (manually input required?),

The sensor must first learn about its nominal acoustic environment, ie. with no leaks present. This is done with a user configurable training session, which essentially specifies how long it should listen to the "quiet" ambient before automatically transitioning to monitor mode.

- sound attenuation (the sound strength you measured),

If the sound level in the target spectral range of 7--11 kHz is sustained, steady, and one standard deviation about the background, a potential leak is identified.

- orientation of the microphone to the leakage position (different angle, different sound strength),

The orientation of the microphone/Helmholtz resonator will affect sound level, but we are only trying to determine if a leak is present, not its point of origin. High frequency acoustics will readily reflect from planar surfaces like walls, floors, and ceilings, which makes triangulation notoriously difficult.

- orifice shape model you use (different shape may lead to different sound strength and frequency),
- flow velocity in the pipe (may impact on frequency and sound strength)
- leakage position (such as the difference between "close to the wall" or "in the center of the room" )

The orifice size and shape, pressure, geometry, will all certainly affect the strength of the signal. For example, a smooth pinhole drilled in a pipe will create a Laminar jet that produces a much weaker signal than a loose fitting. Again, we are trying to detect the existence of weak leak signals with reasonable reliability.

Honestly, I can imagine such tools can somehow work but it must be poor in accuracy.

Accuracy is not a goal or even part of the design. The aim is to identify the presence of very weak leak signals and flag them with high reliability.

[microphonon]

I should also mention that there are devices available that can measure flow rates with reasonable accuracy. They attach directly to the pipe and may or may not be in contact with the moving fluid.

As I describe above, the acoustic sensor does not measure or quantify flow. The idea is to flag the possible presence of leaks from stand-off distances. It can augment/complement other sensor tech.

[Tigerwoods]

I see, thanks.
Then it should be quite useful.
Good luck!

[microphonon]

Update: The Crowd Supply campaign for the acoustic leak detector was successful and a production run has completed. Units are being shipped to project backers in 17 US States and 7 countries.