One of our most fun experiment so far. We buried shakers and microphones in the ground and built a cloud-based telemetry solution analysing changes in sound timbres. This delivered a reliable and automated monitoring system for turf management. It involved out-of-the box thinking, physics, audio engineering, soldering & cabling, software programming, cloud infrastructure, landscaping equipment and a fabulous team of stakeholders and helpers. Best of all - it worked !

The Challenge

We needed to find a way to measure the condition of a large patch of turf, by means of an automated, reproducible and standardised way.

Current Technology and Constraints

Currently, assessment of sports grounds, playgrounds, turf or sand surfaces is typically performed via a penetrometer or Clegg hammer. Both engage some mechanical interaction with a tiny patch of the ground top layer to assess the overall condition of the surrounding area. Put simply, a piece of metal is dropped on the ground or stuck into it. The amount of bounce or resistance is then used to derive the condition of the surface at the point of impact. This result is extrapolated to the entire surface in one way or another. Such extrapolating from a few square centimetres to a few square meters can be a significant source of error, as a quick experiment confirmed.

In applications where more standardisation is required, the error margin resulting from such variation may be intolerable.

Requirements

The solution required a fundamentally different approach to the measurement of a relatively large area. It had to take into account that realistically, ground is not perfectly homogeneous. It also had to employ technology that does not interfere with the use of the ground and that allows for automation - without having to invent robots or new types of sensors.

Finally, the application in our case also required to measure below the surface, as the ground condition to a depth of approximately 50cm was highly relevant to its use.

Solution

Put simply, we developed a solution that measures the ground condition by means of its influence on sound-waves.

Physical models of sound travelling through solids consider parameters such as elasticity, density, moisture and temperature. One could therefore expect that the overall condition of ground may have some influence on the behaviour of underground sound-waves. If we were to play back a given, fixed test sound, the current conditions would alter the test sound in a specific way as it travels through the ground. When the ground condition changes, one could expect that the influence on sound waves would change also. Knowing that our technology is capable of detecting even the smallest changes in sound timbre, we should therefore be able to relate sound modulations to ground conditions.

Proof of Concept

A proof-of-concept run over a few weeks proved these assumptions to be correct. Weather, maintenance and other factors changed the condition of the ground. As the ground changed, the sounds arriving at the underground microphones showed a change in timbre. During periods of constant ground conditions, the timbres showed no change.

Prototype System

The underground measurements were conducted via a Sound Loop Analysis Process (SLAP). Several sounds with known timbre were played back via an array of transducers burried underground. A corresponding array of microphones deployed in underground air chambers recorded the underground sounds in some distance from the transducers.

The system employed two equidistant arrays of transducers and microphones at 1m intervals, covering an area of 7m length. We used Earthquake 1000W transducers which due to their ruggedness and reliability proved to be a perfect choice. Behringer ECM8000 reference microphones provided sufficient detail at considerable cost. Each array was controlled by Internet-enabled telemetry units. Measurements were automatically performed on an hourly base, 24/7 for several months.

The system published the updated penetrometer value to a cloud infrastructure triggering the necessary data processing and publication of current conditions via a web interface.

Data Analysis

The design did not require interpretation of the timbre (or its changes) recorded by the underground microphones. Instead we conducted a calibration phase during which a relationship between a set of different timbres and known conditions (obtained via traditional methods) was established. Subsequent measurements then used these references to map from current timbre to current ground condition.

The timbres of reference signals and all measurements were stored and processed by means of unique image-based representations. Simple image-matching algorithms delivered the metric required to compare current with calibration timbres.

The Wrap

At the end of the project, the prototype had gathered 50,000 audio recordings. The condition forecasts were within the same accuracy of the manual measurements performed by maintenance personnel. Measuring the condition of ground (or any other solid for that matter) by means of timbre analysis works.

Credits

Sunics is bound to confidentiality but would like to sincerely thank all stakeholders and helpers involved in this project. You know who you are.