Leaking pipes from water distribution systems are a huge issue which has maintained worldwide
attention. A common method of detecting leaks is through the cross correlation of leak vibroacoustic emission signals, created as the water discharges through the hole/crack and recorded
by acclerometers or hydrophones placed either side of a suspected leak location. It is thought
that a number of factors influence leak signals, however there has been little comprehensive study
in controlled conditions evaluating how and what way these factors influence the characteristics
of leaks signals. Moreover, during the process of leak noise correlation, accelerometers and hydrophones are recording information about the leak contained in the leak signal that is currently
not understood. Knowledge of these factors would be highly beneficial to water companies in order
to allow for prioritisation of leak repair.
The research presented herein aimed to derive a method in order to predict the flow rate,
area and shape of a leak using the vibo-acoustic emission signal recorded when performing cross
correlation. A unique methodology was developed which allowed for the isolation of individual
physical variables and how this can influence a leak signal. Specifically, the study focused on
developing a fundamental understanding of how leak flow rate, area, shape, pipe material and
backfill type influenced the characterstics of leak signals.
The results showed that the leak flow rate, shape, backfill types and pipe material all influence
the leak signal. The influence of leak area on the leak signal appeared negligible when leak flow
rates were standardised. Signal processing and machine learning algorithms were applied to the
leak signals and the results showed that it was possible to predict leak flow rate regardless of leak
area, leak shape and backfill type. Moreover, alternate algorithms showed that it was possible
to predict leak shape and leak area from the vibro-acoustic emission signal. This research has
therefore presented a useful and valid tool to predict leak flow rate, leak area and leak shape
which allows water companies to prioritise leak repair and maintenance activities, providing an
opportunity to reduce the volume of water lost through leaks by repairing the larger flow rate leaks
first. Whilst this method shows effective results, it does not provide an exhaustive comparison of
the number of algorithms and techniques which may also make similar predictions.
