Acoustic Machine Learning

Leakage is costing District Energy Companies both financially and economically due to product loss and the expense of running top up systems. The tide is turning on old systems as spectrogram pattern recognition of sound files allows leaks to be identified with a high level of confidence.

The solution is easy to retrofit to existing pipelines and ensures only genuine leaks are followed up for repair.

The technology is successful in finding previously quiet leaks whilst also identifying and labelling non-leak sounds which would otherwise cause false alarms e.g. electrical interference or control valve operation. The technology is now deployed worldwide and has been developed and trained against over 100,000 sound files.