See Anomaly detection, matching and learning in action.Live Demos
Mun Hean provides electrical power monitoring solutions to commercial buildings across South East Asia. These include Hotels (Marina Bay Sands, Singapore), Universities and Government Offices.
They wanted to proactively seek out and replace temperamental infrastructure components according to deterioration signals in their monitoring profiles. Threshold alerts had failed to flag these events as anomalies, so Anomify was brought in.
Mun Hean leveraged the supervised nature of Anomify’s technology to define patterns of expected behavior where no deterioration signals were present. Gradual deterioration signals don’t look like anomalies to the untrained eye, and don’t deviate significantly enough to be detected by threshold based alerts.
Following a period of evaluation the Mun Hean engineering team were confident that Anomify was only picking up the anomalies that mattered to them. No true positive anomalies were missed and false positive anomalies accounted for less than 1/3 of all alerts, and continues to improve.
Predictive maintenance relies on spotting nuanced signals.
Energy monitoring professionals need to know when hardware is about to fail so that components can be replaced. IOT data loggers at hardware sites send health data to the cloud, but there isn’t time to be pouring over graphs all day to determine faults. They need a service that will reliably inform them when health changes in an unexpected way. In the Anomify platform domain experts collaborate with the ML brain behind Anomify to define patterns of normal behavior which informs it of signals to look out for. Certain signals are early signs of faults so it’s critical they are alerted on, in order that the faulty component can be replaced before it fails.