Schedule a meeting
The power of Automation and Accuracy
Comparison between traditional handling of abnormal behaviour and the automated Asset monitoring of Eneos
Abnormal behaviour undetected
Equipment damage
Water, energy or production loss
No localization possible
No alarm or False alarms
Meter-data underused
High maintenance/config. effort
Detect relevant anomalies
Prevent damage
Minimized losses
Localized alerts
Targeted contextual alarms
Virtual data-analyst on every meter!
Fully automatic roll-out
What?

An anomaly is a deviation from what is standard, normal, or expected. Simply put: anything that goes wrong or that is unusual. To accurately detect anomalies, you need to know how the asset should behave every moment in time, in any context (e.g. a cold sunny day with a certain wind speed and direction combined with a certain production phase that is running).
How we do it
Eneos builds incredibly accurate models, with limited amounts of data!
(No need for incredibly dense metering, configuring loads of rules or complex "manual" models such as BIM models)
Just send us the main metering data on the aspect you want to model (e.g. Energy consumption, Steam usage, production efficiency, etc.). Our proprietary AI technology will collect and derive context from it and create the best possible model, fully automatic.
This model mimics the normal behavior of a building/process/site, considering the interfering variables.
On top of that we have built a statistical layer that analyses if a deviation is probably just a glitch or an actual anomaly. This way you only get notified when needed, limiting distractions by false alarms.
Overall our system is more accurate & way more economical than a big team of data scientists, is scalable by design = fast deployment, requires almost no configuration and will give you ONE overview off all of your assets.
And as with everything we do: fully integrable with your existing EMS/data collection platform. No need to change platforms
Accuracy of model
The more accurate the model is, the better it can mimic the normal behavior. This way it will also better detects deviations from it. Our accurate models avoid the following:
- Lots of False Positives: There are a lot of alarms for e.g. Leakages, but often there is nothing wrong. This creates overhead and people start ignoring notifications.
- Missing Real Abnormalities: Anomaly remains detected - results in losses and damage
The anomaly model accuracy can be compared with following two statistical parameters:
True Positive Rate (TPR)
This gives an indications of the correct predictions, the times an alarm was generated, when there really was an anomaly.

In this formula following therms were used:
- Nti: the amount of anomalies which were detect
- Ni: the total amount of anomalies
False Positive Rate (FPR)
This gives an indications of the "False Alarms", the times an alarm was generated, without there being an anomaly.

In this formula following therms were used:
- TPR: True Positive Rate
- Ni: the total amount of anomalies
- N: amount of detections (True and False)
Detect problems faster, prevent damage and waste less
- No configuration –fully self-learning
- Easy to roll out across meters/installations
- Applicable on multiple streams (Energy, Process, Water,…)
- High accuracy, few false alarms
- Detects abnormal behavior in complex conditions
The faster an issue is detected, the faster it can be mitigated. The awareness of a fault in the system is key here: be able to react fast!
Detection of water leakage
Below a visual representation of how Eneos Anomaly Detection has detected a leakage 6 months earlier than it would have been without our software. As one can see: the data in red (the raw metering data) do not show any clear pattern indicating the issue to the naked eye!
Want more information? Feel free to request a meeting or contact us for our paper on this use case!


+32 456.14.12.73
info@eneos.cloud
Rolwagenstraat 63
2018 Antwerpen