Monitoring tags via limits or condition monitoring is not always sufficient. Machine learning-based anomaly detection is much more flexible and powerful.
“Show me your past and I’ll tell you who you are” – this could be the maxim of Machine Learning algorithms. Unlike simple limit value monitoring, history and, above all, tendency play a major role in machine learning-based anomaly detection. Thus, one and the same value can mean a completely plausible value in one case and an anomaly in the other.
Anomaly detection with a few clicks in the cloud
With just a few clicks, the anomaly detection of a tag can be activated. A prerequisite for activation is recorded data in the cloud so that the AI algorithm is able to identify patterns and regularities. The data does not have to be evaluated in advance and can be processed without further intervention. This is also referred to as unsupervised machine learning.
The configuration is very simple: Only the sensitivity has to be set – the rest is done by the machine learning service in the cloud. If an existing anomaly is not detected, the sensitivity can be increased. A graphical display visualizes the results of the algorithm during configuration, enabling quick monitoring.
Once the anomaly detection has been set up, all anomalies found are automatically displayed in already configured diagrams. The diagram is also the tool of choice – often the problem does not show up at the value that led to the anomaly, but in the course of the curve before.
Live monitoring
A logical tag is created for each activated anomaly detection. Depending on whether the value of the tag is active or inactive, an anomaly is currently present. The tag can be monitored via alerting and users can be notified via email, push notification, SMS or voice call.
The detection of anomalies has primarily an informative character. No automated action is performed. Ultimately, the user of the cloud project decides whether there is actually a need for action or not. The goal is to be able to bring about the decision as quickly as possible.
Currently, the Machine Learning service of AnyViz Cloud is still in testing phase. Feel free to contact us if you want to learn more about it or try out the AI feature in your cloud project.