Case study

Anomaly Detection

The Challenge

Monitoring industrial systems, like wastewater treatment plants, is becoming more and more complex, due to the rising amount of sensors and available data.

Smart algorithms are needed to automatically identify abnormal behaviour and support human operators.

The Solution

COMBINING MACHINE LEARNING WITH HISTORICAL DATA

‍

1. Using historical data a machine learning model learns to extract the normal behaviour of the system out of the sensors data

‍

2. The model identifies anomalies by reconstructing the original signal and calculating a reconstruction error

‍

3. Monitoring a real system, the model is able to detect abnormal behaviour in real time and alert a human operator

Benefits

Monitor large amounts of sensors in real time
Improve the detection rate of defects
Detect anomalies before they lead to severe damages

Further Use Cases

Fraud Detection: Detect abnormal user behavior
Network Systems: Protect your infrastructure from potential intruders
Predictive Maintenance: Identify abnormal machine behavior

... and many more

Have we sparked your interest?

Let's talk!

Back to Our Work