UReason's anomaly detection platform monitors data from multiple sensors, establishing a sense of what is 'normal', and raising the alarm when something unexpected happens. When appropriate, its 'plume tracking' capability can direct mobile sensors towards the source of the anomalous signal.
Overview
The anomaly detection platform comprises four main parts – feature extraction, anomaly detection, anomaly classification and anomaly location.
Feature sensors collect data and perform pre-processing to extract features. These are fed into the anomaly detection engine, which learns a sense of ‘normality’ and spots any deviations from normal operating parameters – anomalies. The anomalies generated can be classified to establish whether several anomalies are related, or compared against pre-existing data. Finally, we are working on integration with mobile sensor platforms allowing the system to direct sensors towards the location of the source of the anomaly.
Each part of the platform is fully pluggable, allowing a range of different sensors to be used, providing different types of data; and a selection of different algorithms to be used for the anomaly detection, classification and location.
