Anominer — detection of anomalies in time series
Anominer from Knowtion is a software system for the detection of anomalies in time series.
Here, anomalies include all changes in the data that occur due to disturbances, outliers, drift over time, and further external influences. Anominer detects anomalies in time series by comparing the data with a learned model that describes the normal behavior of the data.
The software employs two phases, the so called learning phase, where the model of the normal behavior is learned, and the evaluation phase, where Anominer monitors the data for changes.
Learning Phase: Anominer learns dependencies over time and over multiple sensors from input data. Learning is performed on-line without delay. The result is the model of the normal behavior of the data.
Evaluation Phase: During evaluation phase, Anominer continuously compares the received data with the model of normal behavior. This evaluation is also performed online allowing Anominer to detect and signal deviations immediately to the user. The result of evaluation is the so-called anomaly indicator, which describes the extent of deviation from the model of normal behavior. The anomaly indicator is calculated for each measurement and is immediately available.
Anominer uses standardized Restful web services for data input and configuration. The imput data consists of the measurement values and a time stamp:
support for up to 20 variables or sensors
processing of up to 5000 data points per second, depending on processor performance
Anominer is especially suitable for repeating processes, such as surveillance of machines, production processes, or test stands. Due to the continuous on-line surveillance, Anominer can be employed within scenarios that require immediate intervention in the case of an error.
Setting up Anominer is considerably simplified due to the automatic learning phase. Complex scenarios, which are difficult to monitor due to their interdependencies and cannot be surveilled with simple thresholds, can easily be monitored by evaluating the learned model.
Use Case Filling Plant
Filling plant error detection: This use case shows the deployment of Anominer for online monitoring of a filling plant. It is shown that Anominer detects changes and disturbances during the filling process immediately.
Use Case DC Motor
DC motor error detection: This use case shows the employment of Anominer for on-line monitoring of a DC motor.
On-line monitoring of process data allows a continuous surveillance of your facility.
Anominer allows the early detection of deviations and disturbances. Cost-intensive manufacturing errors can be detected in a timely manner.
Simple and quick setup is possible due to the automatic learning phase. After product changes, Anominer’s monitoring capability is quickly restored.
Knowtion supports you with extensive services and technical support.
Customer specific adaptions are always possible. New interfaces and for your application specialized algorithms can be implemented according to your needs.
Anominer is being developed as part of a ZIM cooperation project together with the Intelligent Sensor-Actuator-Systems Laboratory (ISAS) at the Karlsruhe Institute of Technology (KIT). This project is funded by the Federal Ministry of Economics and Technology based on a decision of the German Bundestag.