Anomaly Detections

Description of Process

Anomaly detection is examining and analyzing specific data points and detecting rare occurences that seem suspicious because they're different from the established pattern of behaviors. In terms of IFS Cloud solution offering, these anomalies can be detected for a specified period as defined by the asset prediction configuration.

Depending on the pre-configured settings, these anomalies can be either auto/manually reported or auto/manually ignored. Once these anomalies are reported, it will automatically create a work order for the specific anomaly and notify the relevant parties, prompting them to take appropriate action.

Prerequisities

For the Anomaly Detection setup to work properly, the following needs to be ensured.

• Relevant service user to be assigned with permissions to the sites where the target objects are.
• Objects for which time-series forecasting and anomaly detection are performed need to be defined in IFS Cloud as functional objects. The user can create a parameter template that can be applied to functional objects instead of creating the Test points and Parameters manually in the TESTPOINTS/PARAMETERS tab in the Functional Object page.
• Once the ML Data Service is deployed, the returned connection parameters need to be entered in IFS Cloud to ensure the connection to the ML Data Service for the purposes of Anomaly Detection and Time-Series Forecasting.
• To run an Asset Prediction Configuration and obtain results for Time-Series Forecasting; to predict 1 hour or day ahead, a minimum of 228 history data points are required, to predict 24 hours or days ahead, a minimum of 320 history data points are required. Any deviations from these data requirements might result in inaccurate or no results.

Configure and Schedule Asset Predictions

For Anomaly Detections to work, its required to configure & schedule the asset predictions. This involves setting up the configuration data including targets, parameters, time series data, anomaly detection data, site data, schedule data, etc. and can be configured through the “Asset Prediction Configuration” page in IFS Cloud.

Once a configuration is set & saved, the detections will run on the pre-set schedules, detecting anomalies for the past data.

Handle Anomaly Detections

Once anomalies are detected, unless they are set to be reported automatically, there will be a set of anomalies in the ‘Open’ section, where you can report or ignore an anomaly.

When the anomalies are moved to the ‘Reported’ section either automatically or manually, a work order will be automatically created to notify the relevant parties.

During the configuration stage, if a period is set to automatically ignore anomalies when they are detected, the system will automatically move these anomalies to the ‘Ignored’ section after the specified time has elapsed.

Ignored Anomalies cannot be retrieved back to open or reported anomalies.

View Asset Performance Details

Once anomalies are detected, these can be viewed from the ‘Asset Performance Details’ page in IFS Cloud.

The ‘Anomaly Detection’ tab will allow users to view Anomaly data by filtering according to the time-period and the anomaly status.

Similarly, the ‘Forecast’ tab will allow users to view the Time-Series Forecast data by filtering according to the time-period. Additionally, this tab also allows users to get a graphical view of the time-series prediction by filtering according to the time-period, anomaly status, test point and parameters.

View Anomaly Analysis Dashboard

The ‘Anomaly Analysis Dashboard’ in IFS Cloud will enable users to visualize the behavioral patterns of the detected anomaly data in a graphical view. Users can hover over the anomaly graphs to visually pinpoint which parameter values are responsible for the anomaly, facilitating easy identification of contributing factors.

Users can filter according to date, site, parameter, asset type, asset ID wise options to further enhance data visualization clarity.