Guided Analytics

TwinThread Guided Analytics is an implementation of supervised machine learning which enables the creation of intelligent models that provide greater analytics on your assets and processes. Currently, Guided Analytics supports a guided model: TwinThread Process Anomaly Detection. For detailed information, see Select an algorithm.

The Anomaly Detection model can be configured to monitor a set of tags and analyze real-time data for deviations outside of the normal operating range that the model is trained with, based on a customizable sampling of historical data.

When a model is created, AVEVA Insight creates a new tag under the asset which tracks the anomaly score. Anomaly scores are an indicator of how much the current value deviates from the normal operating conditions. This tag uses the format $Analytics.ModelName.AnomalyScore.

When an anomaly is detected, a story is generated in the News Feed. Note that if an asset configured with Guided Analytics is deleted, any previously generated news stories will remain without an active link back to the asset.

Note: To use Guided Analytics, your AVEVA Insight solution name must be less than 50 characters and cannot contain any of the following special characters:
!@#$%^&*()1234567890{}[]\|:;"'<,>.?/*-+~`

View anomalies in a chart view

News stories generated by Guided Analytics can be selected to load a chart view for further analysis. The default chart view will include all tags from the model, though a maximum of three tags decided by anomaly score will be active on the chart when initially viewed. The chart view will also include the tag created to track the anomaly score.

See updates about your published data on a line chart. Select the news icon to display the news item inline with the chart.

To remove news from a line chart, go to your account details and turn off the Enable News Items on Trends setting.

Model filter and Operational Mode Indicator

Anomaly Detection models can be trained with a filter. By applying a filter to a given model, you train a model to not report on expected deviations from a given asset. For example, if a tag reports a specific value when no data is being reported by the data source, this value can be filtered out of anomaly reporting.

Anomaly Detection models also support an Operational Mode Indicator. The Operational Mode Indicator is a tag that can be used to train the model on variances that can occur when an asset changes modes. The Operational Mode Indicator tag must be a discrete or string type, it can not be a float.

For example, the Operational Mode Indicator tag can capture how much power is being delivered to the asset. By including this tag in the model definition, the model can account for deviations such as temperature spikes and increased fan speeds based on a "full power" operating mode. The Operational Mode Indicator can also indicate which batch process an asset is set for and account for deviations between processing jobs.