Anomaly detection with Anomaly Detect Type - Detector

When Batch Tracking receives a detection signal that is associated with an Anomaly Detect configured as a Detector, the following deadband windows are observed:

1. Normal Detection window: This window is defined using an upstream normal detection volume (using a positive value) and a downstream normal detection volume (using a negative value).

2. Early/Late Detection window: This window is defined using an upstream early detection volume (using a positive value) and a downstream late detection volume (using a negative value).

The following diagram depicts the two detection windows that can be configured for an adetect record:

Note: The definition of upstream and downstream always aligns with the direction of flow. Using the diagram above, if the flow in the pipeline is moving from N1 to N3, then the area to the left of N2 is considered upstream, and the area to the right of N2 is considered downstream. For a reversed flow moving from N3 to N1, the area to the right of N2 is considered upstream, and the area to the left of N2 is considered downstream.

Using this configuration, the following scenarios are possible:

Scenario 1: The detection occurs when the simulated location of the anomaly is within the Normal Detection window. An anomaly detected within this window upstream of the adetect is considered upstream normal. An anomaly detected within this window downstream of the adetect is considered downstream normal. In both cases, the anomaly is moved to the adetect location and adjusted by an offset to account for the time elapsed since the detection has occurred. An alarm will be raised to indicate this occurrence.

Scenario 2: The detection occurs when the simulated location of the anomaly is outside the Normal Detection window but within the Early/Late Detection window. An anomaly detected outside of the Normal Detection window but still within the upstream early detection volume is considered upstream early. An anomaly detected outside of the Normal Detection window but still within the downstream late detection volume is considered downstream late. In both cases, the anomaly is moved to the adetect location and adjusted by an offset to account for the time elapsed since the detection has occurred. An alarm will be raised to indicate this occurrence.

The following diagram depicts an example of an upstream early detection for anomaly A1. The anomaly will be automatically moved, and the same behaviour applies for downstream late detection.

Scenario 3: The detection occurs when the simulated location of the anomaly is outside the Normal Detection window and the Early/Late Detection window. If an anomaly is detected upstream of the Early/Late Detection window, it is considered upstream too early. If it is detected downstream of the Early/Late Detection window, it is considered downstream too late. In both cases, Batch Tracking will not perform any automatic adjustments to the simulated location. An alarm will be raised to indicate this occurrence. The Controller can then decide if the detection signal does in fact correspond to the simulated anomaly and make adjustments accordingly.

The following diagram depicts an example of an upstream too early detection of anomaly A1 outside of the Early/Late window. The anomaly will not be automatically moved, and the same behaviour applies for downstream too late detection.

Scenario 4: The detection occurs but there is no simulation equivalence of the anomaly in the pipeline. In this case, an alarm will be raised. The Controller can then choose to manually create an anomaly or ignore the alarm in case the detection was trigged by a false signal.

Scenario 5: A simulated anomaly leaves the downstream end of the Early/Late Detection window without a detection signal being received. In this case, an alarm will be raised to capture this occurrence, and the anomaly will continue to move down the pipeline. The Controller can then choose to keep the anomaly in case the detection signal was missed, or delete it in case the simulated anomaly was erroneously created or is no longer in the pipeline.

Scenario 6: The anomaly is detected more than one node downstream of the adetect record.

Anomaly detection records make use of anomaly ETA records to determine an anomaly’s position relative to the node that the anomaly detection record is attached to. For more information on ETAs, refer to the AVEVA Pipeline Operations for Liquids Controller documentation.

While ETA records are generated for all nodes downstream of an anomaly that don’t have ETA calculation disabled, they are only generated for the nearest upstream node to the anomaly.

Because of this, if the Downstream Normal Detection Volume or Downstream Late Detection Volume of an anomaly detect record are beyond one node from the node at which the anomaly detect record resides, then they will not have any effect. If these volumes are configured beyond one node, then if a detection signal is received and the anomaly is beyond one node downstream of the anomaly detection record, then an ADET_MISSING_ETA alarm will be raised.

Additionally, even if the Downstream Normal Detection Volume or Downstream Late Detection Volume are configured to be within one node of the location of the anomaly detect record, if the detection signal for the anomaly is received when the anomaly is beyond one node downstream of the anomaly detect record then an ADET_MISSING_ETA alarm will be raised rather than an ADET_DNS_TOO_LATE alarm.