The Alerts administration page is used to setup alerts. The list of alerts automatically generated based on these definitions are displayed on the Active Events page.
To create an alert, the use picks the "+" button in the menu below which is located at the top left of the page.
This creates a new row in the Alerts config table. Highlighting a row enables the user to copy, edit or remove a row. Picking edit, enables the user to set up the alert.
The sensor or zone providing the values to trigger alerts. For more information regarding triggering alerts for zones vs sensors, please click here.
The measurement of the signal.
Defines the method that is used to evaluate the alert criteria. The user can define 4 different types of alerts:
Pattern based alarms are evaluated using the pattern as a reference (pattern is generated based on some historical data and machine learning and statistical equations). For this type, alerts are only generated if all below conditions are simultaneously met:
An alert informs about an event that may be occurring in the network. Once a pattern based alert is generated, volume lost and duration of the event are automatically computed (more information on the Active Events help page). The event automatically ends whenever real values return to normal, this means values return to the pattern. For high type alerts this means below percentile 80th (values are color coded with a black circle). For low type alerts this means above percentile 20th (values are color coded with a black circle).
Pattern based alarms are useful to trigger and detect anomaly or extraordinary events that don't occur often, and therefore do not make part of the normal behavior/pattern of the zone or sensor. Some examples are events like bursts and leaks, fire events, or a flow meter or PRV malfunctioning (please see this article for some tips how to configure bursts alerts).
The figure below illustrates a pattern based alert, that uses the pattern as a reference.
Figure 1 - Pattern based alert
Absolute based alarms are evaluated directly by comparing real time series values with the user defined threshold. For this type, alerts are generated if all below conditions are simultaneously met:
An alert informs about an event that may be occurring in the network. Once an absolute based alert is generated, volume lost and duration of the event are automatically computed (more information on the Active Events help page). The event automatically ends whenever real values are below (for hight type) or above (for low type) the absolute numerical value defined by the user.
Absolute based alarms do not look at the sensor patterns and are useful to identify situations that do not depend on the sensor or zone pattern, like for example problems related with pressure service requirements - pressure below the minimum or maximum required by law - or tank overflows (see this article for some tips how to configure absolute type of alerts).
Figure below illustrates an absolute based alert that uses the zero reference.
Figure 2 - Absolute based alert.
A No Data alert is triggered whenever there is no data at all for the specified sensor for the minimum duration defined by the user.
A Flat Reading alert is triggered whenever there are a specified number of consecutive flat readings, greater than the minimum duration defined by the user.
Figure below illustrates a Flat Reading alert.
Figure 3 - Flat Reading alert.
The user picks High or Low to indicate if the trigger is a high or low alarm.
If the alert is pattern based (please see above), a high alert means that real time series values need to be above the pattern bands (percentile 95th) and a low alert means that real time series values need to be below the pattern bands (percentile 5th).
For absolute based alarms (please see above), a high alert means that real time series values need to be above the absolute numerical threshold defined while low alert means that real time series values need to be below the absolute numeric threshold defined.
The user defines the real time series where all trigger configurations will be assessed. Available options are the 15 minutes time series, the daily minimum series and the daily average series.
If the user defines 15 minutes, all trigger conditions defined by the user will be checked every 15 minutes, whenever new data arrives to the software, based on the 15 minutes time series. If the user chooses daily minimum or daily average, all trigger conditions defined by the user will be checked every day (midnight) based on the daily minimum or daily average time series.
The 15 minutes time series are good to track real time events in the system like for example big and medium bursts, flushing, fire events, meter failures, PRV failures, pressure problems, tank overflows and other events. The daily minimum time series are for example good to track leakage in the network, by triggering alarms based on the minimum nightly flow (whenever applied to the measurement type flow). And one example of the benefits of using the daily average time series is to track the increase in volume lost (due to leakages) in big transmission mains along a longer period, where sometimes the daily minimum analysis is not useful - for example because it accounts for tanks draining and filling during the night, etc.
This option is only available for pattern based alerts. The user can choose the time period that will be used to serve as reference to generate the alerts:
The user can analyses the differences between the patterns before configuring the alerts, by entering in the sensor details page, and then changing the Pattern dropdown option.
The user defined value from which an alarm can be automatically generated. For pattern based type, the threshold value corresponds to the difference between the real time series value and the pattern (percentile 50th), while for the absolute based type it corresponds to the absolute reference value that can trigger the alarm (for low type it means a real time series value below the threshold; for high type it means a real time series value above the threshold). For burst detection, this threshold usually defines what is the leak/burst flow needed to trigger the alarm.
Note: The Value/Threshold units are those defined in the Digital Twin settings, under Administration.
Pattern based alert
Absolute based alert
It defines the minimum number of consecutive outliers needed to generate the alarm.
For pattern based alerts, this corresponds to the minimum duration in which all time series values are above percentile 95th - for high type - or below percentile 5th of the pattern - for low type (and will be represented as red dots in the graphs).
For absolute based alerts, this corresponds to the minimum duration in which all time series values are above the user defined threshold - for high type - or below the user defined threshold - for low type.
Note: Minimum duration variable is used simultaneously with the threshold variable in order to trigger the alarms. For pattern based, alarms will be generated when all time series values are above or below percentile 95th or 5th of the pattern, for the minimum duration defined, and at least one of the outliers value has a difference relative to the percentile 50th of the pattern higher than the threshold defined (if the user does not insert any threshold value, zero will be assumed). For absolute based, alarms will be generated when all time series values are above or below the user defined threshold, for the minimum duration defined.
Alarms are only generated for sensors that have an active configuration. The user has the option to turn off alarms generation for a specific sensor by changing this option to inactive.
For more information about WaterSight, please go to OpenFlows WaterSight TechNotes and FAQ's.