Set up time series alerting

This page will walk you through creating a time series alerting automation that identifies time periods when the pressure loss for machines on a hypothetical Machine root object type is above 120 kPa.

1. Select a root object

In Quiver, select the relevant object type (Machine in this example), and choose an individual root object from the list using the Object selector card. Below, we selected the Machine 1 object.

2. Select time series properties as inputs

In our example below, the Machine object type has two time series properties (TSPs): Inlet pressure and Outlet pressure. Add your desired TSPs to your Quiver canvas with the Object time series property card, or pop out the time series property directly from the object set card. These TSPs will become the inputs to our calculations in the next step.

3. Apply time series transforms (optional)

Quiver offers a wide range of time series transforms, such as derivatives and rolling averages. All transformations that are supported in derived series are supported in time series alerting automations. See the list of supported operations.

In our example, we will use a Time series formula card to add a transformation that calculates the difference between the inlet and outlet pressure of the Machine 1 object.

4. Apply a time series search

Add a Time series search card to identify periods of interest in a time series. In our example, we will identify periods when the change in pressure was greater than 120 kPa.

5. Save the time series search logic as an automation

From the Time series search card configuration panel, navigate to the Automations section and modify the maximum lookback window as needed. The maximum lookback window defines how far back the automation will look when incrementally running your search. We recommend setting this window to be the maximum of any time windows you define in the underlying series of the search (for example, the window configuration of a rolling aggregate series). If your maximum lookback window is too large, we will only "look back" 50 transactions.

Next, select Add automation. A new window will open in the Automate application where you will be guided through saving the time series alerting automation.

6. Modify the automation scope (optional)

By default, the time series search logic will be applied to all objects in the selected root object type. In our example, the pressure drop of 10 machines will be analyzed. You can preview the logic on a different root object by selecting the dropdown menu at the top of the logic tree.

If you wish to limit the scope of your automation, select Edit Logic and then Edit scope. Remember to also select Apply changes after adding your filter.

7. Select time series alert object type

Now, you need to choose an object type where new events will be added as objects. From the edit view of the time series condition page, navigate to the Select time series alert object type section. If you do not already have an object type that supports time series alerting automations, choose to Create new object type. You must have permissions to perform Marketplace installations and create object types with backing datasets to create new object types. Additionally, your object type and the time series alerting automation must be stored in the same Project.

If you create a new object type, you must wait for it to fully index into the Ontology before you can proceed with setting up your automation. We recommend waiting up to ten minutes.

An existing object type can be used if it has properties with the following type classes:

• timeseries:alert_event_id
• timeseries:automation_rid
• timeseries:automation_version
• timeseries:root_object_pk
• timeseries:root_object_type_api_name
• timeseries:start_time
• timeseries:end_time
• timeseries:generation_time

8. Modify evaluation frequency as desired

As mentioned in the overview, the time series alert object type can store alerts from one or many automations. This object type relates to exactly one evaluation job. Therefore, one job can relate to many automations. A job is a batch Spark job and can be viewed in the Builds application. Specifically, the job outputs a dataset which backs the alert object type, where each row is an alert. By default, the job will run whenever there is new data in the datasets backing the time series used in your alerting logic. However, after you save your automation, you can put the evaluation job on a time-based schedule if you prefer.

The evaluation frequency mentioned at this stage in the automation setup is not the same as the evaluation job schedule. The evaluation frequency defines how often the automation will trigger any included effects. As effects are optional, it is possible that this evaluation frequency will have no effect on your automation.

8. Modify remaining fields as desired

The automation can be saved without any effects or custom settings. Modify any of the remaining configuration sections as desired and select Next until you reach the Summary page.

9. Create your automation

Name your automation and select a save location. Your automation must be saved in the same Project as your selected alerting object type. Fix any issues highlighted in the Security settings section. Finally, choose Create automation.

At this point, the logic backing your automation has been fully separated from the Quiver analysis in which you started. Modifying the Quiver analysis will have no effect on the automation, and vice versa. To modify the logic backing your automation, navigate to the Automate application.