V2/Concepts/Expression

Expression

User can make use of Senfi's flexible autonomous intelligence engine to be alerted of abnormality in the system they are monitoring.

As introduced, an Expression is a tool that you can use to monitor your assets and be notified of abnormal situations or events. You can configure an expression to trigger an event to send email or notify your operators in the Digital Twin. Examples of use cases include: a fire is detected in a building, a lift has broken down, or when a machine is overdue for maintenance. Expression editor is the tool you can use to configure these conditions.

An Expression in Senfi comprises of:

• Inputs - Assets, Metric & Tag values
• Conditions - Data transformation and/or logic
• Result - False to True / True to False
• Event - Trigger Event when Result changes. e.g. fire detected
• Expression execution type - Configures when a Expression is evaluated

Inputs are evaluated by Conditions to generate a single Result. When value of Result changes (from False to True or True to False), an Event can be triggered. Actions defined in the Event such as email and Digital Twin alerts, are executed when the Event is triggered. Expression execution type allows you to configure how the Expression is executed. e.g: how often the Expression is evaluated, whether it's immediate or on poll.

Inputs --> Conditions --> Result

Result Changed --> Event --> Actions

Example

Using our Smart Weighing Scale measurement from earlier, we can form one simple Expression as follows:

• Input: iot_weighingscale_v1.battery_level
• Condition: < 20
• Event: Battery level low
• Action: Send email notification to remind changing of battery
• Expression execution type: Immediate (means evaluate this rule whenever there is any incoming measurement)

Inputs, Conditions and Result

Before adding a Rule, you should already have an event or situation that you wish to be notified about. The event or situation should be one that can be detected by monitoring data received by Senfi.

Select, as Inputs, the Metrics and Measurements from which the event or situation can be detected from. Then, construct the set of Conditions that will evaluate to true when event or situation occurs. The result of said Conditions is the Result of the Rule.

Actions

Whenever Result changes (from true to false and vice versa), Actions may be executed. Actions includes sending an email, request to a webhook or triggering an alert in the dashboard.

Content of Actions

Depending on the Action, different content options are available. Some examples is the title & body of an email, and body content for webhook request. You will be able to describe the event or situation that has happened to the recipient of the notification in the content of an Action

To further aid you in describing the event, you can embed the value or tag of an Input metric in the content of an Action. You can also embed the Rule's Name, Description and Severity in the content of an Action.

To embed:

• metric's value, use ${<measurement_code>.<metric_code>}. e.g: ${iot_temperature_sensor_v1.temperature}
• metric's tag, use ${<measurement_code>.<tag>}. e.g: ${iot_temperature_sensor_v1.site_id}
• Rule's name, use ${NAME}.
• Rule's description, use ${DESCRIPTION}.
• Rule's severity, use ${SEVERITY}.

Note: embedding Rule's name, description and severity is case sensitive. You must use all capital letters.

Take a look at this more detailed example.

Rule execution options

By setting Evaluate Type to Immediate or Polling, you can choose to execute a Rule when new values are received for any Input Measurements, or periodically, regardless of whether any Input Measurements have changed.

If Rule is set to run periodically (Polling), you must specify how often a Rule is executed by setting Intervals in seconds.

Misc

Name and Description

A Rule also has a Name and Description. The Name and Description allows you to describe the event or situation that the Rule detects, such that recipients of the Rule's alert/notification knows exactly what happened and is able to react to it accordingly.

• The Rule's Name is displayed in the CMS' list of Rules.
• The Rule's Name and Description is displayed in the Dashboard.

You can also embed the value or tag of Input metrics in the Description. The syntax is the same as embedding in the content of an Action, as described here.

Description is also the default content of an Action. You can describe the event or situation once in the Description, and reuse the information in all the Actions of the Rule.

Severity

You can also set the severity of the Rule. Default severity is Critical. Severity allows users of the Dashboard to filter Alerts by severity and attend to more severe Alerts first.

Detailed Example

The event that you would like to be notified about is potential fire in a building. There are temperature sensors throughout the building. When a reading from any temperature sensor exceeds 60°C, a fire has probably broken out near the sensor. You would like to be notified by email of such an occurrence.

• The Input in this example would be the metric and measurement that readings from temperature sensor is sent to. e.g: iot_temperature_sensor_v1.temperature
• Condition is iot_temperature_sensor_v1.temperature > 60
• Action is email, with body text:

"Potential fire detected in room: ${iot_temperature_sensor_v1.room}. Temperature in the room is ${iot_temperature_sensor_v1.temperature}°C"

With the above Rule configured, when iot_temperature_sensor.temperature goes from 59 to 61, an email will be sent with the body text:

"Potential fire detected in room: #03-00. Temperature in the room is 61°C"

Note: "room" is a Tag of the iot_temperature_sensor_v1 measurement that indicates which room the sensor is in.

Constructing Condition

This section will guide you to construct a condition using the Rule Editor when adding/editing a Rule in the CMS. The objective here is to tell Senfi how to detect the event or situation that you wish to be notified of.

Rule Editor

The Rule Editor is a graph based "Drag and Drop" editor. Each node in the Editor represents either an Input, Data Transformation, Logic, or an Action. Arrows are used to link the nodes and control the flow of data between nodes.

Input & Output of a node

Arrows coming into the left of a node indicates the node's input,
and arrows going out the right of a node indicates the node's output.

The graph must always start from Input nodes and ends in the Result or Action nodes (Data originates from Input nodes, is transformed by Data Transformation, Logic nodes, and is consumed by Result & Action nodes).

Input nodes

Metric - outputs the value/s of a metric. can be set to output the last known value, or a range of historical values

Logic Nodes

Compare - comparison operators. e.g. <, >=, ==, !=

Logic - perform "AND" or "OR" between 2 or more boolean values

Data Transformation nodes

Filter - used to filter a range of values and output a subset number of values. Can only be used after a Metric node in "Range of Value" mode and before an Aggregate

Aggregate - used to aggregates a range of values into a single value. E.g. sum, average, median. Can only be used after a Metric node in "Range of Value" mode

Function - value-wise transformation. Transform each value, e.g ABS, Floor, Round etc. Can be used after a Metric node and before a Compare node. Outputs the same number of values as its input.

Result node

Result - Result node. Must have 1 and only 1 node. Specifies the Result of a Rule

Action nodes

Action - Action node. Input must always be from the "Result" node. Can have 0 or more "Action" nodes. Instructs Senfi to do something when Result changes.

Basic Steps

1. Drag a node above working area into the working area to create a node.
2. Create "Metric" nodes, one for each Input metric.
3. Select the correct measurement and metric for each "Metric" node. Read: how to configure a node
4. (Optional) If you need to compare metric value against a constant value, use the "Constant" node.
5. Create "Operator" and "Logic" nodes to construct necessary logic
   • logic graph should output a single boolean (true or false) value
   • output from "Metric" node must go to input of a "Operator" node.
   • "Operator" and "Logic" nodes can be nested.
   • You can attach more than 2 inputs to a "Logic" node
6. Create a "Result" node and attach output of logic graph as its input
7. Create "Action" nodes, one for each desired Action and attach output of "Result" node as its input
8. Save when done

Example

Example 1

Example 1: A simple Rule detecting when a Lift has tripped (safetyTrip == 1).

Example 2

Example 2: Detects when a Lift tripped (safetyTrip == 1), AND passengers are trapped in the Lift (load > 0)

Working with historical values

When historical values is required to detect an event, the "Range", "Filter" and "Aggregate" nodes are needed.

By default, a "Metric" node will output the Last Known value of the metric. To obtain historical values of a metric, attach the output of "Metric" node to the input of a "Range" node. The output of the "Range" node will be historical values from the metric.

The configuration options of the "Range" node allows you to change the range of historical values to output. E.g: You can configure it to output metric values from 1 hour ago till now, or the last 10 received metric values.

After obtaining the list of historical values, you can:

• Filter the list using the "Filter" node.
• Transform each value in the list using the "Function" node.

The "Filter" and "Function" nodes can be nested.

Finally, the list of historical values must be aggregated using the "Aggregate" node before its output can be used by a "Logic" or "Operator" node.

Example using historical values

Historical Example 1

The example above detects when a Lift is stopped at Level 2 for over a minute.

Configuring Nodes

Each node contains attributes & options which you can modify. To modify the attributes of a node, move your mouse over the node and click on the "Gear" button that appears near the top right corner of the node (Beside the "X" button). A window will popup to allow you to configure the node.

Aggregate node

This section explains what each of the options for "Aggregate" node does.

sum

the sum of all input values

count

the number of input values

average

the average of input values

median

the median value of input values

max

the largest value

min

the smallest value

mode

the value that appears most often

range

the difference between the largest and smallest value (max - min)

stddev

standard deviation

percentile

percentile. This option accepts an additional "Percentile Value" parameter.

every

outputs true if every value fulfills the condition. This option accepts an additional "Condition" parameter.

any

outputs true if at least 1 value fulfills the condition. This option accepts an additional "Condition" parameter.

Function node

This section explains what each of the options in "Function" node does.

abs

absolute value

round

round to nearest integer

ceil

round up

floor

round down

map

This option accepts an additional "function" parameter. Using the custom "function" parameter, transforms each input value

Next: Computed Measurement

Pre-requisites

A rule requires design and logical programing in order for it to work correctly. If this is your first time creating a rule, please read this guide first.

You should also be familiar with the basic measurements and its concepts. Please read this section if you require a refresher.

You should

• Know what event the Rule should detect.
• Know what you would like Senfi to do (Action) when the event is detected
• Have already created the Input measurement & metrics.
• Know the thresholds or values which indicates occurrence of event.
• Be able to send records to Input measurement & metric.
• (Optional) Be able to send simulated records to Input measurement & metric (to simulate occurrence of event or situation)