273: Getting Data From Your Assets¶

Where this course fits¶

Before you can do anything useful in Kelvin, you need data. Data comes from your assets through Connections. This course covers everything you need to know about setting up and managing that pipeline.

1. 270: Exploring the Kelvin UI for Platform Administrators — the Administration area overview
2. 271: Setting Up Platform Infrastructure — clusters, nodes, and users
3. 272: Setting Up Assets and Data Streams Quickly — creating assets in bulk
4. This course — connecting those assets to real data
5. 274: Asset Data Assurance — data quality and guardrails

What is a Connection?¶

A Connection (also called a Connector) is a software component that runs at the edge and bridges the gap between your physical assets and the Kelvin Platform.

Your asset speaks its own language. A PLC might speak Modbus. A SCADA system might speak OPC UA. Kelvin speaks its own language: Assets and Data Streams. The Connection is the translator between the two.

There are two types of Connection:

Import Connections are the live link between Kelvin and your physical assets — exactly like how a SCADA system communicates with a PLC. They read real-time data from your assets continuously, and can also write commands or setpoints back to them. This is the connection type you will use for your assets.

Export Connections take data that is already in Kelvin and push it to a third-party system — for example, sending historical process data to Databricks for ML model training, or to an Azure Data Lake for long-term storage. Export Connections are always custom-built by a developer.

This course focuses entirely on Import Connections using OPC UA. Export Connections are briefly introduced at the end.

How an Import Connection Works¶

When you deploy an Import Connection, it runs as a workload on your edge cluster. It polls your asset at a configurable rate, reads the data, and stores it temporarily on the node and then sends it to the cloud.

Data also flows the other way. When a SmartApp generates a recommendation and an Operations Engineer approves a control change, Kelvin sends the new setpoint back down through the same Import Connection to the physical asset. This is the write direction — just like a SCADA system sending a new target value to a PLC.

The four things you configure when creating a connection:

Available Protocols¶

Kelvin ships with built-in connectors for the most common industrial protocols. You do not need to write any code to use these.

Creating a Connection: Step by Step¶

This walkthrough uses OPC UA as the example. The process for Modbus and Emerson ROC follows the same steps with different configuration parameters.

Before you start¶

You will need:

• A running edge cluster with at least one node (covered in Course 271)
• Assets and Data Streams already created in Kelvin (covered in Course 272)
• The OPC UA endpoint URL for your asset (for example: opc.tcp://192.168.1.100:48010)
• The Node IDs for the data points you want to read

Step 1: Open the Connections page¶

Go to Connections in the left sidebar under the Connectors section.

You will see a list of any existing connections. Click the Create Connection button.

Step 2: Select connection type¶

You will see two options. Select Import Data and Control Setpoints and click Next.

Step 3: Select the protocol¶

A list of available protocols appears. Select OPC UA, choose a version, and click Next.

Step 4: Configure the connection¶

This is where most of the setup happens. All sections must show a green check before the Create button becomes active.

Section A: Information¶

Section B: Configuration¶

The key parameters for OPC UA:

Section C: IO Mapping¶

This is the most important section. It maps each physical data point on your asset to a Kelvin Asset and Data Stream.

For each data point, you need:

Section D: Cluster¶

Select which cluster to deploy the connection to. The cluster must have direct network access to your OPC UA server.

You can optionally select a specific node. If you do not, the platform assigns one automatically.

Step 5: Create the connection¶

When all sections show a green check, the Create button turns blue. Click it.

The connection will appear in your Connections list. Its status will move through:

1. Pending — created, waiting to deploy
2. Deploying — being installed on the edge node
3. Running — active and collecting data

After a few minutes, your connection is live. You can verify it is working by navigating to the Data Explorer and selecting your assets and data streams to view the live data.

Managing Your Connections¶

Once a connection is running, you have four management actions available.

Starting and Stopping¶

You can pause a connection without removing it from the edge. The workload stays deployed on the node — you can restart it at any time without redeploying.

From the Connection list:

Go to Connections. Click the Start/Stop button on the relevant row.

From Connection Details:

Click on the connection name to open its detail page. Click the Start/Stop button at the top.

Updating a Connection¶

You do not need to delete and recreate a connection to make changes. Four things can be updated on a deployed connection:

Update the version:

Go to Connections. Click the upgrade icon on the row, or click the connection name and use the Update Connection button. Select the new version from the list.

Update the configuration:

Click the connection name. Go to the Configuration tab. Edit the settings in the UI, YAML, or JSON view. Click Apply and Redeploy when ready.

Update the deployment:

Click the connection name. Go to the System tab. Adjust environment variables as needed. Click Redeploy.

Update the IO Map:

Click the connection name. Go to the Connection IO tab. Click Edit IO Map. Add, edit, or remove assets and data stream mappings. Click Save.

Deleting a Connection¶

Deleting a connection is permanent. It removes the workload from the edge entirely. You cannot undo this.

From the Connection list:

Go to Connections. Click the Delete icon on the relevant row. Follow the confirmation prompt.

From Connection Details:

Click the connection name. Click the hamburger menu at the top right. Select Delete Connection. Confirm.

Building a Custom Connection¶

The built-in OPC UA, Modbus, and Emerson ROC connectors cover most industrial use cases. But if your asset uses a different protocol — or you need to pull data from an API, database, or other custom source — you can build a custom connection using the Kelvin SDK.

Building a custom connection is a developer task. As a Platform Administrator, you do not need to write the code. Your responsibility starts once the developer has uploaded the connection to the App Repository.

The four stages for a developer building a custom connection:

Once a custom connection is deployed, you manage it exactly the same way as any other connection — start, stop, update, and delete all work identically.

For full developer documentation:

• Create an Import Connection
• Create an Export Connection
• Deploy an Import Connection
• Deploy an Export Connection

A Brief Note on Export Connections¶

Export Connections take data that is already stored in Kelvin and push it to a third-party system for further processing. They are always custom-built by a developer — there are no out-of-the-box export connectors.

The most common use is sending data to platforms like Databricks for ML model training and advanced analytics, or to cloud storage like Amazon S3 or Azure Data Lake for long-term archiving.

Export Connections are a developer responsibility. Once built and uploaded to the App Repository, you deploy and manage them the same way as import connections.

What Comes Next: Data Quality and Guardrails¶

Getting data into Kelvin is only the first step. The next question is: can you trust the data?

Course 274 covers Asset Data Assurance — how to configure data quality rules and guardrails that protect your platform from bad data and runaway control changes.

A few things to look forward to in Course 274:

• Setting data quality metrics per data stream
• Defining guardrail rules to catch out-of-range values
• Using static value thresholds and dynamic data stream thresholds
• Understanding what happens when a guardrail rule is breached

Quick Reference¶

Quick Quiz¶

Summary¶

A Connection is the bridge between your physical assets and the Kelvin Platform. Import Connections communicate with your assets — reading live data from them and writing commands or setpoints back, just like SCADA talks to a PLC. Export Connections push data from Kelvin outward to third-party systems like Databricks or data lakes for further processing.

The key things to remember:

• Kelvin ships with built-in connectors for OPC UA, Modbus TCP/RTU, and Emerson ROC. Use these for standard industrial assets.
• When creating a connection, the four key things to configure are: Information (name), Configuration (protocol settings), IO Mapping (which data points map to which Kelvin assets and data streams), and Cluster (where to deploy).
• For large IO Maps, use the CSV download/upload feature to save time.
• Once deployed, you can start/stop, update, or delete a connection. Stop is temporary — delete is permanent.
• You can update a connection's version, configuration, deployment settings, and IO map without deleting and recreating it.
• Custom connections are built by developers using the Kelvin SDK, but deployed and managed by Platform Administrators in the same way as built-in connectors.
• Course 274 covers the next step: data quality rules and guardrails to ensure the data coming in from your connections is trustworthy.

For full documentation, visit:

• Connectors Concept
• Connect to Data Quick Start
• Update a Connection
• Start/Stop a Connection
• Delete a Connection
• Develop Custom Connections

Good luck and Happy Learning !