Digital Twin Architecture or Information Flow - Digital Twin Basics

July 03, 2025

We can divide Digital Twin Architecture or the Flow of Information of a digital twin into different layers. Broadly speaking, there are three layers:

Physical Layer (sensors/edge)
Middle Layer (cloud/ML)
Frontend (visualization/control)

But some additional layers are also mentioned to clear the Digital Twin Architecture Picture.

Digital Twin Information Flow - Digital Twin Architecture
1. Physical Layer (Real World)

This is the physical layer of the Digital Twin. It's where the actual processes happen, like

Machines
Conveyor belts
Robotic arms etc
Embedded with Sensors (e.g., Temp, Speed, Weight, Vision Sensors)

2. Edge Layer (Optional but Recommended)

The data from sensors typically goes through an edge device or IoT gateway, which might preprocess the data, handle protocol conversion.

Edge Devices / Gateways:

Example: Raspberry Pi, Siemens IoT2040, Azure IoT Edge
Functions:

Aggregate sensor data
Filter or pre-process
Secure and compress transmission
Convert protocols (e.g., OPC-UA → MQTT)

Example: A temperature sensor on a packing robot sends readings every second to a Raspberry Pi (edge device) that relays it via MQTT to a central broker or backend.

3. Communication Protocols

Communication Protocols are the method to transmit or receive data. Data is transmitted via:

MQTT (lightweight IoT protocol)
WebSockets (for real-time communication)
REST APIs (for request-response)
Kafka (for enterprise-scale pipelines and batch processing)

4. Backend / Cloud Server Layer

This is where the magic happens! Data is ingested, stored, and processed with real-time analytics and machine learning, enabling alert systems and deeper insights.

Data Ingestion and Storage

AWS IoT, Azure Digital Twins, Custom Node.js/Python server
Ingested via MQTT broker, Kafka topics, etc.

Processing Layer

Real-time analytics or ML
Alert systems

5. Digital Twin Frontend (Unity or Web App)

Built using Unity3D or Three.js, Unreal, etc.
Pulls or receives data (REST API /WebSocket/MQTT/Kafka)
Updates the 3D model or visualization in real-time

6. Digital Twin Output

Now, on the virtual side of the digital twin, we have the Output. We have different ways to show the output. It can be a 2D format or it can be in 3D format for an immersive view. It shows how is the digital twin state currently or you can even see old logs/stats and simulate new conditions or test.

3D visualization (robot arms moving)
Dashboards (temperature chart, anomaly alerts)
User Interactions (control interface, simulations)

🔁 Optional: Feedback/Control Loop

From digital twin → server → device
Example: Adjust machine speed based on analytics

✅ Example in Action:

A robotic packing arm has:

A load sensor
A vision camera

The load sensor sends data to a Raspberry Pi, which publishes to an MQTT broker.
A Unity3D app connects via WebSocket or subscribes to the MQTT topic.
The 3D model in Unity shows:

The real-time weight of the product
Animation of the arm movement
If overloading is detected, an alert appears in the UI.