Digital Twin Series · Part 3
Digital Twin Architecture: Deep Dive
What sits between the real machine and the screen, and what does each layer do?
A digital twin is not one product. It is a stack: the physical asset, the sensors that capture its data, an edge device near the machine, the network, a data platform, the virtual model and the applications people use. Each layer has one job and one set of technologies, and a weak layer limits everything above it. This page walks the stack layer by layer, shows how it maps to the ISO 23247 reference architecture, adds up the delay from sensor to action, and ends with the design points and the implementation steps.
Select a layer. Data moves from left to right, and decisions and commands move back to the asset through the feedback loop. The ISO 23247 view groups the layers into the entities of the manufacturing digital twin framework. The standard describes the entities and does not prescribe how to build them.
A simplified delay budget. Both paths start with the sampling interval and the link to the edge device. On the edge path the analysis and the action stay local. On the cloud path the data travels to the cloud and the answer travels back. Real values depend on the network, the services and the load, so measure them on your system.
| Protocol / standard | Purpose | Typical use |
|---|
| Area | What it does | Common examples |
|---|
Examples only: the list is not a recommendation. Product names change, so check each vendor.
| Use case | What it does | Benefit |
|---|
Layer names differ between vendors and papers: some show five layers, some seven. The content is the same. The delay calculator is a teaching model. Secure every layer and every connection, and follow the standards that apply to your site (for example IEC 62443 for industrial cybersecurity).