Tag binding is 70% of a digital twin project
The rendering is a fortnight. Naming 340 assets consistently, mapping them to tags that were named by six people over eleven years, and keeping the mapping alive through plant changes — that is the project.
Every digital twin proposal underestimates the same thing. Not the 3D, not the telemetry infrastructure, not the front end. The mapping.
A typical plant has tag names accumulated over a decade by several people with different conventions, in at least two languages, with abbreviations that made sense to someone who left in 2019. Some assets have four tags in three systems. Some have none.
We now start every twin engagement with a naming exercise, and we are explicit that it is the majority of the effort. The output is a registry: stable node IDs mapped to tag paths, with an owner and a review date. It outlives the twin.
OPC-UA helps because address-space browsing turns discovery into a mapping exercise rather than a typing exercise. Sparkplug B helps more, because birth certificates announce what exists. Bare MQTT topics are the hardest case and also the most common.
The test for whether the registry is real: add a machine to the line and see how long it takes to appear in the twin. If the answer involves a developer, the mapping is code, not data, and it will rot.
Written by the industrial team at 3D.MBA, Ahmedabad.
Disagree? Tell us