Questions this answers
- What is an Asset Administration Shell and why is it needed?
- How do OPC-UA and AAS differ?
- What kind of standard is KS X 9101?
- In what order should manufacturing data standardization proceed?
The most common reason a manufacturing AI project runs long is not that the modeling is hard. The time goes into getting the data into a usable state.
If someone has to check by hand whether the value machine A sends as “temperature” is the same as what machine B sends as “TEMP,” then adding a single line means starting that work over.
Standardization removes that repetition.
Three standards, three different jobs
These are easy to confuse, but they are not competitors — they sit at different layers.
OPC-UA — how it is exchanged
An industrial communication standard for how equipment and systems exchange data: what protocol connects them, in what form values are passed, and how security and authentication are handled.
Think of it as the layer that reconciles the communication methods each equipment maker does differently.
AAS — how the asset is described
The Asset Administration Shell is a standard model for describing an asset digitally — a machine, a component.
It expresses, in a fixed structure, what the asset is, what properties it has, what values it emits and what documents accompany it: manuals, certificates, drawings.
If OPC-UA is “how it gets there,” AAS is “what it is.” A connection that works but carries values nobody can interpret is of no use, so both are needed.
AAS earns its keep most clearly in the supply chain. Receiving asset information in AAS form along with the equipment itself substantially reduces the work of connecting it to your systems.
KS X 9101 — Korea’s manufacturing data scheme
A Korean Industrial Standard covering the collection, storage and use of manufacturing data. If government support programs or supply chain integration in Korea are in view, aligning to it early makes later work easier.
What happens if you skip it
In the moment, standardization looks like an unnecessary detour. “Collect the data first, tidy it up later” is the natural choice.
The catch is that the later cost is not linear.
Adding a line — the mapping work starts again for each new line, and accumulates.
Replacing equipment — swap in a machine with the same function but a different data format and every analysis connected to it breaks.
Looking at historical data — if a field was renamed partway with no record, you can no longer tell which period of data is usable.
Integrating with the supply chain — exchanging data with another company eventually requires a standard, and at that point everything gets reconciled anyway.
The stages
1. Assess. List what equipment emits which values, and where and under what name those values are stored. It sounds routine, but this is where most of the time goes. Documentation and reality frequently disagree.
2. Define the data (AAS). Describe each asset’s properties and values against the standard model. You do not need every machine at once — start with the core equipment you intend to analyze.
3. Connect the equipment (OPC-UA). Unify the communication layer. Older machines often do not support OPC-UA directly, so a gateway goes in front of them.
4. Storage and use (KS X 9101). Load the collected data into the standard scheme so it can be found and reused.
5. Operate. Establish a procedure for registering new equipment against the standard. Without it, you are back where you started within a few years.
You do not have to do all of it
Framing standardization as an enterprise-wide programme makes it hard to start. The practical approach is to standardize what you intend to analyze first.
If five machines are in scope for predictive maintenance, put those five into standard form. Widen the scope once the result is confirmed. We have seen a number of projects try to standardize a whole plant from the outset and stall at stage two.
IMPIX analyzes equipment characteristics and shop-floor data, then builds collection, storage and utilization schemes on AAS, OPC-UA and KS X 9101. If you would like to start by assessing the current state of your data, get in touch.