Skip to content

Blog · Protocols

OPC UA for inspection: when a semantic protocol beats raw I/O.

Updated July 2026 · 6 min read · Adente Vision Engineering Team

OPC UA suits inspection when the line needs structured, self-describing results at the MES or SCADA layer, not a single pass/fail bit. As the standard's information model, it carries named per-part data such as variant, confidence and millimeter measurements. Reach for raw I/O when one reject bit is all you need.

What does OPC UA add over discrete I/O and classic fieldbus?

OPC UA adds meaning to the data an inspection unit sends. A discrete output is one bit whose meaning lives only in your wiring diagram, and a classic fieldbus carries numbers whose interpretation is agreed off-line. OPC UA instead exposes named, typed values in a browsable information model, so a consumer can read "Confidence" or "MeasurementMillimeters" by name rather than by memorizing which register or which wire it is.

In industrial automation that difference is what the OPC Foundation standard is built for. The information model is self-describing: a client can browse the server, discover which values exist, read their data types and units, and subscribe to changes, all without a hand-maintained tag list on both ends. That is why OPC UA is the usual carrier at the MES and SCADA layer, where the consumer is a database or a plant system rather than a robot waiting for a single bit.

Adente Vision is an edge-AI visual inspection unit built by ADENTE Advanced Engineering Technologies, part of the Aden Group, sold through automation system integrators. It supports OPC UA alongside the other four protocols, so an inspection result can be published as structured data rather than reduced to a lamp on a panel.

When does a pass/fail bit stop being enough?

A pass/fail bit stops being enough the moment someone downstream needs to know why a part passed or failed, or by how much. Traceability, statistical process control and MES record-keeping all want the per-part detail behind the decision, not just the verdict.

That is where the richer result matters. The unit can report per-part dimensional measurement in millimeters, feature coordinates and the model's confidence, and OPC UA is the natural way to carry those to a plant system as named values. A quality manager building a traceability record wants the measured dimension and the confidence attached to each serial number, which a single reject output cannot express.

The unit also exposes results on a web-based dashboard for remote monitoring, so an operator or engineer can watch the same per-part data live without wiring into the fieldbus at all. The dashboard is for people; OPC UA is for the systems that store and act on the data at scale.

OPC UA vs discrete I/O for inspection results

The choice is not which protocol is better, but what the result has to carry and who consumes it.

DimensionDiscrete I/O (4-in / 4-out, 24V)OPC UA
What it carriesOne or a few pass/fail bitsNamed, typed values: variant, confidence, mm measurement
Data modelMeaning lives in the wiring diagramSelf-describing, browsable information model
Typical consumerRobot or PLC reacting to a bitMES, SCADA, historian, traceability database
SetupWire and assign the outputsModel the nodes once, clients browse them
Best whenA reject action must fire nowStructured per-part data must be recorded

How does an edge unit act as the OPC UA server?

The unit itself is the OPC UA server, so there is no separate PC or gateway between the inspection and the plant system. It runs inference on-device, forms the result, and publishes it from the same enclosure, which keeps the data path short and keeps raw imagery on the line.

Edge-first hosting matters for data sovereignty. Because the server runs on the unit, the images never leave the line: only the structured result, the confidence, the measurement, the variant, is published to MES, while the frames stay on-device. For a plant that will not send raw part imagery off site, that is the difference between a reportable result and a compliance problem.

It also simplifies the architecture. A client on the MES side browses the unit's model, subscribes to the values it cares about, and receives updates per part, without a middle server translating between the camera and the plant. The inspection unit is a first-class participant on the network rather than a sensor bolted onto one.

When is OPC UA overkill for an inspection cell?

OPC UA is overkill when the only thing the result must do is fire a reject in real time. If the cell needs a fail bit to drive an air blast or hold a robot, a discrete output or a real-time fieldbus does that with less setup and tighter timing than modeling and browsing a node.

The common design is to use both. Carry the time-critical reject on discrete I/O or a real-time bus so the actuator fires immediately, and publish the per-part record over OPC UA for MES and traceability in parallel. The reject does not wait on the reporting layer, and the reporting layer does not compromise the reject.

So the decision reduces to two questions: does the result need to fire an action now, or be recorded and queried later, and who is the consumer. For the recording-and-query job, OPC UA earns its place. For choosing among the real-time carriers, see the sibling post on rule-based versus AI machine vision for the decision-first mindset, and the note on EtherCAT for high-speed cells. The full method sits in the pillar guide, and the connectivity behind this sits on the Adente Vision system page.

Frequently asked questions

Need per-part results in your MES, not just a reject bit?

Schedule a walkthrough and we show the OPC UA model, the per-part measurement and the confidence the unit publishes, before quoting. See how Adente Vision reports structured inspection data from the edge.