Services
PLC programmingCondition monitoringSCADA & HMIBuilding automationEnergy managementIoT & electronicsSystem support
Technologies
EVON XAMControlBeckhoff TwinCATSiemens TIA PortalCOPA-DATA zenon
Industries
Machines & plantsPharma & life sciencesEnergy & infrastructureBuildings & services
BlogAbout MiRADA Contact

Insights / Automation

Modern PLC architectures: what really matters in planning

Selecting a control platform is the easiest decision in an automation project. The architecture of the program is not. It determines whether a plant will still be extensible in two years – and whether the next external service provider can find their way around in it.

The following five rules come from practice. They do not sound spectacular, but they decide whether an automation project lasts a year or fifteen years.

1. Separate process logic and device coupling

The most common source of errors in legacy plants is the mixing of process logic and hardware access. A signal is read, checked and calculated – all in the same function. As long as the plant runs unchanged, this goes unnoticed. At the first sensor replacement or the switch to a different controller environment, the structure falls apart.

The recommendation from practice: hardware access lives in a dedicated function block, the process logic receives internal signals only. Replacing a device type then affects exactly one place in the program – and not half the project.

How to recognise it in practice

  • Addresses or channels appear in several functions.
  • A change to the sensor has an effect in many places.
  • Replacing hardware drags changes into the actual logic along with it.
  • Searching for a specific terminal assignment takes longer than the change itself.

2. Functional blocks instead of spaghetti

Functions that belong together technically belong in one function. A sequence, an interlock, an error value handling – one unit each with a clear task. Named blocks with meaningful identifiers take more time to write and save considerably more during maintenance.

The decisive point is the naming. A block called FB_100 prevents no error – but it prevents the error from starting with the search. Meaningful names like FB_FillLevel_Ventilation cost seconds and save hours with every troubleshooting session.

3. Choose data structures deliberately

A data structure is often used because it happens to fit – not because it suits the task. An array that always grows in the same place eventually becomes a bottleneck. Anyone who aligns the structure with the task from the start saves themselves a later migration, which in a running plant rarely stays on schedule.

  • Fixed size only when the task is actually limited.
  • For growing quantities: dynamic lists with a clear sort order.
  • Structures with meaningful element names instead of a bare index.
  • A shared data model for visualisation, diagnostics and remote maintenance.
  • A data point catalogue that keeps designations, units and scalings in one place.

4. Plan diagnostics from the start

Diagnostic functions are the part of automation that gets cut first – and the first thing you need in case of a fault. A well-thought-out concept made of status values, flags and a fixed diagnostics block costs little during the project and saves days during a fault.

A proven minimum: one uniform status channel per assembly, a central flag field for faults and messages, a diagnostics block at the end of the call chain, and a log that shows the last events before the fault. With that, troubleshooting can often begin without additional diagnostics hardware.

5. The interface to the level above

When a SCADA or visualisation is connected, the data point catalogue determines how well it works later. Anyone who keeps designations, units and scalings consistent in the control program gets a visualisation that works without rework.

  • Unit and scaling at the signal, not in the visualisation.
  • Consistent designations that remain unambiguous years later.
  • A clear boundary between process value and operating parameter.
  • A continuous naming standard across all assemblies.

Checklist for planning

  • Are process logic and hardware access separated?
  • Does every technical task have exactly one block with a meaningful name?
  • Are the data structures aligned with the task rather than with coincidence?
  • Is there a central diagnostics block and an event log?
  • Is there one central place for designations, units and scalings?
  • Could a different team take over the plant without us?

Conclusion

A good PLC architecture is not an end in itself. It is the prerequisite for a plant remaining changeable – and for changes remaining safely manageable even when they happen under operating conditions. The decisions that help are unspectacular: clean interfaces, modular functions, deliberate data structures, planned diagnostics and a consistent data point catalogue.

Is it worth reimplementing existing programs?

That depends on the condition. For programs that mix logic and hardware access, reimplementation is often cheaper than incremental tidying. After the stocktake we tell you honestly which variant we recommend.

How much effort does clean architecture cost?

A little more in the project, considerably less in maintenance. It usually pays for itself at the first major process change – that is, exactly when the structure has to be touched anyway.

All articles