From alarm code to operator action: designing industrial HMIs that help
11 min read · By the neexo engineering team · Denmark
Published: 12 November 2026
When a machine stops, an alarm list should help the operator decide what to check next. A useful industrial HMI connects the event to its machine location, consequences, verified conditions and approved recovery procedure. Better diagnosis starts in the control software and event model. The HMI must never replace machine safety logic or authorize an unsafe restart.
A stopped machine should not become an alarm-code investigation
A packaging machine stops during a transfer. The HMI displays a transfer failure, a station timeout and a cycle-stopped message. None of those messages tells the operator which one matters first. The operator opens a manual, calls a colleague or tries a reset. This is an illustrative scenario, not a reported customer incident.
Many alarms describe how the control software detected a condition: a timer expired, a sensor did not switch, or a module could not continue. That is useful for programmers, but it is an incomplete explanation for the person standing at the machine. The operator needs to know where to look, which conditions can be checked and when to call maintenance.
An alarm screen should support a decision under time pressure. A polished interface with weak event information cannot do that. Conversely, a modest interface can work very well when the PLC provides clear state, location and recovery information.
Three messages may describe one machine stop
Consider a vacuum gripper on the same illustrative packaging line. The controller does not observe the expected vacuum signal before a transfer. That may trigger a transfer error; the waiting downstream station then times out; finally, the machine stops. Three entries appear in the event list, although the operator may need to investigate only one area first.
A useful HMI preserves every relevant event for service while presenting an understandable diagnostic path. It can distinguish a detected initiating condition from later consequences. Where timestamps, signal states or sequencing data are incomplete, the interface must say 'possible cause' rather than presenting an unverified root cause.
Time order alone is not proof. A fault that appears first in the log might be a consequence of something the system does not measure, such as a mechanical restriction or a supply problem. The aim is to narrow the investigation, not to invent certainty.
- Keep the event sequence available for service and troubleshooting.
- Separate a detected condition from an inferred mechanical or process cause.
- Show which station, module or actuator the event relates to.
- Never hide a safety-relevant condition merely to simplify the operator view.
Start with one recurring machine stop. Improve the information and recovery sequence around it before redesigning the entire HMI.
The five questions every recovery view should answer
An operator facing a stop should be able to understand what happened, where it happened, what the consequence is, what can be checked next and how the team can confirm that recovery is complete. These are useful design questions for a machine builder reviewing an existing HMI.
For the vacuum example, the title could read 'Vacuum not reached at transfer gripper'. The screen points to the relevant station and explains that the transfer cannot continue. It then shows a short, role-approved inspection instruction, such as checking the gripper and supply according to the documented procedure.
There is no need to place every technical detail on the first screen. Show the decision first and make signal values, event history and diagnostic data accessible on demand. An inexperienced operator gets a clear first step; a service technician can still drill down to the controller evidence.
- What happened? A precise condition, not only an error code.
- Where? A machine area, station or component.
- What does it affect? The operation or function currently blocked.
- What may I do? An approved inspection or escalation step.
- When is the machine ready again? Verified state and the controller's actual permissives.
Good alarm guidance begins with PLC information, not graphic design
The HMI cannot explain conditions the controller never exposes. If the interface receives only a boolean alarm bit and an error number, it cannot reliably distinguish a failed sensor from a blocked process, or an initiating event from a consequence. Improving the screen may require a small, deliberate change to the PLC diagnostic contract.
For each important stop, define the detected condition, its source, machine location, severity or operational priority, relevant mode and state, expected next check, and recovery prerequisites. Keep the event identity stable enough to connect diagnostics, translations and technical documentation. Version the alarm definition when behaviour changes.
A control project may use PackML to express modes and states such as Stopped or Resetting, but PackML alone does not explain every module fault and does not provide functional safety. The HMI should derive availability and readiness from authoritative controller feedback, not from an operator acknowledging a text message.
The most practical first increment is often a small set of high-impact faults with clear ownership, consistent message templates and reviewable recovery flows. It does not require replacing the PLC architecture or rebuilding the whole user interface.
- Controller: detected facts, state, permissives and fault ownership.
- Event model: chronology, source, priority and relationship to other events.
- HMI: location, consequence, approved response and drill-down.
- Service documentation: deeper tests, repair instructions and escalation criteria.
An alarm is not the same as an event, an acknowledgement or a reset
A status change or an event log entry is not automatically an alarm requiring operator attention. Alarm priority should reflect the required response and consequence; otherwise an interface quickly becomes an archive of messages that compete for attention.
Acknowledge generally records that an operator has seen the alarm. Correcting the fault means the physical or logical condition is no longer present. Reset may clear a latched fault or request a controller transition when prerequisites are satisfied. Restart is a separate controlled action. The precise behaviour belongs in the machine's design and operating instructions.
Safety functions and protective measures must remain authoritative. No guided HMI procedure should bypass a guard, pressure release requirement, lockout procedure or validated safety reset. If a task requires an authorized technician or an isolation procedure, the interface should stop the guidance and escalate. A green button is never proof of a safe machine.
The ISA-101 series provides a lifecycle-oriented framework for HMI design. Alarm-management principles from ISA-18.2 and IEC 62682:2022 can also inform the work, but these alarm standards chiefly address process-industry systems. Apply the useful practices thoughtfully rather than treating them as mandatory in full for every packaging machine.
See the design pattern in context
Our related case explains a cause-to-action alarm hierarchy and how operator guidance can be structured without weakening safety boundaries.
Explore the alarm-to-action caseUse a real operating scenario to test the interface
Choose a recurring stoppage with a known expected response. Ask an operator who did not design the machine to locate the fault, explain the consequence, choose a permissible next step and describe when assistance is needed. The objective is to discover missing information, unsafe assumptions and confusing navigation before rolling the pattern out to every module.
Run the same scenario with the existing HMI and a proposed recovery view. Record what the person can identify without verbal hints, what information they search for, and where they hesitate. If metrics are used, measure the actual test. Do not promise a percentage reduction in downtime based on an illustrative mock-up.
For machine builders, digital tests can verify whether the HMI reflects the correct PLC states and interlocks. Operator comprehension still needs realistic user testing; physical recovery and any safety-related behaviour require their own appropriate verification.
- Fault appears in normal production mode and with a blocked downstream station.
- The initiating condition clears while a secondary timeout remains active.
- Operator acknowledges an alarm while the fault condition persists.
- Reset is requested before prerequisites are met.
- The same recovery flow is checked in the three supported HMI languages.
Start with the worst alarm, not a complete HMI redesign
For an engineering manager, the useful decision is where a modest software change can remove the greatest operator uncertainty. Review the five stops that create the most support questions or the longest investigations, then select one with an understandable event sequence and a documented recovery method.
Agree the diagnostic contract between controls, HMI, service and the people responsible for safety. Build one reusable pattern with a location, clear condition, consequence, approved next action and escalation route. Test it before replicating it across the machine family.
A better HMI makes the machine easier to operate because the control software and user interface tell the same story. neexo works across PLC, machine software and industrial UI/UX, which makes it possible to address the information behind the screen as well as the screen itself.
Questions machine builders often ask
Do we need to replace the HMI platform to improve fault recovery?
Usually not for a first improvement. Define the alarm information and approved response, then prototype one recovery view within the existing architecture. A wider platform change is a separate decision based on maintainability, performance and integration needs.
Can AI tell the operator exactly why the machine stopped?
AI can assist with retrieval and interpretation when approved documentation and trustworthy machine data are available. It must distinguish confirmed controller facts from suggestions. Do not let an AI-generated explanation authorize repair, safety reset or motion.
Is the first alarm always the root cause?
No. It is the first recorded event, which may be a consequence of an unobserved physical condition. Use timestamps, sequence states and relevant signals to form and test a hypothesis.
Should every machine fault be an alarm?
No. Diagnostics, state information and events can be valuable without demanding immediate operator action. An alarm should have a meaningful consequence, a response expectation and a level of priority suited to that response.
How can we verify the design before deployment?
Use a reproducible fault scenario against the control logic and test the interface with intended users. Verify state mapping, translations, interlocks and escalation. Physical machine recovery and safety functions need separate validation.
Take one alarm from your current machine. Write down the controller's detected condition, the machine location, the required operator decision and the verified reset prerequisites. If you cannot complete the four fields, improve the diagnostic information before drawing another screen.
Discuss an HMI recovery flow