How-to
Designing HMI User Access Levels for Operators and Maintenance
Short answer
Build the HMI around three access classes: operators get start, stop, status and recipe selection; supervisors add process parameter and recipe editing; maintenance gets clearly labeled overrides at the highest level. Commission all three with the data bindings by logging in as each class, proving one permitted and one prohibited action, and confirming every override is shown on the overview until returned to normal.
On this page
- How many user access levels does a machine HMI need?
- What should operators be able to change?
- How do recipes keep the operator screen simple?
- How much control should supervisors get?
- Where do maintenance overrides and test modes belong?
- What must alarms and help do at every access level?
- How do you commission the access structure?
- When is a login on the HMI not enough?
- Questions asked
Structure a machine HMI around three access classes — operator, supervisor and maintenance — and commission that structure together with the data bindings, not after them. HMI work typically accounts for 8–12% of total machine or process-system cost: higher on a small machine with extensive HMI requirements, lower on a plant-wide automation system. The access design decides how much of that share becomes control the customer can actually live with, because it draws the line between a machine that is easy to run and one that invites the wrong setting at the wrong login.
How many user access levels does a machine HMI need? #
Three classes cover nearly every machine: operator for routine production, supervisor for process parameters, maintenance for overrides and test modes. Define each class by the damage it must prevent, and the level count follows from the tasks: operators must not be able to ruin product or damage the machine with a bad setting, supervisors must not be able to damage the machinery itself, and maintenance must be able to do abnormal things deliberately.
The ISA-101 series frames the whole job the same way — design the interface around what each user needs to accomplish, with safety, usability and efficiency as the measures — and access levels are how that principle becomes enforceable on a panel. The mechanics of accounts, groups and passwords live in your engineering software's user-management editor; for one platform's concrete walk-through, see how to set up granular user access on a CODESYS touch panel.
| Class | Primary need | Typical access | Where it lives on screen |
|---|---|---|---|
| Operator | Routine daily production | Start, stop, status, recipe selection, limited adjustments | Startup, overview and production screens |
| Supervisor | Process and recipe control | Broad parameter and recipe editing, no machinery-damaging functions | Login-gated objects on operator screens, or dedicated screens |
| Maintenance | Diagnosis and abnormal test operation | Raw sensor data, component controls, overrides, maintenance modes | Hidden, access-protected detail screens |
What should operators be able to change? #
Start, stop, status, recipe selection and a few limited adjustments — nothing an incorrect setting can use to damage the machine or ruin the product — with detail available on request but off the primary screens. The limit is error prevention, not distrust: every parameter on an operator screen is one more the operator must understand and one more that can be set wrong during a shift.
- List the tasks performed during a normal production cycle.
- Map each task to the smallest set of controls and readouts that accomplishes it.
- Move raw sensor data, detailed settings and overrides to station-detail or maintenance screens.
- Run a normal cycle from the primary screens only; if the operator needs a protected page to get through the cycle, the split is wrong.
What belongs on those screens — color, hierarchy, object choice — is its own discipline, and designing HMI screens for clear operator decisions covers that half of the work.
How do recipes keep the operator screen simple? #
The operator picks the part or process from a stored list and the control system applies the approved set of process settings — one selection replaces every individual parameter entry and removes the experiment. Recipe selection is the strongest simplification available to an operator interface, because the machine receives a complete preprogrammed value set instead of a series of manual entries.
Commission the selection and its result together: a correct recipe name with an incorrect applied value set is a failed selection, not a partial pass. Compare every applied setting of every recipe against its approved definition before delivery, because the operator will trust the name on the list.
How much control should supervisors get? #
Supervisors get every setting that can affect the product — process parameters, recipe editing, tuning — and stay locked out of the settings that can damage the machinery itself, which is the line that separates them from maintenance. Two configurations both work, and the choice between them is a counting decision, not a preference.
When the supervisory edits fit on one or two screens, keep the normal operator screens and enable the extra objects after supervisor login: one set of graphics, no duplication. Once supervisory functions need a third screen of their own, build dedicated supervisor screens instead and keep the operator screens clean — the duplication costs less than the clutter.
Where do maintenance overrides and test modes belong? #
At the highest access level, on hidden and clearly labeled screens, with any active override shown on the operator overview until it is returned to normal — maintenance sometimes has to run the machine abnormally to test it. Those functions can damage the machine or make scrap if performed incorrectly, so the labeling carries the warning: each override names the device it affects, the state it commands, and the fact that the mode is abnormal.
Clutter is tolerated on these screens because they are used rarely. What is not tolerated is an invisible override. One workable overview pattern is a strip per station: red marks a station fault, yellow marks an active manual override, so a technician cannot leave the machine looking normal while automatic logic is bypassed somewhere inside it.
What must alarms and help do at every access level? #
A plain-language banner on every screen while an error is unacknowledged, and a log that records three timestamps per event: when it occurred, when it was acknowledged, and when it cleared. The banner is global because the operator may be anywhere in the interface when the fault occurs; the log is permanent because a vague memory of last night's alarm is not troubleshooting data.
Write each message at the operator's level of skill: name the device and the condition in plain language and include the first troubleshooting step, with the goal that routine faults are cleared without a call to maintenance or a supervisor. Pair each message with a help button that opens instructions specific to that fault, reachable only from its own message so the help always matches the condition. Add an always-available help section with concise topics — and the manuals themselves when the HMI is PC-based — for everything else.
How do you commission the access structure? #
Log in as each of the three classes, attempt one permitted and one prohibited action everywhere, then select every recipe and compare the applied settings with the approved definition — on the physical panel, not only in the simulator.
- Create one account per class, and individual accounts within a class wherever the platform supports them.
- Log in as the operator, visit every screen, and confirm that navigation to supervisor and maintenance pages is refused.
- At each of the three logins, attempt one permitted and one prohibited action on every screen that offers one.
- Select every commissioned recipe and compare each applied setting with its approved definition.
- Log out and confirm every protected control is unavailable without a fresh login.
- Log in as maintenance, activate each override in turn, and confirm the overview identifies the affected station.
- Return each override to its normal state and confirm the indication clears with it.
- Repeat the pass on the panel itself: when HMI user management works in simulation but not on the panel explains why the simulator result does not carry over.
Size the effort with a worked example using illustrative numbers: an 18-screen project carrying 12 recipes of 8 settings each. Action checks come to 3 logins × 18 screens × 2 tests = 108. Recipe checks come to 12 recipes × 8 settings = 96 value comparisons. The entire access structure therefore closes on 108 + 96 = 204 explicit, signable checks — the kind of list two people run through in a delivery session, which is exactly why it should be written down before delivery instead of reconstructed after the first lockout call.
When is a login on the HMI not enough? #
Whenever machine protection or personnel safety is at stake: login controls who can operate an object, but only controller-side permissives and interlocks decide what the controller executes, and they must reject an unsafe request at any login level. A hidden maintenance button is still hazardous if a stale session, a duplicated object or a wrong write binding can issue the same command, so the rejecting conditions stay in the controller and run regardless of who is logged in. How an HMI and a PLC divide the work in a control system lays out that boundary in detail.
When actions must be attributable, write the active user and access level to controller tags at login; sending the active HMI user to the PLC with login macros shows one working method. One limit to state plainly: HMI access control is not a safety function — personnel protection belongs to the machine's safety system, and no login scheme substitutes for it.
Questions asked
What does the ISA-101 series cover for HMI design?
It provides a framework for both designing and managing HMIs in process automation, with emphasis on safety, usability and efficiency. In practice that means the level definitions, color conventions and override indications should be written into the project documentation, so the next engineer maintains the structure instead of re-inventing it.
Should data logging be part of the access design?
Yes. Even when the customer asks for no reports, configure logging of key process points at design time; the logs earn their keep the first time service or engineering troubleshoots a fault on the running machine. Raw comma-delimited files are enough to start, and they can be formatted into reports later for maintenance planning or quality documentation.
Why does one maintenance password also open supervisor buttons?
Because access levels are ordered: a control set to level 2 accepts any login at level 2 or higher, so the maintenance account passes every check below it. That is convenient, but it means you audit each object against the class that should own it rather than assuming level numbers keep the classes apart.
Should a whole shift share one operator login?
No. The alarm log records who acknowledged each event and the change history shows who entered a value, and both are only useful when a login maps to one person. Give each class its own account at minimum, and individual accounts wherever the platform supports them.
Can operators open the station detail screens at all?
Yes, read-only. Operators occasionally need deeper data for process troubleshooting, so let them open the station detail screens with every entry object and button on those screens gated to the higher levels. Read-only drill-down costs nothing in risk and saves a supervisor call; the commissioning test is that every control on those pages refuses the operator login.