An alarm list is the most-read screen on any machine, and usually the least designed. The default outcome — a numbered list of internal fault bits — turns every stop into a phone call.
Name the device, the condition and the limit
“Fault station 3” is not an alarm, it is a notification that something is wrong somewhere. Compare it with: “Station 3 clamp did not reach closed position within 2 s”. The second version tells the operator which device to look at, what it failed to do, and how long it was given. That is often enough to clear the fault without escalating.
The information is already in the program — the timer, the expected sensor, the position. Writing it into the alarm text costs nothing at build time.
Separate what stopped the machine from what did not
Faults stop production. Warnings do not. Mixing them in one list trains operators to ignore the list, because most entries do not require action. Two categories, visually distinct, restores the meaning of a red entry.
Show what is active, not what has ever happened
The main alarm screen should show currently active alarms, with acknowledged and cleared entries moving to a history screen. An operator dealing with a stop does not want to scroll past this morning’s cleared faults.
Timestamp everything and keep the history
The first question after an unexplained stop is what happened immediately before it. A history with real timestamps answers it. Without one, you are reconstructing events from memory, and memory is unreliable about sequence.
Keep enough history that a Monday morning investigation can look at Friday night. A few thousand entries is not expensive.
Suppress the cascade
One root fault often triggers a dozen consequential alarms. If the list shows all of them equally, the operator has to work out which came first. Suppressing known consequential alarms while their root cause is active — or at minimum ordering strictly by time — makes the first entry the useful one.
Write the text in the operator’s language
Not the engineer’s. If the machine ships to three countries, build the text library from the start with a language column. Adding a fourth language then costs a spreadsheet rather than a redesign.
Test it with the people who will use it
Show the alarm list to an operator before commissioning ends and ask what they would do about each entry. The ones they cannot answer are the ones to rewrite. This takes an afternoon and it is the highest-value hour of HMI work you will do.