Designing HMI Screens for the Worst Ten Minutes

An operator screen is designed on a quiet afternoon by someone with time to think, and it is read during the worst ten minutes of a five-year period by someone who has none. Almost every criticism of industrial HMI design follows from that mismatch, and almost every improvement comes from designing for the second condition rather than the first.

Start with colour, because it is where most of the damage is done. On a typical legacy screen, colour carries no information: pipes are blue because water is blue, motors are green because green looks healthy, the background is a photograph of the plant because somebody had one. When everything is coloured, colour cannot signal anything. The high-performance convention inverts this — a low-contrast grey background, equipment drawn in muted grey outlines, and saturated colour reserved exclusively for abnormal conditions. On such a screen a single red element is visible from across the control room, which is the entire point. And because roughly one in twelve men has some form of colour vision deficiency, colour must never be the only channel: pair it with shape, position or text.

The second offender is decoration. Three-dimensional pipes with gradient fills, bevelled tank graphics, animated rotating fans and photorealistic valves all consume screen area and attention while conveying nothing that a flat symbol would not. The useful discipline is to ask of every pixel what decision it supports. A drop shadow supports none.

Then reconsider what a value should look like. A number tells the operator the measurement; it does not tell them whether the measurement is acceptable, and that is the actual question. A horizontal bar with the normal operating range marked, the alarm limits marked, and the current value shown against them answers it at a glance and without arithmetic. Add a small trend line beside it and the operator also sees direction and rate, which is what determines whether they need to act now or watch. Operators reason about slopes far more than absolute values, and screens that hide the slope in a separate trending application force them to hold history in their heads.

Structure the screens as a hierarchy and be strict about it. A level one overview covers the whole process at a glance and should make an abnormality obvious without reading a single number. Level two covers a unit with enough detail to diagnose. Level three is the detail of individual equipment, and level four is diagnostics and support information. The failure mode is a flat set of thirty screens with no relationship between them, in which finding the relevant one during an upset is itself a task.

Consistency is worth more than elegance. The same equipment should sit in the same place on every screen that shows it, navigation should be predictable, and no control should be more than two actions away. Operators build motor memory over years, and a redesign that moves things for aesthetic reasons destroys it — which is why an HMI refresh should be treated as a change requiring training, not as a cosmetic upgrade.

Some practical details separate screens that survive from screens that get worked around. Always show units, and show them next to the value rather than in a legend. Match significant figures to instrument accuracy: five decimal places from a transmitter with half a per cent accuracy is a lie that erodes trust in everything else on the screen. Keep consequential controls physically distant from routine ones — an acknowledge button adjacent to a line-stop button will eventually be pressed by mistake, and the fault will be recorded as operator error. Confirm actions that are hard to reverse, but do not confirm everything, because a confirmation dialogue that appears fifty times a shift is dismissed reflexively and protects nothing.

Update rate deserves attention it rarely gets. A screen whose values refresh every three seconds teaches operators that the screen lags reality, and once they believe that they will start walking to the plant to check. Sub-second updates on the values that change quickly are not a luxury; they are what makes the screen a source of truth rather than an indication.

Design for the room as it is. Displays are read at two metres, at night, sometimes with reflected glare, by people who may be wearing gloves if the panel is a touchscreen. Font sizes, contrast ratios and touch target sizes all follow from those facts rather than from what looks right on a desktop monitor during development.

Finally, test the screens the way they will be used, which almost nobody does. Sit with an operator, simulate an upset, and measure how long it takes them to detect that something is wrong, identify what, and find the control they need. That number is the design specification. ISA-101 provides the framework and vocabulary, but the measurement is what tells you whether the screen works, and it is the one part of the exercise that cannot be delegated to a graphics package.