The Fault-Finding Interview Question: How to Answer It Properly

The question

It comes in a few forms:

• "A conveyor will not start. Walk me through what you check."
• "An operator says the machine stops randomly. What do you do?"
• "A transmitter reads zero. Is it broken?"

They all test the same thing, and it is not knowledge. It is method. The interviewer wants to know whether you narrow the problem down with measurements or guess and start swapping parts.

The wrong answer

"I would check the program."

Or worse, a list: "I would check the sensor, the wiring, the PLC, the contactor, the motor." A list is not a method. It tells the interviewer you know what the parts are, which they assumed.

Jumping straight to the program is a particular tell, because it is what somebody who has only worked in software does. The program is the least likely thing to have changed on a machine that ran yesterday.

The right shape

1. Ask what changed.
"Has it ever worked? What changed recently, and when did it last run?" A machine that never worked is a commissioning problem. A machine that stopped working yesterday is a failure, and something caused it. Maintenance work, a part replacement, a program download, a power event.

This question alone marks you out, because it is what an experienced engineer asks first.

2. Ask what the machine says.
Alarms, HMI messages, fault codes, indicator lights, the PLC's own diagnostic buffer. Modern equipment usually tells you something. Reading it before touching anything is free.

3. Halve the problem.
This is the core of the answer. State it explicitly:

The command starts at the operator and ends at the motor. I would go to the middle first: is the PLC output being set in the program? If yes, the fault is downstream, in the output card, wiring, contactor or motor. If no, the fault is upstream, in the inputs, permissives or the logic.

That single sentence is the answer they are listening for. Everything else is detail.

4. Say where you would put the meter.
Concretely. "I would check for 24 V at the contactor coil terminals. If it is there and the contactor is not pulling in, the coil or the mechanism is at fault. If it is not there, I work back to the output card."

Naming the measurement is the difference between someone who has done this and someone who has read about it.

5. Confirm before concluding.
"Once I found it, I would prove it, by forcing the output or by operating the device, rather than replacing something and hoping the symptom goes away."

6. Say what you would do about the record.
"Then I would note what it was and tell whoever needs to know, because an intermittent fault that happened once will happen again."

The intermittent-fault version

Harder, and asked of more senior candidates. The right instincts:

• Intermittent usually means mechanical, thermal or electrical noise, not software. Software fails consistently.
• Correlate with events. Does it fail when a big motor starts? At a particular point in the cycle? After running for an hour? When it rains?
• Use the historian. Trend the relevant signals and look at what happened immediately before, rather than relying on reports of what happened.
• Do not fix and walk away. If you cannot reproduce it, you have not found it. Say that honestly rather than claiming a fix.

Saying "I would trend it rather than rely on the operator's description" is a strong answer, because it acknowledges that eyewitness reports of machine behaviour are unreliable without being rude about operators.

The analog version

"A level transmitter reads zero. Is it broken?"

The answer is the current reading. Below 3.6 mA is a fault indication rather than a low level, 0 mA is an open circuit, and 4 mA is a genuine empty tank. Then: check the loop supply and fuse before suspecting the transmitter, measure at two points to halve the problem, and check the card configuration and software scaling before condemning the field device.

Full detail in 4 to 20 mA loop troubleshooting.

Practise saying it out loud

This is a spoken answer, and it is different from a known one. Rehearse it until the shape is automatic: what changed, what does it say, halve it, where does the meter go, prove it, record it.

Our free interview question prep tool has this question and the rest of the commissioning set, with what a good answer contains and where candidates typically lose it.

Where the instinct comes from

You cannot fake this convincingly, because the follow-up questions go one level deeper than a memorised answer can reach. It comes from having stood in front of a panel with something broken.

On the Automation Engineer Program faults are introduced deliberately, on real hardware, and trainees diagnose them under time pressure. That is the closest thing to the real experience that training can provide.

Related: I/O loop checking, Portfolio projects that get interviews.