Production monitoring: knowing what the line is doing while it is still doing it

Running or stopped is harder to define than it sounds

A machine can be cutting, idling between parts, waiting for an operator, starved of material, blocked by a full downstream buffer, in setup, or faulted. Collapsing these into a simple binary state loses the distinction that matters, since a machine starved by the process before it is not a machine with a problem. Someone has to define the state model, decide the threshold below which a brief pause is not a stoppage, and settle whether setup counts as production. Those definitions belong to manufacturing engineering and should be written down, because they will otherwise be re-argued every time the figures are questioned.

Reasons captured in the moment beat reconstruction afterwards

Ask an operator at the end of a shift why the machine stopped for a period six hours ago and you receive a plausible answer rather than an accurate one. Prompting at the machine while the stoppage is live, with a short list of causes that reflect how the team actually speaks, produces something usable. The prompt has to be quick and forgiving, because the operator is busy solving the problem. Unattributed downtime will always exist; the useful target is keeping it small enough that the attributed portion still supports a decision, rather than pursuing complete coverage and getting invented answers.

Escalation is the feature that changes outcomes

A stoppage that lasts beyond a defined threshold should reach a named person, then a second one if it continues, with the route depending on the reason. Material shortage goes to logistics, a fault goes to maintenance, a quality issue goes to the quality engineer covering that area. Plants that install monitoring without this end up with accurate records of long waits caused by nobody being told. The threshold itself is a management decision and differs by area, since a minute on a fast packaging line is more serious than several minutes on a machining cell with a long cycle.

Live figures change behaviour, and not always in the intended direction

Once a counter is visible, people respond to it. Sometimes that is the point, with a team pacing themselves against a target and resolving small problems earlier. Sometimes it produces behaviour nobody wanted: running through a condition that should have prompted a stop, splitting a stoppage into short pieces that fall below the threshold, or arguing about the count instead of the process. The way monitoring is introduced determines which of these follows. Presenting it as a tool the team uses to get help faster, rather than as a measurement of the team, is the difference operators notice immediately.

What live monitoring cannot tell you

The signal is thin on its own. A stoppage becomes diagnostic only when you know which order was running, which material lot, which tool, which operator and what conditions applied, and that context comes from other systems. Monitoring will show that a machine stops frequently on Thursday nights; it will not tell you that Thursday nights are when a particular grade is run. Plants that treat monitoring as the finished answer accumulate downtime records without explanations, then conclude the data is useless, when the missing element was the context it was never given.

Frequently asked questions

Can we monitor machines that have no digital connection?

Yes, and many plants start that way. Output can be counted with a simple sensor, machine state inferred from power draw or a lamp signal, and stoppage reasons entered by the operator on a small terminal at the station. That gives most of the operational benefit without touching the control system, which also avoids a lengthy discussion with whoever supports the machine. The limitation is that inferred states are approximate, so define them conservatively and expect to refine after watching the first weeks of data.

Where should the display go?

Where the person who can act will see it without going to look. For an operator that means at the station or on the line; for a shift leader it means somewhere on the route they already walk; for maintenance it usually means a notification rather than a screen, since they are rarely standing still. Screens placed in offices flatter management and change nothing on the floor. If a display cannot be tied to somebody doing something differently, it does not need to exist.

Is monitoring worth it on a low-volume, high-mix shop?

The economics differ from a high-volume line. Utilisation matters less than knowing where each job actually is and how long setups really take, so the useful outputs are job progress and setup duration rather than a rate against a target. Many job shops find the strongest case is capturing genuine setup and run times to improve future quoting, since estimates in that environment are usually based on recollection and tend to be optimistic on exactly the work that hurts most.