Insigtful Blog Pages of Thread in Motion

Machine Downtime Tracking: See Real Stop Reasons With Vision AI

Written by Aycan Karadağ | Nov 17, 2026, 1:30:31 AM

Machine downtime tracking only helps if it captures the real reason a machine stopped, including on older machines that send no data at all. A fixed camera with vision AI can read the machine's motion, material flow, stack light and operator presence, and classify each moment as running, waiting, setup or fault. Machine utilization monitoring in TIM INSPECT uses this to record stop reasons automatically and alert the right person when a stop begins.

Many plants already have some form of downtime tracking, yet the same questions keep coming back: why does the report say the machine was running when the operator was waiting for material? Why is "other" the largest stop category? This guide explains where conventional downtime data breaks down and how camera-based classification fills the gaps.

Why machine downtime data is often wrong

Three patterns come up again and again on the shop floor.

  • "Running" that is really waiting. A machine can report itself as running while the cutting head is idle because no material is reaching the table. The signal is technically correct and operationally misleading.
  • Stop reasons that end up as "other". When an operator has to pick a code from a long list at the end of a stop, the quickest choice wins. Over time the most common reason becomes the least useful one.
  • Machines with no data at all. Older presses, lathes or welding cells may have no network interface. They are often excluded from utilization reports simply because nobody can measure them.

The result is a utilization picture that looks precise but does not explain where time is actually lost, and improvement priorities set on that basis can miss the real problem. Frameworks such as overall equipment effectiveness (OEE) are only as good as the stop data behind them.

Conventional downtime tracking methods and their limits

Machine signals and PLC data

Where available, machine data gives accurate run and stop times. It does not, however, see what happens around the machine: material shortages, an empty operator position or a die change in progress.

Operator-entered reason codes

Manual entry captures context, but it depends on discipline and timing. Short stops are skipped and long ones are coded hastily.

Retrofitting sensors

Adding sensors to legacy machines is possible, but it means touching the machine, wiring and sometimes downtime for installation. It also measures only what the sensor was chosen to measure.

Walking the floor

Supervisors see a lot during rounds, but only the moment they pass by. The waiting that happened twenty minutes earlier is gone.

How machine downtime tracking works with a camera

The camera-based approach does not ask the machine for anything. It watches the machine and its surroundings, the way an experienced supervisor would, and does so continuously.

Watch the machine and its area

A fixed camera covers the machine, its infeed and outfeed and the operator position. No data is expected from the machine, so older machines are covered the same way as new ones.

Classify the status

The vision model assesses machine motion, material feed and part output, whether the operator is at the machine and the colour of the signal light together. From these it classifies the status into four classes: running, waiting, setup and fault. Waiting for material, operator absence, die changes and setup time, and idle moments are tracked separately.

Alert the right person

When the machine stops unexpectedly or runs out of material, the person who can act is notified on AXION: the operator, the maintenance technician or the material handler. A notification such as "Machine 3 is waiting for material" goes to the material handler, not to everyone. Who receives which situation is defined during setup.

Record the reason

Stop reasons and durations are recorded automatically, without depending on a code picked under time pressure. Planned production time and quantity come from the MES, and utilization reports are built by combining the two. Video on its own is not enough for utilization figures; it explains the time, while the MES supplies the plan.

The machine utilization scenario gives you one record per machine: how long it ran and waited during the day, the reason for each stop and the setup time spent on die changes. Improvement priorities can then follow recorded reasons rather than the loudest complaint.

What to plan before installation

The camera view

The camera must see the moving part of the machine, the material entry and exit and, ideally, the stack light. Guards, enclosures and nearby racks can block the view, so the scope is finalised around the angle from which the camera actually sees the machine.

Defining stop categories

Agree beforehand what counts as setup versus fault, and which waiting reasons matter to you. The clearer the categories, the more useful the report.

Routing notifications

Map each situation to a role. A material shortage should reach the material handler; a fault should reach maintenance. Sending every alert to everyone quickly creates alert fatigue.

Privacy for operators

The camera assesses whether someone is at the machine, not who that person is. Zones and actions are observed, images are filtered and only the event moment is stored, with the design aligned to GDPR and local rules from the start.

MES data

Decide which plan and quantity fields will be taken from the MES and for which machines, so utilization can be calculated consistently.

For manual stations rather than machines, the companion cycle time and microstops scenario measures every cycle and brief stop. Both run on the TIM INSPECT platform.

FAQ

Is the camera alone enough to calculate utilization?

No. Video shows when the machine ran, waited and why it stopped. Planned time and quantity come from the MES, and utilization reports combine both.

Can it be used on older machines that provide no data?

Yes. The camera reads the signal light, the machine's motion and the material flow, so no data connection to the machine is required.

Which statuses are told apart?

Running, waiting, setup and fault, with material waits, operator absence, die changes and idle time tracked separately.

Who receives the notification?

The operator, the maintenance technician or the material handler, depending on the situation. The routing is defined together during setup.

Next step

A good way to start is with a single machine and its real stop reasons: request a demo and we will review the machine, the camera view and the plan data together.