Why Does My Production Line Keep Stopping?

The typical line has dozens of sensors, but they were installed based on the OEM's default I/O list. That list covers the obvious stuff: main drive current, hydraulic pressure, discharge temperature. It rarely covers secondary failure modes.

Consider a gearbox on a conveyor drive. The OEM monitors oil temperature and output speed. But the most common failure mode for that gearbox is bearing wear on the input shaft. Without a vibration sensor on that bearing housing, the first sign of trouble is a seized bearing and a dead line.

The fix isn't buying more sensors blindly. Walk your downtime Pareto. Pull the last 12 months of unplanned stops. For each one, ask: was there a sensor that could have caught this 30 minutes earlier? If not, that's a blind spot.

Common blind spots we see repeatedly:

• Motor winding temperature on auxiliary drives (not just the main spindle)
• Vibration on gearbox input bearings (not just output)
• Pressure differential across filters that get changed on a calendar, not on condition
• Ambient temperature near VFDs in enclosed panels

Retrofitting a wireless vibration sensor costs $200-400 per point. Compare that to the cost of one unplanned stop. The math usually takes about five seconds.

Every rotating component has a rated life. Servo motors, spindle bearings, ball screws, pneumatic cylinders. OEMs publish these numbers, but they end up in a binder on a shelf.

The problem isn't that plants ignore maintenance. It's that hour-based or cycle-based replacement intervals aren't connected to the actual running hours of the machine. A press that runs two shifts accumulates hours twice as fast as the same press on one shift. The calendar-based PM schedule treats them the same.

Here's what typically happens. A servo motor rated for 20,000 hours is installed. The PM schedule says "replace every 3 years," which assumes single-shift operation. The plant adds a second shift six months later. Nobody updates the PM interval. The motor hits 28,000 hours in 2.5 years and fails.

The fix is tracking actual run hours per component, not per machine. Most PLCs already count cycles or run hours internally. The data is there. It's just not connected to the maintenance system.

Start with your top 10 downtime contributors. For each one, find the rated life of the wear components. Compare that to the actual accumulated hours. You'll likely find several components running 30-40% past their rated life. Those are your next unplanned stops.

A packaging line runs perfectly for three days, then suddenly starts jamming every 20 minutes. The maintenance team checks everything mechanical. Nothing is wrong. Then the material lot changes and the problem disappears.

This pattern is common in food and beverage, pharma, plastics, and any process where raw material properties matter. Film thickness varies by 5-10 microns between rolls. Resin melt flow index shifts between suppliers. Powder moisture content changes with warehouse humidity.

The line was tuned for a specific material spec. When the material drifts, the line parameters are no longer optimal. But because the material change isn't tracked alongside machine performance, nobody connects the two.

What helps:

• Log material lot numbers against production runs in your MES or even a shared spreadsheet.
• Track incoming material certificates of analysis. Note the actual values, not just pass/fail against spec.
• When you get a run of stoppages, check if the material lot changed in the last 2-4 hours.
• Build a simple scatter plot of reject rate vs. key material properties over 30 days.

You'll often find that 80% of your "random" stops correlate with material lots from specific suppliers or specific property ranges. That's actionable. You can tighten incoming specs, adjust line parameters per material batch, or set up automatic recipe changes based on incoming test results.

Shift handoffs are a known risk. The data confirms it. Plants that track scrap and downtime by time of day consistently see spikes in the first 30 minutes after a shift transition. The pattern is predictable, but most plants treat it as unavoidable.

What goes wrong during handoffs:

• The outgoing operator adjusted a temperature setpoint 2 degrees up to compensate for a material batch. They didn't mention it. The incoming operator sees the non-standard setpoint and changes it back.
• A machine had an intermittent fault that the outgoing operator learned to work around. The incoming operator doesn't know the workaround and calls maintenance.
• The outgoing operator ran the last 30 minutes of their shift at reduced speed to avoid a known jam point. The incoming operator starts at full speed.

Plants that reduce shift-change losses do it with structured handoff procedures. Not a clipboard that nobody reads. A 5-minute standing meeting at the machine with both operators present.

The handoff should cover three things: what's different from standard right now, what problems came up and how they were handled, and what's queued for the next shift. Some plants use a digital shift log displayed on the HMI. Others use a whiteboard at the line. The format matters less than the habit.

One approach that works well: the incoming operator shadows the outgoing operator for 10 minutes before the official shift change. They run the machine together. Problems that would have taken 20 minutes to diagnose take 2 minutes because the context is right there.