Tablet HMIs on the Forming Floor: What Actually Changes

It's 0600 on a Tuesday at a US Midwest container glass plant and the outgoing hot-end operator is trying to read his own handwriting on a laminated traveller that's been wiped with acetone so many times half the ink has lifted off the page. The incoming operator can't tell if the mould temperature reading is real or guessed. On most lines I've audited, this exact handover misses the night-shift swabbing data more often than anyone likes to admit.

That's the gap tablet HMIs are supposed to close. Not by being smarter than the operator, but by making the job card, the gauge reading and the defect alert visible to everyone on shift at the same time, instead of trapped in one person's notebook or a clipboard nobody can find at 0555.

What a laminated traveller actually costs you

A paper traveller isn't stupid. It's just slow, and it only exists in one place. If the hot-end superintendent wants to know whether cavity four is running hot, someone has to walk over, find the sheet, and read a number that might be forty minutes old. On a busy shift with a job change running, that lag is where the scrap comes from.

Cavity-to-cavity gob weight spread on an IS machine typically sits around ±0.6g when you're reading an hourly strip-chart recorder. Put live per-cavity weight trend lines on a tablet in front of the hot-end operator and that spread commonly tightens to around ±0.2g, because drift gets caught inside a cycle or two instead of at the next paper log entry. That's not a controls upgrade. That's a visibility upgrade.

Digital job cards do something similar for early-campaign scrap. Wrong-cavity-count and wrong-gob-weight data entry errors typically account for 15-20% of the reject rate on a freshly started job, and most of that traces back to someone transcribing a number off a laminated card instead of confirming it against a locked recipe.

Where the tablet earns its keep

In 2021 I worked a US Midwest plant running an eight-section Emhart AIS machine with 1990s relay logic behind the panel. Job changes ran 70-90 minutes, mostly radio call-outs between the mechanic and the mould shop confirming cavity counts by shouting over the compressor noise. We paired a tablet HMI to the existing job card system, pushed SMED checklists and go/no-go gauge confirmations to the hot-end operator and the IS machine mechanic at the same time, and watched changeover time drop under 25 minutes within six weeks.

Twenty-three minutes. That was the number that stuck on the board by week eight, and it held.

The tablet also gave us named defect alerting instead of a shift-end guess. Checks (cold-contact cracks from mould temperature drift), birdswing (a glass tail trapped against the blank), choke marks (neck-ring drag lines), overpress flash and spike marks from plunger drag all got flagged against a live threshold instead of showing up as an unexplained cold-end reject count three hours later. Hot-end reject rate on that line moved from a 4-6% baseline to under 3% inside 90 days. Not because the furnace got better. Because the crew could see the drift before it became scrap. Vendors like to quote +3 to +5 OEE points within two to three months of a rollout like that, and on this line the driver wasn't furnace efficiency at all. It was faster changeovers and faster root-cause isolation.

Where it fails on day one

Vendor-led rollouts usually get sold as a screen swap. New glass, new login, same paperwork underneath. Not a controls problem. A workflow problem. That's the OEM-affiliated consultancy pitch, and it misses the actual work, which is redesigning the shift-handover ritual so the data on the tablet is the data the next shift trusts.

And it fails specifically at the 0600 handover, every time, if nobody rebuilt that ritual around the new tool. A tablet that just digitises the old paper traveller changes nothing. Look, a dashboard login count feels like progress. It isn't. It just gives management something to celebrate while the reject rate sits exactly where it was.

Generic Lean Six Sigma boutiques make a related mistake. They'll talk OEE and DMAIC in the abstract and never once say birdswing or gob weight CV, which tells you fast whether the person in the room has actually run a forming line. If your consultant can't name the defect mode, don't let them near your job change process.

A tablet doesn't fix a bad process. It makes a bad process visible to everyone in real time, which is either the best thing that happens to your shift or the worst, depending on what you were hiding before.

Role-based logins change who owns the alert

The value isn't the screen. It's the routing. A properly configured tablet HMI sends different alert sets to different logins. The Hot End Operator gets section-by-section gob loading and timing. The IS Machine Mechanic gets plunger, mould and blank-arm drift. The Forming Supervisor holds job sign-off authority and sees nothing starts without it. The Quality Technician gets go/no-go gauge checks correlated against cold-end inspection data instead of a separate paper log that reconciles two days later.

• Gob temperature target 1,050-1,100°C, flagged against the active job card
• Blank and mould temperature window 480-520°C, shown red, amber or green
• Mould cooling wind pressure setpoints per section
• IS machine speed in bottles per minute tied to the job that's actually loaded, not the job loaded an hour ago

Get the roles wrong and the tablet becomes noise. Everyone sees everything, nobody owns anything, and the superintendent ends up doing what the paper traveller did anyway, just on a screen (and yes, I know your controls fitter will swear the alert routing is fine, check the login map yourself before the first live job).

The market pressure most consultants don't mention

None of this happens in a vacuum. O-I Glass has spent several years idling and closing furnaces under its Fit to Win consolidation programme (confirmed in the company's own investor updates), and that raises the stakes on changeover speed and first-ware yield at the furnaces still running. Under USMCA's rules of origin, container glass moves largely tariff-free across the US-Mexico border, so a plant in Ohio isn't benchmarked against some abstract industry average. It's benchmarked directly against a plant in Nuevo León making the identical SKU. A 60-minute job change is a competitive disadvantage now, not just an inefficiency.

That's the argument for building tablet rollout into a systemised approach rather than a hardware purchase. Our Live Execution layer maps the checklist to the nine-stage Job Change Lifecycle, so the tablet isn't replacing paper with paper on glass. It's replacing tribal knowledge with a locked recipe and a role-based screen. Pair that with the KPI trending under digitalisation and reporting and you've got a defect-mode-specific reject rate you can defend at a board review, not a login count.

Whether your plant sits in Ohio, Valdosta or somewhere further afield, the test is the same. Not whether the crew logged in. Whether the reject rate moved, and whether you've got a vendor-neutral container glass consultant looking at the workflow, not just the screen you bought.