SMED for Container Glass Is Not SMED for Automotive

Every Lean Six Sigma trainer who walks into a container glass plant brings the same slide. Shingo. Toyota. The 1,000-tonne Danly press die change cut from four hours to three minutes. The implication is clear: apply the same thinking to your IS machine changeover and expect similar compression. You can apply the thinking. But the targets on those slides are physically impossible on a hot-end forming line, and if your programme chases them, you will spend a year demoralising a crew that is doing nothing wrong.

What Shingo proved, and what a press cannot tell you

Shingo's 1985 SMED study produced a 97% compression on a stamping press because a stamping die is thermally inert. You bolt it in, align it, and run. No conditioning requirement. That is a genuine external-setup problem with a genuine external-setup solution.

An IS machine section is not a die. A single section carries 80-120 kg of cast iron at process temperature. The blank mould, finish mould, neck ring, and plunger assembly all need to reach working temperature before the glass will form correctly. Install a blank mould body below 350°C and you will get check cracks (surface micro-fractures at the shoulder or heel) within the first eight to 15 minutes of production. The target body temperature before installation is 420-480°C. Pre-heat racks charged externally reduce thermal ramp-up rejects by 40-60%, and that is the correct SMED principle correctly applied. But you cannot eliminate the conditioning window itself.

And once the sections are running, gob weight drift adds a constraint no automotive SMED template accounts for. After a job change, gob weight variance of ±3-5 g on blown ware takes 8-18 minutes per section to settle within the ±2 g process specification. During that window, every bottle goes to cullet. Every single one.

The constraints generic SMED training doesn't see

The hot-end defect spike after a section restart is predictable and thermally driven. Reject rates climb from a baseline of 2-6% to 25-45% in the first 10-20 minutes after each section restart. The dominant failure modes are blowback (parison fails to seat in the finish mould), unfilled finish (thread profile incomplete from a cold neck ring), and settle wave (corrugated sidewall from incorrect blank-open timing). None of these respond to faster crew choreography. They respond to temperature management and recipe precision.

A generic Lean boutique running value-stream mapping on a day shift will capture bolt sequences and crew movement. What they won't capture: forehearth profile stability in the hour before changeover, feeder needle wear state, or gob conditioning quality. OEM-affiliated consultancies share the same blind spot. They optimise within the tooling ecosystem they supply and leave upstream process conditions outside their scope. That gap accounts for 20-35% of transition-time waste their programme won't address.

A job change that straddles a crew change adds a statistically predictable 25-45 minutes of transition time. Most SMED programmes never capture it because the consultant observed day shift and was gone before the crew changed.

The incoming Gaffer doesn't own the section state. The Hot-End Inspector on the new shift hasn't seen the first-ware run. The lehr operator is reading a temperature profile from scratch. Day-shift-only observation makes all of this invisible in the analysis, and the root cause never appears on the improvement action list.

Where the time actually hides in a real changeover

In 2017 I was reviewing changeovers at a two-furnace, six-line plant running a mixed SKU slate across wine, beer, and food containers. The plant had completed SMED training the year before. Average job-change time had not moved. When we mapped elapsed time from line-down to stable pack across 40 job changes, the bulk of the overrun sat in three places: IS machine timing file re-entry, lehr re-profiling, and shift-boundary handover. None of those three appeared in the SMED training outputs. The programme had been perfecting the bolt sequence on a problem that wasn't the bottleneck. Not a press problem. A glass problem.

Emhart Glass, Bucher-Heye, and Siemens IS timing controllers all allow job timing files to be stored and recalled per container article. Plants without a validated job-recipe library lose 20-45 minutes per changeover to operators manually re-entering section timing values (and yes, I know your experienced operator recalls most articles from memory — check what happens when he's on leave and the cover man last ran this SKU 18 months ago). The single largest avoidable internal-setup activity in a modern container glass plant, and it costs nothing to fix.

Lehr profile switching is the next target. Annealing lehr peak temperature sits at 560-580°C and must be re-profiled for new container weight and wall thickness. Without pre-programmed profile loading, re-profiling carries 25-40 minutes of response lag. With PLC-controlled pre-stored profiles loaded as part of the job recipe, that drops to 4-8 minutes. One of the highest-value external-setup conversions available, and an OEM-tied team running a three-day workshop won't find it because it sits outside their mould-change scope.

Hot-end coating re-parametrisation adds another 10-20 minutes if SnCl2 spray booth parameters aren't pre-stored per job recipe. Incorrect coating weight causes scuffing on cold-end conveyors, typically surfacing 45 minutes after section restart and long after the changeover record has been closed.

With all of that accounted for, a single IS machine section mould change with a two-person crew takes 22-38 minutes. A three-person crew with a pre-staged tooling cart compresses that to 14-20 minutes. That is a 35-45% reduction on the physical swap interval. That is the realistic ceiling for SMED on a container glass forming line. Not 97%.

What a plant-specific approach actually delivers

Across the GCC, SKU proliferation in the non-alcoholic beverage segment has pushed job-change frequency on mixed-product furnaces to 5-7 per month at several plants, up from a historical two. In Europe, under EU ETS Phase IV, the hollow-glass product benchmark sits at 0.5377 tCO2eq per tonne of saleable glass; as free allocation headroom shrinks toward 2030, each tonne of changeover scrap carries a growing uncompensated carbon cost, making reject-rate reduction a quantifiable decarbonisation lever. In North America, O-I Glass's 'Fit for the Future' restructuring permanently closed Ruston, LA and Corsicana, TX, with surviving plants absorbing higher SKU variety and unchanged manning models.

The operating environment across all three regions is the same: more job changes, tighter margins, and less tolerance for overruns. A vendor-neutral container glass consultant approaching the problem without OEM ties will map elapsed time across real job changes, identify where the bottlenecks actually sit, and build improvement targets around what the physics allows.

Look, you can run SMED training every quarter and not move the number if the programme is built on automotive assumptions. The methodology has genuine value in container glass. The automotive compression targets don't transfer.

The Job Change Tool Lean Glass built is a systemised methodology with a digital execution layer covering the full changeover lifecycle, from recipe lock-in and pre-staged tooling through first-ware sign-off and post-mortem. It is vendor-neutral by design, which matters on mixed-OEM fleets where outside consultants have often been tied to equipment suppliers. A forming audit is the right starting point if you want to understand where your changeover time actually sits before committing to a programme. Or read the SMED glossary entry for a working definition of where the methodology holds and where it hits its structural limits on a hot-end forming line.