Container Glass Colour Change Downtime: The Real Bottleneck

Every SMED consultant who walks into a container glass plant says the same thing: split the changeover into internal and external work, do what you can offline, and you'll cut the downtime in half. It works on a stamping press. It falls apart on a furnace running a colour change, because there is no offline. The melt doesn't stop, the chemistry doesn't pause, and half the 'setup' is actually a multi-hour migration happening inside a tank that's still pulling glass.

That's the gap I keep finding when I sit in on colour-change post-mortems, from a two-furnace flint/amber plant in northern Spain to a multi-forehearth site in the American Midwest. Everyone times the mould change. Almost nobody times the redox reversal that's actually driving the reject rate.

Why flint-to-amber clears faster than amber-to-flint

In 2017, on a two-furnace amber/flint plant in northern Spain, we ran an amber-to-flint changeover that the schedule said would take thirty hours. It took fifty-four. Twenty-three hours of that overrun sat in one place: waiting for residual iron and chromium in the refractory and cullet to dilute back down to flint spec. Nobody had built that into the plan, because everyone was still watching the forming end.

Twenty-three hours. That's what the log showed, and it's the direction that catches plants out every time. Flint-to-amber is comparatively forgiving. You're adding colourant and reductant into a batch that's already clean, and the chemistry moves the way you push it. Amber-to-flint asks the furnace to purge iron, chromium and sulphur contamination that's soaked into the refractory throat and spout over a full campaign, sometimes weeks long, before flint spec is even reachable. On an older single-forehearth design (I've seen this on a Sorg fish-tail forehearth running a 1990s control set) that purge happens across the whole melt, not just the channel, and full-furnace colour changes on that kind of fleet routinely run 24-72 hours depending on the colour pair.

The redox reversal is the real bottleneck, not the machine changeover

Amber glass is held at a target redox ratio, Fe2+ against total iron, of roughly 40-55%, driven by carbon or coke reductant dosing in the batch. Flint needs near-zero iron colourant and oxidising conditions. Reversing that redox state through a continuously operating melt is the actual bottleneck. Not a mould problem. A chemistry problem, and the forming end is usually the last thing to feel it.

Cord is the tell. It shows up as chemical striae running through the ware, and nine times out of ten on a colour changeover it traces back to incomplete batch homogenisation during the redox reversal, not to gob weight or mould temperature drift downstream. Baseline reject rate on a stable colour run sits at 2-3%. During the first hours of a new colour it commonly spikes to 15-30%+, driven by that cord, by colour streaking, and by seed and blister defects from the redox shift itself.

Three roles carry that window, and none of them is the mechanic changing moulds:

• The batch house operator adjusts colourant and reductant dosing in real time as the melt migrates
• The IS machine operators watch gob weight and mould temperature drift as glass chemistry shifts underneath them
• Quality lab runs spectrophotometer colour checks, CIE Lab and dominant wavelength, every 30-60 minutes until the new spec holds stable

(and yes, I know your batch house will tell you the dosing curve is dialled in, pull the spectro trend anyway).

A colour change is a chemistry event that happens to end with a mould change. Treat it like a mechanical setup and you'll optimise the fifteen minutes that don't matter and miss the fifteen hours that do.

Recipe lock per colour family is where plants lose control

The hot-end superintendent owns recipe lock. The operator doesn't move a set point without sign-off, and that sounds obvious until you sit through a 0600 handover. On most lines I've audited, that handover misses the night shift's spectro data on residual chromium contamination more than half the time, because it's written in a logbook nobody's reading against the next shift's plan.

This is exactly the gap the Job Change Tool was built to close. It's a versioned recipe per colour family, locked in a SKU Library rather than a night-shift operator's notebook, so the amber recipe you run in March is the same amber recipe, on the same mould set and forming spec, that you run in November. Zaid Hassoneh, who founded Lean Glass after running the floor at O-I Brisbane from 2005 and making plant manager by 2019, won O-I's Most Improved Job Change Plant, Globally award in 2016 for closing exactly this kind of gap on a live production fleet. It's not a hardware fix. It's a discipline fix, and it maps cleanly onto the 9-stage Job Change Lifecycle, from plan and prep through first ware and post-mortem.

What Europe's carbon rules actually change about colour campaigns

Under EU ETS Phase IV, the free-allocation benchmark for glass manufacturing tightens by a 4.3% linear reduction factor a year through 2030 (European Commission DG CLIMA), and that squeezes how much non-saleable melt-through time a furnace can absorb during a full colour change. But CBAM, whose definitive phase started 1 January 2026, doesn't currently list container glass in its Annex I covered goods, unlike cement, steel and aluminium (European Commission Taxation and Customs Union). That's a distinction I've seen more than one generic consultancy get backwards when they pitch carbon-cost savings on faster changeovers.

FEVE, the European Container Glass Federation, reports EU-wide average recycled cullet content above 52%, with member states like Belgium and Germany collecting over 90%. A high-cullet furnace flushes through a colour transition differently than a virgin-batch-heavy one, because the recycled glass changes how fast contamination dilutes out. It's also why the fleet matters more than the recipe. FEVE members run over 150 furnaces across the EU, the bulk on multi-forehearth designs that allow a colour switch in a single forehearth, feeder and conditioning channel only, without disturbing the main melt. That cuts changeover to as little as 4-8 hours against 48-plus for a full-furnace change on a single-forehearth fleet. It's a scheduling and campaign-planning decision far more than a capex one, and it's rarely framed that way in an OEM's pitch deck.

In the US it's a permit problem as much as a chemistry one. Title V air permits routinely cap annual selenium and chromium emissions tied to colourant use under EPA's Glass Manufacturing NESHAP, which in practice limits how many amber-to-flint changeovers a furnace can legally run in a year without a permit modification. With O-I and Ardagh idling North American furnace capacity through 2023-2024, the changeovers left on the remaining assets carry more weight, and fewer of them can go wrong.

Cullet discipline is what separates a changeover from a scrap event

Pack-to-melt commonly drops 15-25 percentage points during the changeover window as off-spec ware gets diverted to cullet. On a 200-400 tonnes/day furnace, a full-furnace colour change can cost an estimated $150,000-$500,000 per event in lost saleable production, scrapped cullet and sustained energy input during a non-saleable melt-through period, and most plants never model that against refractory campaign life, typically 8-15 years, when they're deciding how often to run the change at all.

Cull discipline means the off-spec cullet from a green or amber campaign doesn't quietly re-enter the next flint batch through a shared return line. Chromium residue from a green run, dosed at 0.3-1.0% Cr2O3, is one of the hardest contaminants to purge before flint, and it doesn't care whether it came from the furnace refractory or from a cullet bin someone didn't segregate properly. This is one of the first things we check on a hot end audit, whether a plant's cullet segregation is actually preventing cross-contamination during a colour change or just moving the problem downstream to the next campaign.

And this is where a generic Lean boutique or an OEM-affiliated shop tends to stop short. They'll pitch a faster feeder or an automated dosing system and call it solved. An independent, vendor-neutral container glass consultant looks at the melt chemistry, the permit constraints and the refractory wear together, because no single piece of hardware fixes a discipline gap on its own.

Fifty-four hours against a thirty-hour plan, on a furnace that was still legally allowed several more chromium changeovers that year before hitting its permit cap. So what's the real number on your last amber-to-flint, and does anyone in your plant actually know it without pulling three logbooks to find out?