Blank Mould Cooling: The Recipe You Keep, The Dials You Don't Touch

It's 0600 on a 10-section IS machine and the outgoing shift leaves a note: "Sec 4 running warm, bumped the air up a touch." No number. No timestamp. No sign-off. By the time the day hot-end tech reads it, section 4 has been running off job-card for four hours and nobody can tell you what "a touch" means in bar.

That note is how most check defects and half your unplanned mould regrinds actually start. Not on the floor. In the handover.

Why cooling drift hides for two job changes before it shows up

Blank mould surface temperature in steady-state blow-and-blow sits in a 450-550°C band. Push past roughly 580°C and you get sticking, seam pick-up, and washboard marks on the parison surface. Drop under about 420°C and you get cold-glass lap and parisons that won't distribute evenly into the blank cavity. That's the window. It's not a wide one, and lightweighted NNPB ware narrows it further because thinner glass sections are more sensitive to thermal shock in the first place.

Here's the part that catches plants out: a cooling-profile drift rarely shows up as a defect on the job it happened on. I've seen check rates spike two, sometimes three job changes later, once the mismatch between mould dwell time and cooling air flow has compounded across several SKU changeovers. Standard in-the-moment SPC charts miss this completely because they're built for discrete manufacturing, where cause and effect sit on the same line of the log. Glass forming doesn't work that way. The lag is real and if your root-cause process doesn't correct for it, you'll blame the wrong shift every time.

The header pressure number nobody's watching

Cooling air header pressure to the blank mould jackets should hold within ±0.2 bar of set point, section by section. On a 10- or 12-section machine, a ±0.5 bar swing across sections is one of the most common root causes of weight and wall-thickness variation you'll find, and it's invisible on a single-section reading. You have to look across the whole machine, section to section, same shift, same sample window. Most plants I've audited check individual sections and never plot them side by side.

Store it, recall it, don't touch it on the floor

Here's where I part ways with a lot of the advice floating around hot-end forums. The instinct when you see a marginal defect is to reach for the cooling air setpoint and nudge it. Don't. The cooling profile is a recipe, not a dial. It gets set once per SKU, tested, locked, and then recalled identically every time that job comes back onto the machine.

This is exactly the discipline our Job Change Tool is built around. Every mould set has a versioned cooling profile sitting in the SKU Library alongside its recipe and mould spec, so the night-shift guy isn't reconstructing a setpoint from memory or a notebook. It's on the job-card, it's the same profile every single time that SKU runs, and nobody's guessing.

Settle-blow delay is a good example of what happens when that discipline slips. It's specified to within ±0.05 second on the job-card. I've watched operators shift that timing off-recipe to knock out a cosmetic defect on the floor, and it doesn't fix the cooling imbalance underneath. It just moves the reject somewhere else, usually onto the check detector downstream, where it shows up as an unrelated-looking problem three stations later.

The cooling profile isn't a knob you turn to make today's problem go away. It's a number you protect so tomorrow's problem doesn't exist.

Who actually owns the setpoint

On a well-run line the split is clear. The IS machine operator monitors mould temperature every shift, typically via a fixed-location IR pyrometer reading, and logs it against a documented action limit — commonly ±15°C from job-card target. If it's outside that band, the call goes to the hot-end or forming technician, who has the authority to re-baseline air distribution. The operator does not touch the cooling air setpoint. That's not a bureaucratic line, it's the thing that keeps one shift's fix from becoming next shift's mystery defect.

In 2014 I was running a three-line plant in Brisbane where we tracked exactly this gap. We pulled six months of unexplained job-to-job defect drift and found the pattern wasn't random at all — it clustered around three operators who'd developed their own "feel" for cooling air on marginal jobs, all with good intentions, none of it on a job-card. Once we locked the profiles and moved adjustment authority to the technician tier only, section-to-section check rate on our worst line dropped noticeably inside two job changes. Not a furnace fix. A governance fix.

Swabbing interval is the other place this bites people. Over-frequent spray lubricant application, usually done to mask a marginal cooling setup, drops effective mould surface temperature transiently and produces intermittent cord and stone-adjacent defects that inspection teams then blame on batch glass quality. It's cooling, dressed up as a raw material problem.

Regional context most transplanted recipes ignore

If you're running a European-trained cooling profile on a Gulf plant, check your assumptions at the door. GCC shop-floor ambient conditions commonly run 38-45°C in summer, and that pushes blank mould and cooling-air system design specs higher than the European baseline the recipe was probably written against. I've audited more than one plant in the region running an imported job-card that never accounted for that gap, and it shows up as unexplained regional variance that gets blamed on operators instead of on the spec.

Mould tooling lead times compound the problem. Red Sea shipping disruption has stretched commissioning timelines for imported IS machine tooling and mould sets into Gulf plants well into 2025 and 2026, which means fewer spare mould sets on hand to swap out when a cooling-driven regrind comes due early. Mould life on cast-iron blank moulds normally runs 400,000 to 900,000 cycles, and uneven cooling is the single biggest driver of premature replacement ahead of straight wear. Get the profile locked and you protect the mould as much as the ware.

And the energy angle isn't going away either. US plants under pressure from industrial electricity and gas price volatility are looking hard at cooling-air blower duty cycles as a place to cut cost. It's a legitimate lever, but cut it without modelling the reject-rate and mould-life trade-off and you'll bank a good energy KPI for one quarter and pay it back in OEE over the next two or three. Same instinct, different region, same mistake.

What this actually costs you

A cooling recipe that isn't locked, versioned, and enforced by role isn't really a recipe. It's a suggestion that degrades one undocumented adjustment at a time. That's the process-adherence gap most outside consultants miss entirely, mainly because the fix costs nothing and doesn't sell hardware.

If you're not sure whether your blank mould cooling profiles are actually locked or just written down somewhere, that's worth finding out before your next job change, not after the check rate tells you. Our forming audit is where we usually find this first — section-by-section, shift by shift, against the job-card that's supposed to be running.