Furnace campaign life: reading the signals before rebuild

In 2014 I walked the furnace crown at a two-furnace container plant in Queensland with the melter on shift, and he pointed at a hot spot near port five that hadn't been there six months earlier. Nobody had logged it as a trend. It just showed up on a Tuesday, and by Thursday the plant manager was fielding calls from the CFO about a rebuild quote that wasn't anywhere in this year's capex plan.

Campaign life is a curve, not a countdown

Most plants treat furnace campaign life as a number on a spreadsheet: twelve years, fifteen years, whatever the OEM quoted at commissioning. That number is a design target, not a promise. The real campaign life is written in a wear curve that starts moving from day one and rarely moves in a straight line. Look, the wear curve doesn't care what's in this year's capex plan.

Crown temperature creep is the cleanest early signal you have. A crown running +1°C a month above its commissioning baseline, sustained over a full quarter, tells you the refractory is thinning faster than the design curve assumed. Bottom wear tells the same story on a slower clock, typically 3-4mm a year against a 12-15 year design campaign on a conventional regenerative furnace. Track both against baseline, not against last month. Drift hides in month-to-month noise.

• Crown temperature creep against commissioning baseline, not against last month
• Bottom wear rate in millimetres per year against the design campaign
• Cullet ratio variance week to week, not just the average
• Gob weight CV drift in the back half of a campaign
• Defect rate trend by section, read against the crown temperature log

Twenty months, not two years. That was the real gap between what the wear curve said and what the annual inspection report said, on a furnace I audited in 2019.

Cullet ratio is quietly setting your wear rate

Cullet ratio gets treated as an energy and emissions lever, and it is. It's also a refractory chemistry lever, and most hot-end teams don't track it that way. Held in a steady 40-60% band with clean, well-sorted cullet, alkali vapour attack on a silica crown stays inside the design curve. Let contamination or colour-mix variance push sulphate carryover up, and you accelerate the same crown wear you're trying to budget around.

And it's not just contamination. A cullet ratio that swings 15-20 points week to week, chasing whatever's available on the yard, puts the same thermal and chemical stress on the furnace as a batch chemistry problem you'd never let happen on purpose. I've seen plants manage batch chemistry to three decimal places and let cullet ratio run wild, because nobody owns it as a refractory variable. It usually sits with procurement, not with the hot-end superintendent, and that's the gap.

On a 1990s silica-crown regenerative furnace, alkali attack is the dominant wear mechanism, and cullet ratio matters more than almost anything else in the batch house. On a newer furnace running an AZS fused-cast crown with oxy-fuel assist, the wear mechanism shifts and cullet matters less for the crown and more for melter temperature control. Know which furnace you're actually running before you copy someone else's wear curve.

Quality drift is the early warning system everyone ignores

By the time a furnace shows visible refractory damage, the ware has usually been telling you for months. Stones from crown spalling. Cords from alkali vapour attack working into the glass. Blister counts creeping up in the back half of a campaign as trapped gas in worn refractory starts to reboil into the melt. None of these show up as a single dramatic failure. They show up as a slow lift in the defect rate that QA logs as within spec, trending, for a year before anyone escalates it.

Gob weight CV drifting past 0.4% late in a campaign is often a refractory story before it's a forming story, because throat and spout wear change flow characteristics the feeder can't fully compensate for. Not a furnace problem. A quality system problem, if nobody's connecting the dots between the crown temperature log and the defect trend on the same section (and yes, the standard answer is 'we adjust set points to compensate,' which is exactly how a campaign that should run fifteen years quietly turns into eleven).

The 0600 handover log on most lines I've audited carries forward gob weight and section reject counts without fail. It rarely carries forward the crown temperature trend or the cullet ratio for the shift, so the two data sets that would show the drift together never actually sit side by side.

What a hot end audit catches that the annual inspection doesn't

The annual refractory inspection is a snapshot. It tells you condition on the day someone walked the furnace with a torch and a tape measure. What it doesn't do, in most plants I've audited, is connect that snapshot to six months of crown temperature trend, cullet ratio variance and section-level defect data sitting in three different systems that don't talk to each other.

A proper hot end audit pulls those threads together into one wear picture, and it's usually uncomfortable reading, because it turns a vague 'the furnace seems fine' into a dated forecast with a number attached. That's the whole point. A rebuild you can plan for eighteen months out costs a fraction of one you're forced into on six weeks' notice, in capex, in lost production, and in the panic-buy premium on refractory and contractor scheduling that always follows an unplanned cold repair.

In Europe that runway matters even more under EU ETS Phase IV and the incoming CBAM cost exposure, where every month of unplanned downtime and every tonne of extra fuel burned chasing a tired furnace back to temperature carries a real carbon cost on top of the production loss. Zaid Hassoneh, who built Lean Glass after running hot ends from O-I Brisbane through the $220M USD Arglass Yamamura greenfield build, has seen the same pattern on three continents: the wear curve was always available, nobody was reading it as one file.

This is where an independent, vendor-neutral read matters. An OEM-tied inspection has a natural incentive to round toward 'you're fine for now, call us next year.' A vendor-neutral container glass consultant has no furnace, no refractory and no rebuild contract to sell you, so the read is just the data. Lean Glass runs this alongside strategic advisory on capex timing for plants weighing a full rebuild against a partial hot repair, because the two decisions are connected and shouldn't be made by two different teams reading two different reports.

A furnace doesn't fail on the day of the rebuild quote. It fails on the day nobody was reading the curve.

That hot spot near port five in 2014 didn't need a miracle fix. It needed someone reading the crown log against the defect trend six months earlier, so the rebuild was a planned line-down instead of a CFO phone call. Most campaigns end the same way yours will, unless somebody's job is to watch the curve.