Glassworks: furnace temperatures drift, rejects and gas follow

Bottom and throat temperatures drift

What happens. The furnace bottom and throat temperatures wander over days, although the burners and boost run as usual. The shift adjusts fuel or boost when the trend looks wrong, each shift a little differently. By the time a change in bottom temperature shows, the glass that caused it is long gone.

What it costs. Seeds, stones and cord appear at the cold end, and the furnace runs hotter than needed to stay on the safe side. That means more gas, more electric boost and faster refractory wear.

Why the usual fix fails. Standard controllers on single crown or bottom temperatures react to what has already happened and cannot handle the furnace’s long delays. Hand corrections depend on who is on shift.

How a fixed-price remote diagnosis checks it. From your furnace history (crown, bottom and throat temperatures, fuel and boost, pull, batch and cullet) I check how well the bottom and throat temperatures can be predicted hours ahead. The fixed-price remote diagnosis shows whether predictive temperature control could hold them steadier and at a lower mean, and what it would need.

Background: Predictive control of process plants

Cullet share and colour change from load to load

What happens. The share of internal and external cullet changes with what is available, and external cullet arrives in mixed colours and qualities. The batch recipe stays as planned while the actual mix in the furnace shifts. Operators notice it when the glass colour moves or foam builds up.

What it costs. Colour off target, foam that blocks heat transfer, and redox swings that take a day to correct. Each correction burns fuel and can produce off-spec glass for hours.

Why the usual fix fails. More cullet analysis helps, but samples cover a tiny part of the delivery. Fixed recipe corrections are tuned for one cullet quality and do not fit the next one.

How a fixed-price remote diagnosis checks it. From batch weights, cullet deliveries and analyses, furnace data and colour and redox measurements I check whether the real composition of the melt can be estimated from what goes in and what comes out. The fixed-price remote diagnosis tells you whether an estimated batch could adjust the recipe in time, and what it would take.

Background: Control that adapts to drift

The redox state is known only from the lab

What happens. The redox state of the glass decides colour and fining, but it is measured in the lab, a few times a day at best. In between, the furnace runs on fuel ratios and batch carbon set from the last result.

What it costs. When the redox drifts, colour defects and bubbles appear before the next lab value confirms it. For coloured and flint glass alike, every hour of drift means glass that is sorted out or downgraded.

Why the usual fix fails. Sampling more often adds lab work but still looks back. Online analysers for the melt are expensive and hard to keep running in the heat.

How a fixed-price remote diagnosis checks it. From your lab redox values together with batch, cullet, combustion and furnace data from the same periods I check whether the redox state can be computed continuously from signals you already have. The fixed-price remote diagnosis shows how well such a computed value would have warned of past drifts, and whether it is good enough to act on.

Background: Computed measurements

Gob temperature is not uniform across the forehearth

What happens. The forehearth zones are set by experience, and the glass reaching the spout is hotter in the middle than at the sides, or warmer on one side than the other. When the pull or the article changes, the zones are adjusted again by hand.

What it costs. Uneven gobs form unevenly: thin walls, wrong weight distribution, checks and other forming defects. Rejects rise, and the machine runs slower than it could.

Why the usual fix fails. Each zone has its own standard controller, but the zones influence each other and the glass needs time to pass through. Tuning one zone shifts the next, so operators settle for a compromise.

How a fixed-price remote diagnosis checks it. From the forehearth thermocouples, burner and cooling settings, pull and article data, and the reject records I check how strongly the zones interact and how uniform the gob temperature could be. The fixed-price remote diagnosis answers whether model-based zone control would improve the uniformity, and what it needs.

Gob weight drifts with glass temperature

What happens. The gob weight changes when the glass temperature or viscosity changes, even with the same plunger and tube settings. The weight control corrects after the weighed articles show the drift, or the operator corrects by hand.

What it costs. Underweight articles are rejected; overweight ones give away glass and slow the forming. Each article change brings a new period of searching for the right weight.

Why the usual fix fails. A weight controller that only reacts to measured weights is always one step behind the temperature. Wider weight limits hide the problem in giveaway.

How a fixed-price remote diagnosis checks it. From the weight records, forehearth and spout temperatures and the feeder settings I check how much of the weight variation follows the glass temperature and how early it shows. The fixed-price remote diagnosis tells you whether adaptive gob weight control using the temperature could keep the weight tighter, and what it would take.

Background: Control that adapts to drift

Mould swabbing and machine timing depend on the operator

What happens. On the forming machines the operators swab the moulds and adjust the timing of each section from what they see and feel. An experienced operator keeps the defects down; on another shift the same article runs worse.

What it costs. Defects that come and go with the shift, articles sorted out at the cold end long after the cause, and training that takes years. When good operators are missing, quality drops.

Why the usual fix fails. Written settings and timing sheets do not capture why a change was made. Automatic swabbing helps with the routine but not with the decisions.

How a fixed-price remote diagnosis checks it. From machine timing, mould temperatures, swab cycles where recorded, and the inspection results per section I check which process signals predict the defects before inspection does. The fixed-price remote diagnosis shows whether a quality estimate from process signals could guide the operators, and how reliable it would be.

Annealing lehrs run fixed temperature profiles

What happens. The lehr runs a temperature profile set once for the heaviest article, and it stays the same for lighter ones. Nobody knows the glass temperature inside the articles as they pass; the profile is the only guide.

What it costs. Gas for heat the lighter articles do not need, and a lehr that is longer or slower than necessary. When the profile does not fit, residual stress leads to breakage later in filling or use.

Why the usual fix fails. Changing the profile by hand per article takes trials and stress measurements, so most plants keep a safe profile for everything. A few polariscope checks show the result, not the margin.

How a fixed-price remote diagnosis checks it. From lehr zone temperatures, burner settings, belt speed, article data and your stress measurements I compute how the glass temperature develops through the lehr for each article. The fixed-price remote diagnosis tells you how much margin each article has and whether a computed glass temperature could set the profile.

Float glass thickness swings when the line speed changes

What happens. When the line speed or the target thickness changes, the ribbon thickness overshoots or lags before it settles. The operators adjust top rolls and temperatures by hand and wait.

What it costs. Glass outside the thickness tolerance during every change, cut out and returned as cullet. Changes are avoided or made slowly, which costs flexibility in the order book.

Why the usual fix fails. Separate controllers on speed, rolls and bath temperatures work against each other during a change. Ramping more slowly reduces the swings but lengthens the transition.

How a fixed-price remote diagnosis checks it. From the line history around past changes (speed, top roll settings, bath and lehr temperatures, thickness profile) I check how the thickness responds and how well it can be predicted. The fixed-price remote diagnosis answers whether predictive control could shorten the transitions and keep the ribbon in tolerance, and what that would take.

Background: Predictive control of process plants

Fibre glass bushing temperatures interact

What happens. Each bushing has its own temperature control, but neighbouring bushings, the glass level and the cooling all affect each other. When one position is adjusted, the others move. Operators keep the temperatures where filaments break least, by feel.

What it costs. Filament breaks interrupt production and need restringing. Each break means lost output, waste and staff time, and fibre diameter varies more than it should.

Why the usual fix fails. Tuning single loops tighter makes the interaction worse. Keeping wide margins on temperature reduces breaks but costs throughput and fibre quality.

How a fixed-price remote diagnosis checks it. From bushing temperatures and power, glass level, cooling settings and the break records I check how strongly the positions interact and which conditions come before a break. The fixed-price remote diagnosis shows whether multivariable control of the bushings could reduce breaks, and what it needs.

Regenerator reversal runs on a fixed timer

What happens. The firing side changes over on a fixed timer, whatever the pull, the chamber temperatures or the fuel. The timer was set years ago and nobody has dared to change it.

What it costs. Less heat recovered from the waste gas, so more fuel for the same glass. Uneven chamber temperatures also strain the checkers and can shorten their life.

Why the usual fix fails. Changing the timer by trial is risky on a running furnace and the effect is hard to see in day-to-day noise. So the old setting stays.

How a fixed-price remote diagnosis checks it. From chamber and waste-gas temperatures, fuel and air flows, and the reversal times I compute how much heat each cycle recovers and how that changes with timing and pull. The fixed-price remote diagnosis tells you whether an optimised reversal would recover more heat safely, and how to test it.

Background: Economic predictive control

Oxygen and fuel are dosed with a wide safety margin

What happens. In oxy-fuel or air-fired furnaces the oxygen or air excess is set high enough to be safe on the worst day. It stays there, although fuel quality, pull and furnace condition change.

What it costs. Excess oxygen or air heats up and leaves through the stack, which costs fuel and, in air-fired furnaces, raises nitrogen oxides. Too little would harm the glass and the redox state, so nobody wants to go lower.

Why the usual fix fails. A fixed ratio cannot follow the changing conditions. Waste-gas analysers help, but their readings lag and drift, and a standard controller on them alone is slow.

How a fixed-price remote diagnosis checks it. From fuel and oxygen or air flows, waste-gas analyses, crown temperatures, pull and glass quality data I check how much excess is really needed under which conditions. The fixed-price remote diagnosis shows whether optimised combustion control could run closer to the need, and what it would require.

Electric boost runs at a fixed share whatever power costs

What happens. The electric boost delivers a fixed share of the melting energy, set when the furnace was commissioned or when prices were different. Power prices now change every hour, and gas prices move too, but the split stays.

What it costs. Boost runs at full power in expensive hours and is not used more when power is cheap. The bill for melting energy is higher than it needs to be.

Why the usual fix fails. Shifting energy between boost and gas by hand affects glass temperatures, convection and quality, so operators do not touch it. A price signal on its own says nothing about what the glass can take.

How a fixed-price remote diagnosis checks it. From boost and gas history, furnace temperatures, quality data and your price contracts I check how far the split could move with prices without harming the glass. The fixed-price remote diagnosis answers whether running the boost by price and glass quality would pay off, and what it would need.

Background: Economic predictive control

The furnace behaves differently as the campaign goes on

What happens. Over a furnace campaign the refractory wears, insulation changes and the furnace reacts more slowly or differently to the same settings. Settings that worked in the first years no longer fit, and the shift corrects more often.

What it costs. More fuel for the same glass, more temperature drift and more defects later in the campaign. Controllers tuned at start-up become sluggish or unstable.

Why the usual fix fails. Retuning controllers now and then helps for a while. Without a picture of how the furnace has changed, each retuning is a guess.

How a fixed-price remote diagnosis checks it. From several years of furnace data I check how the furnace’s response has changed over the campaign and whether a model can follow that change. The fixed-price remote diagnosis tells you whether adaptive models could keep control and fuel use steady as the furnace ages.

Background: Control that adapts to drift

Job and pull changes upset the furnace

What happens. When the article or the number of machines changes, the pull changes, and the furnace needs hours to settle. The operators adjust fuel and boost after the temperatures have already moved.

What it costs. Defects and colour or seed problems during the transition, and the habit of running hotter just in case. Job changes are planned around the furnace rather than the orders.

Why the usual fix fails. Feedback control only starts once the temperatures move. Hand corrections ahead of the change depend on one person’s experience.

How a fixed-price remote diagnosis checks it. From the history around past job changes (pull, fuel, boost, temperatures, quality) I check how the furnace responds to a pull change and how far ahead the right correction can be computed. The fixed-price remote diagnosis shows whether feedforward for pull changes could shorten the upset, and what it needs.

Hydrogen firing trials bring a new flame and new heat transfer

What happens. A trial with hydrogen in the burners is planned or under way. The flame is shorter and radiates differently, the water vapour in the furnace rises, and heat goes to different places in the melt. Nobody knows in advance how the glass will react.

What it costs. Uncertain glass quality during the trial, refractory and redox questions, and the risk of stopping the trial early without useful results.

Why the usual fix fails. Trials run on the existing controls with conservative settings, which shows little about real operation. Learning by trial and error on a production furnace is slow and expensive.

How a fixed-price remote diagnosis checks it. From your furnace data, burner layout and the trial plan I check which quantities will change most and what the control must watch. The fixed-price remote diagnosis answers whether a furnace model and an adapted control concept for the trial are worthwhile, and what they would take.

Inspection rejects are not traced back to their cause

What happens. The inspection machines at the cold end count and sort out defects by type. Why they rose this morning is discussed in the meeting, but the link to the furnace, forehearth or forming data is made by memory.

What it costs. Glass melted, formed and annealed only to be sorted out. Without a clear cause, the same defects return.

Why the usual fix fails. Each department looks at its own data. Checking dozens of signals by hand against hours of delay between melting and inspection rarely leads to a firm answer.

How a fixed-price remote diagnosis checks it. I align your inspection records with the furnace, forehearth and machine data, taking the delays into account, and look for the signals that come before each type of defect. The fixed-price remote diagnosis shows which causes the data supports and whether a permanent root-cause check would be worth building.

Thermocouples in hot zones fail and leave the furnace blind

What happens. Crown, bottom and forehearth thermocouples fail or drift in the heat. Replacing them often has to wait. Meanwhile the control runs on the remaining sensors or the operators run blind on that spot.

What it costs. Loops switched to manual, temperature drift nobody sees, and defects that follow. A drifting thermocouple may go unnoticed for weeks and pull the control the wrong way.

Why the usual fix fails. More thermocouples cost money and fail too. Simple substitution rules break when conditions change.

How a fixed-price remote diagnosis checks it. From your temperature history I check how well each measurement can be computed from the others and from fuel, boost and pull, and how early a drifting sensor would show. The fixed-price remote diagnosis tells you whether fault-tolerant estimation could cover failed sensors, and what it needs.

Background: Computing unmeasured states for control

Furnaces, compressors and fans are not run together

What happens. The furnace, compressed air, cooling air, combustion fans and the cold end each run on their own settings. Compressors and fans often run at fixed speed or with throttles. Nobody looks at the site as one energy system.

What it costs. Power for air nobody uses, peaks that push up grid charges, and waste heat that is not used where it could be.

Why the usual fix fails. Single measures (a new compressor, a variable-speed drive) help locally. Without coordination the savings in one place are partly lost in another.

How a fixed-price remote diagnosis checks it. From meter data, compressor and fan logs, furnace and production data I build an energy picture of the site and look for the largest losses and the links between the consumers. The fixed-price remote diagnosis shows whether site-wide optimisation would pay off and where to start.

Background: Economic predictive control

Colour changes are planned by habit

What happens. The order of colour changes and campaigns follows habit and the planner’s experience. Transitions are run the way they have always been run, with the same batch steps and the same waiting time.

What it costs. Days of glass that is neither the old nor the new colour and has to be recycled or sold cheaply. Long transitions also block capacity for urgent orders.

Why the usual fix fails. Shortening a transition by trial risks off-colour glass in the new campaign. Planning software usually does not know what a given transition really costs.

How a fixed-price remote diagnosis checks it. From past colour changes (batch changes, colour measurements over time, losses) and your order data I check how the transitions could be run faster and in which order campaigns cost least. The fixed-price remote diagnosis tells you whether campaign scheduling and planned transitions would pay off, and what they need.

Background: Optimised schedules and setpoint plans

Furnace experts retire and take their know-how with them

What happens. A few senior people know how the furnace reacts, which settings to touch and which to leave alone. They are retiring, and younger staff are hard to find. Much of the knowledge has never been written down.

What it costs. Slower reactions to upsets, more defects and more fuel when the experienced people are not on site. The risk of a costly mistake on the furnace grows.

Why the usual fix fails. Handover documents list settings but not the reasons. Training on the job takes years that the plant may not have.

How a fixed-price remote diagnosis checks it. With the experts and the furnace history I check which of their rules can be confirmed in the data and captured in a documented furnace model. The fixed-price remote diagnosis shows how much of the know-how a model can hold, and how it could support the operators who follow.

Waste heat must be reported, yet nobody uses it

What happens. Under the German Energy Efficiency Act, companies using more than 2.5 gigawatt hours of final energy a year must report their waste heat to a federal platform every year, for sources above a minimum size, with quantity, temperature and availability over the year (Bundesstelle für Energieeffizienz). In glassworks that heat leaves through the stack and the cooling air.

What it costs. Reporting effort for heat that stays unused, and a growing expectation to avoid or reuse it. The fuel behind that heat is already paid for.

Why the usual fix fails. The report is filled in from estimates once a year. A recovery project needs real load profiles and a control that matches the furnace’s operation, which the report does not give.

How a fixed-price remote diagnosis checks it. From waste-gas temperatures and flows, cooling-air data and the furnace’s operating history I compute how much heat is available, when and at what temperature. The fixed-price remote diagnosis shows where heat recovery could be integrated and how its control would have to run with the furnace.

Hybrid furnaces with a large electric share behave differently

What happens. Glassmakers are building hybrid melters that can take up to 70 % of their energy from electricity, and some are planning large electric melting tanks; the same report notes that electric furnaces tend to wear out about twice as fast as gas furnaces (Canary Media). Convection, bubbles and refractory wear behave differently than in a gas furnace.

What it costs. A new furnace with little operating experience, exposure to power prices, and the risk of quality problems in the first years.

Why the usual fix fails. Control concepts are copied from gas furnaces with boost. They do not use the freedom of the gas and electric split, and they are not built for prices that change by the hour.

How a fixed-price remote diagnosis checks it. From the furnace design, the planned operation and any trial or boost data I check which quantities the control has to watch and how the split could follow power prices within the limits of the glass. The fixed-price remote diagnosis tells you whether a melt model and predictive control of the split are worthwhile, and what they would need.

Background: Economic predictive control

Hydrogen in the natural gas changes the flame from day to day

What happens. As hydrogen is blended into the gas grid, the gas arriving at the furnace changes in composition. Even where its heating value stays within limits, flame length, radiation and nitrogen oxides change with the hydrogen share. The burners are set for one gas.

What it costs. Efficiency, glass quality and emissions drift with the gas, without anyone in the plant seeing why. Operators add margin, which costs fuel.

Why the usual fix fails. Burner settings and fuel-air ratios are fixed at commissioning. A gas analyser shows the composition but does not tell the furnace what to do about it.

How a fixed-price remote diagnosis checks it. From gas quality data where available, burner and furnace data, waste-gas analyses and quality records I check how strongly your furnace reacts to gas composition. The fixed-price remote diagnosis answers whether combustion control from computed gas quality could keep the furnace steady, and what it would take.

No reduced-gas operating mode is ready for a tight winter

What happens. German gas storage stood at about 53 % in early September 2026, the lowest for the time of year in fifteen years of records (Clean Thinking). A glass furnace cannot simply stop. If gas becomes scarce or very expensive, the plant has to decide quickly how far it can reduce gas.

What it costs. Price peaks, possible supply limits, and decisions made under pressure that can damage the glass or the furnace.

Why the usual fix fails. Emergency plans list what to switch off but not how far gas can be reduced while the glass stays sellable, or how boost, cullet share and pull should change together.

How a fixed-price remote diagnosis checks it. From your furnace history and quality data I compute in advance which reduced-gas and fuel-switch operating modes are possible and what they cost in output and quality. The fixed-price remote diagnosis shows which modes are realistic and what preparation they need.