Scaling a Cheese Sauce from 20 L to 500 L
A 20 L pilot recipe for Mornay sauce uses 1.2 kg butter, 1.2 kg flour, 20 L milk, 3 kg grated Gruyère, 0.4 kg Dijon mustard and salt. It is scaled 25× to a 500 L steam-jacketed kettle with a bottom-mounted paddle agitator and a 1.5 m² jacket. On the first industrial run the operator used 6 bar steam and the same sequence as the pilot. The batch took 95 minutes to reach 85°C instead of the expected 35 minutes, scorched about 2 mm on the jacket surface, and had a viscosity of 1,900 cP against a target of 5,000–5,500 cP at 65°C. The cheese did not fully melt and the sauce tasted floury. Production wants to add 30% more starch to correct the viscosity. The next run in 48 hours must supply 12,000 portions. The kettle cannot be fitted with a scraped-surface agitator before that run, but the jacket pressure can be controlled and the order of addition changed.
1. Diagnose the scale-up failure using heat transfer, mixing and starch/cheese chemistry. Why did the batch take 95 minutes and scorch?
Read the worked answer
The pilot recipe was scaled by mass, but heat transfer and mixing do not scale linearly.
Heat-transfer area per unit volume falls sharply. The pilot pot might have roughly 0.25 m² of heated surface for 20 L, or about 0.0125 m²/L. The kettle has 1.5 m² for 500 L, or 0.003 m²/L - about four times less jacket area per litre. The same steam pressure therefore gives a much lower heat flux per unit volume, so the come-up time stretches. The 95-minute time is consistent with a surface-area limitation, not with a faulty steam supply.
The high jacket pressure makes the problem worse, not better. Steam at 6 bar is about 165°C. Against a product at 20–85°C, the wall temperature is far above the point where milk proteins and starch burn. A 2 mm scorched layer on the jacket is an insulating fouling layer, which further reduces heat transfer and extends the batch time. The paddle agitator does not scrape the wall, so the boundary layer at the hot surface is stagnant and overheats.
Starch and cheese chemistry: dry flour added to hot fat can form lumps; in a 500 L batch the flour may not hydrate evenly before the viscosity rises. Incomplete gelatinisation gives a floury taste and low viscosity. Adding cheese at 85°C coagulates the cheese proteins and squeezes out fat, so the cheese does not melt smoothly. The lower surface-to-volume ratio also means less evaporation than in the open pilot pot, leaving the sauce wetter and thinner than expected.
2. Propose a corrected process for the next 48-hour run without a scraped agitator. Include jacket pressure, addition sequence, mixing and holding.
Read the worked answer
Do not add 30% more starch. That would raise viscosity by making a gluey, pasty sauce, mask the incomplete gelatinisation and change the recipe and yield without fixing the cause.
Corrected process:
- Pre-blend the flour into cold milk to make a smooth slurry, or make a roux in a smaller kettle and dilute it into the main batch. A slurry hydrates the starch before it meets high heat and reduces lumping.
- Reduce jacket pressure to about 2 bar (approximately 134°C) and control product temperature, not steam pressure alone. Lower the jacket temperature further as the product approaches 80°C.
- Heat to 90–92°C with the paddle at maximum practical speed, and keep the product moving over the jacket. Without a scraper, accept a longer come-up but avoid a hot stagnant wall. If the kettle has an intermittent agitation mode, do not use long pauses.
- Hold at 90–92°C for 10–15 minutes to complete starch gelatinisation, then check viscosity. If it is still low, adjust with a small amount of pre-gelatinised starch or reduce the milk charge; do not dump in 30% starch.
- Cool to below 70°C before adding the cheese and mustard. Add cheese in stages with gentle agitation; hold at 65–70°C for no more than 10 minutes before filling.
- Adjust for evaporation: the 500 L kettle loses proportionally less water than the pilot pot, so start with about 3–5% less milk or add back only measured water.
Record the actual jacket pressure, product temperature curve, agitation speed and viscosity at 65°C for the batch record.
3. Define the quality checks and scale-up validation for the next run and ongoing batch release.
Read the worked answer
The next run is a validation batch, even though it must also supply production. Build the checks into the process.
In-process checks: product temperature at the centre and near the wall at 5-minute intervals; jacket pressure; agitation speed; time at 90–92°C; visual inspection of the jacket surface after emptying for burn-on; viscosity at 65°C using a fixed Brookfield spindle and speed; colour; pH; salt; and a starch-gelatinisation check (iodine test or microscopy) if available.
Sensory and functional checks: floury taste, grittiness, cheese flavour, gloss, cling and heat stability. Hold a sample at 65°C for 30 minutes and at 4°C for 24 hours to check viscosity recovery and fat separation.
Microbiological checks: total viable count, Bacillus cereus and Clostridium perfringens if the product is held warm, plus Listeria and Staphylococcus aureus as required by the HACCP plan.
Validation acceptance: three consecutive batches within the viscosity target, no detectable burn-on, sensory approval against the pilot reference, and a temperature history that meets the safety limit. Establish critical limits for jacket pressure, hold temperature and hold time. Ongoing release should require the batch record, viscosity, temperature curve and sensory check. Trend the viscosity data; if it drifts, investigate the flour lot, milk solids or kettle fouling before changing the recipe.