Agricultural Engineering · Post-Harvest, Food & Process Engineering · also Dairy and Food Engineering, Dairy Engineering
Unit operations of food and milk plants: balances, fluid flow, heat exchange, thermal processing, evaporation, separation, freezing and hygienic design.
In the exams: GATE AG Section 7 (~9 marks a year 2019–26; thermal processing in all 8 papers); ICAR AICE subject 38; ICAR AIEEA-PG Unit VI; NCEES PE processing
Food and dairy engineering applies the unit operations of chemical engineering to milk and other foods. A dairy or food plant is a chain of such operations: materials are received and balanced, pumped through pipes, heated and cooled in heat exchangers, pasteurised or sterilised, concentrated, dried, homogenised, separated, filtered, frozen and, between every run, cleaned. The engineer's job is to size and operate that chain so that the product is safe, meets its compositional standard, and is made with the least energy, water and loss.
The subject matters because it sits where public health meets engineering arithmetic. A holding tube a few metres short, a regenerator with a pinhole on the wrong side of the pressure balance, or a retort process that ignores the cold point can put a pathogen into thousands of packs. Equally, the difference between a single-effect and a mechanical vapour recompression evaporator, or between 80 % and 92 % regeneration, is a large share of a plant's energy bill.
Worldwide the same core is taught under the names Dairy Engineering, Food Process Engineering and Dairy and Food Engineering. In India it forms Section 7 of GATE Agricultural Engineering, where a thermal-processing calculation has appeared every year, and subject 38 (Dairy Engineering) of the ICAR PhD entrance, and it sits in Unit VI of the ICAR PG entrance. In the United States it appears in the Processing Systems area of the NCEES PE Agricultural and Biological Engineering exam; in the Philippines in the bioprocess part of the ABE licensure; in Canada's food process engineering syllabus; and in the Nepal and Bangladesh registration and cadre syllabi. Legal time-temperature standards, compositional standards and hygiene rules differ between countries; this manual gives the engineering principles and the widely recognised reference values, and flags where the local regulation must be checked.
Every calculation in a food or dairy plant rests on two conservation laws. Mass is neither created nor destroyed, and energy is neither created nor destroyed. For a steady-state process, which is how most continuous plants are designed and operated, the general balance reduces to: what goes in equals what comes out. For a batch process, or for the start-up of a continuous line, accumulation must be added: input minus output equals accumulation.
The working method is the same whatever the operation. Draw a block diagram and mark every stream crossing the system boundary. Choose a basis, usually 100 kg or 1000 kg of feed, or one hour of operation. Write the total mass balance and then one component balance for each independent component: water, total solids, fat, solids-not-fat (SNF) or a dissolved salt. With n unknown streams you need n independent equations. A tie component, one that passes through a unit unchanged in mass, such as the solids in an evaporator or the dry air in a dryer, gives the shortest route to the answer.
For two streams of known fat content, A and B, mixed to give a product of fat content C, the fat and total balances give the mass of A needed: m_A = M (C − B)/(A − B), where M is the product mass. The Pearson square is simply a graphical form of these two equations. Example: to make 1000 kg of milk at 3.0 % fat from whole milk at 4.5 % and skim milk at 0.05 %, m_whole = 1000 × (3.0 − 0.05)/(4.5 − 0.05) = 662.9 kg, and skim = 337.1 kg.
The same algebra governs cream separation. From 1000 kg of whole milk at 4.0 % fat, separated into cream at 40 % and skim at 0.05 %, cream = 1000 × (4.0 − 0.05)/(40 − 0.05) = 98.9 kg. A student who ignores the fat lost in the skim gets 100 kg, a small but examinable error. In practice in-line standardisation is done continuously by measuring the fat of the cream and skim streams and remixing a controlled cream flow into the skim, with surplus cream sent to butter or cream products.
An energy balance for a steady flow process with negligible kinetic and potential energy change states that the heat added equals the change in enthalpy of the streams. For a liquid heated without phase change, Q = m cp (T2 − T1), where m is the mass flow rate (kg/s) and cp the specific heat (kJ/kg K). Specific heat of milk is close to 3.9 kJ/kg K and falls as the solids content rises; for foods in general it can be estimated from composition (Choi–Okos and Siebel type equations are the common textbook correlations).
Worked example. Heating 2000 kg/h of milk from 4 °C to 72 °C with cp = 3.9 kJ/kg K requires Q = (2000/3600) × 3.9 × 68 = 147.3 kW. Forgetting to convert per hour to per second gives 530,400 kJ/h, a correct energy rate but in a unit that fails if the answer is asked in kW. Energy balances also expose savings: regeneration in a pasteuriser, vapour recompression in an evaporator and exhaust-air heat recovery in a spray dryer are all justified by comparing the enthalpy of streams leaving the plant with the energy purchased.
Liquid foods are moved through stainless-steel sanitary pipework by pumps. The Reynolds number, Re = ρvD/μ, compares inertial and viscous forces. For flow in a circular pipe, Re below about 2100 is laminar and above about 4000 is turbulent, with a transition range between; textbooks differ slightly on the exact limits. Milk at 2 m/s in a 50 mm bore pipe, with density 1030 kg/m³ and viscosity 2 mPa s, has Re = 1030 × 2 × 0.05/0.002 = 51,500, well into turbulence. Viscous products such as concentrates, cream, ice-cream mix and purées often flow laminar.
The Darcy–Weisbach equation gives frictional head loss: h_f = f (L/D) (v²/2g), where f is the Darcy friction factor. In laminar flow f = 64/Re, which reproduces the Hagen–Poiseuille result Δp = 32 μ L v/D². In turbulent flow f depends on Re and relative roughness and is read from the Moody chart or the Colebrook equation. Some books use the Fanning factor, which is one quarter of the Darcy factor; mixing the two is a classic error. Fittings, valves and plate heat exchangers add minor losses expressed as K v²/2g or as an equivalent length.
Two families of pumps are used. Centrifugal pumps add energy through an impeller; they suit low-viscosity liquids such as milk, whey and CIP solutions, give smooth flow, and deliver less as the system head rises, following the pump characteristic curve. The operating point is where the pump curve crosses the system curve (static head plus friction losses that rise with the square of flow in turbulent flow). Centrifugal pumps are poor at viscous products and can damage fat globules or incorporate air if run at high speed with an open suction.
Positive-displacement pumps (rotary lobe, gear, piston, progressive cavity and diaphragm types) move a fixed volume per revolution, so flow is nearly independent of discharge pressure. They handle viscous and particulate foods gently and act as timing pumps: in an HTST pasteuriser the timing pump fixes the maximum flow and therefore the minimum holding time. Because a positive-displacement pump will build pressure until something fails, it needs a relief or bypass arrangement.
Pump power follows directly from the energy equation: hydraulic power = ρ g Q H, and shaft power = ρ g Q H/η. For 10 m³/h of milk (1030 kg/m³) against a total head of 25 m at 65 % efficiency, shaft power = 1030 × 9.81 × (10/3600) × 25/0.65 = 1.08 kW. Multiplying by efficiency instead of dividing gives 0.46 kW, the most common slip. Net positive suction head available must exceed that required, especially for hot liquids near their boiling point, or cavitation occurs.
Rheology is the study of how materials deform and flow. A Newtonian fluid has shear stress proportional to shear rate, τ = μ γ̇; milk, clarified juices, dilute sugar solutions and oils are close to Newtonian. Many foods are not. The power-law (Ostwald–de Waele) model, τ = K γ̇ⁿ, describes shear-thinning (pseudoplastic, n < 1) fluids such as fruit purées, concentrated milk and many sauces, and the rarer shear-thickening (dilatant, n > 1) fluids such as some concentrated starch suspensions. K is the consistency coefficient (Pa sⁿ) and n the flow behaviour index.
A Bingham plastic needs a yield stress τ₀ before it flows, then flows linearly: τ = τ₀ + μ_p γ̇. The Herschel–Bulkley model combines both: τ = τ₀ + K γ̇ⁿ. Time-dependent behaviour is also seen: thixotropic foods (some yoghurts) thin with time of shearing, rheopectic ones thicken.
For a power-law fluid the apparent viscosity is μ_a = K γ̇ⁿ⁻¹. With K = 5 Pa sⁿ, n = 0.4 and γ̇ = 100 s⁻¹, μ_a = 5 × 100⁻⁰·⁶ = 0.32 Pa s. Given two points on a flow curve, n = log(τ₂/τ₁)/log(γ̇₂/γ̇₁); for τ = 10 Pa at 10 s⁻¹ and 40 Pa at 100 s⁻¹, n = 0.60 and K = 2.5 Pa sⁿ. Viscosity of liquids falls as temperature rises, so product temperature must be specified when pumps and heat exchangers are sized. Instruments include capillary, rotational (concentric cylinder, cone-and-plate) and falling-ball viscometers.
For a thin flat wall, 1/U = 1/h₁ + x/k + 1/h₂ + R_f, where h₁ and h₂ are the film coefficients on each side, x/k the wall conduction resistance and R_f the fouling resistance. With h₁ = 3000 and h₂ = 5000 W/m² K and a 0.6 mm steel plate of k = 16 W/m K, U = 1752 W/m² K; adding a fouling resistance of 0.0002 m² K/W drops U to 1297 W/m² K, a fall of about a quarter. For a tube, the resistances are referred to one surface area and the wall term becomes logarithmic.
Fouling is the defining problem of dairy heat exchange. When milk is heated above about 70 °C, whey proteins (chiefly β-lactoglobulin) denature and deposit together with calcium phosphate. Deposits grow fastest in the hottest sections, raise the pressure drop, lower U and eventually force a stop for cleaning. Run length between cleans is therefore a design variable, and it is why UHT plants carry an extra fouling allowance and are cleaned more often.
Because the temperatures change along the exchanger, the correct mean is the log mean temperature difference: LMTD = (ΔT₁ − ΔT₂)/ln(ΔT₁/ΔT₂), where ΔT₁ and ΔT₂ are the temperature differences at the two ends. Worked example: hot water cools from 80 to 60 °C while milk warms from 10 to 55 °C. In counter-current flow the end differences are 25 and 50 K, so LMTD = 36.1 K. In co-current (parallel) flow they are 70 and 5 K, so LMTD = 24.6 K. The arithmetic mean, 37.5 K, overstates the driving force. For 1 kg/s of milk (cp 3.9) the duty is 175.5 kW, so with U = 2000 W/m² K the counter-current area is 2.43 m², the parallel-flow area 3.56 m². Counter-current flow also allows the cold outlet to exceed the hot outlet temperature, which is impossible in parallel flow; here the milk leaves at 55 °C, just below the water outlet, and the parallel arrangement is barely feasible.
When outlet temperatures are unknown, the effectiveness–NTU method is quicker. Effectiveness ε is the actual heat transfer divided by the maximum possible, C_min (T_h,in − T_c,in). NTU = U A/C_min, and C_r = C_min/C_max. For counter-current flow with C_r = 1, which is nearly the case in a milk-to-milk regenerator, ε = NTU/(1 + NTU). When one fluid condenses or boils, C_r = 0 and ε = 1 − e^(−NTU) for any arrangement. Multi-pass and cross-flow arrangements use a correction factor F applied to the counter-current LMTD.
The plate heat exchanger (PHE) is the workhorse of liquid dairy processing. Corrugated stainless-steel plates, sealed with elastomer gaskets and clamped in a frame, form thin alternate channels for product and service fluid. The corrugations induce turbulence at low Reynolds numbers, giving high U values, a close temperature approach and a compact, expandable unit that opens for inspection. Several sections (regeneration, heating, cooling, chilling) share one frame, separated by connector plates. Limits are pressure, the temperature tolerance of gaskets, and particulates or fibres that block narrow gaps.
Tubular exchangers (tube-in-tube, shell-and-tube or multi-tube) tolerate higher pressures and temperatures and pass pulps and small particles, so they are preferred for UHT milk with long run times, juices with pulp and viscous concentrates. Their U is lower and they need more area, and they cannot be opened for visual inspection.
The scraped-surface heat exchanger (SSHE) has a cylinder with a rotating dasher carrying blades that continually scrape product off the heat-transfer wall. It handles very viscous, sticky or particulate foods and crystallising products: margarine and shortening, ice-cream freezing, heating of purées and sauces, and aseptic processing of particulates. Heat transfer is kept high by the scraping rather than by fluid turbulence, at the cost of high capital and power. Choosing among the three is a trade-off among viscosity, particles, pressure, fouling tendency, cleanability and cost.
Heat kills microorganisms and spores in a pattern that, for design purposes, is treated as first order: at constant temperature the number of survivors falls by the same fraction in each equal time interval. Plotting log N against time gives a straight line, the survivor curve. Its negative reciprocal slope is the decimal reduction time, D: the time at a stated temperature to reduce the population by 90 %, that is one log cycle. Thus log(N₀/N) = t/D, or t = D log(N₀/N).
Example. If a count falls from 10⁶ to 10² in 8 minutes at constant temperature, four log reductions took 8 minutes and D = 2 min. The first-order rate constant is related by k = 2.303/D. Real survivor curves may show shoulders (activation or clumping) and tails (resistant sub-populations); these are why processes carry safety margins and why validation uses the actual product.
D depends on the organism, the medium (pH, water activity, fat, salt) and temperature. Values are therefore always quoted as D_T, with the temperature, and should be taken from validated literature for the specific product. The standard reference organism for low-acid canned foods is Clostridium botulinum, whose D at 121.1 °C is commonly quoted in textbooks as about 0.21 min.
The z value is the temperature change that changes D by a factor of ten. Plotting log D against temperature (the thermal death time curve) gives a line whose negative reciprocal slope is z: z = (T₂ − T₁)/log(D₁/D₂). If D is 5 min at 110 °C and 0.5 min at 120 °C, z = 10 °C. With z known, D at another temperature follows from D_T = D_ref × 10^((T_ref − T)/z); for D₁₁₅ = 4 min and z = 10 °C, D₁₂₁.₁ = 0.98 min.
Bacterial spores typically have z values around 10 °C, while quality attributes such as colour, vitamins and texture have much larger z values, commonly 25–50 °C, though the figure varies by attribute and food. That difference is the physical basis of high-temperature short-time (HTST) and UHT processing: raising the temperature accelerates spore destruction much more than quality loss, so a short hot process gives the same sterility with less damage.
Two related measures appear in examinations. The temperature quotient Q₁₀ is the ratio of rates for a 10 °C rise, Q₁₀ = 10^(10/z); z = 10 °C gives Q₁₀ = 10, z = 25 °C gives about 2.5. The Arrhenius model, k = A exp(−E_a/RT), is used by chemists and in research; over the narrow ranges used in processing, the z and Arrhenius descriptions give similar predictions, but they are not identical over wide ranges.
The F value is the equivalent time at a reference temperature that delivers the same lethality as the actual process. For sterilisation of low-acid foods, the reference is 121.1 °C (250 °F) with z = 10 °C, and the F value under these conditions is written F₀. The lethal rate at any temperature is L = 10^((T − T_ref)/z), so one minute at 111.1 °C is worth 0.1 min at 121.1 °C. Thirty minutes at a constant 115 °C gives F = 30 × 10^((115 − 121.1)/10) = 7.4 min. Getting the sign of the exponent wrong gives 122 min, an answer that should look absurd.
For a real process the product temperature changes with time, so F = ∫ L dt, evaluated numerically from heat-penetration data by the trapezoidal or Simpson rule (the general method, Bigelow). Ball's formula method uses heating-rate parameters f_h and j instead; f_h is the time for the heat-penetration curve to cross one log cycle.
The 12D concept (botulinum cook) requires a process that would reduce C. botulinum spores by twelve log cycles: F₀ = 12 × 0.21 ≈ 2.5 min. Commercial processes for low-acid foods typically deliver more than this because spoilage organisms are more heat resistant than C. botulinum. Spoilage probability follows from the same equation: N = N₀ × 10^(−F/D). With 10³ spores per can, D = 1 min and F = 6 min, N = 10⁻³, meaning about one can in a thousand is expected to spoil, or 50 cans in a batch of 50,000. Acid foods (pH at or below 4.6) do not support growth of C. botulinum and receive milder processes.
Pasteurisation is a heat treatment that destroys vegetative pathogens and reduces spoilage organisms so that the product is safe and keeps for a limited time under refrigeration. It does not produce a sterile product and does not destroy bacterial spores. For milk, the historical target organisms were Mycobacterium tuberculosis and later Coxiella burnetii, the more heat-resistant of the two.
The two classic minimum combinations, recognised in the Codex Alimentarius code of hygienic practice for milk and in most national codes, are 63 °C for 30 minutes (low temperature, long time: LTLT or batch, vat or holder pasteurisation) and 72 °C for 15 seconds (high temperature, short time: HTST or continuous flow). National codes list further higher-temperature, shorter-time equivalents and add margins for cream, flavoured and high-solids products; the engineer applies the local legal requirement. Plants commonly run a little above the minimum to allow for control variation.
Adequacy of pasteurisation of milk is checked by the alkaline phosphatase test: this natural enzyme is slightly more heat resistant than the target pathogens, so a negative result shows the minimum treatment was reached. Lactoperoxidase is inactivated only at higher temperatures, around 80 °C; a positive lactoperoxidase result shows the milk has not been over-heated.
A typical HTST plant has a balance tank with a float valve to hold a constant head on the pump; a regeneration section in which cold raw milk is warmed by hot pasteurised milk; a heating section with hot water; a holding tube; a flow diversion valve; the cooling side of the regenerator; and a final cooling section with chilled water or glycol. A separator, standardiser and homogeniser are usually placed between regeneration and heating, where the milk is warm.
Safety devices enforce the legal minimum. A positive-displacement timing pump or a flow controller with a meter limits the maximum flow and therefore guarantees the minimum holding time. A recording and controlling thermometer at the end of the holding tube operates the flow diversion valve: if the temperature falls below the set point, the valve diverts milk back to the balance tank instead of forward. In the regenerator, the pasteurised milk must be kept at a higher pressure than the raw milk (a booster pump or a back-pressure arrangement is used), so that any pinhole leak in a plate carries pasteurised milk into raw, never the reverse. The holding tube slopes upward in the direction of flow so that air is purged and the tube runs full.
Regeneration efficiency is the temperature rise achieved in regeneration divided by the total temperature rise needed: raw milk entering at 5 °C and leaving regeneration at 60 °C, with 72 °C pasteurisation, gives (60 − 5)/(72 − 5) = 82 %. Modern plants reach over 90 %, cutting the heating and cooling loads to about a tenth of what they would be without regeneration.
The holding tube must hold the fastest-moving particle for the full legal time, not just the average. Mean velocity is v = Q/A. For laminar Newtonian flow the maximum velocity is twice the mean; for turbulent flow textbooks use a ratio of about 1.2 (often expressed as mean/maximum ≈ 0.82). Required length L = v_max × t.
Example. A viscous mix flows laminar at 10,000 L/h through a tube of 47.5 mm bore, to be held for 15 s. Flow = 2.778 × 10⁻³ m³/s, area = 1.772 × 10⁻³ m², mean velocity = 1.57 m/s. The fastest particle moves at 3.14 m/s, so L = 47.0 m. Designing on the mean velocity gives 23.5 m, which would under-process the centre of the stream by half. For thin milk in turbulent flow in the same tube the length is about 28.7 m.
In practice holding time is verified at commissioning and by regulators using a salt or dye injection test, timing the first appearance of the tracer at the tube outlet at the maximum pump capacity. Any change of tube length, pump speed, impeller or pipe diameter requires re-verification. Heat treatment also continues in the regenerator and cooler after the tube, but this is not credited in the legal holding time.
Sterilisation in the container is the classical canning route. Food is filled and sealed in cans, glass jars, retort pouches or bottles, then heated in a retort, a pressure vessel, with saturated steam, steam–air mixtures or water, commonly at 110–130 °C. The process is called commercially sterile: no microorganisms capable of growing in the product under normal storage remain, though a small number of dormant spores may survive.
Heat reaches the slowest-heating point, the cold point, by conduction in solid packs (the cold point is near the geometric centre) and by convection in liquids (the cold point is lower on the axis). Heat-penetration tests with thermocouples at the cold point are used to calculate F₀. Temperature-distribution tests confirm that all positions in the retort reach temperature. Venting removes air, since an air–steam mixture is cooler than pure steam at the same pressure. Agitating (rotary) retorts shorten processes for liquid and semi-liquid products by forced convection. Hydrostatic towers and continuous rotary cookers are used for high throughput. Sterilised milk made in the bottle shows cooked flavour and browning from the Maillard reaction, which is why continuous UHT largely replaced it.
Ultra-high temperature (UHT) treatment heats a liquid continuously to a high temperature for a few seconds, then packs it aseptically. Typical milk conditions are around 135–150 °C for one to several seconds, with the legal minimum depending on the country. Because spore destruction has a small z value and chemical changes a large one, UHT reaches commercial sterility with far less browning and cooked flavour than in-container sterilisation.
Indirect UHT uses plate, tubular or scraped-surface exchangers. It allows high regeneration, often around 85–90 %, but heating and cooling are relatively slow and fouling limits run time.
Direct UHT brings the product into contact with steam: steam injection (steam into product) or steam infusion (product falls through a steam-filled vessel). Heating is almost instantaneous. The condensed steam dilutes the product and must be removed by flash cooling in a vacuum vessel, which evaporates an equal mass of water, provided the vacuum is set so that the outlet temperature equals the inlet temperature before steam contact. Steam used must be culinary quality. Energy balance: heating milk from 75 to 140 °C (cp 3.9 kJ/kg K) with steam giving up about 2200 kJ/kg needs 3.9 × 65/2200 = 0.115 kg steam per kg milk, about 10 % dilution. Direct systems give a milder flavour but recover less heat.
UHT gives a sterile product only if it is packed aseptically: the product, the package and the environment are sterile at the moment of filling. Packaging material is commonly sterilised with hydrogen peroxide followed by heat to remove residues, or by other chemical, heat or radiation methods; the filling zone is kept under sterile overpressure. Packages are typically paperboard–polymer–aluminium laminates, bottles or pouches that exclude light and oxygen.
Evaporation concentrates a liquid food by boiling off part of the water, usually as a step before drying (milk powder, whey powder) or to make a concentrate (condensed milk, tomato paste, juice concentrate). It is done under vacuum so the product boils at a low temperature, often 40–70 °C, which protects heat-sensitive components.
Mass balance. Feed F with solids fraction x_F becomes concentrate P with fraction x_P and vapour W. Solids balance: F x_F = P x_P. For 5000 kg/h of milk at 12 % total solids concentrated to 45 %: P = 1333 kg/h and W = 3667 kg/h. A common error is to subtract percentages directly, which gives 1650 kg/h.
Energy balance. Steam S condensing in the calandria gives up latent heat λ_s. It must heat the feed to the boiling point and supply latent heat of vaporisation λ_v: S λ_s = F cp (T_b − T_F) + W λ_v (heat of concentration and losses neglected). With the feed entering at its boiling point, λ_v = 2358 kJ/kg and λ_s = 2202 kJ/kg, S = 3667 × 2358/2202 = 3926 kg/h, and steam economy (kg water evaporated per kg steam) = 0.93. If the feed enters at 20 °C and boils at 60 °C (cp 3.9), S rises to 4281 kg/h and economy falls to 0.86. A single effect therefore always has an economy just below one.
Heat transfer area follows from Q = U A ΔT, where ΔT is the steam temperature minus the boiling temperature. Boiling-point elevation, the rise of boiling point caused by dissolved solids, reduces the available ΔT; it is small for milk but large for concentrated sugar and salt solutions.
Steam economy is improved by reusing the vapour. In multiple-effect evaporation the vapour from the first effect heats the second, which runs at lower pressure and so boils at lower temperature, and so on. Each kilogram of steam evaporates roughly one kilogram of water in each effect, so the economy is approximately (but always somewhat less than) the number of effects. The total temperature difference between the live steam and the last effect's boiling point is divided among the effects, so the area per effect grows as effects are added. The optimum number is a balance between steam cost and capital.
Feed arrangements: forward feed (feed and steam flow in the same direction, easy and common for heat-sensitive dairy products because the most concentrated product sees the lowest temperature), backward feed (useful for viscous concentrates, but needs pumps between effects), and mixed or parallel feed.
Vapour recompression raises the pressure, and therefore the condensing temperature, of the vapour so it can heat the same effect. Thermal vapour recompression (TVR) uses a steam jet ejector with high-pressure motive steam; it adds roughly the effect of one extra stage. Mechanical vapour recompression (MVR) uses an electrically driven fan or compressor; it can reach very high equivalent economies and is common in modern milk and whey plants, particularly where electricity is cheap relative to steam.
The falling-film evaporator is the standard for milk and whey. Liquid is distributed at the top of long vertical tubes and flows down as a thin film while boiling; residence time is short (seconds to a minute or so per pass), the temperature difference can be small, and there is no hydrostatic head to raise the boiling point. Even distribution is critical: dry patches foul quickly.
Spray drying turns a liquid concentrate into a powder by atomising it into a stream of hot air. It is the main route for milk, whey, infant formula bases, coffee and many flavours. The stages are: atomisation, contact of spray with hot air, evaporation of moisture, and separation of powder from air.
Atomisers are of two types. Rotary (centrifugal) atomisers throw liquid from a wheel or disc spinning at high peripheral speed; they tolerate variable feed rates and viscous feeds and give a wide chamber. Pressure nozzles force liquid through a small orifice at high pressure; they give coarser, denser particles and suit tall, narrow chambers. Two-fluid nozzles use compressed air and are common in small driers. Air and spray may flow co-currently, which is normal for heat-sensitive foods because the wettest droplets meet the hottest air, or counter-currently, or in mixed flow.
While a droplet still has free water at its surface, it evaporates at close to the wet-bulb temperature of the air, far below the inlet air temperature; this evaporative cooling is why milk entering at air temperatures near 200 °C is not scorched. Once a crust forms, the particle temperature rises towards the outlet air temperature, so the outlet air temperature, not the inlet, largely governs final moisture and heat damage. Inlet and outlet temperatures used for milk vary between plants and products; values of roughly 180–220 °C inlet and 70–95 °C outlet are commonly quoted for single-stage milk drying.
Powder is separated from the exhaust by cyclones, bag filters or wet scrubbers. Many modern plants use two- or three-stage drying: the chamber leaves the powder at a higher moisture, and an internal or external fluid bed completes drying at lower temperature and allows agglomeration. Agglomerated, lecithin-coated powders wet and disperse readily and are sold as instant powders. Fine powder is a dust-explosion hazard, so chambers carry explosion vents or suppression.
A spray drier balance uses the dry solids and the dry air as tie components. Example: 1000 kg/h of concentrate at 48 % total solids is dried to a powder of 3.5 % moisture (wet basis). Solids = 480 kg/h; powder = 480/0.965 = 497.4 kg/h; water evaporated = 502.6 kg/h. The wrong-basis error treats 3.5 % as dry basis and gives 496.8 kg/h, close but not the same.
If inlet air humidity is 0.010 kg water per kg dry air and outlet humidity 0.060, each kilogram of dry air picks up 0.050 kg of water, so the dry-air flow is 502.6/0.050 = 10,052 kg/h. The heat to warm this air from 20 to 200 °C is found with the humid heat c_s = 1.005 + 1.88 H = 1.024 kJ/kg K: Q = 10,052 × 1.024 × 180/3600 = 515 kW, about 3700 kJ per kg of water evaporated. That is well above the latent heat of water, because the exhaust air leaves hot; spray drying consumes much more energy per kilogram of water than evaporation, which is why milk is concentrated in a multiple-effect evaporator as far as practical before the drier.
In drum drying, a thin film of liquid or paste is applied to the outside of a slowly rotating, steam-heated metal drum and scraped off by a doctor blade after most of a revolution. Single-drum and double-drum types exist, fed by dipping, splashing, rollers or nip feeding between two drums. Operated at atmospheric pressure, the product reaches temperatures close to the drum surface, so drum-dried milk shows more heat damage: lower solubility, more browning and a cooked flavour. It is now used where these properties are acceptable or useful: some animal feeds, instant cereals, mashed-potato flakes, and roller-dried milk for chocolate, whose caramelised flavour is valued. Vacuum drum driers reduce heat damage at higher cost.
Mass balance works the same way: 200 kg/h of feed at 40 % solids dried to 4 % moisture gives 83.3 kg/h of product and removes 116.7 kg/h of water. Drying capacity is set by drum area, film thickness, speed and steam pressure.
Milk fat exists as globules, mostly between about 1 and 10 µm in diameter, each surrounded by a natural milk fat globule membrane. Because fat is less dense than the serum, globules rise; at low temperature they also cluster (cold agglutination), so a cream layer forms. Homogenisation reduces the globule size, commonly to below about 1 µm, so creaming becomes too slow to matter during shelf life.
Stokes' law explains why. The rising velocity of a sphere is v = d² (ρ_s − ρ_f) g/(18 μ). Velocity depends on the square of diameter, so reducing diameter from 4 µm to 1 µm cuts creaming speed sixteenfold. Homogenisation also creates a large new fat surface, which is covered mainly by casein micelles and whey proteins. The result: whiter appearance, fuller body and mouthfeel, no cream line, better stability of cream liqueurs, recombined milks, infant formula, ice-cream mix and emulsions generally.
There are costs. Homogenised milk is more sensitive to light-induced off-flavour, and homogenised raw milk undergoes rapid lipolysis because lipase gains access to the new, unprotected fat surface. For this reason milk is homogenised hot and pasteurised immediately, or homogenised after pasteurisation under aseptic conditions for UHT. One widely cited test of homogenisation efficiency, from US regulation, is that after 48 hours of quiescent storage the fat content of the top 100 mL of a quart container does not differ from that of the remainder by more than 10 % of itself; other countries use laboratory methods based on centrifugation or particle size distribution.
A dairy homogeniser is a high-pressure piston pump (three or more plungers) that forces liquid through a narrow adjustable gap between a valve seat and a forcer, then against an impact ring. Pressure energy becomes kinetic energy in the gap; the disruption of globules is explained by turbulence, cavitation and shear, with turbulent eddies generally considered dominant under industrial conditions.
Two-stage homogenisation is standard for milk and cream. The first stage, at the main pressure, breaks the globules; the second stage, at a much lower pressure, typically a small fraction of the first, breaks up clusters of newly formed globules that share casein on their surfaces. Clustering is worst in high-fat products, so two stages matter most for cream and ice-cream mix. Operating pressures for market milk are commonly in the range of about 10–25 MPa, depending on product and required stability.
Homogenisation is done warm, above the melting point of milk fat, commonly around 55–80 °C, since solid fat cannot be dispersed. Partial (stream) homogenisation treats only the cream fraction, which saves energy because the homogeniser flow is much smaller.
Power: the hydraulic power is the pressure drop times the volumetric flow. For 10,000 L/h (2.78 × 10⁻³ m³/s) at 20 MPa, hydraulic power = 55.6 kW; with an overall efficiency of 85 % the motor must supply about 65 kW. Nearly all of this energy ends as heat, raising the product temperature by a few degrees; each 10 MPa is commonly said to raise temperature by about 2–2.5 °C.
Centrifugation replaces gravity in Stokes' law by centrifugal acceleration r ω², where r is the radius and ω the angular speed in rad/s. The settling or rising velocity becomes v = d² Δρ r ω²/(18 μ). The ratio r ω²/g is the relative centrifugal force or 'number of g'. At 0.2 m radius and 6000 rpm (ω = 628 rad/s) it is about 8050 g; at 0.15 m and 7200 rpm it is about 8690 g. A 4 µm fat globule that rises at 0.64 µm/s under gravity (Δρ = 110 kg/m³, μ = 1.5 mPa s) moves at about 5 mm/s in the separator field.
Higher temperature lowers the viscosity of skim milk and raises separation efficiency, so milk is usually separated warm (around 45–55 °C in most plants) after passing through the regenerator; cold separation is possible with larger machines and is used where it protects quality.
The disc-stack (disc-bowl) centrifuge is the standard machine. A stack of conical discs, separated by thin spacers, divides the bowl into many shallow settling channels, so particles travel only a few tenths of a millimetre before reaching a disc surface. This multiplies capacity for a given bowl size. Sigma (Σ) theory expresses the capacity as equivalent to a gravity settling tank of area Σ, allowing scale-up between machines.
In a cream separator, milk enters through distribution holes in the disc stack. Skim milk, the denser phase, moves outward and leaves from the periphery; fat globules move inward and leave as cream near the axis. The fat content of cream is set by throttling the cream or skim outlet; outlet pressure control keeps it steady. Separation efficiency is judged by the fat left in the skim milk, which in good practice is a few hundredths of a percent. Small globules, below about 1 µm, escape into the skim regardless.
Bactofugation uses high-speed centrifuges to remove bacterial spores, which are denser than milk. It is used for cheese milk, where spores of Clostridium tyrobutyricum can cause late blowing in hard and semi-hard cheeses, and for milk for powder and extended-shelf-life products. Removal efficiencies are high but not complete, so it is combined with other measures. Fat recovery follows from the balance: from 1000 kg of milk at 4.0 % fat, cream at 40 % and skim at 0.05 %, cream = 98.9 kg and 98.9 % of the fat is recovered. Flow above rated capacity sharply raises skim fat.
Filtration separates solids from a liquid by passing the liquid through a porous medium. In cake filtration the deposited solids form a cake that becomes the effective filter; in depth filtration particles are trapped within the medium. Flow through the cake follows Darcy's law, and for constant-pressure filtration the volume collected V grows so that time t varies with V² plus a term for the medium resistance (the Ruth equation). A practical result: if medium resistance is negligible, doubling the volume filtered takes four times as long. Compressible cakes, common in foods, have resistance that rises with pressure, so raising pressure gives diminishing returns. Filter aids such as diatomaceous earth keep compressible cakes porous.
Membrane processes separate by size and charge, using pressure as the driving force and cross-flow operation, in which the feed sweeps along the membrane surface to limit build-up. The feed splits into a permeate that passes through and a retentate (concentrate) that does not.
Microfiltration (MF), with pores of roughly 0.1–10 µm, retains bacteria, spores, somatic cells and fat globules, and at the fine end casein micelles; it is used to make extended-shelf-life milk and to separate casein micelles from serum proteins. Ultrafiltration (UF) retains proteins and fat while passing water, lactose and minerals; it is used to concentrate whey protein (WPC), to standardise protein in cheese milk and to make fresh cheeses and high-protein yoghurts. Nanofiltration (NF) passes water and monovalent ions, retaining lactose and divalent ions; it partly demineralises whey. Reverse osmosis (RO) retains nearly all solutes and passes water; it is used to pre-concentrate milk or whey at low energy and to polish evaporator condensate. The operating pressure rises in the order MF < UF < NF < RO because RO must overcome osmotic pressure. The boundaries between classes overlap and pore sizes and cut-offs quoted vary between texts and suppliers.
Key terms: volume concentration factor, VCF = feed volume/retentate volume; permeate flux J in L/m² h; rejection R = 1 − C_p/C_f. From 1000 L of skim milk, removing 800 L of permeate in 4 h through 20 m² gives VCF = 5 and flux = 10 L/m² h. If a solute is at 5.0 % in feed and 0.2 % in permeate, R = 0.96.
Flux falls during operation for two reasons. Concentration polarisation is a reversible build-up of retained solutes in a boundary layer at the membrane; higher cross-flow velocity and temperature reduce it. Fouling is the harder, partly irreversible deposition of proteins, minerals and fat on and within the membrane; it needs chemical cleaning. In UF, flux rises with pressure only until a limiting flux is reached, beyond which more pressure just compacts the gel layer.
Osmotic pressure for dilute solutions follows the van 't Hoff relation π = i C R T. For a 0.1 mol/L non-electrolyte at 25 °C, π = 0.1 × 1000 mol/m³ × 8.314 × 298 K = 248 kPa. Electrolytes dissociate (i ≈ 2 for sodium chloride), roughly doubling the value. RO must operate above the osmotic pressure of the concentrate, which rises as concentration proceeds; this sets the practical upper concentration.
Diafiltration adds water to the retentate during or after concentration to wash out permeable solutes, for example to raise the protein content of a WPC by removing more lactose and minerals. Membrane materials are polymeric (polysulfone, polyethersulfone, polyamide thin-film composites) or ceramic; ceramics tolerate harsher cleaning and temperatures. Spiral-wound modules are compact and cheap; tubular and ceramic modules handle viscous or particle-laden feeds. Chlorine and pH limits of each membrane restrict the cleaning chemicals used.
Freezing preserves food by lowering temperature and by converting water into ice, which lowers water activity in the unfrozen phase. Foods begin to freeze below 0 °C because dissolved solutes depress the freezing point, commonly to between about −0.5 and −3 °C for fruits, vegetables, meat and milk. As ice forms, the remaining solution becomes more concentrated and its freezing point falls further, so freezing occurs over a range of temperatures and some water remains unfrozen even at normal storage temperatures around −18 °C.
A freezing curve at the thermal centre shows cooling of the unfrozen product, often a brief supercooling, a long near-plateau where most latent heat is removed, and then cooling of the frozen product. Fast freezing forms many small ice crystals, largely within cells, giving less drip loss and better texture on thawing; slow freezing forms large extracellular crystals that damage tissue. Recrystallisation during fluctuating storage temperatures coarsens crystals, which is why stable cold-store temperature matters as much as low temperature.
The refrigeration load to freeze a product is the sum of sensible heat above freezing, latent heat and sensible heat below freezing: Q = m[c_u (T_i − T_f) + x_w λ + c_f (T_f − T_e)]. For 1000 kg with 75 % water, cooled from 5 °C to −18 °C with T_f = −2 °C, c_u = 3.6, c_f = 1.9 kJ/kg K and λ = 334 kJ/kg: Q = 25.2 + 250.5 + 30.4 = 306 MJ. The latent term dominates.
Plank's equation estimates the time to freeze a product that starts at its initial freezing point: t_F = (ρ λ/(T_f − T_a)) × (P a/h + R a²/k), where ρ is the density, λ the latent heat of the food (water fraction × 334 kJ/kg), T_f the initial freezing point, T_a the cooling medium temperature, a the characteristic dimension, h the surface heat transfer coefficient and k the thermal conductivity of the frozen food. For an infinite slab P = 1/2 and R = 1/8, with a the full thickness; for an infinite cylinder P = 1/4 and R = 1/16, with a the diameter; for a sphere P = 1/6 and R = 1/24, with a the diameter.
Example. A slab 5 cm thick, ρ = 1000 kg/m³, λ = 266.6 kJ/kg (80 % water), T_f = −1.5 °C, air at −30 °C, h = 30 W/m² K, k = 1.5 W/m K. The first bracket term is 0.5 × 0.05/30 = 8.33 × 10⁻⁴ and the second 0.125 × 0.0025/1.5 = 2.08 × 10⁻⁴; t_F = (1000 × 266,600/28.5) × 1.042 × 10⁻³ = 9744 s ≈ 2.7 h. Using the half-thickness for a with the slab constants gives 1.2 h, a frequent mistake. A sphere of 5 cm diameter under the same conditions freezes in about 0.9 h, because it has more surface per unit volume.
The two terms show the controls: the P a/h term is the surface resistance, reduced by faster air or immersion; the R a²/k term is internal conduction, reduced only by thinner product. Plank's assumptions are that freezing starts with the whole product at the freezing point, latent heat is released at a single temperature, and thermal conductivity of the frozen layer is constant. It ignores precooling and subcooling, so it underestimates total time; modified methods (Nagaoka, Cleland and Earle, Pham) correct for this.
Freezers are grouped by the heat-transfer route. Air-blast freezers (tunnel, spiral, belt) circulate cold air, typically −30 to −40 °C at several metres per second; they are flexible but surface coefficients are moderate. Fluidised-bed freezers suspend small particulate foods such as peas, diced vegetables and shrimps in upward air, giving individually quick frozen (IQF) products. Plate (contact) freezers press flat packages or blocks between refrigerated plates; they give high heat-transfer rates for regular shapes such as fish blocks and meat cartons. Immersion and cryogenic freezers use liquid nitrogen or carbon dioxide; they are very fast, with low capital cost but high running cost for the cryogen, and they are used for high-value or delicate products. Ice-cream is partly frozen in a continuous scraped-surface freezer, which also whips in air (overrun), then hardened. Thawing is slower than freezing because the thawed outer layer conducts heat less well than ice.
Cleaning-in-place (CIP) cleans pipework, tanks, heat exchangers and other closed equipment by circulating cleaning solutions without dismantling. It is reproducible, can be automated and recorded, uses less water and chemicals than manual cleaning, and keeps workers away from hot caustic.
Dairy soils are fat, protein, lactose and minerals. Cold surfaces carry mainly fat and protein films; heated surfaces carry denatured protein and calcium phosphate (milkstone). Alkaline detergents, chiefly sodium hydroxide, saponify fat and dissolve protein; acid detergents, chiefly nitric or phosphoric acid, dissolve mineral deposits. A typical heated-surface cycle is: pre-rinse with water to remove loose soil; hot caustic circulation; intermediate rinse; acid circulation; final rinse; then sanitising, with hot water or a chemical sanitiser, before the next production. Cold-surface circuits may use caustic only, with a periodic acid wash. Concentrations and temperatures depend on the soil, detergent formulation and equipment; caustic around 0.5–2 % at roughly 70–85 °C is widely used in dairies, but the engineer follows the validated procedure for the plant.
The factors of cleaning are summarised in Sinner's circle: chemical action, temperature, mechanical action and time. Reducing one must be compensated by increasing another. In pipes, mechanical action comes from turbulent flow; a mean velocity of at least 1.5 m/s is the widely used rule. For a 48 mm bore pipe that needs 1.5 × π/4 × 0.048² × 3600 = 9.8 m³/h. Tanks are cleaned by spray balls or rotating jet heads, whose coverage and flow rate must be checked; a riboflavin coverage test is common at commissioning.
Making up solution is a mass balance: 2000 L of 1.5 % (w/v) caustic needs 30 kg NaOH, which is 60 kg of 50 % (w/w) liquid caustic, or 39.5 L at a density of 1.52 kg/L.
A single-use CIP system makes fresh solution for each clean and discards it; it suits heavily soiled equipment such as UHT plants and evaporators. A recovery (reuse) system stores caustic and acid in tanks, returns the solutions after each clean, tops up concentration and reuses them, saving chemicals, water and heat; it suits lightly soiled circuits such as milk reception and storage. Many plants combine both. A CIP station has tanks, a supply pump, a heat exchanger, conductivity and temperature sensors, a return pump (often a self-priming or liquid-ring type for tank circuits) and valves that route each circuit.
Conductivity measurement controls concentration and detects the interfaces between water and chemical at changeovers. The control system records time, temperature, flow and conductivity for every clean, forming part of the plant's HACCP verification. Verification of cleaning uses visual inspection where possible, ATP bioluminescence swabs, microbiological swabs and rinse-water tests.
Hygienic design makes equipment cleanable and prevents contamination. Guidelines from the European Hygienic Engineering & Design Group (EHEDG), 3-A Sanitary Standards in the United States and ISO 14159 set out the principles, which are widely followed internationally.
Materials in product contact must be corrosion resistant, non-toxic, non-absorbent and cleanable: austenitic stainless steels (commonly AISI 304 and the molybdenum-bearing 316 grade for more corrosive duties, such as with chlorides) and approved elastomers and plastics. Surfaces must be smooth; a surface roughness Ra of 0.8 µm or better is the commonly specified limit for product-contact surfaces. Welds are ground smooth and crevice-free, and orbital welding is preferred for pipework.
Geometry: no dead legs where liquid stagnates (branches are kept short relative to diameter), no crevices, internal corners with generous radii, self-draining lines sloped to drain, no threads in product contact, and hygienic couplings and valves. Mix-proof (double-seat) valves with a leakage chamber allow product and cleaning solution to be in adjacent lines safely. Equipment must be inspectable and, where CIP cannot reach, dismantled easily for manual cleaning.
Use this subject to learn the method of every process calculation: draw the block diagram, pick a basis and tie component, write balances, then add rates. In laboratory classes you will measure a survivor curve and calculate D, run a plate heat exchanger and calculate U and LMTD, and test a cream separator by the fat in the skim. For GATE and ICAR, drill F and D calculations until the lethal-rate formula is automatic, and practise converting kg/h to kg/s and wet to dry basis. Visit a dairy plant and trace a litre of milk from reception to the filler, naming each unit operation.
A plant engineer five years in runs and troubleshoots these operations daily. Typical tasks: checking that the HTST timing pump, diversion valve and recorder work and passing the holding-time test; investigating falling regeneration or rising pressure drop from fouling; setting CIP cycles and conductivity set points; calculating steam demand and economy after an evaporator change; tuning drier outlet temperature to hit powder moisture; and diagnosing high skim fat from a separator. Decisions are about run length, cleaning frequency, energy use and whether a product deviation must be held or released.
A senior engineer or plant manager with 25 years plans capacity and capital. Decisions include choosing direct or indirect UHT for a new line, adding effects or MVR to an evaporator versus buying steam, sizing a spray drier with fluid beds for instant powder, specifying hygienic design to EHEDG or 3-A in tenders, designing CIP systems with recovery, and setting water and energy targets. They answer to regulators in audits of pasteurisation and aseptic processes, approve thermal process changes, and weigh product quality against safety margins.
A researcher with 25 years of experience works where the textbook models break down: non-log-linear inactivation kinetics (Weibull and other models) and their regulatory acceptance, fouling mechanisms and anti-fouling surfaces, residence-time distribution of particulates in aseptic processing, novel heat treatments such as ohmic and microwave heating and the severity indices (B*, C*) used to compare UHT plants, membrane fouling and new membrane materials, low-energy dewatering, glass transition and stickiness in drying, and life-cycle assessment of dairy processing. They design validation studies, supervise students, and advise standards bodies on process equivalence.
This manual does not yet cover: detailed Ball formula tables and the full procedure for f_h/j process calculation; aseptic processing of particulates (residence-time distribution and particle-centre lethality); non-thermal and novel processes (high pressure, pulsed electric field, ohmic and microwave heating); non-log-linear inactivation models; steam generation, boiler and utility design; detailed psychrometric-chart use for drier design; dairy product manufacture (butter, cheese, ice cream, condensed milk) beyond the unit operations; and country-by-country legal time-temperature and compositional standards, which must be checked in the local regulation. Typical operating ranges quoted (UHT temperatures, drier temperatures, homogenising pressures, CIP concentrations) are indicative textbook values that vary with plant and product.
Original practice questions written to the syllabus and pattern of the exams named in each question's examTags. They are not past papers and do not reproduce any board's questions. Every numerical answer was computed by script; legal time-temperature standards quoted are the widely recognised reference values and must be checked against local regulation.
These notes are free to read here, with no account.
What the Laisha app adds:
Free for ten days, then a subscription. We would rather write that here than have you find it later.