Heat and Mass Transfer cover
Interactive book · Semester 5-6

Heat and Mass Transfer

Find the hottest point in a fuel rod, see why thin insulation can raise heat loss, size a heat exchanger by LMTD and NTU, and watch water climb the boiling curve to burnout

21 chapters, 21 interactives, few words. Every chapter is mapped to the Anna University, VTU, SPPU, Mumbai University and KTU units it covers: modes of heat transfer and thermal resistance, the conduction equation and boundary conditions, cylinders, spheres and the critical radius of insulation, heat generation, fins, lumped and Heisler-chart transient conduction, finite differences, thermal boundary layers, forced convection over bodies and inside tubes, free convection, boiling and condensation, heat exchangers by LMTD and effectiveness-NTU, radiation laws, view factors and shields, and diffusion and convective mass transfer.

  • AUME3691Heat and Mass Transfer
  • VTUBME601Heat Transfer
  • SPPU302042Heat and Mass Transfer
  • MUMEC502Thermal Engineering (Modules 1-4: heat transfer)
  • KTUPCMET503Heat and Mass Transfer

By Skill-Lync · Updated 17 Nov 2026 · 21 chapters · 21 interactives · free, no sign-up

Chapter 1

Modes of heat transfer, thermal resistance and composite walls

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 1SPPU 302042 Unit 1MU MEC502 Module 1KTU PCMET503 Module 1

What are the three modes of heat transfer and the law behind each?

Conduction passes energy from molecule to molecule inside a solid or a still fluid; Fourier's law says the flux is proportional to the temperature gradient, q = −k dT/dx. Convection is conduction plus fluid motion at a surface; Newton's law of cooling lumps it into q = h(Ts − T∞). Radiation needs no medium: every surface emits εσT⁴. A pipe at 100 °C with ε = 0.9 in a room at 30 °C radiates a net 558 W/m², comparable to what natural convection removes, which is why radiation is never ignored for hot bare surfaces.

How much does thermal conductivity vary between materials?

Over four orders of magnitude. Copper conducts at 401 W/mK and carbon steel at 54, but stainless steel 304 manages only 14.4 because its alloying scatters the free electrons that carry heat in metals. Common brickwork sits at 0.6-1.0, water at 0.606, polyurethane foam at 0.03 and still air at 0.026. Good insulators trap air in small cells so it cannot circulate. Thermal diffusivity α = k/(ρc) is the related property for changing temperatures: it tells how fast a temperature change spreads, not how much heat flows.

How does the electrical analogy turn a wall into a circuit?

Treat temperature difference like voltage, heat flow like current, and each layer or film as a resistor: L/(kA) for a layer, 1/(hA) for a film. Layers in series add; paths side by side combine like parallel resistors. The overall coefficient U = 1/ΣR (per m²) collects everything. Then q = U ΔT, and the temperature at any joint is the hot temperature minus q times the resistances passed. In a furnace wall, the insulating brick supplies 57% of the total resistance although it is the thinnest layer.

Composite wall: thermal resistance network, heat flux and interface temperatures for three layers with convection on both sides: interactive loads as you scroll

What is contact resistance, and how is conduction enhanced across joints?

Two pressed surfaces touch only at the peaks of their roughness; the gaps hold air, so heat squeezes through small spots and the temperature steps at the joint. Measured contact resistances mostly lie between 0.000005 and 0.0005 m²K/W. That is nothing beside 100 mm of foam (3.3 m²K/W) but can be 27 times the resistance of a 1 mm steel sheet. Designers reduce it with higher clamping pressure, smoother faces and thermal interface materials (greases, pads), and spread heat with high-k inserts, heat spreaders and fins.

Predict, then check

A cold-store panel is 1 mm steel, 100 mm polyurethane foam and 1 mm steel, with air films on both sides. Which part controls the heat leak?

Worked examples

Three-layer furnace wall · Fire-clay brick 230 mm (k 1.4), insulating brick 115 mm (k 0.15), red brick 230 mm (k 0.8); furnace gas 1,000 °C with h = 50 W/m²K, room 30 °C with h = 10 W/m²K (typical)
  1. Resistances per m²: 1/50 = 0.020, 0.23/1.4 = 0.1643, 0.115/0.15 = 0.7667, 0.23/0.8 = 0.2875, 1/10 = 0.100; ΣR = 1.3385 m²K/W
  2. q = 970/1.3385 = 724.7 W/m²; U = 0.747 W/m²K
  3. Hot face 985.5 °C; firebrick to insulating brick 866.4 °C; insulating to red brick 310.8 °C; outer face 102.5 °C
  4. The insulating brick carries 57% of the resistance, so it takes 556 K of the drop

Answer: q ≈ 725 W/m²; the insulating brick's hot face is at about 866 °C, which its rating must exceed

Heat leak into a cold store · Panels of 1 mm steel (k 54), 100 mm polyurethane foam (k 0.03), 1 mm steel; outside 35 °C, h = 20; store −20 °C, h = 10 W/m²K (typical); 120 m² of wall and roof
  1. ΣR = 0.05 + 0.0000185 + 3.3333 + 0.0000185 + 0.1 = 3.4834 m²K/W
  2. q = 55/3.4834 = 15.79 W/m²; over 120 m²: 1,895 W
  3. Add the upper-end contact resistance 0.0005 m²K/W at both steel-foam joints: q falls by only 0.029%

Answer: About 1.9 kW of heat leaks in, set almost entirely by the foam

A concrete column in a brick wall (parallel paths) · 10 m² of 230 mm wall: 9 m² brick (k 0.8) and 1 m² dense concrete column (k 1.4, middle of 1.0-1.8); outside 40 °C, h = 20; room 24 °C, h = 8 W/m²K (typical); heat taken to flow straight through each path
  1. Brick path ΣR = 0.05 + 0.2875 + 0.125 = 0.4625; column path 0.05 + 0.1643 + 0.125 = 0.3393 m²K/W
  2. Brick: 9 × 16/0.4625 = 311.4 W; column: 1 × 16/0.3393 = 47.2 W; total 358.5 W
  3. An all-brick wall would pass 345.9 W

Answer: The column is 10% of the area but carries 13% of the heat, raising the gain by 3.6%

Chapter 2

The general conduction equation and boundary conditions

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 1SPPU 302042 Unit 1MU MEC502 Module 1KTU PCMET503 Module 1

Where does the general conduction equation come from?

From an energy balance on a tiny cube dx by dy by dz. Heat conducted in through each face minus heat conducted out of the opposite face, plus heat generated inside (q‴ dV), must equal the rate at which the cube stores energy, ρc ∂T/∂t dV. Writing each face flux with Fourier's law and letting the cube shrink gives the equation below. With constant k it becomes ∇²T + q‴/k = (1/α) ∂T/∂t, where α = k/ρc.

What changes in cylindrical and spherical coordinates?

The balance is the same, but the area that heat crosses grows with radius: 2πrL for a cylinder, 4πr² for a sphere. So the radial term is (1/r) ∂/∂r(k r ∂T/∂r) in a cylinder and (1/r²) ∂/∂r(k r² ∂T/∂r) in a sphere, with angle terms carrying 1/r² (and 1/(r² sin²θ) for the sphere's azimuth). For steady radial conduction without generation, r dT/dr is constant in a cylinder, giving a logarithmic profile, and r² dT/dr is constant in a sphere, giving a 1/r profile.

What are the Fourier, Poisson and Laplace forms?

They are the general equation with terms dropped. No generation, unsteady: ∇²T = (1/α) ∂T/∂t, the Fourier (diffusion) equation. Steady with generation: ∇²T + q‴/k = 0, Poisson's equation. Steady without generation: ∇²T = 0, Laplace's equation. In one dimension Laplace's equation says d²T/dx² = 0, so the profile in a plane wall is a straight line, and any curvature you measure in a steady wall means heat is being generated or absorbed inside it.

What boundary and initial conditions does the equation need?

A second-order equation needs two conditions in each direction, and an unsteady problem also needs the initial temperature field. First kind: a fixed surface temperature. Second kind: a fixed heat flux, with insulation as the special case q = 0. Third kind: convection, −k dT/dx = h(Ts − T∞) at the face. A radiating face gives εσ(Ts⁴ − Tsur⁴), often linearised as h_r(Ts − Tsur): about 7.9 W/m²K for ε = 0.8 at 400 K facing 300 K. Flux on both faces never fixes the temperature level.

Boundary conditions of the first, second and third kinds on a slab, and when no steady state exists: interactive loads as you scroll
Predict, then check

A 100 mm slab gets 1,500 W/m² on its left face and its right face is perfectly insulated. What steady temperature does it reach?

Worked examples

Fixed temperature on one face, convection on the other · Slab 100 mm, k = 1.4 W/mK; left face held at 150 °C; right face to air at 25 °C, h = 20 W/m²K (typical)
  1. T = C1 + C2 x; T(0) = 150 gives C1 = 150
  2. At x = L: −k C2 = h (T(L) − 25)
  3. Same as resistances in series: q = 125/(0.1/1.4 + 1/20) = 125/0.1214 = 1029.4 W/m²
  4. T(L) = 25 + 1029.4/20 = 76.5 °C (check: 150 − 1029.4 × 0.1/1.4 = 76.5 °C)

Answer: q ≈ 1029 W/m², right face 76.5 °C

Fixed flux on one face, fixed temperature on the other · Same slab; an electric heater pad supplies 1,500 W/m² to the left face; the right face is held at 30 °C
  1. Second kind on the left: −k dT/dx = 1,500, so the slope is −1,500/1.4 K/m
  2. T(0) = 30 + 1,500 × 0.1/1.4 = 137.1 °C

Answer: The heated face settles at 137.1 °C

Reading a measured temperature field · Steady temperatures in a 50 mm plate (k = 15 W/mK) fit T = 200 − 2,000 x² °C (x in m from the left face, which is insulated)
  1. Poisson form: d²T/dx² = −q‴/k; here d²T/dx² = −4,000 K/m²
  2. q‴ = 15 × 4,000 = 60,000 W/m³: the plate generates heat
  3. At x = 0, dT/dx = 0: consistent with the insulated face
  4. At x = 0.05 m: q = −k dT/dx = 15 × 4,000 × 0.05 = 3,000 W/m², and T = 195 °C
  5. Check: generation per m² of face = 60,000 × 0.05 = 3,000 W/m², equal to the flux leaving

Answer: q‴ = 60,000 W/m³; 3,000 W/m² leaves the right face at 195 °C

Chapter 3

Cylinders, spheres and the critical radius of insulation

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 1SPPU 302042 Units 1, 2MU MEC502 Module 1KTU PCMET503 Module 1

What is the thermal resistance of a pipe wall or a spherical shell?

In a cylinder r dT/dr is constant, so the temperature falls logarithmically and R = ln(r2/r1)/(2πkL). In a sphere the profile goes as 1/r and R = (1/r1 − 1/r2)/(4πk). Composite pipes and tanks are again resistors in series, with films 1/(h 2πrL) or 1/(h 4πr²) at the inner and outer radius. One metre of steel pipe from 25 to 30 mm radius has 0.00054 K/W; 50 mm of glass wool over it has 3.9 K/W, about 7,000 times more.

Why can a thin layer of insulation increase heat loss?

Adding insulation to a small cylinder does two things: its conduction resistance rises, but its outer surface grows, so the film resistance 1/(h 2πr2 L) falls. Below a critical radius the second effect wins. Differentiating the total resistance gives rc = k/h for a cylinder and 2k/h for a sphere. For PVC (k 0.19) in air with h = 10 W/m²K, rc = 19 mm, so PVC on a thin wire helps it shed heat. For glass wool (k 0.04), rc is only 4 mm, so on any real pipe it always cuts the loss.

Critical radius of insulation: heat loss from an insulated pipe, wire or sphere as the insulation gets thicker: interactive loads as you scroll

How are insulating materials chosen, and why does safety matter?

Low k comes from trapped gas: glass and mineral wool (0.04 W/mK), polyurethane foam (0.03), 85% magnesia (0.07), and cellular rubber for chilled lines. The choice also depends on the highest service temperature, moisture (wet insulation conducts far better), fire behaviour, and strength under cladding. Insulation also protects people: a bare 180 °C steam line has a surface near 179 °C, while 50 mm of glass wool brings it to about 38 °C. Cold lines need a vapour barrier so water does not condense inside.

How do engineers decide how thick insulation should be?

Each extra millimetre costs more than the last (more material on a larger circumference) and saves less heat than the last (diminishing returns). The economic thickness minimises insulation cost plus the value of the heat lost over the plant's life. A quick check used in SPPU problems is the payback period: installed cost divided by the money saved each year. Saving = (Q bare − Q insulated) × hours per year × cost of heat. Payback within a year or two is common for hot pipes.

Predict, then check

A bare copper wire of 1 mm radius is 60 K hotter than the air (h = 10 W/m²K). PVC (k 0.19) is added out to 3 mm radius, keeping the wire at the same temperature. The heat loss:

Worked examples

Insulated steam pipe · Steel pipe, radii 40 and 45 mm (k 54), glass wool 50 mm thick (k 0.04); steam 180 °C, inside h = 2,000; air 30 °C, outside h = 10 W/m²K (typical); per metre
  1. Inner film 1/(2,000 × 2π × 0.04) = 0.00199; steel ln(45/40)/(2π × 54) = 0.00035; glass wool ln(95/45)/(2π × 0.04) = 2.9731; outer film 1/(10 × 2π × 0.095) = 0.1675 K/W
  2. ΣR = 3.1429 K/W; Q = 150/3.1429 = 47.7 W/m
  3. Outer surface = 30 + 47.7 × 0.1675 = 38.0 °C
  4. Bare pipe for comparison: Q = 150/(0.00234 + 0.3537) = 421 W/m, surface 179 °C

Answer: 47.7 W/m instead of 421 W/m; the surface drops from 179 °C to 38 °C

Payback period of the insulation (SPPU style) · The pipe above, 100 m long, running 6,000 h a year; heat costs ₹1.5 per kWh and glass wool with cladding costs ₹1,200 per metre installed (illustrative prices); outside h taken as 10 W/m²K for both cases
  1. Heat saved = (421.3 − 47.7) W/m × 100 m = 37.36 kW
  2. Energy saved per year = 37.36 × 6,000 = 2,24,163 kWh
  3. Money saved per year = 2,24,163 × 1.5 = ₹3,36,245
  4. Cost = 1,200 × 100 = ₹1,20,000; payback = 1,20,000/3,36,245 = 0.36 year

Answer: The insulation pays for itself in about 4.3 months

Insulating a wire: more heat loss, cooler wire · Copper wire of 1 mm radius, PVC (k 0.19) to 3 mm radius, air h = 10 W/m²K (typical)
  1. Critical radius rc = 0.19/10 = 19 mm, above 3 mm, so the loss rises with PVC
  2. Same wire temperature (60 K above air): bare 3.77 W/m, insulated 9.64 W/m
  3. Same current (the wire makes 3.77 W/m either way): with PVC the wire is only 3.77 × (0.920 + 5.305) = 23.5 K above the air instead of 60 K

Answer: Below rc, insulation lets the wire run 37 K cooler at the same current

Chapter 4

Conduction with internal heat generation

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 1SPPU 302042 Unit 1MU not in syllabusKTU PCMET503 Module 1

What does uniform heat generation do to the temperature in a plane wall?

Every slice makes heat, so the flux grows linearly from zero at the centre (by symmetry) to q‴L at each face of a slab of thickness 2L. The profile is therefore a parabola, with the maximum at the centre. Two steps give the temperatures: all the heat leaves by convection, so Ts = T∞ + q‴L/h; then conduction adds Tmax − Ts = q‴L²/(2k). Examples are an electric heating element strip, curing concrete (cement hydration releases heat) and a nuclear fuel plate.

How do the results change for a cylinder and a sphere?

The heat made inside radius r has more volume per unit surface as r grows, so the numbers change. For a long cylinder Ts = T∞ + q‴R/(2h) and Tmax − Ts = q‴R²/(4k); for a sphere q‴R/(3h) and q‴R²/(6k). A current-carrying wire, a fuel rod and a cylindrical battery cell are cylinders; a spherical pellet bed or a fruit storing respiration heat are spheres. The centre rise scales with R², so doubling the radius at the same q‴ quadruples it.

Conduction with internal heat generation in a slab, cylinder and sphere: surface and centre temperatures: interactive loads as you scroll

Which is hotter relative to the fluid: the centre or the surface?

Divide the inside rise by the surface rise: for all three shapes (Tmax − Ts)/(Ts − T∞) = Bi/2, with Bi = hL/k (or hR/k). A copper wire in air has Bi near 0.00004, so it is practically uniform and the whole problem is cooling its surface. A ceramic fuel pellet in fast water has a large Bi, so most of the temperature rise is inside the pellet. This tells you where to spend design effort: better cooling or better conduction.

Where does internal heat generation matter in practice?

Joule heating in conductors and busbars (q‴ = I²ρe/A²), nuclear fuel rods, where the centreline must stay well below the melting point (2,865 °C for uranium dioxide, which conducts heat poorly), mass concrete in dams and raft foundations, where hydration heat can crack the core as the surface cools, and lithium-ion cells in an EV pack in Bengaluru traffic, where internal losses heat the jelly roll from inside. In each case the hot spot is in the middle, where you cannot measure it directly.

Predict, then check

A wire's radius is doubled while the heat generated per m³ stays the same. The temperature rise from the surface to the centre:

Worked examples

Copper wire carrying current · Bare copper wire of 2 mm diameter, 30 A, resistivity 1.68 × 10⁻⁸ Ω m, k = 401 W/mK; air at 30 °C, h = 15 W/m²K (typical, still air with radiation)
  1. A = π × 0.001² = 3.1416 mm²; q‴ = I² ρe/A² = 900 × 1.68 × 10⁻⁸/(3.1416e-06)² = 1.532 MW/m³
  2. Ts = 30 + q‴R/(2h) = 30 + 1.5320e+06 × 0.001/30 = 81.1 °C
  3. Tmax − Ts = q‴R²/(4k) = 1.5320e+06 × 10⁻⁶/1,604 = 0.00096 K

Answer: The wire runs at about 81 °C, and its centre is under 0.001 K hotter than its surface

Fuel pellet centreline (illustrative numbers) · Cylindrical pellet R = 4 mm, q‴ = 3 × 10⁸ W/m³, k = 3 W/mK (illustrative: uranium dioxide conducts poorly), surface at 400 °C
  1. Tmax − Ts = 3 × 10⁸ × 0.004²/(4 × 3) = 400 K
  2. Centreline ≈ 800 °C, well below the 2,865 °C melting point
  3. Heat per metre of rod = q‴ πR² = 15.1 kW/m

Answer: Centre ≈ 800 °C; the rod delivers 15.1 kW per metre

Curing concrete slab · Slab 400 mm thick, both faces to air at 30 °C, h = 10 W/m²K; early hydration heat 500 W/m³ (illustrative); dense concrete k = 1.4 W/mK
  1. Half-thickness L = 0.2 m; Ts = 30 + 500 × 0.2/10 = 40 °C
  2. Tmax − Ts = 500 × 0.2²/(2 × 1.4) = 7.14 K
  3. Bi = 10 × 0.2/1.4 = 1.43, so the core rise is Bi/2 = 0.71 of the surface rise

Answer: Core about 47.1 °C, surface 40 °C: a 7.1 K core-to-surface difference

Chapter 5

Fins: temperature profile, efficiency and effectiveness

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 2SPPU 302042 Unit 2MU MEC502 Module 2KTU PCMET503 Module 1

Why add fins, and where are they used?

Heat leaving a surface is h A ΔT. When h is small, as with air, the cheapest way to raise the product is more area. Fins do that: straight rectangular fins on motorcycle cylinders and electronics heat sinks, pin fins on LED and power-module coolers, annular fins on air-conditioner and radiator tubes. They help most where h is lowest, which is why a car radiator's fins are on the air side, not the coolant side. A fin is usually considered worth adding only when its effectiveness exceeds about 2.

What equation governs the temperature along a fin?

Take a slice of fin of cross-section A and perimeter P. Conduction in minus conduction out equals convection from its sides: k A d²T/dx² = h P (T − T∞). With θ = T − T∞ this is θ'' = m²θ, where m = √(hP/kA). The solution is a mix of e^(mx) and e^(−mx), set by the base temperature and the tip condition. m grows with h and falls with k: a copper fin stays hot far further from the base than a steel one of the same size.

What are the infinite, insulated and convective tip cases?

An infinitely long fin ends at the fluid temperature: θ = θb e^(−mx), Q = √(hPkA) θb. An insulated tip gives Q = √(hPkA) θb tanh(mL); since tanh 3 = 0.995, a fin longer than about mL = 3 adds under 1%. A convective tip has an exact but longer formula. The usual shortcut is the corrected length Lc = L + A/P (t/2 for a thin plate fin, d/4 for a pin) used in the insulated-tip formula; for a 5 mm aluminium pin it matches the exact answer to 0.001%.

Fins: temperature along a pin or straight fin, heat rate, fin efficiency and effectiveness for each tip condition: interactive loads as you scroll

What do fin efficiency and effectiveness measure?

Efficiency η compares the fin with an ideal fin whose whole surface is at the base temperature: η = Q/(h A_fin θb). For an insulated tip η = tanh(mL)/(mL). Effectiveness ε compares the fin with the bare base area it covers: ε = Q/(h A θb). For a long pin ε = √(4k/(hd)): a long 5 mm aluminium pin gives 78 in air at 25 W/m²K but a stainless one in boiling water (h 5,000) only 1.5, so it is not worth it. A finned surface combines both: η0 = 1 − (N A_fin/A_total)(1 − η).

Why does a thermometer in a well read the wrong temperature?

A thermometer in a pocket (well) welded into a pipe measures the temperature at the bottom of the pocket. The pocket is a fin: hot gas heats it along its length, while its base conducts heat out into the cooler pipe wall. So the tip sits between the gas and the wall temperatures, (T_tip − T∞)/(T_wall − T∞) = 1/cosh(mL) for an insulated tip. To cut the error, make mL large: a longer, thinner-walled pocket of low-k material, good flow over it, and lagging on the pipe outside.

Predict, then check

An insulated-tip fin already has mL = 3. Doubling its length adds how much heat?

Worked examples

Aluminium pin fin in air · Pin 5 mm diameter, 60 mm long, aluminium alloy 3003 (k 190 W/mK); base 100 °C, air 25 °C, h = 25 W/m²K (typical)
  1. A = π × 0.005²/4 = 19.635 mm², P = π × 0.005 = 15.71 mm; m = √(25 × 0.01571/(190 × 1.9635e-05)) = 10.260 1/m; mL = 0.6156
  2. √(hPkA) θb = 2.8707 W; insulated tip Q = 2.8707 × tanh 0.6156 = 1.5733 W
  3. Convective tip (exact) Q = 1.5988 W; corrected length Lc = 60 + 1.25 = 61.25 mm gives 1.5988 W
  4. η = 1.5988/(25 × 9.621 × 10⁻⁴ × 75) = 88.6%; ε = 1.5988/(25 × 1.9635e-05 × 75) = 43.4; tip at 87.3 °C

Answer: Q ≈ 1.60 W per pin, η ≈ 89%, ε ≈ 43

Error of a thermometer well · Stainless steel 304 pocket (k 14.4 W/mK), 12 mm outside and 10 mm inside diameter, 80 mm long, in hot gas with h = 30 W/m²K (illustrative); pipe wall at the pocket's base 100 °C; the thermometer reads 180 °C
  1. A = π(0.012² − 0.010²)/4 = 34.56 mm², P = π × 0.012 = 37.70 mm
  2. m = √(30 × 0.03770/(14.4 × 3.4558e-05)) = 47.67 1/m; mL = 3.814; cosh mL = 22.67
  3. (180 − T∞)/(100 − T∞) = 1/22.67, so T∞ = (180 × 22.67 − 100)/(22.67 − 1) = 183.7 °C
  4. A carbon steel pocket (k 54) of the same size has mL = 1.97 and would read only 160.8 °C
  5. To bring the stainless pocket's error under 0.5 K it would need L = cosh⁻¹(83.7/0.5)/47.67 = 122 mm

Answer: The gas is at about 183.7 °C, so the reading is 3.7 K low

Is a fin worth it? · Long 5 mm pins: aluminium (k 190) in water with h = 1,000 W/m²K, and stainless steel 304 (k 14.4) in boiling water with h = 5,000 W/m²K (typical h)
  1. For a long pin ε = √(kP/(hA)) = √(4k/(h d))
  2. Aluminium in water: √(4 × 190/(1,000 × 0.005)) = 12.3
  3. Stainless in boiling water: √(4 × 14.4/(5,000 × 0.005)) = 1.52
  4. For comparison, aluminium in air at 25 W/m²K: 78

Answer: The aluminium pin pays (ε 12); the stainless pin in boiling water does not (ε 1.5 < 2)

Chapter 6

Transient conduction: lumped systems, Biot number and time constant

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 2SPPU 302042 Unit 2MU MEC502 Module 2KTU PCMET503 Module 1

When can a body be treated as having one temperature?

Heat leaving a hot body must first be conducted to the surface (resistance Lc/k per unit area) and then carried away by the fluid (resistance 1/h). The Biot number Bi = hLc/k is their ratio. When Bi < 0.1 the inside resistance is small, the temperature inside is nearly uniform, and the whole body can be treated as one lump. Lc = V/As: D/6 for a sphere, D/4 for a long cylinder, half the thickness for a plate cooled on both faces. A 20 mm copper ball in water (h 500, typical) has Bi 0.004; the same ball in glass has Bi 1.7.

How does a lumped body cool, and what is the time constant?

An energy balance says the heat convected away equals the drop in stored energy: −hAs(T − T∞) = ρVc dT/dt. The solution is an exponential: the gap to the fluid temperature shrinks by the same fraction every second. The time constant τ = ρcV/(hAs) sets the pace. After 1τ, 63.2% of the change is done; after 3τ, 95%; after 5τ, 99.3%. The heat given up by time t is ρVc(Ti − T). Change the body, material and h in the interactive and watch τ and the time to a target move.

Lumped system cooling: Biot number, time constant and the time a body or thermocouple takes to reach a temperature: interactive loads as you scroll

What does the Fourier number add?

The Fourier number Fo = αt/Lc² is dimensionless time: how far heat has diffused into the body compared with its size (α = k/ρc). The lumped exponent t/τ can be written Bi × Fo, so the answer depends on those two numbers only. Notice k cancels out of t/τ: within the lumped range, a copper part and a steel part of the same size cool at nearly the same rate, because ρc is 3.49 MJ/m³K for copper and 3.85 for steel. High conductivity only matters for passing the Bi < 0.1 test.

Why does a thermocouple lag, and how is it made faster?

A thermocouple junction is a tiny lump: it reads its own temperature, which trails the gas around it. Its response is the same exponential with τ = ρcD/(6h) for a bead of diameter D. A 0.8 mm bead in gas with h 200 W/m²K (typical) has τ 2.3 s and needs 12 s to come within 1 °C after a 175 °C jump. If the gas warms steadily, the reading lags by (rate × τ). Smaller beads and faster flow (higher h) cut τ; τ is often quoted as the time to 63.2% of a step.

Predict, then check

A 20 mm copper ball and a 20 mm steel ball, both at 200 °C, cool in the same air (both lumped). Which reaches 50 °C first?

Worked examples

Thermocouple response in a hot gas stream · Bead 0.8 mm diameter, taken as copper (ρ 8,940 kg/m³, c 390 J/kgK, k 390 W/mK: illustrative), h = 200 W/m²K (typical gas flow), bead at 25 °C plunged into gas at 200 °C
  1. Lc = D/6 = 0.1333 mm; Bi = 200 × 1.3333e-04/390 = 6.84e-05, far below 0.1: lumped
  2. τ = ρcLc/h = 8,940 × 390 × 1.3333e-04/200 = 2.32 s (63.2% of the step after 2.32 s)
  3. Within 1 °C of 200 °C means θ = 1/175: t = τ ln 175 = 2.324 × 5.165 = 12.0 s
  4. If the gas instead rises at 2 °C/s, the reading settles 4.6 °C behind (2 × τ)
  5. For τ = 0.5 s at the same h: D = 6hτ/(ρc) = 0.17 mm

Answer: τ ≈ 2.3 s; about 12 s to read within 1 °C

Fan cooling of a steel shaft · Long carbon-steel shaft, D = 40 mm (k 43, ρ 7,850, c 490), at 200 °C; fan air at 30 °C, combined convection and radiation h = 60 W/m²K (typical); cool to 50 °C
  1. Lc = D/4 = 10 mm; Bi = 60 × 0.01/43 = 0.0140 < 0.1: lumped
  2. τ = 7,850 × 490 × 0.01/60 = 641 s
  3. t = τ ln[(200 − 30)/(50 − 30)] = 641 × 2.1401 = 1,372 s = 22.9 min
  4. Heat given up per metre = ρc(πD²/4)(200 − 50) = 725 kJ/m

Answer: About 23 minutes; 725 kJ per metre of shaft

Copper ball or glass ball: is lumped allowed? · Two balls of 20 mm diameter at 90 °C dropped into water at 20 °C, h = 500 W/m²K (typical); copper k 390, glass k 0.96, ρ 2,530, c 840
  1. Lc = D/6 = 3.333 mm. Copper Bi = 500 × 0.003333/390 = 0.0043: lumped. Glass Bi = 500 × 0.003333/0.96 = 1.74: not lumped
  2. Copper: τ = 8,940 × 390 × 0.003333/500 = 23.2 s; to 50 °C: t = τ ln(70/30) = 19.7 s
  3. Glass, if wrongly lumped: τ = 14.2 s, so after 60 s the formula claims θ = e^(−60/14.2) = 0.014, the ball at 21 °C
  4. The glass centre is in fact far hotter; chapter 7 works it out with the one-term solution

Answer: Copper: lumped, 20 s to 50 °C. Glass: Bi 1.7, use Heisler

Chapter 7

Transient conduction with internal gradients: Heisler charts and semi-infinite solids

Which syllabus covers thisAU ME3691 Unit IVTU BME601 Module 2SPPU 302042 Unit 2MU not in syllabusKTU PCMET503 Module 1

What exactly is a Heisler chart?

When Bi is not small, the centre lags the surface and the full solution is an infinite series of decaying terms, each e^(−λn²Fo). The higher terms die fast, so once Fo > 0.2 the first term alone is within about 2% (our check against the full series). A Heisler chart is just that first term plotted: centre θ0/θi against Fo, one line per Bi. λ1 is the first root of λ tan λ = Bi (wall), λJ1(λ)/J0(λ) = Bi (long cylinder) or 1 − λ cot λ = Bi (sphere). Here Bi = hL/k uses the half-thickness L or the radius ro, not V/As. The interactive solves these equations directly.

Heisler charts from the one-term solution: centre and local temperature of a wall, cylinder or sphere, and the semi-infinite solid: interactive loads as you scroll

How are the off-centre temperature and the heat removed found?

Within the one-term solution the shape of the profile is fixed, so the local temperature is the centre value times a shape factor: cos(λ1x/L) for a wall, J0(λ1r/ro) for a cylinder, sin(λ1r/ro)/(λ1r/ro) for a sphere. Those are the second set of Heisler charts. The heat given up as a fraction of the most it could give, Q/Q0 with Q0 = ρcV(Ti − T∞), is 1 − θ0 sin λ1/λ1 for a wall, 1 − 2θ0 J1(λ1)/λ1 for a cylinder and 1 − 3θ0(sin λ1 − λ1 cos λ1)/λ1³ for a sphere: the third (Gröber) chart.

When is a body semi-infinite, and what is the solution?

Early on, heat has only soaked a depth of a few √(αt) into a thick body, so the far side does not know anything happened: the body behaves as if it were infinitely deep. For carbon steel √(αt) is about 10.6 mm after 10 s. If the surface is suddenly held at Ts, (T − Ts)/(Ti − Ts) = erf(x/(2√(αt))). If instead the surface meets a fluid with h, (T − Ti)/(T∞ − Ti) = erfc η − exp(hx/k + h²αt/k²) erfc(η + h√(αt)/k), η = x/(2√(αt)). Frost depth in soil and a die touching hot metal are typical uses.

How are short cylinders and rectangular bars handled?

A short cylinder is the overlap of an infinite wall (its length) and an infinite cylinder (its radius); a long rectangular bar is the overlap of two walls; a brick, three. For a uniform start and the same fluid all round, the dimensionless temperature of the combined shape is the product of the one-dimensional ones: θ/θi = (θ/θi)wall × (θ/θi)cylinder. Each factor uses its own Bi and Fo. Because each factor is below 1, the corners and short shapes cool faster than either parent shape.

Predict, then check

A plane wall, a long cylinder and a sphere have the same Bi = 1 (on L or ro) and reach the same Fo = 0.5. Which centre has cooled the most?

Worked examples

Oil quench of a steel plate · Carbon-steel plate 60 mm thick (k 43, α = 43/(7,850 × 490) = 1.118 × 10⁻⁵ m²/s) at 500 °C, quenched in oil at 40 °C with h = 800 W/m²K (typical); find the centre and surface temperatures after 2 min
  1. L = 30 mm; Bi = 800 × 0.03/43 = 0.5581 (> 0.1, so not lumped); Fo = αt/L² = 1.491 (> 0.2, so one term is enough)
  2. Solve λ tan λ = 0.5581: λ1 = 0.6842; C1 = 4 sin λ1/(2λ1 + sin 2λ1) = 1.0768
  3. θ0/θi = 1.0768 × e^(−0.6842² × 1.491) = 0.5359, so T0 = 40 + 460 × 0.5359 = 286 °C
  4. Surface: θ/θ0 = cos λ1 = 0.7749, so Ts = 40 + 460 × 0.4152 = 231 °C
  5. Q/Q0 = 1 − θ0 sin λ1/λ1 = 0.505: 53.6 MJ per m² of plate out of 106.2 MJ

Answer: Centre about 286 °C, surface about 231 °C, 50% of the heat removed

The glass ball of chapter 6, done properly · Glass sphere ro = 10 mm (k 0.96, α = 0.96/(2,530 × 840) = 4.517 × 10⁻⁷ m²/s) at 90 °C in water at 20 °C, h = 500 W/m²K; after 60 s
  1. Bi = h ro/k = 500 × 0.01/0.96 = 5.208; Fo = αt/ro² = 0.2710 (> 0.2)
  2. Solve 1 − λ cot λ = 5.208: λ1 = 2.5899; C1 = 4(sin λ1 − λ1 cos λ1)/(2λ1 − sin 2λ1) = 1.7981
  3. θ0/θi = 1.7981 × e^(−2.5899² × 0.2710) = 0.2919: centre 40.4 °C
  4. Surface: × sin λ1/λ1 = 0.2024, so 24.1 °C; Q/Q0 = 0.862
  5. The lumped formula had claimed 21 °C everywhere

Answer: Centre ≈ 40 °C, surface ≈ 24 °C after 60 s

Heat soaking into a thick steel mould · Thick carbon-steel mould wall at 30 °C; at t = 0 hot metal holds its face at 400 °C (illustrative); how deep is the 100 °C point after 10 s, and what is the temperature 5 mm in?
  1. √(αt) = √(1.118 × 10⁻⁵ × 10) = 10.57 mm; the wall is far thicker, so treat it as semi-infinite
  2. (T − Ts)/(Ti − Ts) = (100 − 400)/(30 − 400) = 0.8108 = erf η, so η = 0.9284
  3. x = 2η√(αt) = 2 × 0.9284 × 10.57 = 19.6 mm
  4. At 5 mm: η = 5/(2 × 10.57) = 0.2364, erf = 0.2619, T = 400 − 370 × 0.2619 = 303 °C

Answer: The 100 °C point is 20 mm deep; 5 mm in it is about 303 °C

Product solution for a short cylinder · Steel billet 60 mm diameter × 60 mm long, same quench as the plate example (500 °C, oil at 40 °C, h = 800), after 2 min
  1. Wall factor (length 60 mm, L = 30 mm): 0.5359, from the plate example
  2. Cylinder factor (ro = 30 mm): Bi = 0.5581, Fo = 1.491; λ1 J1/J0 = Bi gives λ1 = 0.9871, C1 = 1.1261, θ0/θi = 0.2635
  3. Centre θ/θi = 0.5359 × 0.2635 = 0.1412, T = 40 + 460 × 0.1412 = 105 °C

Answer: Centre ≈ 105 °C, far cooler than the plate's 286 °C

Chapter 8

Numerical conduction: the finite difference method

Which syllabus covers thisAU not in syllabusVTU BME601 Module 3SPPU not in syllabusMU not in syllabusKTU PCMET503 Module 1

What does a finite difference model actually do?

It replaces the smooth temperature field by values at nodes Δx apart and writes an energy balance for the small slice around each node: heat conducted in from the left neighbour, plus from the right, plus heat generated, equals heat stored (zero when steady). Conduction between nodes is k(Tm−1 − Tm)/Δx, the same Fourier law in finite steps. For an interior node in steady state this gives the equation below; one equation per unknown node gives a linear system. The error falls as Δx²: halve the spacing and it drops to about a quarter.

How are boundary nodes written?

A node on the surface owns only half a slice, Δx/2 thick. Write the same energy balance for that half slice. Convective face: k(Tm−1 − Tm)/Δx + h(T∞ − Tm) + q̇Δx/2 = 0 (steady). Insulated face: the outside term is zero, which is the same as imagining a mirror node with Tm+1 = Tm−1. Face at a known temperature: no equation is needed, the value is simply fixed. In time-marching, the half slice also stores half the energy, which is why the convective node has its own stability limit.

Explicit or implicit: what is the difference, and why can explicit blow up?

In time, the explicit method computes each new nodal temperature from the old values of the node and its neighbours: simple, no equations to solve. But the old value of the node is weighted by (1 − 2Fo), with Fo = αΔt/Δx². If Fo > ½ that weight goes negative: a hotter neighbour makes the node colder next step and the error grows each step. A convective face node needs Fo(1 + Bi) ≤ ½, Bi = hΔx/k. The implicit method uses the new neighbour values instead, so every step needs a tridiagonal solve, but it is stable for any Fo.

Finite difference conduction: nodal temperatures, explicit and implicit time steps, the Fourier-number stability limit and Gauss-Seidel: interactive loads as you scroll

How is the steady system solved?

For one dimension the equations form a tridiagonal system that the Thomas algorithm solves directly in a few operations per node. For 2-D and 3-D grids, iteration is simpler: Gauss-Seidel sweeps through the nodes, replacing each value by the one its equation gives using the latest neighbours, until the largest change in a sweep is below a tolerance. It always converges for these equations, but slowly on fine grids: the number of sweeps grows roughly with the square of the node count, which is why over-relaxation and multigrid exist.

Predict, then check

An explicit scheme is run with Fo = αΔt/Δx² = 0.6 to save time. What happens?

Worked examples

Steady wall with heat generation by Gauss-Seidel · Steel wall 30 mm thick, k = 43 W/mK, uniform generation 2 × 10⁷ W/m³ (illustrative, e.g. induction heating), left face insulated, right face held at 100 °C; nodes 0-3 at 10 mm spacing
  1. q̇Δx²/k = 2 × 10⁷ × 0.01²/43 = 46.512 K
  2. Insulated node (half slice): T0 = T1 + 23.256; interior: T1 = (T0 + T2 + 46.512)/2, T2 = (T1 + 100 + 46.512)/2
  3. Start at 100 °C. Sweep 1: T0 = 123.3, T1 = 134.9, T2 = 140.7; sweep 2: 158.1, 172.7, 159.6
  4. After 28 sweeps every node is within 0.1 °C of the converged 309.3, 286.0, 216.3 °C
  5. Exact: T = 100 + q̇(L² − x²)/(2k) gives the same values: central differences are exact for a parabola

Answer: T0 = 309.3 °C at the insulated face, 286.0 and 216.3 °C inside

Largest explicit time step · Δx = 5 mm; carbon steel (α = 1.118 × 10⁻⁵ m²/s, k 43) and window glass (α = 4.517 × 10⁻⁷, k 0.96); face cooled by water, h = 500 W/m²K (typical)
  1. Steel interior: Δt ≤ ½Δx²/α = 0.5 × 0.005²/1.118 × 10⁻⁵ = 1.118 s
  2. Steel face: Bi = hΔx/k = 0.0581; Δt ≤ ½Δx²/[α(1 + Bi)] = 1.057 s
  3. Glass interior: Δt ≤ 27.7 s; glass face: Bi = 2.60, Δt ≤ 7.68 s
  4. The face node governs for glass: its low k makes the face Biot number large

Answer: Steel ≤ 1.06 s, glass ≤ 7.7 s

Two explicit steps by hand · Five nodes at 20 °C; at t = 0 the left face (node 0) goes to 100 °C, node 4 stays at 20 °C
  1. Fo = 0.25: node 1 = 20 + 0.25(100 − 40 + 20) = 40 °C, the others stay at 20
  2. Step 2: node 1 = 40 + 0.25(100 − 80 + 20) = 50 °C, node 2 = 20 + 0.25(40 − 40 + 20) = 25 °C
  3. Fo = 0.6: node 1 = 20 + 0.6 × 80 = 68, then 68 + 0.6(100 − 136 + 20) = 58.4 °C: it cooled while heated from one side
  4. The negative weight 1 − 2Fo = −0.2 is the cause; implicit steps would stay smooth

Answer: Fo 0.25: node 1 rises 20 → 40 → 50 °C. Fo 0.6: 20 → 68 → 58.4 °C, an unphysical dip

Chapter 9

Convection: thermal boundary layers, dimensionless numbers and the Reynolds analogy

Which syllabus covers thisAU ME3691 Unit IIVTU BME601 Module 4SPPU 302042 Unit 3MU MEC502 Module 3KTU PCMET503 Module 2

What are the velocity and thermal boundary layers, and where does h come from?

Fluid touching a plate does not slip, so a thin layer is slowed: the velocity boundary layer δ (the Fluid Mechanics book covers it and its transition). If the plate is hotter, a thermal layer δt also grows, across which the fluid goes from Ts to T∞. Right at the wall the fluid is still, so heat crosses by conduction: h = −k(∂T/∂y)wall/(Ts − T∞). A thin thermal layer means a steep gradient and a high h. On a plate both layers thicken as √x, so the local h falls as 1/√x and the laminar average over a length L is twice the local value at L.

What do Re, Pr, Nu, Gr, St and Pe each mean?

Reynolds Re = Ux/ν: inertia against viscous force; laminar on a plate up to about 5 × 10⁵. Prandtl Pr = ν/α: how fast momentum diffuses compared with heat; air 0.71, water at 27 °C 5.8, engine oil hundreds to thousands, mercury 0.025. Nusselt Nu = hL/k: convection against pure conduction through the same fluid layer. Grashof Gr = gβΔTL³/ν²: buoyancy against viscosity (free convection, chapter 12). Stanton St = h/(ρcpU) = Nu/(Re Pr): the fraction of the stream's heat capacity that reaches the wall. Péclet Pe = Re Pr: carried heat against conducted heat.

How does the Prandtl number set the thermal layer?

The velocity layer spreads at the rate ν and the thermal layer at α, so their thickness ratio depends only on Pr: for Pr above about 0.6, δt ≈ δ Pr^(−1/3). Water's thermal layer is about half its velocity layer and engine oil's a tenth, so oils give high h for their low k. Solving the laminar equations (Blasius, Pohlhausen) gives Nu_x = 0.332 Re^½ Pr^⅓, within about 2% of the exact result for Pr 0.7 to 100 (our check). Liquid metals are the opposite: the heat runs far ahead of the velocity layer; use Nu_x = 0.565 (Re Pr)^½, which is about 15% high at Pr 0.025.

Velocity and thermal boundary layers on a flat plate: how the Prandtl number sets their thickness and the local heat transfer coefficient: interactive loads as you scroll

What do the conservation equations say, and what is developed flow?

Three balances on a small fluid element. Continuity: what flows in, flows out. Momentum: the change in momentum carried along equals the pressure and viscous forces. Energy: the heat carried in by the flow plus heat conducted in equals heat carried out (plus viscous heating at high speed). Boundary-layer theory drops the small terms, such as conduction along the flow. Inside a pipe the layers grow from the wall until they meet: after that the velocity profile and the dimensionless temperature profile stop changing along the pipe and h becomes constant. That is fully developed flow (chapter 11).

What are the Reynolds and Chilton-Colburn analogies?

Momentum and heat travel through a boundary layer by the same mechanisms, so wall friction and wall heat transfer are linked. When Pr = 1 the velocity and temperature profiles are identical and St = Cf/2: the Reynolds analogy. For other fluids Chilton and Colburn scaled it with Pr^(2/3), valid for roughly 0.6 < Pr < 60. It lets a measured drag or a friction-factor chart give h without a heat transfer test. For the laminar plate it reproduces 0.332 Re^½ Pr^⅓ exactly, and it is widely used for turbulent flow, but not where pressure drag dominates, such as behind a cylinder.

Predict, then check

Air (Pr 0.71) and water (Pr 5.8) flow over identical plates at the same Reynolds number. Compared with its own velocity layer, whose thermal layer is thinner?

Worked examples

Air over a warm plate · Air at 25 °C (ν 15.52 × 10⁻⁶ m²/s, k 0.02624 W/mK, Pr 0.707) at 5 m/s along a plate at 60 °C; at x = 0.3 m (properties at 25 °C for simplicity; chapter 10 uses the film temperature)
  1. Re_x = 5 × 0.3/15.52 × 10⁻⁶ = 96,649: laminar (transition only at x = 1.55 m)
  2. δ = 5x/√Re = 5 × 0.3/310.9 = 4.82 mm; δt = δ × 0.707^(−1/3) = 5.42 mm
  3. Nu_x = 0.332 × 310.9 × 0.707^(1/3) = 91.9; h_x = 91.9 × 0.02624/0.3 = 8.04 W/m²K
  4. Average over the first 0.3 m = 2h_x = 16.1 W/m²K; heat from one face 0.3 m long and 1 m wide: q = 16.1 × 0.3 × 35 = 169 W

Answer: δ ≈ 4.8 mm, δt ≈ 5.4 mm, h_x ≈ 8.0 W/m²K, about 169 W per metre width

Water and engine oil at the same speed · 1 m/s, x = 0.2 m. Water at 27 °C (ν 0.857 × 10⁻⁶, k 0.6103, Pr 5.85); engine oil at 50 °C (typical: ν 70.1 × 10⁻⁶, k 0.133, Pr 919)
  1. Water: Re = 2,33,427; δ = 2.07 mm, δt = 1.15 mm; h_x = 882 W/m²K
  2. Oil: Re = 2,853; δ = 18.7 mm, δt = δ × 919^(−1/3) = 1.93 mm; h_x = 115 W/m²K
  3. The oil's velocity layer is 9 times thicker, but its thermal layer only 1.7 times: a high Pr keeps the heat close to the wall

Answer: Water about 882 W/m²K, oil about 115 W/m²K

h from friction by Chilton-Colburn · The air flow of the first example; laminar local friction coefficient Cf,x = 0.664/√Re_x
  1. Cf,x = 0.664/310.9 = 2.1358e-03
  2. St = (Cf/2) Pr^(−2/3) = 1.0679e-03 × 1.2600 = 1.3456e-03
  3. ρcp = k Pr/ν = 0.02624 × 0.707/15.52 × 10⁻⁶ = 1195 J/m³K
  4. h = St ρcp U = 1.3456e-03 × 1195 × 5 = 8.04 W/m²K, the same as the direct formula

Answer: h_x ≈ 8.04 W/m²K from the friction coefficient alone

Chapter 10

Forced convection over plates, cylinders, spheres and tube banks

Which syllabus covers thisAU ME3691 Unit IIVTU BME601 Module 4SPPU 302042 Unit 3MU MEC502 Module 3KTU PCMET503 Module 2

How is the average h on a flat plate found, and why the film temperature?

A fluid flowing along a plate builds a boundary layer from the leading edge, so the local h is highest at the edge and falls as x^(-1/2) while the layer is laminar. Averaging the local Nusselt number 0.332 Re_x^(1/2) Pr^(1/3) over the length doubles the constant to 0.664. If the layer turns turbulent at Re = 5×10⁵, h jumps there and the mixed formula subtracts the laminar head start (the 871). Properties vary between the wall and the stream, so they are taken at the film temperature (Ts + T∞)/2.

Forced convection over a flat plate, cylinder, sphere and tube bank: Reynolds number, Nusselt number and h from the correlations: interactive loads as you scroll

What changes when the flow crosses a cylinder?

The flow stagnates on the front, accelerates round the sides and separates near the widest point, leaving a wake behind. Local h is highest at the front stagnation line, falls to a minimum near separation and rises again in the wake. The average is correlated as Nu = C Re^m Pr^(1/3) with C and m changing by Reynolds band (Hilpert's table), or by one formula for all Re (Churchill-Bernstein, for Re Pr ≥ 0.2). The two differ by up to about 25% across the range: a correlation is a fit to scattered data, not a law.

Why does a sphere's Nusselt number never fall below 2?

Even in perfectly still fluid a hot sphere loses heat by conduction into the surrounding fluid, and steady conduction from a sphere of diameter D into an infinite medium gives exactly Nu = hD/k = 2. Whitaker's correlation adds the flow's contribution to this floor: Nu = 2 + (0.4 Re^(1/2) + 0.06 Re^(2/3)) Pr^0.4 (μ/μs)^(1/4), with properties at T∞ and μs at the wall, for 3.5 < Re < 7.6×10⁴. Droplets, prills in a cooling tower and fuel sprays are typical uses.

How do in-line and staggered tube banks differ, and what is the row correction?

Tube banks fill boilers, air heaters and air-cooled condensers. In an in-line bank each tube hides in the wake of the one in front; in a staggered bank the stream weaves and hits each tube more squarely, so h is usually higher for the same pressure budget. Zukauskas correlated the average for deep banks on the velocity in the narrowest gap, Vmax. The first rows see an undisturbed stream and do worse, so banks of fewer than 20 rows are multiplied by a row factor C2 (about 0.63-0.68 for one row, 0.97 for ten). The fluid warms through the bank, so q uses the log-mean temperature difference.

Predict, then check

Air flows over a flat plate and the boundary layer stays laminar throughout. You double the air speed. What happens to the average h?

Worked examples

Air cooling a hot steel sheet · Sheet 0.6 m long in the flow direction at 90 °C; air at 30 °C, 8 m/s; air properties at the film temperature 60 °C (ρ 1.059 kg/m³, μ 2.008×10⁻⁵ Pa·s, k 0.02808 W/mK, Pr 0.7202)
  1. ν = μ/ρ = 1.8961e-05 m²/s; Re_L = 8 × 0.6/ν = 2,53,147: laminar (below 5×10⁵)
  2. Nu = 0.664 × 2,53,147^½ × 0.7202^⅓ = 299.5
  3. h = Nu k/L = 299.5 × 0.02808/0.6 = 14.01 W/m²K
  4. Per metre width: q = h L ΔT = 14.01 × 0.6 × 60 = 505 W
  5. Transition would start at xc = 5×10⁵ ν/V = 1.19 m, beyond this sheet

Answer: h ≈ 14.0 W/m²K; about 505 W per metre of sheet width

Wind across a bare pipe: two correlations · Pipe 50 mm outside diameter at 70 °C; wind at 30 °C, 5 m/s; film 50 °C: ν = 1.963×10⁻⁵/1.092 m²/s, k 0.02735 W/mK, Pr 0.7228
  1. Re_D = 5 × 0.05/ν = 13,907
  2. Churchill-Bernstein: Nu = 0.3 + 0.62 Re^½ Pr^⅓ [1 + (0.4/Pr)^⅔]^(−¼) [1 + (Re/2,82,000)^⅝]^⅘ = 64.9
  3. Hilpert (4,000-40,000 band: C 0.193, m 0.618): Nu = 0.193 Re^0.618 Pr^⅓ = 63.0
  4. h = 64.9 × 0.02735/0.05 = 35.5 W/m²K (Hilpert gives 34.4)
  5. Loss per metre = h π D ΔT = 35.5 × π × 0.05 × 40 = 223 W/m

Answer: About 223 W per metre; the two correlations agree within 3% here

A staggered air heater bank · Air at 20 °C, 4 m/s approach velocity, crosses 8 rows of 25 mm tubes held at 100 °C by condensing steam; staggered, ST = SL = 50 mm. Air at 20 °C: ρ 1.204, cp 1,007, k 0.02514, μ 1.825×10⁻⁵, Pr 0.7309; Prs (100 °C) 0.7111
  1. Gaps: transverse ST − D = 25 mm; the two diagonal gaps total 2(SD − D) = 61.8 mm, wider, so Vmax = ST/(ST − D) × V = 8 m/s
  2. Re_D,max = ρ Vmax D/μ = 13,195; staggered, ST/SL = 1 < 2: C = 0.35, m = 0.6
  3. Nu (20 or more rows) = 0.35 Re^0.6 Pr^0.36 (Pr/Prs)^¼ = 93.4; 8 rows: C2 = 0.9652, Nu = 90.1; h = 90.6 W/m²K
  4. Outlet: (100 − To)/80 = exp(−π D NL h/(ρ V ST cp)) = exp(−0.2349), so To = 36.7 °C
  5. ΔTlm = (80 − 63.3)/0.2349 = 71.3 K; heat per metre of each tube = h π D ΔTlm = 508 W

Answer: Air leaves at about 37 °C; h ≈ 91 W/m²K (properties at the inlet temperature: a first pass)

Chapter 11

Forced convection in tubes: entrance length, laminar and turbulent flow

Which syllabus covers thisAU ME3691 Unit IIVTU BME601 Module 4SPPU 302042 Unit 3MU MEC502 Module 3KTU PCMET503 Module 2

What are the hydrodynamic and thermal entrance lengths?

Fluid entering a tube has a flat velocity profile; the wall slows it, and a boundary layer grows until it fills the tube. Beyond that hydrodynamic entrance length the profile stops changing: fully developed flow (for laminar flow about 0.05 Re D). The temperature profile develops the same way over the thermal entrance length, about 0.05 Re Pr D in laminar flow, so it is Pr times longer: 2.2 times for water at 80 °C, about 500 times for engine oil at 80 °C. In turbulent flow both are short, roughly 10 diameters. Inside the entrance region h is higher than the fully developed value.

What is the mean temperature, and how does it change along the tube?

The mean (bulk) temperature Tm is the temperature you would measure if you collected the fluid at a section in a cup and mixed it: the energy-weighted average. An energy balance on a slice gives ṁ cp dTm/dx = q″ P. With a uniform wall heat flux, such as electric heating, Tm rises in a straight line, and in fully developed flow the wall stays a constant q″/h above it. With a uniform wall temperature, as in a condensing-steam heater, the driving difference Ts − Tm decays exponentially, so the tube's last metres add little heat.

Forced convection in a tube: entrance lengths, laminar and turbulent Nusselt numbers and the mean temperature along the tube: interactive loads as you scroll

Why is laminar Nu a constant, 3.66 or 4.36?

Once both the velocity and temperature profiles are fully developed in laminar flow, their shapes stop changing along the tube, so the wall gradient scales with (Ts − Tm)/D and Nu = hD/k becomes a pure number. Solving the energy equation for the parabolic profile gives 3.66 for a constant wall temperature and 48/11 = 4.36 for a constant wall heat flux. Speeding up the flow does not change h at all (as long as it stays laminar); only a smaller diameter does, since h = Nu k/D.

Which correlation for turbulent flow, and what about non-circular ducts?

Dittus-Boelter, Nu = 0.023 Re^0.8 Pr^n (n = 0.4 when the fluid is heated, 0.3 when cooled), is the classic quick estimate for smooth tubes with Re above 10⁴. Gnielinski's correlation uses the friction factor f = (0.79 ln Re − 1.64)⁻² and holds from Re 3,000 to 5×10⁶; it is the better one, and the two differ by up to about 16% for Re 10⁴-10⁵. For ducts that are not round, use the hydraulic diameter Dh = 4A/P in Re and Nu: a 300 × 150 mm duct behaves like a 200 mm pipe.

Predict, then check

Water flows in laminar, fully developed flow through a tube with a constant wall temperature. You double the flow velocity (it stays laminar). What happens to h?

Worked examples

Length of a steam-heated water tube · Water at 0.8 m/s in a 20 mm tube, heated from 20 to 60 °C by a wall held at 80 °C; properties at the mean bulk temperature 40 °C: ρ 992.1, cp 4,179, k 0.631, μ 0.653×10⁻³, Pr 4.32
  1. Re = ρ um D/μ = 24,309: turbulent; entrance about 10 D = 0.2 m, short
  2. f = (0.79 ln Re − 1.64)⁻² = 0.0249; Gnielinski Nu = 144.4 (Dittus-Boelter, n = 0.4: 133.2)
  3. h = 144.4 × 0.631/0.02 = 4,554 W/m²K
  4. ṁ = ρ um π D²/4 = 0.2493 kg/s
  5. (80 − 60)/(80 − 20) = exp(−π D L h/(ṁ cp)), so L = ṁ cp ln 3/(π D h) = 4.00 m
  6. Heat = ṁ cp (60 − 20) = 41.7 kW

Answer: About 4.0 m of tube; 41.7 kW

Entrance lengths: water against engine oil · 12 mm tube at 80 °C. Water at 0.05 m/s (ν = 0.355×10⁻³/971.8, k 0.670, Pr 2.22); engine oil at 0.4 m/s (ρ 852, μ 0.03232, k 0.138, Pr 499.3)
  1. Water: Re = 1642 (laminar); x_fd,h = 0.05 Re D = 0.99 m; x_fd,t = × Pr = 2.19 m
  2. Oil: Re = 126.5; x_fd,h = 76 mm, but x_fd,t = 37.9 m
  3. Fully developed constant-Ts h = 3.66 k/D: water 204 W/m²K, oil 42.1 W/m²K
  4. In a few-metre oil cooler the temperature profile never develops, so the real h is higher than 3.66 k/D: use a developing-flow correlation for oils

Answer: Water develops in about 2.2 m; oil needs about 38 m

Warm air in a rectangular duct · Duct 300 × 150 mm; air at 30 °C, 5 m/s, losing heat to cooler surroundings (cooled: n = 0.3); ρ 1.164, μ 1.872×10⁻⁵, k 0.02588, Pr 0.7282
  1. Dh = 4A/P = 4 × 0.045/0.9 = 0.20 m
  2. Re = ρ V Dh/μ = 62,179: turbulent
  3. Dittus-Boelter (cooling): Nu = 0.023 Re^0.8 Pr^0.3 = 143.0, h = 18.5 W/m²K
  4. Gnielinski: Nu = 126.4, h = 16.4 W/m²K

Answer: h ≈ 16-19 W/m²K on the inside of the duct

Chapter 12

Free and mixed convection

Which syllabus covers thisAU ME3691 Unit IIVTU BME601 Module 4SPPU 302042 Unit 3MU MEC502 Module 3KTU PCMET503 Module 2

What drives free convection, and what do Gr and Ra measure?

Air next to a hot surface warms, expands and becomes lighter, so it rises and cooler air flows in to replace it: no fan needed. The Grashof number compares the buoyancy force with the viscous force that resists it, using the expansion coefficient β (1/T in kelvin for a gas: 1/323 at 50 °C). The Rayleigh number Ra = Gr Pr plays the role Re plays in forced flow: it sets the Nusselt number and whether the layer is laminar (on a vertical plate, turbulent above about 10⁹). Free-convection h in air is small, typically 2-10 W/m²K.

How are vertical and inclined plates handled?

On a vertical plate the rising layer starts thin at the bottom and thickens upward, like a forced boundary layer turned on its end. Churchill and Chu fitted one formula over all Ra; in the laminar range it sits within about 17% of the older 0.59 Ra^(1/4). For a plate tilted up to 60° from vertical, the hot face looking down is handled by replacing g with g cos θ in Ra, because only the component of gravity along the plate drives the flow there.

Free convection from plates, cylinders and spheres: Grashof and Rayleigh numbers, h and heat loss, with the mixed-convection check: interactive loads as you scroll

Why does a hot plate facing up lose more than one facing down, and what about cylinders and spheres?

Above a hot plate facing up, warm air can rise freely as plumes, and cool air sweeps in from the edges: Nu = 0.54 Ra^(1/4) (to Ra 10⁷), then 0.15 Ra^(1/3). Under a hot plate facing down, warm air is trapped and must creep sideways to the edges before rising: Nu = 0.27 Ra^(1/4), half as much. The length scale is Lc = area/perimeter. A cold plate is the same picture upside down. Horizontal cylinders (steam lines) and spheres have their own Churchill correlations; a sphere again never drops below Nu = 2.

When do both free and forced convection matter, and what about enclosures?

Compare buoyancy with inertia through Gr/Re². Well below 1 (a common rule of thumb is under 0.1) forced convection dominates; well above (over 10) free convection does; in between both matter, aiding or opposing each other. A 0.5 m plate 40 K above the air with a 0.4 m/s draft gives Gr/Re² ≈ 3.8: mixed. In a horizontal enclosure heated from below, such as the air gap of a flat solar collector, the air stays still until Ra (on the gap width) exceeds about 1,708; below that the gap simply conducts, so h = k/L.

Predict, then check

The same 0.5 m square hot plate is laid horizontal in still air. Which way loses more heat?

Worked examples

A hot vertical panel in a room · Panel 0.5 × 0.5 m at 70 °C in still air at 30 °C; film 50 °C: ν = 1.963×10⁻⁵/1.092 m²/s, k 0.02735, Pr 0.7228, β = 1/323.15 K⁻¹
  1. Gr = 9.81 × (1/323.15) × 40 × 0.5³/ν² = 4.697e+08
  2. Ra = Gr Pr = 3.395e+08: laminar (below 10⁹)
  3. Churchill-Chu: Nu = 88.5 (the simple 0.59 Ra^¼ gives 80.1)
  4. h = 88.5 × 0.02735/0.5 = 4.84 W/m²K; Q = h A ΔT = 4.84 × 0.25 × 40 = 48.4 W
  5. With a 0.4 m/s draft along it: Re = 11,126, Gr/Re² = 3.8: mixed convection, so neither result alone is safe

Answer: h ≈ 4.8 W/m²K, about 48 W from one face

A bare steam line in a plant room · Horizontal pipe 100 mm outside diameter at 150 °C; still air and walls at 30 °C; film 90 °C: ν = 2.139×10⁻⁵/0.9718, k 0.03024, Pr 0.7132, β = 1/363.15; emissivity 0.8 (typical oxidised steel)
  1. Ra_D = g β ΔT D³ Pr/ν² = 4.772e+06
  2. Churchill-Chu (cylinder): Nu = {0.60 + 0.387 Ra^⅙/[1 + (0.559/Pr)^(9/16)]^(8/27)}² = 22.8
  3. h = 22.8 × 0.03024/0.1 = 6.88 W/m²K; convection = h π D ΔT = 259 W/m
  4. Radiation = ε σ π D (Ts⁴ − Tsur⁴) = 337 W/m: larger than the free convection

Answer: About 596 W per metre in all (259 by free convection, 337 by radiation): a case for insulation

Face up or face down · The 0.5 × 0.5 m plate at 70 °C laid horizontal in air at 30 °C (same film properties)
  1. Lc = A/P = 0.25/2 = 0.125 m; Ra = 5.305e+06 (between 10⁴ and 10⁷)
  2. Hot face up: Nu = 0.54 Ra^¼ = 25.9, h = 5.67 W/m²K, Q = 56.7 W
  3. Hot face down: Nu = 0.27 Ra^¼ = 13.0, h = 2.84 W/m²K, Q = 28.4 W

Answer: 57 W facing up against 28 W facing down (the vertical panel: 48 W)

Chapter 13

Boiling: pool boiling regimes and critical heat flux

Which syllabus covers thisAU ME3691 Unit IIIVTU BME601 Module 5SPPU 302042 Unit 3MU MEC502 Module 3KTU not in syllabus

What are the regimes of the pool boiling curve?

Plot the heat flux from a heater in a pool of saturated water against the excess temperature ΔTe = Ts − Tsat (Nukiyama's experiment). Up to a few kelvin there are no bubbles: free convection. Then bubbles form at pits and scratches and stir the liquid violently: nucleate boiling, where q rises as ΔTe³. At the critical heat flux bubbles merge into columns and vapour starts to blanket the surface: past it, in transition boiling, q falls as ΔTe rises. At the Leidenfrost point a stable vapour film covers the heater: film boiling, the reason a water drop skates on a very hot pan.

The pool boiling curve for water: natural convection, nucleate boiling, critical heat flux, transition and film boiling: interactive loads as you scroll

How is nucleate boiling predicted, and what is Csf?

Rohsenow's correlation treats each bubble as a tiny pump stirring the liquid. The surface-fluid constant Csf holds everything about how many nucleation sites the surface offers: for water on scored copper 0.0068, polished copper 0.0128, mechanically polished stainless steel 0.0132, brass 0.0060 (n = 1 for water). Because q ∝ (ΔTe/Csf)³, a rougher surface with a lower Csf boils the same flux at a smaller ΔTe. Scatter is large: a factor of two in q is common, but because of the cube that is only about 26% in ΔTe.

What is the critical heat flux, and why does it mean burnout?

The critical heat flux is the most a boiling surface can carry before vapour leaving it chokes the liquid coming back. Zuber derived it from the stability of the vapour columns: about 1.26 MW/m² for water at 1 atm under a large horizontal plate (C = 0.149) and 1.11 MW/m² on a horizontal cylinder (C = 0.131). If the heater's power is fixed, as in an electric element or a nuclear fuel rod, pushing past q″max forces it onto the film-boiling branch at a far higher temperature, often above the melting point: burnout.

How is film boiling estimated, and where is the Leidenfrost point?

In film boiling the heat crosses a smooth vapour layer by conduction and by radiation. Bromley's analysis, like Nusselt's condensation film, gives Nu = 0.62[g ρv(ρl − ρv) h′fg D³/(μv kv ΔTe)]^(1/4) for a horizontal cylinder, with the latent heat raised to h′fg = hfg + 0.8 cp,v ΔTe for the superheated vapour. Above about 300 °C radiation adds noticeably. The Leidenfrost point is the minimum flux, about 19 kW/m² for water at 1 atm by Zuber's estimate with C = 0.09.

What changes in flow boiling inside a tube?

In a boiler or evaporator tube the liquid is pumped through while it boils, so the flow pattern changes along the tube as vapour builds up: bubbly flow, then slug flow, then annular flow with a liquid film on the wall and a vapour core, and finally mist flow once the wall film dries out. Heat transfer is best in the annular region and drops sharply at dryout, the flow-boiling version of burnout. Forced flow also lifts the critical heat flux well above the pool value.

Predict, then check

An electric heater in water is set to a fixed power just above the critical heat flux. What happens?

Worked examples

Nucleate boiling from a polished copper vessel · Vessel base 200 mm diameter, polished copper (Csf 0.0128, n = 1), water at 1 atm, ΔTe = 8 K. Saturated water at 100 °C: ρl 957.9, ρv 0.5978, hfg 2,257 kJ/kg, cpl 4,217, μl 0.282×10⁻³, Prl 1.75; σ 0.0589 N/m
  1. [g(ρl − ρv)/σ]^½ = 399.3 m⁻¹
  2. cpl ΔTe/(Csf hfg Prl) = 4,217 × 8/(0.0128 × 22,57,000 × 1.75) = 0.6673
  3. q″ = μl hfg × 399.3 × 0.6673³ = 75,513 W/m²; h = q″/ΔTe = 75,513/8 = 9,439 W/m²K
  4. Boiling rate = q″ A/hfg = 75,513 × 0.0314/22,57,000 = 3.78 kg/h
  5. A polished stainless base (Csf 0.0132) needs ΔTe = 8 × 0.0132/0.0128 = 8.25 K for the same flux

Answer: About 76 kW/m², boiling off 3.8 kg of water an hour

Critical and minimum heat flux for water at 1 atm · Same saturated properties; Zuber C = 0.149 (large plate), 0.131 (horizontal cylinder); minimum flux C = 0.09
  1. [σ g (ρl − ρv)/ρv²]^¼ = 6.272
  2. Plate: q″max = 0.149 × 22,57,000 × 0.5978 × 6.272 = 1.26 MW/m²
  3. Horizontal tube: × 0.131/0.149 = 1.11 MW/m²
  4. On polished copper the tube reaches it at ΔTe = (q″max/K)^⅓ = 19.6 K (K = q″/ΔTe³ from the Rohsenow example)
  5. q″min = 0.09 ρv hfg [σ g (ρl − ρv)/(ρl + ρv)²]^¼ = 19.0 kW/m²

Answer: q″max ≈ 1.26 MW/m² (plate), 1.11 MW/m² (tube); q″min ≈ 19 kW/m²

Film boiling on a 10 mm tube at 500 °C · ΔTe = 400 K; vapour at the film temperature 300 °C: ρv 0.3831, cp,v 1,997, kv 0.04345, μv 2.045×10⁻⁵; emissivity 0.8 (typical oxidised metal)
  1. h′fg = 22,57,000 + 0.8 × 1,997 × 400 = 28,96,040 J/kg
  2. Bromley: hc = 0.62 [kv³ ρv (ρl − ρv) g h′fg/(μv D ΔTe)]^¼ = 198.2 W/m²K
  3. Radiation: hr = ε σ (773.15⁴ − 373.15⁴)/400 = 38.3 W/m²K
  4. Combined: h^(4/3) = hc^(4/3) + hr h^(1/3), solved by iteration: h = 227.6 W/m²K (the shortcut hc + ¾ hr gives 227.0)
  5. q″ = h ΔTe = 91.1 kW/m², only 8% of the CHF the same tube carries at about 20 K

Answer: About 91 kW/m² at ΔTe = 400 K

Chapter 14

Condensation: Nusselt film theory and dropwise condensation

Which syllabus covers thisAU ME3691 Unit IIIVTU BME601 Module 5SPPU 302042 Unit 3MU MEC502 Module 3KTU not in syllabus

What is the difference between film and dropwise condensation?

Vapour condenses on any surface colder than its saturation temperature. If the liquid wets the surface, it spreads into a continuous film that flows down under gravity; every joule of latent heat must then be conducted across that film, so the film is the main resistance. If the surface repels the liquid, it gathers into droplets that roll off and keep bare metal exposed. NPTEL's notes put dropwise rates at up to 10 times film rates, but promoter coatings wear off, so designers use film condensation values. Air leaking into a condenser is worse than either: it collects at the liquid surface and blocks the vapour, which is why power-plant condensers run air-extraction ejectors.

What did Nusselt assume, and what does the film on a vertical plate look like?

Nusselt took a laminar film, pulled down by gravity against viscous shear, with heat crossing it purely by conduction (a straight temperature line from Tsat to the wall) and constant properties. A mass and energy balance then gives a film that grows as x^(1/4) from the top, so the local h = k/δ falls as x^(−1/4) and the plate average is 4/3 of the value at the bottom. For steam at 1 atm on a 0.8 m plate 10 K below Tsat, the film at the bottom is only 0.13 mm thick and h is about 6,800 W/m²K. Liquid properties are taken at the film temperature (Tsat + Tw)/2.

Nusselt film condensation on a vertical plate, a horizontal tube and a tier of tubes: film thickness, h and condensate rate: interactive loads as you scroll

Why do condensers use horizontal tubes, and what happens in a tube bank?

On a horizontal tube the film drains round the circumference, so the length that matters is the diameter D, not the height. The same theory gives 0.729 in place of 0.943 with D in place of L. A 25 mm tube condenses about 1.9 times as much steam lying horizontal as standing 1 m tall. In a vertical column of N tubes, condensate from the upper tubes drips onto the lower ones and thickens their films; Nusselt's estimate treats the column as one tube of diameter N D, so the average h falls as N^(−1/4): 12 tubes give 54% of the single-tube value.

Why the modified latent heat, and when does the film stop being laminar?

The condensate leaves colder than Tsat, so each kilogram gives up its latent heat plus some sensible heat; the usual correction adds 0.68 cp ΔT to hfg, about 1.3% for steam with 10 K of subcooling. The film Reynolds number Re = 4ṁ/(p μ), with ṁ the condensate leaving and p the wetted width, sets the regime: smooth laminar below 30, wavy-laminar from 30 to 1,800, turbulent above. Ripples thin the film in places, so in the wavy range a separate correlation gives a higher h than Nusselt's: 17% higher for the 0.8 m plate at 1 atm. Long vertical tubes in big condensers can reach turbulence, where other correlations apply.

Predict, then check

A 25 mm tube, 1 m long, condenses steam at 0.1 bar with the wall 10 K below Tsat. Should it stand vertical or lie horizontal?

Worked examples

Steam condensing on a vertical plate · Saturated steam at 1 atm (Tsat 100 °C) on a plate 0.8 m tall held at 90 °C; per metre of width
  1. Film temperature 95 °C: ρl = 961.5 kg/m³, cp = 4,212 J/kgK, k = 0.677 W/mK, μ = 0.297 mPa·s; at 100 °C ρv = 0.5978 kg/m³, hfg = 2,257 kJ/kg
  2. h′fg = 2,257 + 0.68 × 4.212 × 10 = 2285.6 kJ/kg
  3. Nusselt: h = 0.943 [9.81 × 961.5 × 960.9 × 2.286e+06 × 0.677³/(2.970e-04 × 0.8 × 10)]^¼ = 6,801 W/m²K
  4. Check Re = 4 h L ΔT/(μ h′fg) = 321: above 30, so the film is wavy
  5. Wavy correlation: Re = [4.81 + 3.70 L k ΔT (g/ν²)^⅓/(μ h′fg)]^0.82 = 377, h = Re k (g/ν²)^⅓/(1.08 Re^1.22 − 5.2) = 7,989 W/m²K
  6. q = 7,989 × 0.8 × 10 = 63.9 kW per metre; condensate = q/h′fg = 101 kg/h per metre

Answer: h ≈ 7,989 W/m²K (wavy film; Nusselt alone gives 6,801), about 101 kg/h of condensate per metre of width

A column of tubes in a surface condenser (illustrative numbers) · Turbine exhaust steam at 0.1 bar; 25 mm OD tubes, outer wall 10 K below Tsat, 12 tubes in a vertical column; per metre of tube length. Illustrative, not a real plant's data
  1. Tsat at 10 kPa (interpolated in ln P between 45 and 50 °C) = 45.8 °C; hfg = 2,393 kJ/kg; ρv = 0.0684 kg/m³
  2. Film at 40.8 °C: ρl = 991.8 kg/m³, k = 0.6320 W/mK, μ = 0.6436 mPa·s, cp = 4,179 J/kgK; h′fg = 2421.4 kJ/kg
  3. Single tube: h = 0.729 [g ρl(ρl − ρv) h′fg k³/(μ D ΔT)]^¼ = 10,087 W/m²K
  4. 12 tubes: h = 10,087 × 12^(−¼) = 5,419 W/m²K
  5. Heat per metre of the column = 5,419 × π × 0.025 × 12 × 10 = 51.1 kW; condensate = 75.9 kg/h
  6. Film Re at the bottom (each side) = 4 × (ṁ/2)/μ = 66: rippled but far from turbulent

Answer: Average h ≈ 5,419 W/m²K for the column, 76 kg/h of condensate per metre of tube

Vertical or horizontal? · The same 25 mm tube, 1 m long, steam at 0.1 bar, wall 10 K below Tsat
  1. Vertical (plate formula with L = 1 m): h = 0.943 [B/(1 × 10)]^¼ = 5,188 W/m²K
  2. Horizontal: h = 0.729 [B/(0.025 × 10)]^¼ = 10,087 W/m²K (B is the same property group)
  3. Ratio = (0.729/0.943)(1/0.025)^¼ = 1.94

Answer: Horizontal condenses 1.94 times as much steam

Chapter 15

Heat exchangers: types, overall coefficient, fouling and LMTD

Which syllabus covers thisAU ME3691 Unit IIIVTU BME601 Module 5SPPU 302042 Unit 6MU MEC502 Module 4KTU PCMET503 Module 2

What kinds of heat exchanger are there?

A double-pipe exchanger is one tube inside another, for small duties, run in parallel flow (both fluids the same way) or counter flow (opposite ways). Shell-and-tube units put a tube bundle in a shell, with baffles steering the shell fluid across the tubes: the workhorse of refineries and power plants. Plate exchangers stack thin corrugated plates and suit liquids with modest pressures. Compact exchangers, such as a car radiator, add fins on the gas side, where h is low. In cross flow the two streams meet at right angles, each either mixed (free to move sideways) or unmixed (held in tubes or fin channels). A regenerator stores heat in a matrix that the hot and cold gases visit in turn.

How is the overall coefficient U built up?

Heat crosses five resistances in series: the inside film, any inside deposit, the tube wall, any outside deposit and the outside film. Adding them, with each resistance referred to the same area (usually the outside of the tube), gives 1/U. The smallest h controls the answer, which is why gas-side fins help so much. A thin metal wall rarely matters: for a 25 mm brass tube it is a few per cent of the total. Deposits are a different story, as the fouling factors show.

What is a fouling factor, and how large is it?

Scale, silt, corrosion products and biological growth build up on the surfaces in service. Designers allow for them as an extra resistance Rf = 1/U(fouled) − 1/U(clean), in m²K/W. Published values (Engineering ToolBox, TEMA-style tables): steam 0.00009, seawater below 52 °C 0.00009, treated cooling-tower water 0.00018, average river water 0.00053, fuel oil 0.0009. The effect is large when U is high: a clean U of 1,000 W/m²K with 0.0005 of fouling falls to 667 W/m²K, so the exchanger needs 50% more area. Over-generous fouling factors oversize exchangers, which then run with low velocities that foul faster.

Why is the mean temperature difference a logarithmic mean?

Along the exchanger the gap between the fluids changes, and the local heat flow is proportional to that gap. Writing the energy balance on a short slice and integrating shows the gap varies exponentially with area, so the right average is (ΔT1 − ΔT2)/ln(ΔT1/ΔT2), taken between the two ends, and Q = U A ΔTlm. Counter flow keeps the gap more even and has a larger ΔTlm for the same four temperatures, so it needs less area; it also lets the cold fluid leave hotter than the hot fluid leaves, which parallel flow can never do.

Heat exchanger sizing by LMTD: parallel, counter, shell-and-tube and cross flow with the correction factor and the area needed: interactive loads as you scroll

What does the correction factor F do?

Shell-and-tube units with several tube passes and cross-flow units are partly parallel, partly counter flow. Their area is found from the counter-flow ΔTlm times a factor F ≤ 1, read from charts or computed from P = (tc,out − tc,in)/(Th,in − tc,in) and R = (Th,in − Th,out)/(tc,out − tc,in). F = 1 when one fluid changes phase. F falls steeply as the outlets approach a temperature cross (cold outlet above hot outlet); designers usually avoid low F by adding shells in series, because a small error in the data then means a large error in area.

Predict, then check

Oil is cooled from 120 to 60 °C by water warming from 25 to 45 °C. Which arrangement needs the least area?

Worked examples

Clean and fouled U of a condenser tube · Brass tube 25 mm OD, 22.6 mm ID, k = 100 W/mK; river water inside, h = 5,000 W/m²K; condensing steam outside, h = 10,000 W/m²K (h and k illustrative); fouling: river water 0.000529, steam 0.00009 m²K/W
  1. Inside film on outside area: 0.025/(0.0226 × 5,000) = 2.212e-04 m²K/W
  2. Inside fouling: 0.000529 × 0.025/0.0226 = 5.851e-04; wall: 0.025 ln(25/22.6)/(2 × 100) = 1.26e-05
  3. Outside film 1/10,000 = 1.0e-4; outside fouling 9.0e-5
  4. Clean: Uo = 1/(3.339e-04) = 2,995 W/m²K
  5. Fouled: Uo = 1/(1.009e-03) = 991 W/m²K

Answer: U falls from 2,995 to 991 W/m²K: the fouled tube needs 3.0 times the area, mostly because of the river-water deposit

Oil cooler: counter vs parallel flow · Oil 120 → 60 °C, water 25 → 45 °C, Q = 50 kW, U = 300 W/m²K (illustrative)
  1. Counter: ends 120 − 45 = 75 K and 60 − 25 = 35 K; ΔTlm = 40/ln(75/35) = 52.48 K
  2. Parallel: ends 120 − 25 = 95 K and 60 − 45 = 15 K; ΔTlm = 80/ln(95/15) = 43.34 K
  3. A = Q/(U ΔTlm): counter 3.18 m², parallel 3.85 m²

Answer: Counter flow needs 3.18 m², parallel flow 3.85 m² (21% more)

The same duty in a 1-2 shell-and-tube unit · As above, oil in the shell, water in two tube passes
  1. P = (45 − 25)/(120 − 25) = 0.2105; R = (120 − 60)/(45 − 25) = 3.0
  2. F = √(R² + 1) ln[(1 − P)/(1 − PR)] ÷ {(R − 1) ln[(2 − P(R + 1 − √(R² + 1)))/(2 − P(R + 1 + √(R² + 1)))]} = 0.920
  3. A = 50,000/(300 × 0.920 × 52.48) = 3.45 m²

Answer: F = 0.92, area 3.45 m²

Chapter 16

Effectiveness-NTU, shell-and-tube design, heat pipes and electronics cooling

Which syllabus covers thisAU ME3691 Unit IIIVTU BME601 Module 5SPPU 302042 Unit 6MU MEC502 Module 4KTU PCMET503 Modules 2, 4

What are effectiveness and NTU?

Effectiveness ε is the heat actually moved divided by the most that could be moved, Cmin (Th,in − Tc,in), where C = ṁ cp is each stream's capacity rate and Cmin the smaller. Only the Cmin fluid could ever span the whole temperature range. NTU = UA/Cmin measures the exchanger's size in thermal terms, and C = Cmin/Cmax the imbalance between streams. For each arrangement ε depends only on NTU and C. Counter flow is always the best; at C = 1 it gives ε = NTU/(1 + NTU), while parallel flow can never pass 1/(1 + C), that is 0.5 for equal streams, however large it is built.

Effectiveness-NTU: the effectiveness curves for each heat exchanger type and the outlet temperatures for a given NTU: interactive loads as you scroll

When is NTU better than LMTD, and what is special about C = 0?

LMTD suits sizing: all four temperatures are known and the area is wanted. When the exchanger already exists and only the inlets are known (rating), LMTD needs trial outlet temperatures, while NTU gives ε directly and the outlets follow in one step. In a condenser or evaporator one stream changes phase at constant temperature, so its capacity rate is effectively infinite, C = 0, and every arrangement gives ε = 1 − e^(−NTU). Effectiveness curves also show diminishing returns: past NTU of about 3, extra area buys little.

What are the steps in designing a shell-and-tube exchanger?

Fix the duty, flows and temperatures, and the allowed pressure drops. Decide which fluid goes in the tubes: the fouling, corrosive, high-pressure or more expensive-alloy fluid usually does, because tubes are easier to clean and cheaper to make strong. Choose a TEMA type, then tube diameter, length, pitch and number of passes, and baffle spacing. Estimate both film coefficients and the fouling, find U, then the area from Q = U A F ΔTlm. Check the velocities and pressure drops, adjust, and repeat until area and pressure drop both fit. Materials range from carbon steel and copper alloys to stainless steel, titanium and nickel alloys, chosen for corrosion as much as conductivity.

What do TEMA letters such as AES and BEM mean?

TEMA names a shell-and-tube unit by three letters: front head, shell, rear head. Front: A is a channel with a removable cover (tubes can be inspected without disturbing the piping), B a cheaper one-piece bonnet. Shell: E is the common one-pass shell, F has a longitudinal baffle for two shell passes, K is a kettle reboiler, X a cross-flow shell. Rear: L, M, N are fixed tubesheets, S a split-ring floating head, T a pull-through floating head, U a U-tube bundle. So BEM is a fixed-tubesheet unit whose shell side can only be cleaned chemically, AES has a removable bundle that tolerates expansion, and AEU uses U-tubes.

How does a heat pipe work, and what limits it?

A heat pipe is a sealed tube with a little working fluid and a wick. The hot end boils the liquid, the vapour rushes to the cold end and condenses, and the wick pulls the liquid back by capillary action. Moving latent heat this way gives an effective conductivity that can approach 100 kW/mK for long pipes, against about 0.4 kW/mK for copper. Copper-water pipes typically run between 20 and 150 °C; ammonia and methanol cover colder duties. Limits: capillary (the wick cannot return enough liquid, the usual one), entrainment, sonic, boiling and viscous. A 1996 Electronics Cooling survey quotes 3-4 mm notebook pipes carrying 6-8 W from the processor to a finned heat sink.

How are electronics and EV battery packs cooled?

Air cooling uses finned heat sinks and fans: cheap and simple, but air's low h and heat capacity limit the heat flux. Liquid cooling pumps water-glycol through cold plates under processors or between battery modules and carries the heat to a radiator; it is the usual choice for high-power chips and many EV packs, because cells must stay within a narrow temperature band and close to each other in temperature. Phase-change materials absorb heat while melting at a fixed temperature and smooth out short peaks, but they conduct poorly and need fins or a conductive matrix. Augmentation is passive (fins, rough surfaces, twisted-tape inserts) or active (pumps, fans, vibration).

Predict, then check

A balanced counter-flow exchanger (C = 1) has NTU = 3. Doubling its area makes ε go from 0.75 to about:

Worked examples

Rating an existing oil heater by NTU · Counter flow; oil (hot) 1.5 kg/s, cp 2.0 kJ/kgK, in at 110 °C; water 1 kg/s, cp 4.18 kJ/kgK, in at 25 °C; UA = 6,000 W/K (illustrative)
  1. Ch = 3,000 W/K = Cmin; Cc = 4,180 W/K
  2. NTU = 6,000/3,000 = 2.0; C = 3,000/4,180 = 0.7177
  3. ε = [1 − e^(−2 × 0.2823)]/[1 − 0.7177 e^(−2 × 0.2823)] = 0.7288
  4. Q = ε Cmin (110 − 25) = 0.7288 × 2,55,000 = 185.9 kW
  5. Oil out = 110 − 1,85,851/3,000 = 48.0 °C; water out = 25 + 1,85,851/4,180 = 69.5 °C

Answer: ε = 0.729, Q ≈ 186 kW, oil leaves at 48.0 °C, water at 69.5 °C

A condenser as a C = 0 exchanger (illustrative numbers) · Steam condensing at 0.1 bar (Tsat 45.8 °C); cooling water 20 kg/s, cp 4.18 kJ/kgK, in at 30 °C; U = 2,500 W/m²K, A = 100 m² (illustrative)
  1. Cmin = 20 × 4,180 = 83,600 W/K (the steam side is Cmax → ∞, C = 0)
  2. NTU = 2,500 × 100/83,600 = 2.990; ε = 1 − e^(−2.990) = 0.9497
  3. Water out = 30 + 0.9497 × (45.8 − 30) = 45.0 °C
  4. Q = 0.9497 × 83,600 × 15.82 = 1.256 MW; steam condensed = Q/hfg = 1.256 × 10⁶/2,393 × 10³ = 0.525 kg/s

Answer: Water leaves at 45.0 °C, within 1 K of the steam; about 0.52 kg/s of steam condensed

Same NTU, different arrangements · NTU = 2, C = 0.5
  1. Parallel: [1 − e^(−3)]/1.5 = 0.6335
  2. Counter: [1 − e^(−1)]/[1 − 0.5 e^(−1)] = 0.7746
  3. 1 shell pass, 2 tube passes: 2/{1.5 + √1.25 (1 + e^(−2√1.25))/(1 − e^(−2√1.25))} = 0.6931
  4. Cross flow, both unmixed (approximate formula): 1 − exp[(2^0.22/0.5)(e^(−0.5 × 2^0.78) − 1)] = 0.7388; the exact series gives 0.7324

Answer: Counter 0.775 > cross 0.732 > 1-2 shell 0.693 > parallel 0.633

Chapter 17

Thermal radiation: Planck, Wien, Stefan-Boltzmann and Kirchhoff

Which syllabus covers thisAU ME3691 Unit IVVTU BME601 Module 3SPPU 302042 Unit 4MU MEC502 Module 3KTU PCMET503 Module 3

What is thermal radiation, and what is a black body?

Every surface above absolute zero emits electromagnetic waves because of the motion of its molecules and electrons. No medium is needed: radiation crosses a vacuum, which is how the sun heats the earth. The part that matters for heat transfer lies roughly between 0.1 and 100 µm, and includes the visible band, 0.38 to 0.76 µm. A black body is the ideal: it absorbs all radiation falling on it, from every direction and at every wavelength, and no surface at the same temperature can emit more. A small hole in a large hollow cavity behaves almost exactly like one, which is how black bodies are built in calibration labs.

What do Planck's law and Wien's law say?

Planck's law gives the black-body emissive power at each wavelength. As temperature rises, emission grows at every wavelength and the peak moves to shorter waves: λmax T = 2,897.8 µm K (Wien). A sun-like source at 5,800 K peaks at 0.50 µm, in the visible; a room-temperature wall at 300 K peaks at 9.7 µm, in the infrared, which is why thermal cameras image warm objects in the dark. At 1,500 K the peak is at 1.9 µm and only 0.13% of the emission is visible, yet that is enough to glow dull red.

Planck's law: spectral emissive power, Wien's peak, Stefan-Boltzmann total and the fraction of energy in a wavelength band: interactive loads as you scroll

What is the Stefan-Boltzmann law, and how is a band of wavelengths handled?

Integrating Planck's curve over all wavelengths gives the total: Eb = σT⁴, σ = 5.67 × 10⁻⁸ W/m²K⁴. A wall at 300 K emits 459 W/m²; double the absolute temperature and it emits 16 times as much, which is why radiation dominates in furnaces. The share emitted between 0 and λ depends only on the product λT, so one table (or series) of F(0→λT) serves every temperature; the power in a band λ1-λ2 is [F(λ2T) − F(λ1T)] σT⁴. A quarter of the emission lies below the Wien peak.

What are emissivity, absorptivity, reflectivity and transmissivity?

Emissivity ε compares a real surface's emission with a black body's at the same temperature; it can vary with wavelength (spectral ε) and direction. Incoming radiation is partly absorbed (α), reflected (ρ) and transmitted (τ), with α + ρ + τ = 1; most solids are opaque, so α + ρ = 1. A gray surface has the same ε at every wavelength, and a diffuse one the same in every direction: the usual engineering idealisations. Real surfaces can be strongly selective: a solar absorber coating is made to absorb the short-wave sunlight well yet emit poorly in the infrared, so it stays hot.

What is Kirchhoff's law, and how are intensity and emissive power related?

Put a body inside a closed enclosure at one temperature. At equilibrium it must emit exactly what it absorbs, at each wavelength and direction, so ελ = αλ. For a gray surface this becomes ε = α at any temperature, which is what makes radiation networks solvable. Intensity I is the power leaving per unit area normal to the direction, per unit solid angle (steradian). A diffuse surface has the same intensity in all directions, so the power it sends at an angle θ falls as cos θ (Lambert's law); summing over the hemisphere gives E = π I.

Predict, then check

A black surface is heated from 300 K to 600 K. Its emitted power becomes:

Worked examples

A furnace wall seen through a peephole · Interior treated as a black body at 1,500 K
  1. Peak: λmax = 2,897.8/1,500 = 1.932 µm
  2. Total: Eb = 5.67 × 10⁻⁸ × 1,500⁴ = 2,87,044 W/m²
  3. Visible: F(0.76 × 1,500 = 1,140) − F(0.38 × 1,500 = 570) = 0.13%
  4. Band 1-5 µm: F(7,500) − F(1,500) = 0.8215, so 2,35,811 W/m²

Answer: Peak at 1.93 µm, 287 kW/m² in all, 82% of it between 1 and 5 µm and only 0.13% visible

Why a filament lamp is a heater that glows · Black bodies at 5,800 K (sun-like) and 2,800 K (a filament temperature, illustrative); visible 0.38-0.76 µm
  1. 5,800 K: λT from 2204 to 4408 µm K; visible fraction = 44.8%; peak at 0.50 µm
  2. 2,800 K: λT from 1064 to 2,128 µm K; visible fraction = 8.7%; peak at 1.03 µm
  3. At 2,800 K, 91% of the emission is infrared, beyond 0.76 µm

Answer: 45% visible at 5,800 K but only 8.7% at 2,800 K: most of a filament's power leaves as heat

Gray vs selective surface at 400 K · Gray surface ε = 0.6; selective surface (illustrative) ε = 0.9 below 2.5 µm and 0.1 above
  1. Eb = 5.67 × 10⁻⁸ × 400⁴ = 1451.5 W/m²; diffuse black intensity I = Eb/π = 462.0 W/m²·sr
  2. Gray: E = 0.6 × 1451.5 = 870.9 W/m²
  3. F(0 → 2.5 × 400 = 1,000 µm K) = 3.21e-04: almost nothing is emitted below 2.5 µm at 400 K
  4. Selective: E = 1451.5 × (0.9 × 3.21e-04 + 0.1 × 0.99968) = 145.5 W/m²

Answer: The selective surface loses 146 W/m² against 871 W/m² for the gray one: it radiates like an ε = 0.1 surface

Chapter 18

View factors and radiation exchange between black surfaces

Which syllabus covers thisAU ME3691 Unit IVVTU BME601 Module 3SPPU 302042 Unit 4MU MEC502 Module 3KTU PCMET503 Module 3

What is a view factor?

The view factor F12 is the fraction of the radiation leaving surface 1 that lands directly on surface 2. For diffuse surfaces it depends only on geometry: sizes, distance and orientation, not on temperature or emissivity. It comes from adding up, over every pair of small patches, how squarely they face each other (cos θ1 cos θ2) and how far apart they are (1/πR²). Two 1 m squares facing each other 1 m apart have F12 = 0.20: four fifths of what one emits misses the other.

Which rules let you find view factors without integrating?

Reciprocity: A1F12 = A2F21, because the double integral is the same either way round. Summation: in a closed enclosure the factors from any surface add to 1, since every ray lands somewhere. A flat or convex surface cannot see itself, so F11 = 0; a concave one can. Symmetry: equal surfaces placed alike get equal factors. In a cube, opposite faces have 0.1998 and each of the four adjacent faces 0.2000, and 0.1998 + 4 × 0.2000 = 1, which is a neat check on both formulas.

Where do the closed-form view factors come from, and how do you use them?

For common shapes the double integral has been done once and tabulated, for example in J. R. Howell's online catalogue of configuration factors: parallel rectangles, perpendicular rectangles sharing an edge, coaxial discs and dozens more. You compute the dimensionless ratios (sides over gap, radii over gap), read or evaluate F12, then get the reverse factor by reciprocity and the rest by summation. Move the plates closer and F12 climbs towards 1; shrink the receiving surface and it falls. The interactive draws a few sample rays to show the idea.

View factors for parallel rectangles, perpendicular rectangles, coaxial discs and long strips (crossed strings), reciprocity and black-surface exchange: interactive loads as you scroll

How does Hottel's crossed-strings method work for long ducts?

For surfaces that are very long in one direction (ducts, channels, long fins), stretch imaginary strings between the ends of the two surfaces in the cross-section. Then F12 = (sum of crossed strings − sum of uncrossed strings) / (2 × width of surface 1). Two directly opposed strips 1 m wide and 0.5 m apart: crossed strings are 2 × √1.25 = 2.236 m, uncrossed 2 × 0.5 = 1 m, so F12 = (2.236 − 1)/2 = 0.618, which is Howell's √(1 + H²) − H with H = 0.5.

How much heat do black surfaces exchange, and how does a black enclosure balance?

A black surface absorbs everything and emits σT⁴. Of what surface 1 emits, A1F12σT1⁴ reaches surface 2, and A2F21σT2⁴ comes back; by reciprocity the net is A1F12σ(T1⁴ − T2⁴). In an enclosure of N black surfaces, the net heat leaving surface i is the sum of such terms over all j. A furnace heater panel at 627 °C over a tray at 127 °C, with F = 0.51 and 2 m² of panel, sends about 36 kW to it.

Predict, then check

Two 1 m × 1 m plates face each other 1 m apart. Roughly what fraction of the radiation leaving one strikes the other?

Worked examples

Heater panel over a tray · Panel 2 m × 1 m, 0.5 m above a tray of the same size, directly opposed; both treated as black (illustrative); panel 900 K, tray 400 K
  1. X = a/c = 2/0.5 = 4, Y = b/c = 1/0.5 = 2
  2. Howell C-11: F12 = 0.5090; same size, so F21 = F12
  3. Q = A1 F12 σ (T1⁴ − T2⁴) = 2 × 0.5090 × 5.67 × 10⁻⁸ × (900⁴ − 400⁴) = 36,392 W
  4. About 49% of the panel's radiation escapes through the sides

Answer: F12 = 0.509, Q ≈ 36.4 kW

Floor and wall of a room · Floor 4 m × 3 m, wall 4 m long and 2.8 m high, sharing the 4 m edge; underfloor heating at 30 °C, wall at 22 °C, both treated as black (upper bound)
  1. Common edge l = 4 m: W = w/l = 3/4 = 0.75, H = h/l = 2.8/4 = 0.7
  2. Howell C-14: F(floor → wall) = 0.2132
  3. Reciprocity: F(wall → floor) = A_floor F/A_wall = 12 × 0.2132/11.2 = 0.2284
  4. Q = 12 × 0.2132 × σ × (303.15⁴ − 295.15⁴) = 124 W

Answer: F floor→wall = 0.213, wall→floor = 0.228; about 124 W to that one wall

View factors inside a cylindrical cavity · A blind hole of radius 0.1 m and depth 0.2 m: surfaces are the bottom (1), the side wall (2) and the opening (3)
  1. Bottom to opening are coaxial discs r = 0.1 m, 0.2 m apart (C-41): F13 = 0.1716
  2. Bottom is flat (F11 = 0), so F12 = 1 − F13 = 0.8284
  3. Reciprocity: F21 = A1 F12/A2 = (π × 0.01 × 0.8284)/(2π × 0.1 × 0.2) = 0.2071; by symmetry F23 = F21
  4. Summation: F22 = 1 − 2 × 0.2071 = 0.5858 (the curved wall sees itself)

Answer: F13 = 0.172, F12 = 0.828, F21 = F23 = 0.207, F22 = 0.586

Chapter 19

Gray surfaces, radiosity networks and radiation shields

Which syllabus covers thisAU ME3691 Unit IVVTU BME601 Module 3SPPU 302042 Unit 4MU not in syllabusKTU PCMET503 Module 3

What is radiosity, and what are the surface and space resistances?

A gray, diffuse, opaque surface emits εEb and reflects (1 − ε) of the irradiation G falling on it. Radiosity J = εEb + (1 − ε)G is everything that leaves it. The net heat leaving is (Eb − J)/[(1 − ε)/(εA)]: a surface resistance that vanishes for a black body. Between two surfaces the net exchange is (Ji − Jj)/[1/(AiFij)]: a space resistance set by geometry alone. Radiation problems then become circuits, with Eb as the battery voltage and q as the current.

How do you handle two-surface enclosures: parallel plates, concentric cylinders and spheres?

Two surfaces that see only each other make a series circuit: surface resistance, space resistance, surface resistance. Large parallel plates (A1 = A2, F12 = 1) give q = σ(T1⁴ − T2⁴)/(1/ε1 + 1/ε2 − 1) per m². For a long cylinder or sphere inside another, F12 = 1 from the inner one, so q = σA1(T1⁴ − T2⁴)/[1/ε1 + (1 − ε2)/ε2 × A1/A2] with A1/A2 = r1/r2 or (r1/r2)². A small body in a large room (A1/A2 → 0) loses εσA1(T1⁴ − T2⁴), whatever the room's emissivity.

What is a reradiating surface?

An insulated wall in a furnace (refractory lining with negligible loss) absorbs radiation and sends it all back out, so its net heat is zero and J = Eb. In the network its surface resistance carries no current and drops out, but its space resistances form a parallel path between the heater and the load. In a long triangular duct with a 1,000 K heater (ε 0.8) and a 500 K load (ε 0.6), the insulated third side's parallel path lifts the heat transfer from 18,225 W per metre (direct path alone) to 23,625, and the wall settles at 887 K.

How do radiation shields work, and how much do they save?

A shield is a thin, low-emissivity sheet placed between two surfaces. It adds two surface resistances and one more space resistance to the series chain, without blocking anything by conduction. Between large plates with all emissivities equal, N shields cut the heat flux to 1/(N + 1). Low-ε shields do far better: one aluminium-foil shield (ε ≈ 0.04) between oxidised steel plates (ε ≈ 0.79) at 800 K and 350 K cuts the flux by 97%. Multi-layer insulation in cryogenic tanks and spacecraft stacks many such layers.

Radiation between gray surfaces with N radiation shields: the radiosity network, heat flux and shield temperatures: interactive loads as you scroll

Do gases take part in radiation?

Nitrogen, oxygen and dry air are practically transparent to thermal radiation, which is why the networks above ignore the gas between surfaces. Carbon dioxide and water vapour, however, absorb and emit in specific wavelength bands. In a boiler or furnace full of combustion products the gas itself radiates strongly to the tubes, and it is treated with gas emissivity charts. That topic is beyond most semester syllabi and is not covered here.

Predict, then check

One thin shield, with the same emissivity as the two large parallel plates, is slid between them. The heat flux becomes:

Worked examples

Foil shield between two steel plates · Large parallel oxidised steel plates, ε = 0.79 (Engineering ToolBox, typical), at 800 K and 350 K; one aluminium foil shield, ε = 0.04 both sides
  1. σ(T1⁴ − T2⁴) = 5.67 × 10⁻⁸ × (800⁴ − 350⁴) = 22,373 W/m²
  2. No shield: q = 22,373/(1/0.79 + 1/0.79 − 1) = 14,607 W/m²
  3. With the shield: q = 22,373/[(2/0.79 − 1) + (2/0.04 − 1)] = 443 W/m²
  4. Shield temperature: Eb,s = σ × 800⁴ − q (1/0.79 + 1/0.04 − 1), so Ts = 679 K

Answer: 14,607 → 443 W/m², a 97% cut; the foil sits at about 679 K

Heat leak into a vacuum-jacketed liquid nitrogen line · Inner tube 25 mm outside diameter at 77 K inside a 50 mm bore outer tube at 300 K, evacuated gap, both polished stainless steel ε = 0.075 (typical)
  1. A1 per metre = π × 0.025 = 0.07854 m²; r1/r2 = 0.5
  2. q = σ A1 (300⁴ − 77⁴)/[1/0.075 + (1 − 0.075)/0.075 × 0.5] = 1.84 W/m
  3. If both were black: σ A1 (300⁴ − 77⁴) = 35.9 W/m

Answer: About 1.8 W per metre into the nitrogen, 20 times less than with black surfaces

Furnace duct with an insulated wall · Long duct with an equilateral triangular section, 1 m sides: heater 1,000 K (ε 0.8), load 500 K (ε 0.6), third side insulated and reradiating; per metre of length
  1. By symmetry and summation each side sees each other side with F = 0.5, so every space resistance is 1/(1 × 0.5) = 2 m⁻²
  2. Surface resistances: (1 − 0.8)/0.8 = 0.25, (1 − 0.6)/0.6 = 0.667
  3. The direct path (2) in parallel with the path through the reradiating wall (2 + 2): 1.333
  4. q = σ(1,000⁴ − 500⁴)/(0.25 + 1.333 + 0.667) = 23,625 W/m
  5. J3 = (J1 + J2)/2 gives the wall temperature 887 K

Answer: 23,625 W per metre against 18,225 W/m through the direct path alone; the wall runs at about 887 K

Chapter 20

Mass transfer: Fick's law, stationary media and Stefan flow

Which syllabus covers thisAU ME3691 Unit VVTU not in syllabusSPPU 302042 Unit 5MU MEC502 Module 4KTU PCMET503 Module 4

How is the amount of a species in a mixture measured?

Mass transfer moves one species relative to the mixture, either by diffusion (random molecular motion down a concentration gradient) or by convection (carried by bulk flow). Its amount can be a mass density ρA (kg/m³), a molar concentration CA (mol/m³), a mass fraction ωA = ρA/ρ or a mole fraction xA = CA/C. For ideal gases xA = pA/P, the partial-pressure ratio. Humid air at 30 °C and 60% relative humidity holds water vapour at 2.55 kPa: mole fraction 0.025, mass fraction 0.016, 18.2 g/m³.

What is Fick's law, and how big is a diffusion coefficient?

Fick's law is the mass-transfer twin of Fourier's law: the diffusive flux is proportional to the concentration gradient, with the binary diffusion coefficient DAB in place of the thermal diffusivity. NIST's 2024 compilation gives, at 298 K and 1 atm, 0.254 cm²/s for water vapour in nitrogen, 0.166 for methanol and 0.129 for ethanol in air. Gas values grow roughly as T^1.75 and fall as 1/P (kinetic theory gives T^1.5 with a slowly varying collision term). Liquids are about 10⁴ times slower (typical) and solids slower still.

How do you solve diffusion in a stationary medium?

Species conservation on a small volume gives ∂CA/∂t = DAB∇²CA + ṅA (generation by reaction), the same form as the heat equation. Steady, one-dimensional, no reaction: the concentration is linear across a plane wall or membrane and the flux is DAB(CA1 − CA2)/L, with ln(r2/r1) and 1/r forms for cylinders and spheres, exactly like conduction resistances. Unsteady problems borrow the transient conduction solutions; carburising steel is the semi-infinite erf solution, explored in the Engineering Materials book.

Why does evaporation up a tube go faster than Fick's law predicts?

In a Stefan tube, vapour diffuses up from a liquid surface into air. Air diffuses down its own gradient, but it cannot enter the liquid, so a slow upward bulk flow must cancel it exactly. That bulk flow also carries vapour, adding to the diffusive flux. The result replaces the linear Fick difference with a logarithm. For water at 25 °C (x0 = 0.031) the correction is only 1.6%; for methanol at 25 °C (x0 = 0.167) it is 9.4%, and it grows without limit as the liquid nears boiling.

Stefan tube: evaporation through a stagnant gas column, Stefan flow against plain Fick diffusion and the evaporation rate: interactive loads as you scroll
Predict, then check

Water evaporates up a Stefan tube into dry air. Going from 25 °C to 55 °C, the gap between the Stefan-flow answer and plain Fick's law:

Worked examples

Concentrations in humid air · Air at 30 °C, 1 atm, relative humidity 60%; water saturation pressure from the NIST Antoine fit; molar mass of dry air 28.97 g/mol
  1. p_sat(30 °C) = 4.244 kPa, so p_v = 0.6 × 4.244 = 2.546 kPa
  2. Mole fraction x_v = 2546/101,325 = 0.0251
  3. Total molar concentration C = P/(R T) = 101,325/(8.314 × 303.15) = 40.20 mol/m³; C_v = 1.010 mol/m³
  4. ρ_v = 1.010 × 0.018015 = 18.2 g/m³; mass fraction ω_v = ρ_v/ρ = 0.0158

Answer: x_v = 0.0251, ω_v = 0.0158, ρ_v = 18.2 g/m³

Methanol evaporating in a Stefan tube · Methanol at 25 °C, 150 mm of still air above the liquid, dry air sweeping the top, 1 atm; D = 0.166 cm²/s (NIST); liquid density 791 kg/m³; tube bore 10 mm
  1. Antoine (Ambrose-Sprake): p_sat = 16.94 kPa, x0 = 0.1672
  2. C = 101,325/(8.314 × 298.15) = 40.87 mol/m³
  3. Stefan: N = (40.87 × 1.66 × 10⁻⁵/0.15) ln(1/(1 − 0.1672)) = 0.8274 mmol/m²s
  4. Fick only: C D x0/L = 0.7562 mmol/m²s, 9.4% low
  5. Level drop = N M/ρ = 8.274e-04 × 0.032042/791 × 3.6 × 10⁶ = 0.121 mm/h; mass rate 7.5 mg/h

Answer: N ≈ 0.827 mmol/m²s; the level falls about 0.12 mm per hour

Water vapour through a still-air gap · A 20 mm layer of still air at 25 °C, 1 atm, between air at 80% and 30% relative humidity; D = 0.254 cm²/s; vapour is dilute, so plain Fick's law applies
  1. p_sat(25 °C) = 3.167 kPa: x1 = 0.02500, x2 = 0.00938
  2. J = C D (x1 − x2)/L = 40.87 × 2.54 × 10⁻⁵ × 0.01563/0.02 = 0.811 mmol/m²s
  3. Per day: 8.112e-04 × 18.015 × 86,400 = 1,263 g/m²
  4. D at 50 °C by T^1.75: 0.254 × (323.15/298.15)^1.75 = 0.292 cm²/s (NIST fit: 0.294); diffusing 10 mm then takes about L²/D ≈ 3.4 s

Answer: About 0.81 mmol/m²s, or 1.26 kg of water per m² per day

Chapter 21

Convective mass transfer and the heat-mass transfer analogy

Which syllabus covers thisAU ME3691 Unit VVTU not in syllabusSPPU 302042 Unit 5MU MEC502 Module 4KTU PCMET503 Module 4

What is a convective mass transfer coefficient?

When air flows over a wet surface, a concentration boundary layer forms, just as a thermal one does over a hot surface. The evaporation rate is written like Newton's law of cooling: mass flux = hm × (vapour density at the surface − vapour density in the free stream). The coefficient hm has units of m/s. The surface value is the saturation density at the liquid temperature. Its dimensionless form, the Sherwood number Sh = hmL/D, plays the part of the Nusselt number.

What do the Schmidt and Lewis numbers tell you?

The Schmidt number Sc = ν/D compares momentum diffusion with mass diffusion, as Pr = ν/α does for heat. The Lewis number Le = α/D = Sc/Pr compares heat with mass diffusion. In laminar flow the layer thicknesses scale roughly as δ/δc ≈ Sc^⅓ and δt/δc ≈ Le^⅓. For water vapour in air near 30 °C, Sc ≈ 0.61 and Le ≈ 0.86, so the three layers are almost the same thickness. For a solute in water, Sc is in the hundreds or more and the concentration layer is very thin.

How can a heat transfer coefficient give the mass transfer coefficient?

The boundary-layer equations for temperature and concentration have the same form, so a heat transfer correlation Nu = f(Re) Pr^⅓ becomes Sh = f(Re) Sc^⅓ with the same f. Dividing the two gives the Chilton-Colburn analogy hm = h/(ρcp Le^⅔): measure or compute h, and the evaporation coefficient follows. Colburn's j-factors put it as jH = jD = f/2 for turbulent duct flow. For air over a 1 m wet plate at 3 m/s, h = 6.8 W/m²K gives hm = 6.4 mm/s.

Convective mass transfer: evaporation from a wet plate, Sherwood number and the heat-mass transfer analogy: interactive loads as you scroll

Which mass transfer correlations are used for plates, pipes and spheres?

Flat plate: Sh = 0.664 Re^½ Sc^⅓ (laminar) and, with transition at 5 × 10⁵, Sh = (0.037 Re^0.8 − 871) Sc^⅓. A fully turbulent form, 0.036 Re^0.8 Sc^⅓, is also quoted and overestimates on plates that are partly laminar. Pipe (Linton-Sherwood): Sh = 0.023 Re^0.83 Sc^⅓. Sphere (Ranz-Marshall): Sh = 2 + 0.6 Re^½ Sc^⅓, where 2 is pure diffusion into still air. A 1 mm water drop moving at 2 m/s through 25 °C air has Re ≈ 129 and Sh ≈ 7.8.

Why does a wet surface cool below the air temperature?

Evaporation carries latent heat away from the water. At steady state the heat convected in from the warmer air exactly supplies it: h(T∞ − Tw) = hm hfg(ρv,s − ρv,∞). With the analogy, T∞ − Tw = hfg(ρv,s − ρv,∞)/(ρcp Le^⅔), independent of the air speed. Air at 35 °C and 40% relative humidity (a Chennai afternoon, illustrative) gives Tw ≈ 23.3 °C. This is the wet-bulb idea, and it is why desert coolers and cooling towers work best in dry air.

Predict, then check

Air flows over a wet plate in laminar flow. You double the air speed. The evaporation rate:

Worked examples

Evaporation from a tray of water · Water surface 1 m long at 28 °C; air at 32 °C, 50% relative humidity, 3 m/s; dry-air properties at the 30 °C film temperature (Engineering ToolBox); D from NIST TN 2279
  1. At 303.15 K: ν = 1.598 × 10⁻⁵ m²/s, D = 0.2615 cm²/s; Re = 3 × 1/ν = 1,87,721 (laminar)
  2. Sc = ν/D = 0.611; Sh = 0.664 √Re Sc^⅓ = 244.1; hm = Sh D/L = 6.38 mm/s
  3. Check by the analogy: h = 6.78 W/m²K, Le = 0.864, hm = h/(ρcp Le^⅔) = 6.38 mm/s
  4. ρv,s = 27.2 g/m³ (saturated, 28 °C); ρv,∞ = 16.9 g/m³ (50% of saturation at 32 °C)
  5. m″ = hm Δρv = 0.0658 g/m²s = 0.237 kg/m²h; latent heat 160 W/m² against 27 W/m² of sensible gain

Answer: About 0.24 kg/m² per hour; the water keeps cooling, because it loses 160 W/m² and gains only 27 W/m²

Drying a long wet sheet: turbulent flow and the analogy · Wet sheet 2.5 m long at 30 °C; air at 40 °C, 30% relative humidity, 8 m/s (illustrative dryer conditions)
  1. Film 35 °C: Re = 12,15,130, above 5 × 10⁵, so mixed: Nu = (0.037 Re^0.8 − 871) Pr^⅓ = 1653
  2. h = 17.76 W/m²K; Le = 0.866; hm = h/(ρcp Le^⅔) = 16.96 mm/s
  3. Δρv = 30.3 − 15.3 g/m³; m″ = 0.917 kg/m²h
  4. The fully turbulent form 0.036 Re^0.8 Sc^⅓ would give 43% more, because it ignores the laminar front part

Answer: About 0.92 kg/m² per hour; latent 619 W/m² against sensible 178 W/m²

Wet-bulb temperature from the analogy · Air at 35 °C, 40% relative humidity, 1 atm; properties at the film temperature; solve T∞ − Tw = hfg(ρv,s(Tw) − ρv,∞)/(ρcp Le^⅔)
  1. ρv,∞ = 0.4 × saturation at 35 °C = 15.82 g/m³
  2. Guess Tw, evaluate the right side with properties at (35 + Tw)/2, and adjust by bisection
  3. Converges at Tw = 23.3 °C, where ρv,s = 20.91 g/m³
  4. The air speed has cancelled out: h and hm rise together

Answer: Tw ≈ 23.3 °C, about 12 °C below the air

Keep going

Recommended reading

  • Fundamentals of Heat and Mass Transfer, T. L. Bergman, A. S. Lavine, F. P. Incropera, D. P. DeWitt
  • Heat and Mass Transfer: Fundamentals and Applications, Y. A. Cengel, A. J. Ghajar
  • Heat Transfer, J. P. Holman
  • Fundamentals of Engineering Heat and Mass Transfer, R. C. Sachdeva
  • Fundamentals of Heat and Mass Transfer, C. P. Kothandaraman
  • Heat Transfer, S. P. Venkateshan

How this book was made

Chapters follow the official syllabi listed above. Every interactive's formulas and every worked-example number were checked by script. Explanations, animations and examples are original to Skill-Lync. Topics informed by MIT OpenCourseWare 2.51 (Intermediate Heat and Mass Transfer), CC BY-NC-SA. All explanations, animations and examples are original to Skill-Lync; MIT does not endorse this page.

From the exam to a core job

Talk to a counsellor.

Ask how CFD and thermal engineers use heat and mass transfer on real projects, and what the Post Graduate Program in CFD covers.

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.

Talk to a career counsellor

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.