Shell-and-Tube Heat Exchanger

Why this design dominates

Walk into any oil refinery, power station or large chemical plant and you will see hundreds of shell-and-tube heat exchangers. Long, cylindrical, often painted silver or grey, mounted horizontally on saddles or stacked vertically in the structural steel. They are by far the most common heat-exchanger architecture in heavy industry, and they have held that position for more than a century. The reasons are practical rather than thermodynamic.

First, the shell-and-tube design handles essentially any combination of pressures, temperatures, fluids and duties that the rest of the plant can produce. Pressures from full vacuum to several hundred bar. Temperatures from cryogenic to well over 600 °C. Single-phase liquids, gases, two-phase boiling and condensing flows. You will see a 200-bar high-pressure feedwater heater on a power plant and a 2-bar atmospheric kettle reboiler on a crude column built to the same generic recipe — bundle of tubes in a shell — with the same family of mechanical standards.

Second, the geometry is mechanically robust and inspectable. The pressure boundary is two simple shapes — a cylinder and a flat tubesheet — that classical pressure-vessel codes know how to analyse, weld and inspect. The tubes can be cleaned on the inside by mechanical scrapers; the outside can be cleaned by pulling the bundle and water-blasting. Failed tubes can be plugged in service without scrapping the whole unit.

Third, the design is codified to the point of being a commodity. TEMA (the Tubular Exchanger Manufacturers Association) publishes a hundred-page rulebook that every major fabricator builds to. A "TEMA AES exchanger, 24 inch shell, 16 foot tubes, 0.75 inch OD on 1 inch triangular pitch" tells anyone in the industry essentially everything about the mechanical design, and several dozen shops worldwide can quote against it. There is competition without bespoke engineering.

Plate-and-frame, brazed-plate, double-pipe, spiral, finned-air-cooled and printed-circuit exchangers all have niches — usually higher U or lower volume — but for the large, high-pressure, high-temperature duties that drive heavy industry, shell-and-tube is still the default answer.

Anatomy of the exchanger

A shell-and-tube exchanger consists of five major mechanical assemblies.

• The shell. The outer cylindrical pressure vessel. Diameter typically 0.2 m to 1.5 m for process exchangers, up to several metres for surface condensers. It contains the shell-side fluid and supports the tube bundle.
• The tube bundle. Tens to thousands of small-diameter tubes (typically 19 mm or 25 mm outer diameter — three-quarter inch or one inch in U.S. practice), arrayed on a square or triangular pitch. The bundle is the heat-transfer surface and the most expensive single component.
• Tubesheets. Thick flat plates drilled with the tube pattern. Each tube is roller-expanded or welded into a tubesheet hole. The tubesheets separate the tube-side and shell-side fluids and absorb the longitudinal pressure load on the bundle.
• Channel heads (front and rear). Caps on the ends of the shell that distribute and collect the tube-side fluid. A 1-2 exchanger has a front head split by a pass partition so that the tube-side fluid enters one half of the bundle, returns through the rear head, and exits the other half — both nozzles on the front. Various TEMA codes label the head styles A through D and P through W.
• Baffles. Plates inside the shell that support the tubes and force cross-flow. The most common is the segmental baffle, a flat plate cut to 25 percent of the shell diameter so 75 percent of the cross-section is blocked and the shell-side fluid is forced through the 25 percent window before being redirected across the bundle to the next baffle, alternating sides.

Auxiliary parts include impingement plates (to protect the tubes nearest the shell-side inlet from erosion by high-velocity feed), sealing strips (to prevent shell-side bypass between the bundle and the shell wall), and tie rods and spacers (to hold the baffles in position).

How heat actually moves

Heat flows from the hot fluid to the cold fluid through three resistances in series.

1/U = 1/h_i + R_fi + (d_o ln(d_o/d_i))/(2 k_tube) + R_fo + 1/h_o    (per outer-area basis)

Term by term:

• 1/h_i — convective film on the tube-side fluid. h_i typically 1 000 to 10 000 W/m²K for liquid water at industrial velocities; 50 to 500 for low-pressure gas; 20 000 plus for condensing or boiling.
• R_fi — tube-side fouling resistance.
• (d_o ln(d_o/d_i))/(2 k_tube) — conduction through the tube wall. For a thin-walled metal tube this is usually small; for a thick-walled tube or a low-conductivity material it matters.
• R_fo — shell-side fouling resistance.
• 1/h_o — convective film on the shell side. h_o is set by the baffle geometry, the bundle layout and the shell-side velocity; typically 500 to 5 000 W/m²K for cross-flow over a tube bank.

U for clean, well-designed exchangers ranges from about 100 W/m²K for gas-gas exchangers up to 3 000 W/m²K or more for water-water service. Once you know U, the LMTD, and the duty Q, you compute the required area from Q = U·A·LMTD·F and round up to the nearest standard bundle.

The log-mean temperature difference

The driving force for heat transfer is the local temperature difference ΔT between the two streams, but ΔT varies along the length of the exchanger. The right average is the log-mean.

LMTD = (ΔT_1 − ΔT_2) / ln(ΔT_1 / ΔT_2)

where ΔT_1, ΔT_2 are the temperature differences
at the two ends of the exchanger.

For pure counter-flow the LMTD reaches its maximum: the hot stream cools from its inlet to its outlet, the cold stream heats from its inlet to its outlet on the opposite end, and ΔT can be roughly constant along the length. For pure parallel-flow the two streams approach a common temperature; ΔT_2 (at the outlet end) tends toward zero and LMTD collapses. For the same end temperatures, counter-flow LMTD always exceeds parallel-flow LMTD — usually by tens of percent.

Multipass exchangers (1-2, 2-4, and so on) are partly counter-flow and partly parallel-flow. The trick is to compute LMTD as if the exchanger were pure counter-flow, then multiply by a dimensionless correction factor F < 1 that depends on two ratios of the four terminal temperatures. F is read from a chart or computed from a closed-form expression; a well-designed exchanger has F > 0.75. If you find yourself with F below 0.75, you usually add a second shell in series rather than accepting the penalty.

Worked example: counter-flow versus parallel-flow

Suppose hot oil enters at 150 °C and must be cooled to 90 °C. Cold water enters at 30 °C and is allowed to warm to 70 °C. The duty Q is fixed by the flow rates and the heat capacities — say 500 kW for this example.

Counter-flow. Hot in 150, hot out 90, cold in 30, cold out 70. The two streams pass in opposite directions, so the hot inlet meets the cold outlet at one end (ΔT_1 = 150 − 70 = 80 K) and the hot outlet meets the cold inlet at the other end (ΔT_2 = 90 − 30 = 60 K). LMTD = (80 − 60)/ln(80/60) = 20/0.2877 = 69.5 K.

Parallel-flow. Same inlets, same outlets. The two streams enter at the same end: ΔT_1 = 150 − 30 = 120 K. They exit at the same end: ΔT_2 = 90 − 70 = 20 K. LMTD = (120 − 20)/ln(120/20) = 100/1.7918 = 55.8 K.

For the same duty Q and the same U, the parallel-flow area is 69.5/55.8 = 1.25 times the counter-flow area — a 25 percent penalty. That is why counter-flow is the default. (Note also that parallel-flow cannot deliver the cold outlet of 70 °C if the hot outlet is 90 °C — both streams are approaching a common temperature; the cold outlet cannot exceed the hot outlet. This particular case worked only because the temperature ranges left enough room.)

Configurations: 1-1, 1-2, 2-4 and beyond

Multipass arrangements trade thermodynamic efficiency for mechanical convenience and shell-side velocity control.

Configuration	Shell passes	Tube passes	Heads / bundle	Typical use
1-1 (true counter-flow)	1	1	Both ends nozzled; fixed tubesheets or floating head	Maximum LMTD, no F penalty; long shells
1-2	1	2	U-tube bundle or split-ring floating head; nozzles on one end	The industrial default — clean, removable, F ≈ 0.85
1-4, 1-6, 1-8	1	4, 6, 8	Multiple pass-partition plates in front head	Higher tube-side velocity for fouling control
2-4	2 (longitudinal baffle in shell)	4	Two-pass shell (F shell) with U-bundle	Closer temperature approach; recovers some LMTD
Split-flow (G shell)	1 (split into two halves)	1 or 2	Central shell-side inlet, two outlets	Horizontal thermosiphon reboilers
Kettle reboiler (K shell)	1 (expanded shell)	1 or 2	Large vapour space above bundle	Boiling on shell side at column bottoms

For most clean refinery and chemical-process duties the 1-2 with a U-tube bundle is overwhelmingly the default. The U-bundle absorbs thermal-expansion differences between tubes and shell, the bundle can be pulled as a single piece for cleaning, and there is only one channel head to bolt and unbolt. The cost is that tube-side mechanical cleaning is only possible up to the U-bend; for severely fouling tube-side fluids a straight-tube design with floating head is preferred.

Baffles in detail

Baffle design is one of the few areas where the shell-side designer has real freedom. The choice trades pressure drop against shell-side heat-transfer coefficient and fouling tendency.

• Segmental, 25 percent cut. The standard choice. A circular plate with a horizontal or vertical chord cut out to give a window equal to 25 percent of the shell area. Spacing between baffles is typically 0.2 to 1.0 shell diameters. Lower spacing increases h_o and pressure drop; higher spacing lets the flow run longitudinal and h_o falls. For boiling and condensing services where pressure drop must be minimised, baffles are spaced at the maximum permitted by tube vibration limits.
• Double-segmental. Two cuts per baffle, one centered and one annular. Roughly halves pressure drop at the cost of some h_o.
• Helical (Smith) baffles. Spiral-shaped plates that wind the shell-side flow around the bundle continuously. Smith helical baffles eliminate the dead zones at baffle tips that drive shell-side fouling in segmental designs, and reduce pressure drop by 30 to 60 percent for the same h_o. Common in modern crude-preheat trains.
• Longitudinal baffles. Run along the shell axis to create multiple shell passes (the F, G and H shells). They impose a sealing requirement against the shell wall — leakage past a longitudinal baffle short-circuits the shell-side flow and ruins the efficiency.
• Rod baffles. No solid plates at all; tubes are supported by rod grids that allow shell-side flow parallel to the bundle. Used for clean services where pressure drop must be very low. Tube-vibration risk is also lower.

Materials of construction

Tube material	Typical service	Why
Carbon steel	Clean hydrocarbons, steam, condensate	Cheap; adequate for non-corrosive, non-oxidising service to ~450 °C
Low-alloy steel (1¼ Cr, 2¼ Cr)	High-temperature hydrogen service	Resists hydrogen attack; refinery hydrotreaters
Stainless 304 / 316	Mildly corrosive chemistry; food and pharma	Cleanability; resistance to mineral acids and chlorides (within limits)
Duplex 2205 / Super-duplex	Chloride-bearing brines; sour service	Higher strength and pitting resistance than 316
Cu-Ni 90-10	Seawater coolers (10-30 °C)	Best general-purpose seawater alloy; resists biofouling
Cu-Ni 70-30	Hotter seawater; higher velocities	Better velocity and temperature limits than 90-10
Titanium grade 2	Hot brine, chlorinated water, aggressive acids	Essentially immune; high cost limits use to where stainless fails
Hastelloy / Inconel	Hot strong acids, hot caustic, sour hydrogen	Last-resort alloy for the most aggressive duties

The shell, tubesheet, baffles and bonnets are chosen on similar grounds. A common cost-saving move is to put the corrosive fluid on the tube side and use cheap carbon-steel shell with a more expensive tube alloy — only the tubes and tubesheet need the upgrade.

Where they show up

• Oil refining. Crude-preheat trains (dozens of shell-and-tube exchangers in series, recovering heat from refined products to preheat incoming crude before the furnace). Reboilers at the bottom of every distillation column. Overhead condensers at the top. Feed-effluent exchangers on hydrotreaters and reformers. A typical large refinery operates several hundred shell-and-tube exchangers.
• Power generation. Steam-cycle feedwater heaters (LP and HP heaters that warm the boiler feed using extracted turbine steam). Surface condensers (the steam-cycle's giant heat sink — a single condenser may have 30 000 tubes and a duty of 1 GW thermal). Lube-oil coolers, generator hydrogen coolers, makeup-water heaters.
• Chemical processing. Reactor jackets, intermediate coolers between stages of compression, vapour-recovery condensers, polymer coolers. Anywhere process heat needs to move between streams.
• HVAC and refrigeration. Flooded and dry-shell chiller evaporators. Large building condensers. Industrial chillers and heat pumps.
• Marine and offshore. Engine jacket-water coolers, lube-oil coolers, fuel-oil heaters, central seawater coolers. Cu-Ni tubes are standard.
• Cryogenic and LNG. Specialised brazed-aluminium exchangers handle the very cold end, but shell-and-tube is still the workhorse for moderate-temperature LNG service.

Worked sizing example

Process oil cooler. Hot oil 80 m³/h, in 130 °C, out 80 °C. Cooling water 60 m³/h, in 25 °C, out 50 °C. Find the required heat-transfer area.

Step 1 — duty. For oil ρ ≈ 850 kg/m³, c_p ≈ 2.1 kJ/kgK. Mass flow 80 × 850 / 3600 = 18.9 kg/s. Q = 18.9 × 2.1 × (130 − 80) = 1 980 kW. (Cross-check on water side: 60 × 1000 / 3600 = 16.7 kg/s, Q = 16.7 × 4.18 × (50 − 25) = 1 745 kW. The two estimates straddle 1 850 kW; use the average for the design.)

Step 2 — LMTD (counter-flow). ΔT_1 = 130 − 50 = 80 K, ΔT_2 = 80 − 25 = 55 K. LMTD = (80 − 55)/ln(80/55) = 25/0.374 = 66.9 K.

Step 3 — choose configuration and F. 1-2 with U-bundle, F ≈ 0.88 read from the standard chart for these end temperatures.

Step 4 — assume U. Water on the tube side, oil on the shell side. From standard tables for clean oil-to-water service with reasonable velocities, U ≈ 350 W/m²K including a tube-side R_fi = 0.0002 m²K/W (cooling water) and shell-side R_fo = 0.0004 m²K/W (process oil).

Step 5 — solve for A. A = Q/(U·LMTD·F) = 1 850 000 / (350 × 66.9 × 0.88) = 89.9 m².

Step 6 — sanity check. 19 mm tubes, 6 m length, gives ≈ 0.358 m² per tube. 89.9 / 0.358 = 251 tubes. On a 25 mm triangular pitch in a 0.4 m diameter shell with eight passes of geometry, that comfortably fits. The next step is to compute pressure drops, check velocities for fouling and erosion limits, choose baffle spacing, run a tube-vibration check and issue the data sheet to a TEMA fabricator.

Common pitfalls

• Picking the wrong fluid for the tube side. Rules of thumb: the corrosive fluid on the tube side (cheaper shell). The high-pressure fluid on the tube side (smaller-diameter pressure boundary). The fouling fluid on the tube side (easier mechanical cleaning). Sometimes these conflict, and the designer has to choose.
• Ignoring tube vibration. Shell-side cross-flow can drive tubes into flow-induced vibration if velocities exceed a critical limit. Vibration fatigue can destroy a bundle in months. Standard checks: Connors fluid-elastic instability, vortex shedding lock-in, acoustic resonance.
• Underestimating fouling. The single most common cause of underperformance. Crude-oil exchangers in particular can foul to 50 percent of clean U within a year. Designers add R_f generously and plan for cleaning cycles — clean during turnaround, monitor U trend in service.
• Multipass with low F. If F drops below 0.75 the configuration is wasting LMTD; either add shells in series, switch to a true counter-flow design, or live with a much bigger exchanger.
• Wrong inlet impingement protection. High-velocity shell-side feed can erode the tubes nearest the inlet within years. An impingement plate or annular ring is standard.
• Mixing dissimilar metals at the tubesheet. Galvanic corrosion at the tube-to-tubesheet joint is a classic failure when the tube alloy and tubesheet alloy are not metallurgically compatible. Specify carefully or use clad tubesheets.