Engineering & Technology · Heat Transfer & Thermal Management

Convection, boundary layers and empirical correlations

Air passing a warm surface has almost the surface temperature immediately near the wall and approaches a different temperature farther away. The near-wall variation matters even when the bulk flow looks well mixed. Convection models summarise the combined transport through this changing fluid region.

Thermodynamics & Energy Systems provides state, energy, heat and boundary vocabulary. Material Structure, Processing & Properties supports conductivity and property conditions. Symbols, ratios, resistance and exponential-response notation are explained here; calculus is optional. All numerical materials, contacts, operating records, curves and limits are constructed teaching cases. Named physical constants retain their ordinary meanings. No equipment operating procedure is given.

Back to the chapter

After reading, you can

Explain what a convection coefficient represents and when an empirical correlation can support it.

Separate the surface from the surrounding fluid

A common model is Q̇=hA(Ts−T∞), where Ts is surface temperature, T∞ a specified reference-fluid temperature, A the exchanging area and h the convection coefficient in W/(m²·K). Positive Q̇ here means heat from the warmer surface into the cooler fluid. The relationship defines the coefficient for the stated temperature difference and averaging convention.

The coefficient h is not a universal material constant like a tabulated density at a specified state. It depends on flow, geometry, fluid properties and thermal boundary conditions. A local h at one wall position and an area-averaged h answer different questions. In an internal flow the reference may be a bulk mean fluid temperature that changes downstream.

Boundary layers explain the dependence

At a conventional stationary solid wall, a viscous fluid commonly satisfies no slip: the fluid velocity at the wall matches the wall. Velocity changes away from it through a boundary layer. A thermal boundary layer contains the corresponding temperature variation. Its thickness need not match the velocity-layer thickness. Conduction still carries heat through the fluid immediately next to the wall.

Forced convection uses imposed bulk motion; natural convection develops through buoyancy associated with density differences. Both may matter together. Surface orientation, fluid properties and upstream conditions influence the pattern. The schematic shows growing near-wall variation but has no physical scale; it does not calculate h or establish whether the flow is laminar or turbulent.

Dashed line indicates a schematic boundary-layer extent growing downstream; orange marks indicate near-wall variation. Geometry is conceptual, with no quantitative length or temperature scale. The worked heat rates use Qdot=hA(Ts−Tinf), not measurements from this sketch.

A correlation carries an evidence range

Empirical correlations organise measured or modelled convection using dimensionless groups. Reynolds number compares inertial and viscous effects, Prandtl number compares momentum and thermal diffusivities, and Nusselt number expresses the convection coefficient in dimensionless form using fluid conductivity and a length scale. The chosen geometry and length definition are essential.

Before applying a correlation, check its flow regime, property-evaluation temperature, local versus average form, development length, geometry and allowed parameter range. A correlation for flow along a plate is not automatically appropriate inside a tube. Boiling, condensation and strong property changes can require different models. Scatter and uncertainty remain even inside a published validity range; extrapolation needs additional justification.

Turn a compatible dimensionless correlation into a coefficient

Define NuD=h̄D/kfluid for a tube, where D is its internal diameter, kfluid is fluid thermal conductivity in W/(m·K), and h̄ is the coefficient averaged over the heated surface in W/(m²·K). This length and conductivity must match the correlation definition. Reynolds number is ReD=ūD/ν, using mean speed ū and kinematic viscosity ν in m²/s. Pr=ν/α compares momentum and thermal diffusivities, with α the thermal diffusivity in the same units.

As a published example, the Dittus–Boelter teaching relation for fully developed turbulent flow in a smooth circular tube is NuD=0.023 ReD^0.8 Pr^n. The inspected NPTEL course specifies n=0.4 when the fluid is heated and n=0.3 when it is cooled, properties at average fluid bulk temperature, 0.7≤Pr≤160, ReD≥10000 and D/L<0.1. It also requires that wall-to-bulk temperature differences not cause substantial property changes. L is the heated length. These are compatibility screens for this approximate correlation, not a guarantee for every tube or a substitute for evidence about roughness, entrance flow, buoyancy or phase change.

Suppose a constructed compatible record supplies D=0.010 m, L=0.20 m, ReD=20000, Pr=1 and kfluid=0.030 W/(m·K), with the fluid being heated and small property variation. The average bulk-property temperature is supplied as 330 K. The ratio D/L=0.05 meets the stated length screen. This is an illustrative input set, not measured performance or a real fluid specification. The tube model and its diameter definition must not be transferred automatically to the preceding external-plate schematic.

Change the assumed coefficient without changing the material

For a constructed0.10 m² surface, let Ts=60 °C and T∞=40 °C. An assumed area-averaged h=10 W/(m²·K) gives Q̇=10×0.10×20=20 W. If a different flow condition gives h=25 with the same area and temperatures, the rate becomes50 W.

The surface and fluid temperature differences were held fixed in this comparison. For a component releasing only20 W at steady state with h=25, the model instead gives Ts−T∞=20/(25×0.10)=8 K. A changed coefficient does not itself specify which boundary condition stays fixed.

Applying that separate tube record gives NuD=0.023×20000^0.8×1^0.4≈63.47, hence h̄=NuD kfluid/D=63.47×0.030/0.010≈190.4 W/(m²·K). The heated area is A=πDL≈0.006283 m². If a compatible mean wall-to-bulk difference of 20 K is supplied and h̄ is treated as constant, Q̇≈190.4×0.006283×20≈23.93 W into the fluid. This estimates a rate for the stated mean conditions; it has not solved an inlet/outlet energy balance or validated the correlation for a device.

A changed record at ReD=5000 is outside the cited turbulent-range screen. Substituting 5000 into the same powers still produces a number, but that arithmetic does not supply supporting evidence. One must obtain a model/evidence appropriate to that regime and its development conditions before interpreting a new coefficient. Likewise, replacing diameter with length in NuD changes the definition and cannot be repaired by keeping the old numerical Nu.

Try the reasoning

Is h a property of the solid material alone?
Show the answer and reasoning
No. It represents a specified fluid/geometry/flow/thermal condition and averaging convention.
Does a velocity boundary layer mean conduction disappears at the wall?
Show the answer and reasoning
No. Conduction through the near-wall fluid remains part of the transport mechanism.
Can a correlation be trusted solely because its equation accepts the given numbers?
Show the answer and reasoning
No. The geometry, regime, definitions and validity range must also match.
A different compatible tube result supplies average NuD=40, D=0.020 m and kfluid=0.030 W/(m·K). Convert to average h̄. Would a value from a local external-plate correlation be interchangeable?
Show the answer and reasoning
h̄=40×0.030/0.020=60 W/(m²·K). A local plate coefficient uses another geometry, length and averaging definition. Its numerical Nu cannot simply replace the compatible tube average.
A proposed Dittus–Boelter case has ReD=20000 and Pr=1 but D/L=0.25. Which check fails, even though both dimensionless flow values are in the cited ranges?
Show the answer and reasoning
The heated-length screen D/L<0.1 fails. A short/developing flow needs appropriate supporting analysis or a different correlation. Passing the Re/Pr tests alone is insufficient.

What to carry forward

Treat h as a conditional model supported by flow and thermal evidence. Identify the reference temperature before interpreting a convection result.

References

Teaching resources and research supporting this lesson’s explanations and examples.