Nusselt Number Calculator

The Nusselt number is the ratio of convective to conductive heat transfer across a fluid-solid boundary. Think of it as a multiplier: if Nu = 50, then forced fluid motion across the surface moves heat 50 times more effectively than a stagnant film of the same fluid would. Every heat exchanger designer in the world uses some form of it.

The formula:

Nu = h·L / k

Where h is the convective heat transfer coefficient (W/m²·K), L is a characteristic length (often pipe diameter or plate length, m), and k is the thermal conductivity of the fluid itself (W/m·K). Nu is dimensionless. The same formula rearranges to give h directly when Nu is known from a correlation: h = Nu·k / L.

Wilhelm Nusselt and why this exists:

Nusselt published the concept in 1915 while at Munich’s Technical University. The problem he was solving: how to predict h for a given system. Direct measurement is possible but tedious; theory from first principles (solving the boundary-layer equations) is intractable except for the simplest geometries. The solution that worked, and is still used today: dimensional analysis collapses the problem into a relationship Nu = f(Re, Pr), where Re is Reynolds number (flow regime) and Pr is Prandtl number (fluid property). Pick the right correlation for your geometry and flow regime, plug in Re and Pr, get Nu, then back out h.

Common correlations:

• Dittus-Boelter (turbulent flow in a smooth pipe): Nu = 0.023 · Re^0.8 · Pr^n, where n = 0.4 for heating the fluid, 0.3 for cooling. Valid for 0.6 ≤ Pr ≤ 160 and Re ≥ 10⁴. The single most-used correlation in undergraduate textbooks.
• Sieder-Tate (turbulent pipe flow with significant viscosity variation): Nu = 0.027 · Re^0.8 · Pr^(1/3) · (μ/μ_wall)^0.14. The viscosity ratio matters for fluids whose viscosity changes a lot with temperature (oils).
• Churchill-Bernstein (cross-flow over a cylinder): more complex but valid across a huge Re range.
• Free convection (vertical plate): Nu = 0.59 · Ra^(1/4) (laminar) where Ra is Rayleigh number = Gr·Pr. Different family entirely; doesn’t use Re because no forced flow.

Worked example, air over a flat plate:

A heated plate, 0.5 m long, in airflow with measured h = 25 W/m²·K. Air thermal conductivity k = 0.026 W/m·K. Find Nu.

Nu = h·L / k = 25 · 0.5 / 0.026 = 12.5 / 0.026 = 481

This is in the typical forced-convection range (a few hundred to a few thousand). Convection is moving heat about 480× better than pure conduction through stagnant air would.

Worked example, predicting h from Re and Pr:

Water flowing in a 25 mm diameter pipe at Re = 10⁵, Pr = 7 (water at 20 °C), being heated. Using Dittus-Boelter:

Nu = 0.023 · (10⁵)^0.8 · 7^0.4 = 0.023 · 10,000 · 2.18 = 501

Water’s k ≈ 0.6 W/m·K, so h = Nu·k/L = 501 · 0.6 / 0.025 = 12,024 W/m²·K. That’s an order of magnitude higher than the air case, which is why water-cooling beats air-cooling for high-density heat loads.

Interpreting Nu values:

• Nu = 1: heat transfer is exactly what conduction alone would give. The fluid is stagnant or moving so slowly it doesn’t help. Equivalent to a still air gap in insulation calculations.
• Nu ≈ 5-20: weak laminar convection. Typical for natural convection from a small heated surface in still air.
• Nu ≈ 50-500: moderate forced convection. Air being pushed across a surface by a fan, or water in slow pipe flow.
• Nu ≈ 500-10,000: strong forced convection. Turbulent pipe flow with water or other dense fluids.
• Nu > 10,000: extreme convection, two-phase flow, or boiling/condensation regimes where phase change adds enormous heat transport.

Why dimensionless numbers exist:

You could try to memorize h-values for every fluid/geometry/flow combination, but the catalog would be infinite and you’d still miss the case in front of you. Dimensionless analysis lets a single Re-Pr correlation cover water at room temperature, oil at 100 °C, supercritical CO₂, and any other fluid in the same flow regime, as long as the assumptions hold. You measure dimensionless groups, look up the matching correlation, and the result transfers across fluid choices. This is how thermal engineering scales from undergraduate homework problems to designing actual industrial heat exchangers.

When Nu predictions go wrong:

• Surface roughness: smooth-pipe correlations under-predict h for rough surfaces (sometimes by 2-3×).
• Phase change: boiling and condensation produce h-values 10-100× higher than single-phase forced convection. Different correlations entirely.
• Entrance effects: pipe flow near the inlet has not yet developed a steady boundary layer; Nu is higher than fully-developed correlations predict. Significant for short pipes (L/D < 60).
• Non-Newtonian fluids: standard correlations assume constant viscosity; thixotropic and shear-thinning fluids violate this badly.

Pick the right correlation for your geometry, double-check Re falls in its valid range, and remember that a 10-20% error in predicted h is normal even for well-characterized systems.