What the simulator shows
Heat does not cross a wall all at once. The face nearest the hot fluid warms first, and the heat soaks inward while the material stores part of it. Only when the wall has filled up does the heat leaving one face match the heat entering the other. That is steady state, the case the heat transfer calculator and the building heat loss calculator work with.
The colour band shows the wall, or a pin fin, with arrows at the faces sized by the heat flux. The profile chart shows the temperature across the thickness now, with fainter lines for earlier times and a dashed line for the steady state. The two lower charts follow the surface temperatures and the heat flux at each face over time. In fin mode the dashed line is the textbook analytic solution, so you can watch the numerical fin warm up onto it.
Things to try
- Watch the wall fill with heat. Keep the defaults: 100 mm of concrete at 20 °C, gas at 200 °C with h = 50 W/(m²·K) on the left, room air at 20 °C with h = 10 on the right. After 1 hour about 2 510 W/m² enters on the left but only 255 W/m² leaves on the right; the difference is stored in the concrete. At steady state both faces carry 940 W/m², the surfaces sit at 181 °C and 114 °C, and the wall gets there within 5 % in about 4.6 hours.
- Add insulation, then move it. Choose 2 layers with 50 mm of expanded polystyrene as layer 2. The steady flux falls from 940 to 125 W/m², almost all of the temperature drop sits across the polystyrene, and the cold surface drops to 32.5 °C. Now swap the layers so the polystyrene faces the hot gas. The steady flux is the same 125 W/m², but the concrete now stays below 42 °C, so it stores less heat and the wall settles in 4.8 hours instead of 7.9.
- Make a lumped mass. Pick 1 layer of carbon steel, 10 mm thick, with h = 10 on both faces. The Biot number is 0.0016, the profile is flat and the status says the wall behaves as a lumped mass. It follows Newton's law of cooling with a time constant ρcL/(2h) of 1 700 s and reaches 95 % of its rise to 110 °C in 85 minutes, three time constants, as the Newton's law of cooling calculator predicts.
- See heat soak into a thick solid. Set 500 mm of concrete, a left face held at a fixed 100 °C and an insulated right face. Turn auto speed off and set 600 simulated seconds per second. The surface flux is 1 270 W/m² after 1 hour and half that, 634 W/m², after 4 hours: it falls as 1/√t. The heated zone reaches about 0.2 m after 1 hour and 0.4 m after 4 hours, the 4√(αt) of a semi-infinite solid.
- Find where a fin stops paying. Switch to the pin fin: aluminium, 5 mm by 100 mm, h = 50, base at 100 °C in 20 °C air. It carries 4.17 W at an efficiency of 66 %, with the tip at 60.6 °C. Triple the length to 300 mm and the heat rate rises only to 4.83 W while the efficiency falls to 26 %: the far end sits near the air temperature. Change to stainless steel at 100 mm and the fin carries just 1.21 W.
The model behind it
The wall. Heat flows in one direction, x, through up to three layers in perfect contact. Each layer has a constant conductivity k, density ρ and specific heat c. The heat equation holds in each layer, and each face exchanges heat with its fluid by convection:
ρ c ∂T/∂t = ∂/∂x (k ∂T/∂x) left face: −k ∂T/∂x = h_L (T∞,L − T), right face: k ∂T/∂x = h_R (T∞,R − T) at an interface: T and k ∂T/∂x are the same on both sides
A fixed surface temperature is the limit h → ∞; an insulated face has h = 0. The wall starts at one uniform temperature.
The numerical method. The wall is cut into about 60 finite volumes, at least 6 per layer. Each cell stores heat C = ρ c Δx per square metre. Neighbours are joined by the series conductance of their two half cells, so the flux and the temperature stay continuous across an interface, and the steady profile is exact. A face film is the convection resistance in series with half a cell:
C_i dT_i/dt = G_(i−1) (T_(i−1) − T_i) + G_i (T_(i+1) − T_i) G_i = 1 / (Δx_i/(2k_i) + Δx_(i+1)/(2k_(i+1))), G_film = 1 / (1/h + Δx/(2k)) backward Euler: (C/Δt + A) T(n+1) = (C/Δt) T(n) + f (tridiagonal, Thomas algorithm)
Backward Euler is stable and free of oscillations for any time step. The page takes many small steps per frame: no longer than 1/150 of the slowest time constant, and early on no longer than 1/40 of the elapsed time, so fast runs stay accurate. The heat through each face is summed with the same steps, so the energy stored always equals the net heat in.
Steady state and the numbers shown. The steady flux is the temperature difference over the sum of the resistances, both films included. The Biot and Fourier numbers use the full thickness L, an effective conductivity and the mean heat capacity:
q = (T∞,L − T∞,R) / (1/h_L + Σ L_j/k_j + 1/h_R) k_eff = L / Σ(L_j/k_j), α = k_eff / (ρc)_mean, Bi = h L / k_eff, Fo = α t / L² lumped (Bi < 0.1): T − T∞ = (T_i − T∞) exp(−t/τ), τ = ρ c V / (h A)
With full thickness, Bi < 0.1 is a cautious test for a wall heated on both faces, where the textbook length is L/2. The time to 95 % of steady state is found by running the model ahead from the uniform start: it is the time when every cell has covered 95 % of its largest initial distance from its steady temperature.
The checks. The model is tested against the one-term series solution for a plane wall with convection (within 2 % for Fo > 0.2), the semi-infinite solid with a step in surface temperature, the lumped exponential and the composite wall:
plane wall: θ* = C₁ exp(−ζ₁² Fo) cos(ζ₁ x*), ζ₁ tan ζ₁ = Bi, C₁ = 4 sin ζ₁ / (2ζ₁ + sin 2ζ₁) semi-infinite: (T − T_s)/(T_i − T_s) = erf(x / (2√(αt))), q_s = k (T_s − T_i) / √(π α t)
The pin fin. A straight rod of diameter D and length L, with its base held at T_b in a fluid at T∞. The temperature is taken as uniform over each cross-section, which holds when h(D/2)/k is well below 0.1. The analytic steady solution for an adiabatic tip is:
m = √(h P / (k A_c)) = √(4h / (k D)) θ/θ_b = cosh(m(L − x)) / cosh(mL), q = √(h P k A_c) θ_b tanh(mL), η = tanh(mL) / (mL) convective tip: θ/θ_b = [cosh m(L−x) + (h/mk) sinh m(L−x)] / [cosh mL + (h/mk) sinh mL]
The numerical fin uses the same finite volumes with a side loss h P Δx (T − T∞) in each cell. It starts at the fluid temperature with the base suddenly at T_b, and its steady profile lands on the analytic curve. Efficiency is the heat rate over h times the fin surface (side, plus tip when it convects) times θ_b; effectiveness is the heat rate over what the bare base area would lose.
Limits. Properties are constant, taken near 300 K. Radiation is not modelled separately; include it in h as a linearised coefficient. There is no contact resistance between layers and no heat generated inside. The convection coefficient is constant: real values depend on the flow and on the temperature difference (see the pipe insulation calculator for surface coefficients). Changing a fluid during a run carries on from the current temperatures; changing a layer, a material or the start temperature restarts from t = 0.
Material data. Conductivity from Incropera Appendix A as in the heat transfer calculator; density and specific heat from Tables A.1 and A.3, with c from the specific heat calculator where it lists the material. The glass fibre specific heat, 835 J/(kg·K), is a typical value for glass fibre insulation.
Sources. F. P. Incropera, D. P. DeWitt, T. L. Bergman and A. S. Lavine, Fundamentals of Heat and Mass Transfer, 7th ed.: Section 3.3 (composite wall), Section 3.6 and Table 3.4 (fins, Example 3.9), Sections 5.1 to 5.2 (lumped capacitance), Section 5.5 and Table 5.1 (plane wall, one-term solution), Section 5.7 (semi-infinite solid), Appendix A. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, 1980, Chapter 4 (finite volumes, interface conductance, implicit scheme).
FAQ
Why is the heat flux at the hot face so high at first?
At the start the cold wall sits right against the hot surface, so the temperature gradient there is steep. As the wall warms, the gradient eases and the flux falls towards its steady value.
What does the Biot number tell me?
It compares the resistance inside the solid, L/k, with the resistance of the fluid film, 1/h. Below 0.1 the solid stays nearly uniform and one exponential describes it. Well above 1 the surface follows the fluid and the inside lags behind.
Why does the order of the layers matter if the steady flux is the same?
The steady flux depends only on the sum of the resistances. The time to get there depends on how much heat each layer must store, and that depends on how hot each layer ends up.
Is a longer fin always better?
No. Once mL passes about 2, tanh(mL) is above 0.96 and extra length adds almost nothing. The far end sits near the fluid temperature and only adds weight and cost.
How accurate is it?
The automated checks hold the model within 0.5 % of the exact series solution for a plane wall, on the erf solution for a semi-infinite solid, on the lumped exponential, and within 0.1 % of the analytic fin. Its limits are those of the model: constant properties, constant h and one-dimensional heat flow.