Objectives
• Present a simple 1D Transient heat conduction problem
• We discretize the domain into 5 grid spacings
• We consider a single time step and solve the problem using FDM (FTCS)
• Vary grid spacings and time steps and re-solve the problem and obtain
solutions
Tend1 = 100 deg.C Tend2 = 1000 deg.C
L=1m
Copper Bar
General Heat Conduction Equation in 3D Cartesian Coordinates
𝜕 2𝑇 𝜕2𝑇 𝜕 2𝑇 𝑔 1 𝜕𝑇
• + + + = (1) ; T = T(x,y,z,t); x,y,z – spatial coordinates, t - time
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2 𝑘 α 𝜕𝑡
𝑚2
• α − 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑣𝑖𝑡𝑦,
𝑠
• α =𝑘/ ρ∗𝑐
• k – thermal conductivity of the material , W/(m K)
• ρ - density of the material, kg/m3
• c – specific heat capacity of the material, J/(kg K)
• g – volumetric rate of internal heat generation, W/(m K)
Assumption of material thermal conductivity
• kx, ky, kz does not vary along each of x, y or z axis homogeneous
• kx = ky = kz = k (isotropic)
For 1D steady-state heat conduction with no heat generation,
equation (1) reduces to a simpler form
• We assume temperature does not vary significantly along y, z directions when
compared with x direction.
• g = 0 (no heat generation);
• Temperature is independent of time (steady state)
𝜕2𝑇
• = 0 (2); T = T(x); This PDE can be further simplified to an ODE.
𝜕𝑥2
𝑑2𝑇
• = 0 (3); T = T(x);
𝑑𝑥2
• Eq (3) is a second order ODE with constant coefficients and can be easily solved.
• Integrating (3) w.r.t x once, we get
𝑑𝑇
• = A (4)
𝑑𝑥
• Integrating (4) w.r.t x once again, we get
• T = Ax + B (5); A, B are constants of integration.
• Eq (5) represents a BVP (boundary value problem);
• To solve Eq (5), we need 2 BCs (boundary conditions)
• (which we have in this case).
• At x = 0, T = Tend1 = 100 deg.C
• At x = 1m, T = Tend2 = 1000 deg.C
• Substituting the BCs into (5), we get
• 100 = A(0) + B;
• 1000 = A(1) + B;
• Solving the above equations, we get
• B = 100; A = 900;
• Eq (5) becomes
• T(x) = 900*x + 100 (6)
• The above problem is a simple problem and hence can be solved
analytically. Many complex problems cannot be easily solved
analytically and we need to resort to numerical methods.
• Here we utilize Finite Difference Method to solve the above heat
conduction problem.
• What does Finite difference method do. To put it briefly,
• FDM converts the differential equations to algebraic equations.
• Solve the algebraic equations to obtain solutions.
• Before we use Finite Difference Method, we need to understand
Taylor series expansion of continuous functions. See below some very
brief info on Taylor series which is an infinite series expansion of any
function.
• f(x+Δx) = f(x) + f’(x) * Δx + f”(x) * Δx^2 / (2!) + …… (7)
• f(x-Δx) = f(x) - f’(x) * Δx + f”(x) * Δx^2 / (2!) - …….. (8)
• Eq (7) is called forward Taylor series expansion and
Eq (8) is called backward Taylor series expansion.
• Note Δx is a fraction and has a small value. Δx2, Δx3.. etc are even smaller
and hence the terms associated with them can be ignored in general.
• Adding Equation (7) and (8) and truncating after the 2nd derivative terms,
we get
• f(x+Δx) + f(x-Δx) = 2*f(x) + f”(x) * Δx^2
• f”(x) = (f(x-Δx) – 2*f(x) + f(x+Δx)) / Δx^2 …..(9)
• Eq (9) is called the centered-difference approximation of the 2nd derivative
term obtained using the Taylor series approximation
• Now lets get back to our governing equation i.e., Eq (3)
• T”(x) = 0; T = T(x); T (x = 0) = Tend1; T(x=L) = Tend2;……..(10)
• Replacing the second order derivative using Finite Difference
Approximation, we get,
• (Ti-1 – 2*Ti + Ti+1 ) / Δx^2 = 0; Simplifying, we get
• (Ti-1 – 2*Ti + Ti+1 ) = 0;………(11)
• Equation (11) is the finite difference approximation of the original
equation, which we were trying to solve.
• Here i represents the node location on the discretized domain.
• The finite difference stencil is given below.
i-1 i i+1
• Now let us discretize the 1D domain into say 5 segments / grid
spacings (equally spaced) as shown below.
1 2 3 4 5 6
• Note Temperatures at Node 1) and Node 6) are known (BCs).
• To apply equation 11), we need to consider the interior nodes 2 to 5.
• Let i = 2, Eq 11) becomes
• T1 – 2T2 + T3 = 0;
• Similarly for i = 3, 4, 5, we get
• T2 – 2*T3 + T4 = 0;
• T3 - 2*T4 + T5 = 0;
• T4 – 2*T5 + T6 = 0;
• T1 and T6 are known and can be moved to the right side of the
equations. The equations are re-arranged as below.
• -2*T2 + T3 = -T1
• T2 - 2*T3 + T4 = 0
• T3 - 2*T4 + T5 = 0
• T4 - 2*T5 = -T6
• The above equation can be arranged in a matrix form as below
−2 1 0 0 T2 −100
1 −2 1 0 T3 0
• =
0 1 −2 1 T4 0
0 0 1 −2 T5 −1000
• The above equation can be solved using Thomas Algorithm (TA) as the
matrix we are dealing with, is a Tri-diagonal matrix. Iterative methods such
as Gauss-Seidel, Successive Over Relaxation (SOR) methods can also be
used.
• Solving the above equation using TA on MATLAB produces the following
results.
• T2 = 280 °C, T3 = 460 °C, T4 = 640 °C, T5 = 820 °C
• In this case, the exact solution obtained analytically matches the finite
difference method solution as the original equation is linear.
• Graphical results are presented using MATLAB for this case.
• Using MATLAB or other software, we can develop codes for a general
case where the number of grid spacings is, say m and solution
obtained accordingly.
• We can then change the number of grid spacings to say 1000 and
results are presented graphically.
• In future videos, we can explore more challenging problems.
• This concludes this presentation.
• Pls. provide comments / suggestions.
• Thank you.
