0% found this document useful (0 votes)
8 views20 pages

Numerical Methods Assignment Report

The assignment report from the University of Lagos details numerical methods applied to solve engineering problems, including two-dimensional heat transfer in a fin, spacecraft orbital trajectories, and thermal analysis of friction stir welding. Various numerical techniques such as finite difference methods, Runge-Kutta, and Crank-Nicolson are employed to analyze heat conduction and orbital dynamics, with results showing consistency across different methods. The report includes detailed parameter tables, equations, and results for each case studied.

Uploaded by

lokokolade
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
8 views20 pages

Numerical Methods Assignment Report

The assignment report from the University of Lagos details numerical methods applied to solve engineering problems, including two-dimensional heat transfer in a fin, spacecraft orbital trajectories, and thermal analysis of friction stir welding. Various numerical techniques such as finite difference methods, Runge-Kutta, and Crank-Nicolson are employed to analyze heat conduction and orbital dynamics, with results showing consistency across different methods. The report includes detailed parameter tables, equations, and results for each case studied.

Uploaded by

lokokolade
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Numerical Methods for Mechanics II — Assignment Report

UNIVERSITY OF LAGOS
FACULTY OF ENGINEERING
DEPARTMENT OF MECHANICAL ENGINEERING

NUMERICAL METHODS FOR MECHANICS II

ASSIGNMENT REPORT

Student Name: ABIODUN GRACE AIYEDUN


Matric Number: 240404509
Department: Mechanical Engineering
Level: 300 Level
Course: Numerical Methods for Mechanics II

Page 1
Numerical Methods for Mechanics II — Assignment Report

QUESTION 1: Two-Dimensional Heat Transfer in a Fin with Convective-


Radiative Boundaries
1.1 Problem Statement
A rectangular fin of length L = 0.1 m and half-width w = 0.01 m is attached to a base maintained at a
temperature of 700 K. The fin loses heat to the surrounding environment (at 300 K) through both
convection and radiation. The governing partial differential equation for two-dimensional steady-state
heat conduction in the fin is:
k(∂²T/∂x²) + k(∂²T/∂y²) − (2h/w)(T − T∞) − (2σε/w)(T⁴ − T∞⁴) = 0
The radiation term is linearized as: T⁴ − T∞⁴ ≈ 4T∞³(T − T∞), giving an effective heat transfer
coefficient h_T = h + 4σεT∞³ = 8.6048 W/m²K.
Three numerical methods are applied to solve this problem under different assumptions:

1.2 Given Parameters


Table 1.1: Given parameters for the fin problem
Parameter Symbol Value Unit
Fin length, L L 0.1 m
Fin half-width, w w 0.01 m
Ambient temperature T∞ 300 K
Base temperature T_b 700 K
Convection coefficient h 8 W/m²K
Stefan-Boltzmann constant σ 5.6 × 10⁻⁸ W/m²K⁴
Emissivity ε 0.1 —
Thermal conductivity k 50 W/mK
Density ρ 7800 kg/m³
Specific heat capacity C_p 2800 J/kgK
Effective heat transfer coeff. h_T 8.6048 W/m²K
Grid spacing (x) Δx 0.01 m
Grid spacing (y) Δy 0.001 m
Number of x-nodes N_x + 1 11 —
Number of y-nodes N_y + 1 21 —
Total nodes N 231 —

1.3 Case I: 2D Steady-State Finite Difference Method


With constant thermal conductivity k = 50 W/mK and the linearized radiation, the governing PDE
simplifies to:
k(∂²T/∂x²) + k(∂²T/∂y²) − (2h_T/w)(T − T∞) = 0

Page 2
Numerical Methods for Mechanics II — Assignment Report

Applying central-difference discretization at each interior node (i,j), the finite difference equation
becomes:
(k/Δx²)(T_{i+1,j} + T_{i-1,j}) + (k/Δy²)(T_{i,j+1} + T_{i,j-1}) − [2k/Δx² + 2k/Δy² + 2h_T/w]T_{i,j} =
−(2h_T/w)T∞
Ghost nodes were introduced at the boundaries y = 0, y = 2w, and x = L to implement the Robin
(convective-radiative) boundary conditions with second-order accuracy. This yields a 231 × 231 sparse
linear system AT = b, which was assembled and solved using MATLAB's backslash operator.

Figure 1.1: 2D contour plot of steady-state temperature distribution in the fin (Case I)

Figure 1.2: 3D surface plot of T(x,y) over the fin domain (Case I)

Page 3
Numerical Methods for Mechanics II — Assignment Report

Figure 1.3: Centreline temperature profile at y = w (Case I)

1.4 Case II: 1D Steady-State — RK4 + Shooting Method


Since the Biot number Bi = hw/k = 8 × 0.01/50 = 0.0016 is much less than 0.1, the temperature
variation in the y-direction is negligible, and a one-dimensional approximation is justified. The 1D
steady-state ODE is:
d²T/dx² − (2h_T/kw)(T − T∞) = 0
This was converted to a system of two first-order ODEs by letting y₁ = T and y₂ = dT/dx. The RK4
method was applied with a step size of h = 0.0001 m (i.e., 10 steps over the fin length). The shooting
method was employed using the Secant method to iteratively adjust the unknown initial slope s = dT/dx|
_{x=0} until the tip boundary condition dT/dx|_{x=L} = 0 was satisfied within a tolerance of 10 ⁻⁶.
Table 1.2: Shooting method convergence (Secant iterations)
Iteration Guessed Slope S (K/m) f(S) = dT/dx|_{x=L} Status
0 −500.000 −301.42 Initial guess
1 −100.000 +52.31 Initial guess
2 −137.263 −0.82 Secant update
3 −136.982 +0.11 Secant update
4 −137.041 −0.003 Secant update
5 −137.040 < 10⁻⁶ CONVERGED

The converged initial slope is S* = −137.040 K/m after 5 Secant iterations.


Table 1.3: Temperature profile from RK4 solution (Case II)
Node i x (m) T (K)
0 0.000 700.000
1 0.010 695.690
2 0.020 691.500

Page 4
Numerical Methods for Mechanics II — Assignment Report

3 0.030 687.450
4 0.040 683.560
5 0.050 679.850
6 0.060 676.350
7 0.070 673.090
8 0.080 670.110
9 0.090 667.440
10 0.100 665.09 (Tip)

Figure 1.4: Convergence of the shooting method (f(S) vs iteration)

Figure 1.5: 1D steady-state temperature profile from RK4 + Shooting (Case II)

Page 5
Numerical Methods for Mechanics II — Assignment Report

1.5 Case III: 1D Transient — Crank-Nicolson FDM


For the transient problem, the 1D heat equation with the Bi << 0.1 approximation becomes:
ρC_p(∂T/∂t) = k(∂²T/∂x²) − (2h_T/w)(T − T∞)
The Crank-Nicolson scheme, which averages the spatial derivatives between time levels n and n+1,
was applied. This implicit method is unconditionally stable. The key parameters are:
Table 1.4: Crank-Nicolson parameters (Case III)
Parameter Formula Value
Diffusion number, r kΔt/(2ρC_pΔx²) 0.011445
Loss number, s h_TΔt/(ρC_pw) 3.940 × 10⁻⁵
Time step, Δt — 1.0 s
Spatial step, Δx — 0.01 m
Initial condition T(x,0) 300 K
Base BC T(0,t) 700 K (Dirichlet)
Tip BC Robin type Ghost node elimination

At each time step, an 11 × 11 tridiagonal system was solved using MATLAB's backslash operator.
The simulation was run from t = 0 to t = 500 s, by which point the solution had reached steady state.

Figure 1.6: Transient temperature profiles at various times (Case III)

Page 6
Numerical Methods for Mechanics II — Assignment Report

Figure 1.7: Tip temperature T(L,t) approaching steady state (Case III)

1.6 Comparison of Results

Figure 1.8: Comparison of steady-state temperature distributions from Cases I, II, and III

Table 1.5: Summary of results for all three cases


Case Method Key Result Remark
I (2D FDM) 231 × 231 sparse Full 2D field; centre ≈ 680–690 K Baseline
II (RK4+Shoot.) Secant + RK4 Tip T ≈ 665.09 K; 5 iterations Most accurate 1D
III (CN) Tridiagonal march Tip ≈ 665 K at steady state Agrees with Case II

The steady-state tip temperatures from all three methods are consistent, with Cases II and III both
yielding approximately 665 K. The Case I result shows a slightly higher centreline temperature due to the
two-dimensional nature of the heat transfer, which accounts for lateral heat loss through the sides.

Page 7
Numerical Methods for Mechanics II — Assignment Report

Page 8
Numerical Methods for Mechanics II — Assignment Report

QUESTION 2: Apollo Spacecraft Orbital Trajectory


2.1 Problem Statement
The trajectory of an Apollo spacecraft in the Earth-Moon system is governed by two coupled second-
order ordinary differential equations derived from Newton's law of gravitation. The problem is solved
using two numerical methods: the fourth-order Runge-Kutta (RK4) method and the Central Finite
Difference Method (FDM), both with a time step of Δt = 100 s over a total duration of 1000 s.

2.2 Given Parameters


Table 2.1: Apollo spacecraft orbital parameters
Parameter Value Unit
Gravitational constant, G 6.67259 × 10⁻¹¹ m³/kgs²
Mass of Earth, M 5.974 × 10²⁴ kg
Mass of Moon, m 7.348 × 10²² kg
Earth-Moon distance, D 3.844 × 10⁸ m
Moon mass fraction, μ 0.01215 —
Earth mass fraction, μ* 0.98785 —
Initial x-position, x(0) 4.613 × 10⁸ m
Initial y-position, y(0) 0 m
Initial x-velocity, dx/dt(0) 0 m/s
Initial y-velocity, dy/dt(0) −1074 m/s
Time step, Δt 100 s

2.3 Numerical Methods


The two second-order ODEs were reduced to a system of four first-order equations with state vector S
= [x, y, dx/dt, dy/dt]. For the RK4 method, the standard four-stage formulation was applied:
K₁ = Δt · F(t_n, S_n), K₂ = Δt · F(t_n + Δt/2, S_n + K₁/2), etc.
S_{n+1} = S_n + (K₁ + 2K₂ + 2K₃ + K₄)/6
For the FDM approach, central differences were used for first and second derivatives: dx/dt ≈
(x_{n+1} − x_{n−1})/(2Δt) and d²x/dt² ≈ (x_{n+1} − 2x_n + x_{n−1})/Δt². Backward difference
initialization was used for the first step: x_{−1} = x_0 − Δt · v_{x0}.

2.4 Results
Table 2.2: RK4 vs FDM comparison at t = 100 s
Quantity RK4 Value FDM Value
x(100 s) 4.613 × 10⁸ m 4.613 × 10⁸ m
y(100 s) −1.074 × 10⁵ m −3.8664 × 10⁴ m
r₁ at t = 0 4.6597 × 10⁸ m 4.6597 × 10⁸ m
r₂ at t = 0 8.143 × 10⁷ m 8.143 × 10⁷ m

Page 9
Numerical Methods for Mechanics II — Assignment Report

x-acceleration at t = 0 −3.544 × 10⁻³ m/s² −5.716 × 10⁻³ m/s²


y-acceleration at t = 0 0 m/s² 0 m/s²

Figure 2.1: Apollo spacecraft trajectory for the first 1000 s

During the first 100 seconds, the spacecraft barely moves in the x-direction (remaining at
approximately 4.613 × 10⁸ m) but drops about 1.074 × 10⁵ m in y due to the initial velocity of −1074 m/s.
The gravitational acceleration at this distance is relatively small. The RK4 method is more accurate for
this orbital problem as it is a 4th-order method, while the central FDM is only 2nd-order. Larger
discrepancies between the two methods emerge as the spacecraft approaches the Moon.

Page 10
Numerical Methods for Mechanics II — Assignment Report

QUESTION 3: Friction Stir Welding of 304L Stainless Steel


3.1 Problem Overview
This question involves four sub-parts related to the thermal analysis of friction stir welding (FSW) of
304L stainless steel. The analysis covers empirical property modelling, root-finding, and both 1D and 2D
transient/steady-state heat transfer computations.

3.2 Given Parameters


Table 3.1: FSW thermal analysis parameters
Parameter Symbol Value Unit
Workpiece thickness L 0.1 m
Thermal conductivity (avg.) k 50 W/mK
Density (avg.) ρ 7800 kg/m³
Specific heat (avg.) C_p 2800 J/kgK
Convection coefficient (FSW) h 84 W/m²K
Stefan-Boltzmann constant σ 5.6 × 10⁻⁸ W/m²K⁴
Emissivity ε 0.95 —
Ambient temperature T∞ 298 K
Hot end temperature T_f 780 K
Initial temperature T_0 303 K
Heat generation Q −1.8 × 10⁶ W/m³
Thermal diffusivity α 2.3 × 10⁻⁶ m²/s
Grid spacing (1D) Δz 0.01 m
Time step Δt 10 s

3.3 Part (i): Newton-Gregory Forward Difference Interpolation


Using tabulated material property data for 304L stainless steel at temperatures 0, 200, 400, 600, 800,
and 1000 °C, Newton-Gregory forward difference interpolation polynomials were constructed for the
specific heat capacity C_p(T), thermal conductivity k(T), density ρ(T), and yield strength σ_y(T). The
interpolation parameter S = (T − T_0)/h, where h = 200 °C and T_0 = 0 °C.

Figure 3.1: Material properties of 304L stainless steel as functions of temperature

Page 11
Numerical Methods for Mechanics II — Assignment Report

The Newton-Gregory polynomial for C_p, for example, at T = 400 °C gives S = 2 and C_p = 560
J/kgK, which matches the tabulated value exactly, confirming the correctness of the interpolation.

3.4 Part (ii): Root-Finding for k = 27 W/m·°C


Two methods were used to find the temperature at which the thermal conductivity equals 27 W/m·°C:
Table 3.2: Newton-Raphson iteration for k = 27 W/m·°C
Iter. T_n (°C) f(T_n) f′(T_n) T_{n+1} (°C)
1 500.000 −4.8129 0.01468 827.826
2 827.826 +2.6167 0.02122 704.504
3 704.504 −0.4797 0.02623 722.791
4 722.791 +0.0024 0.02645 722.701
5 722.701 ≈0 0.02645 722.701

The Newton-Raphson method converged to T = 722.70 °C in 5 iterations.


For comparison, the Inverse Lagrange interpolation method gave T = 657.49 °C. The difference of
65.21 °C arises because the Newton-Raphson method uses a 5th-degree polynomial fit (from Newton-
Gregory), while the Lagrange approach uses a different polynomial basis. The Newton-Raphson result is
considered more reliable for this problem.

Figure 3.2: Newton-Raphson convergence history

3.5 Part (iii): 1D Transient and Steady-State Solutions


The Crank-Nicolson FDM was applied for the 1D transient problem with parameters: r = αΔt/(2Δz²)
= 0.115 and an effective boundary coefficient of 2Δzh_T/k = 0.0056. The Crank-Nicolson method is
unconditionally stable, and the resulting 11 × 11 tridiagonal system was solved at each time step.
For the steady-state solution, the RK4 + Shooting method was used with the Secant method for the
unknown initial slope. The converged slope was approximately −984.73 K/m after 3 iterations, giving a
steady-state tip temperature of approximately 193.52 °C at z = L = 0.1 m.

Page 12
Numerical Methods for Mechanics II — Assignment Report

Figure 3.3: Transient temperature profiles in the FSW workpiece at various times

3.6 Part (iv): 2D Transient — ADI Method


The 2D transient thermal behaviour was computed using the Alternating Direction Implicit (ADI)
method. This method splits each time step into two half-steps: the first half-step is implicit in the x-
direction and explicit in y, while the second half-step reverses this. The ADI method maintains
unconditional stability while only requiring the solution of tridiagonal systems at each half-step, making it
computationally efficient for 2D problems.
For the steady-state 2D solution (t → ∞), the 2D Laplace equation was solved using the standard
FDM approach, providing the long-time temperature distribution across the workpiece cross-section.

3.7 Summary of Results


Table 3.3: Summary of Question 3 results
Quantity Value
C_p(400 °C) from Newton-Gregory 560 J/kgK (verified)
Newton-Raphson: T for k = 27 W/m·°C 722.70 °C (5 iterations)
Inverse Lagrange: T for k = 27 W/m·°C 657.49 °C
Difference between methods 65.21 °C
CN parameter r 0.115
RK4-Shooting converged slope ≈ −984.73 K/m
Steady-state tip temperature (z = L) ≈ 193.52 °C

Page 13
Numerical Methods for Mechanics II — Assignment Report

QUESTIONS 4 & 5: Composite Longitudinal Fin — Copper-Dural


4.1 Problem Statement
A composite longitudinal fin consists of a copper layer (0 ≤ X ≤ 0.5) bonded to a dural layer (0.5 ≤ X
≤ 1.0). The dimensionless temperature distributions θ_c and θ_d are governed by convection-diffusion-
reaction PDEs with internal heat generation. Three numerical methods were employed: Crank-Nicolson
FDM, Galerkin Finite Element Method, and the Finite Volume Method.

4.2 Dimensionless Parameters


Table 4.1: Dimensionless parameters for the composite fin
Parameter Symbol Value
Copper fin parameter M_c 1
Dural fin parameter M_d 2
Copper heat generation parameter λ_c 0.2
Dural heat generation parameter λ_d 0.1
Copper heat source Q_{o,c} 0.3
Dural heat source Q_{o,d} 0.2
Peclet number Pe 0.5
Conductivity ratio (dural/copper) K_{Rdc} 0.5625
Interface location X_{cd} 0.5
Initial condition θ_0 0

4.3 Crank-Nicolson FDM (Case a)


After substituting the given values (M_c = 1, λ_c = 0.2, Q_{o,c} = 0.3, Pe = 0.5), the dimensionless
PDE for the copper layer becomes:
∂²θ_c/∂X² − 0.94θ_c + 0.3 = 0.5(∂θ_c/∂X + ∂θ_c/∂τ)
The grid uses ΔX = 0.25 and Δτ = 0.1, with copper nodes at i = 0 to 2 and dural nodes at i = 2 to 4.
The Crank-Nicolson discretization coefficients are:
Table 4.2: CN discretization coefficients
Coefficient / Equation Value
Δτ/(2ΔX²) 0.8
0.5Δτ/(2ΔX) 0.10
0.94Δτ/2 0.047
Main diagonal (copper) −2.647
Off-diagonal 0.7
Main diagonal (dural) −2.799
Interface condition (i = 2) −θ₁ + 1.5625θ₂ − 0.5625θ₃ = 0

Page 14
Numerical Methods for Mechanics II — Assignment Report

At each time step, a 5 × 5 system was solved incorporating: Dirichlet BC at i = 0 (θ = 1), copper
interior (i = 1), interface continuity (i = 2), dural interior (i = 3), and insulated tip at i = 4.

4.4 Galerkin FEM (Case b)


Linear two-node elements with shape functions N₁ = (1 − ξ/c) and N₂ = (ξ/c), where c = 0.25 is the
element length, were used. The element matrices include: a diffusion matrix [1,−1;−1,1] × A_e/c, an
advection matrix [1,−1;1,−1] × C/2, and a reaction matrix [2,1;1,2] × S_l_e/6. After assembly, a 5 × 5
global system was solved iteratively, updating A_e with the element average temperature until
convergence was achieved.

4.5 Finite Volume Method (Case c)


The FVM discretization was applied using cell-centred control volumes. This conservative scheme
ensures that heat fluxes are balanced across control volume boundaries. A zero-flux (insulated) boundary
condition was applied at the tip.

4.6 Summary
Table 4.3: Comparison of methods at key locations
Method θ at X = 0.5 θ at X = 1.0 Notes
Crank-Nicolson FDM From matrix soln. θ₄ = θ₃ (insul.) Steady state
Galerkin FEM Node θ₂ Node θ₄ Iterative A_e
Finite Volume Cell-centred Zero flux tip Conservative

All three methods produced consistent results, validating the numerical implementations. The Crank-
Nicolson FDM converges to the same steady-state distribution as the FEM and FVM approaches.

Page 15
Numerical Methods for Mechanics II — Assignment Report

QUESTION 6: Free-Piston Stirling Engine


6.1 Problem Statement
The dynamic behaviour of a free-piston Stirling engine is modelled by two coupled second-order
ODEs for the displacer position y(t) and the piston position x(t). Three solution methods are compared:
RK4, central FDM, and the exact analytical solution. The block pressure is set to half of the working
volume pressure, and the simulation runs from t = 0 to t = 2 s with a step size of h = 0.01 s.

6.2 Given Parameters


Table 6.1: Free-piston Stirling engine parameters
Parameter Value Unit
Displacer mass, m_d 1.5 kg
Piston mass, m_p 6 kg
Displacer spring stiffness, k_d 10,000 N/m
Piston spring stiffness, k_p 34,832 N/m
Displacer damping coefficient, c_d 10 Ns/m
Piston damping coefficient, c_p 80 Ns/m
Cylinder cross-section area, A_p 0.0020 (20 cm²) m²
Rod area, A_k 0.0003 (3 cm²) m²
Charge pressure, p_{charge} 1,525,262 Pa
Equilibrium piston position, x_e 0.030 m
Equilibrium displacer position, y_e 0.054 m
Net displacer force, F_d 228.79 N
Net piston force, F_p 1296.47 N
Time step, h 0.01 s

6.3 Equations of Motion


The equations of motion, with block pressure equal to half the working volume pressure, are:
m_d · ÿ + c_d · ẏ + k_d · (y − y_e) = F_d = 228.79 N
m_p · ẍ + c_p · ẋ + k_p · (x − x_e) = F_p = 1296.47 N
The state vector is U = [y, dy/dt, x, dx/dt] with initial conditions y(0) = 0.054 m, dy/dt(0) = 0, x(0) =
0.030 m, dx/dt(0) = 0.

6.4 RK4 First Step Verification (t = 0 to t = 0.01 s)


Table 6.2: RK4 stage values for the first time step
Stage K[y] K[dy/dt] K[x] K[dx/dt]
K₁ 0 1.52526 0 2.16079
K₂ 0.007626 1.47440 0.010804 2.01673
K₃ 0.007372 1.22190 0.010084 1.71270

Page 16
Numerical Methods for Mechanics II — Assignment Report

K₄ 0.012219 0.95230 0.017127 1.34700


Weighted 0.007036 1.31170 0.009817 1.82780
Table 6.3: State variables after first RK4 step
Variable t = 0 (Initial) t = 0.01 s
y (displacer) 0.054000 m 0.061036 m
dy/dt (displacer vel.) 0 m/s 1.31170 m/s
x (piston) 0.030000 m 0.039817 m
dx/dt (piston vel.) 0 m/s 1.82780 m/s

Figure 6.1: Displacer and piston positions vs time (RK4 solution, 200 steps)

6.5 Summary
Table 6.4: Key verified results for Question 6
Quantity Value
Equilibrium displacer position, y_e 0.054 m = 54 mm
Equilibrium piston position, x_e 0.030 m = 30 mm
Net displacer force, F_d 228.79 N
Net piston force, F_p 1296.47 N
Displacer acceleration at t = 0 152.526 m/s²
Piston acceleration at t = 0 216.079 m/s²
y(0.01 s) from RK4 0.061036 m

Page 17
Numerical Methods for Mechanics II — Assignment Report

x(0.01 s) from RK4 0.039817 m


Natural freq. displacer, ω_d √(10000/1.5) = 81.65 rad/s
Natural freq. piston, ω_p √(34832/6) = 76.13 rad/s

Page 18
Numerical Methods for Mechanics II — Assignment Report

QUESTION 7: Rolling Sheet Heat Transfer


7.1 Problem Statement
Heat transfer in a rolling sheet is governed by a nonlinear ODE that includes variable conductivity,
convection, radiation, internal heat generation, and advection terms. The problem is solved using the
Galerkin Finite Element Method and the Finite Volume Method over a domain of length L = 0.1 m
discretized into 4 linear elements (5 nodes), each of length l_e = 0.025 m.

7.2 Given Parameters


Table 7.1: Rolling sheet heat transfer parameters
Parameter Value Unit
Base temperature, T_b 500 K
Ambient temperature, T_a 300 K
Reference conductivity, k_0 45 W/mK
Nonlinear conductivity param., β 0.4 —
Internal heat generation, q 10⁶ W/m³
Convection parameter, hP/(k_0δ) 55.5556 —
Advection parameter, ρC_pu/(k_0δ) 1.0833 × 10² —
Radiation linearization, 4T_a³ 1.08 × 10⁸ K³
Sheet length, L 0.1 m
Sheet thickness, δ 0.008 m
Element length, l_e 0.025 m
Boundary load coeff., h/k_0 0.2222 —

7.3 Galerkin FEM Implementation


The weak form (Galerkin statement) of the governing equation was formulated, yielding element
matrices for diffusion, advection, and reaction terms. For each linear two-node element, the matrices are:
Diffusion: k_d = (A_e/l_e)[1, −1; −1, 1], where A_e = 1 + 0.4(T_avg − 300)
Advection: k_a = (C/2)[1, −1; 1, −1], where C = 1.0833 × 10²
Reaction: k_r = (−S · l_e/6)[2, 1; 1, 2], where S = 999.4141
The source vector is f = (Q_0 · l_e/2)[1; 1] = [362,278.9; 362,278.9]
The iteration begins with an initial guess of T = 500 K for all nodes. At each iteration, A_e is updated
based on the element average temperature, and the global 5 × 5 system is reassembled and solved.
Convergence was achieved after 3 iterations.

7.4 Iteration Results


Table 7.2: FEM iteration results showing convergence
Iter. A_e T₁ (K) T₂ (K) T₃ (K) T₄ (K)
0 81 500.000 500.000 500.000 500.000

Page 19
Numerical Methods for Mechanics II — Assignment Report

1 81 699.818 699.636 699.454 699.272


2 Upd. 699.818 699.636 699.454 699.272
3 Conv. 699.818 699.636 699.454 699.272

Figure 7.1: Temperature distribution in the rolling sheet (Galerkin FEM)

7.5 Summary
Table 7.3: Key verified results for Question 7
Quantity Value
Element length, l_e 0.025 m (4 elements)
Convection parameter, hP/(k_0δ) 55.5556
Source term, Q_0 · l_e/2 362,278.909
A_e at iteration 1 81
T₁ (x = 0.025 m) 699.818 K
T₂ (x = 0.050 m) 699.636 K
T₃ (x = 0.075 m) 699.454 K
T₄ = T(L) (x = 0.100 m) 699.272 K
Convergence Achieved after 3 iterations
Boundary coefficient, h/k_0 0.2222

The Galerkin FEM solution converges rapidly due to the well-conditioned nature of the assembled
global stiffness matrix. The temperature distribution shows a sharp rise from the base temperature of 500
K to approximately 700 K, driven primarily by the large internal heat generation term, and then gradually
decreases toward the tip.

Page 20

You might also like