Me526 hw2 Report

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

ME526

Simulation of Physical Processes

Assignment 2:
Thermal Study

Group 11
Victor Salguero, AbdelRahman AbdelRahman
Dancho Ivanov, Dan Kenney
Introduction 2

Theoretical Analysis 3
Problem 1: 3
Problem 2: 4
Problem 3: 5

COMSOL Results 13
Problem 1: 13
Problem 2: 17
Problem 3: 21

Discussion 23

Conclusion 26

Appendix 27
Assignment Description 27

1
Introduction
The purpose of this exercise is to verify the numerical accuracy of the COMSOL program
in simple thermal problems. The geometries simulated are relatively simple, minimally stressing
the limitations of COMSOL. To yield its results, COMSOL uses FEA techniques that mimic the
corresponding numerical methods which would be used in analytical solutions.
We will be comparing the results of the analytical and FEA derived solutions. Finding
results using both methods helps establish confidence in the results, especially when a low
percent error is found. Correct COMSOL results also enable us to upgrade our models and
accurately analyze more complex problems, some of which could not be solved analytically.
Problems 1 & 2 involve thermal studies conducted on a simple rectilinear parallelepiped.
Each of those two problems will have different boundary conditions. Problem 3 will analyze the
thermal behavior of a sphere with set initial thermal conditions when steady state is met.

2
Theoretical Analysis

Problem 1:
Problem one concerns a rectilinear parallelepiped extending in all Cartesian coordinate
directions (x, y, and z) with lengths Lx, Ly, and Lz. The surface at z = Lz, or at the top of the
parallelepiped, has temperature T2. The bottom surface of the shape, or the one at z = 0, has
temperature T1. Given K, the thermal conductivity, as well as ρ, the density, and c, the specific
heat, Equation 1 below is used to solve for the temperature T. All given values are assumed to
be constant.

Temperature equation derivation:

𝜕𝑇(𝑥, 𝑡) 𝐾 2
= 𝛻 𝑇(𝑥, 𝑡)
𝜕𝑡 𝑐𝜌
Equation 1: Time rate of change of the temperature.

Assuming the system is at steady state:

𝜕𝑇(𝑥,𝑡)
= 0
𝜕𝑡

Therefore it can be said:

𝐾 2
𝛻 𝑇(𝑥, 𝑡) = 0
𝜌𝑐
It is known that K, ρ, and c, are constants and are not zero implying:

𝛻 2 𝑇(𝑥, 𝑡) = 0

𝜕2 𝜕2 𝜕2
𝛻 2 𝑇(𝑥, 𝑡) = ( 2 + 2 + )𝑇 = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧 2

Above, the Laplacian equation for the heat conduction, in terms of the temperature T, was derived.
The equation was set to equal zero due to the assumption that the system is in steady state. From
the set up of the problem, it can be said that the temperature does not change with respect to x
or y, so the first and second terms of the Laplacian equation above can be negated. Therefore
the only remaining term is:

3
2
𝜕 𝑇
= 0
𝜕𝑧 2
To find T, the equation is integrated twice. The first integration yields:

𝜕𝑇
= 𝐶1
𝜕𝑧
The above equation is integrated again to obtain an equation for the temperature as a function of
the z-axis position, T(z):

𝑇(𝑧) = 𝐶1 𝑧 + 𝐶2
In order to find the constants and solve the equation for the temperature, boundary conditions
(B.C’s) must be utilized. These are found from the problem statement:

𝑇(𝑧 = 0) = 𝑇1
𝑇(𝑧 = 𝐿𝑧 ) = 𝑇2
Plugging in the two B.C’s and solving for the constants yields:

𝑇2 − 𝑇1
𝐶1 =
𝐿𝑧

𝐶2 = 𝑇1
Substituting the boundary condition values into the above equation for temperature yields
Equation 2, which specifies the temperature accounting for the boundary conditions:

𝑇2 − 𝑇1
𝑇 = 𝑧 + 𝑇1
𝐿𝑧
Equation 2: Temperature equation.

COMSOL was then utilized to simulate the problem. Some data was then exported out of
COMSOL such that it could be utilized by MATLAB to calculate the error between the analytical
solution and the simulated solution.

4
Problem 2:
Problem 2 concerns a similar case as Problem 1 with the main deviation being a thin band
of width w with temperature T3 placed halfway up each side of the parallelepiped. It is to be noted
that T3 is greater in magnitude than T2 and T1. The introduction of this different temperature
creates problems when trying to solve this problem analytically, as the differential equation is now
a partial differential equation that is much harder to solve. This difficulty shows itself in the
simplification of the Laplacean:

𝜕2 𝜕2 𝜕2
𝛻 2 𝑇(𝑥, 𝑡) = ( 2 + 2 + )𝑇 = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧 2

Now, due to the effects of the thin slice in the middle of the parallelepiped, it can no longer be
assumed that the temperature does not change in the x, or y direction. In other words:

2𝑇 2𝑇
𝜕

𝜕𝑥 2
≠ 0 &
𝜕

𝜕𝑦 2
≠ 0

This means that:

2
𝜕 𝑇 𝜕2𝑇 𝜕2𝑇
+ + = 0
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2
Equation 3: Laplacian equation for Problem 2.

This is a partial differential equation that is much more difficult to solve than the previous ordinary
differential equations. To obtain results for the PDE above, COMSOL was used to simulate and
solve the problem.

Problem 3:
Problem 3 analyzes the temperature of a sphere with a hollow center. The solid region of
the sphere has a radius of Router and the material within that region is assumed to have the same
properties as the material in Problems 1 and 2. Namely, the density, ⍴, the thermal conductivity,
K, and the heat transfer coefficient, c, are the same as the previous two problems. The hollow
portion of the sphere has a radius Rinner. For the interface between the hollow center of the sphere
and the solid sphere around it (i.e. the boundary between the two regions), a Dirichlet boundary
condition was assumed. For the outside boundary of the solid sphere, a Neuman boundary
condition was assumed.
The inner temperature is given by Tinner, while the outer temperature, Touter, is given by
the outer heat flux, defined as:

5
𝜕𝑇
𝑞 = −𝐾
𝜕𝑟
Equation 4: The heat flux equation in Problem 3.

From the heat flux equation above, the goal was to solve this problem analytically using spherical
coordinates, expressing T in terms of Tinner, q, K, r, Router, and Rinner. The first step in the process
was converting Cartesian coordinates into spherical coordinates using the following three
equations:

𝑥 = 𝑟𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃)
𝑦 = 𝑟𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃)
𝑧 = 𝑟𝑐𝑜𝑠(𝜃)
The next step was to obtain the derivative of each of the functions above with respect to their
independent variables, r, 𝜑, and 𝜃:

𝜕𝑥
= 𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃)
𝜕𝑟
𝜕𝑥
= 𝑟𝑐𝑜𝑠(𝜑)𝑐𝑜𝑠(𝜃)
𝜕𝜃
𝜕𝑥
= −𝑟𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃)
𝜕𝜑
𝜕𝑦
= 𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃)
𝜕𝑟
𝜕𝑦
= 𝑟𝑠𝑖𝑛(𝜑)𝑐𝑜𝑠(𝜃)
𝜕𝜃
𝜕𝑦
= 𝑟𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃)
𝜕𝜑
𝜕𝑧
= 𝑐𝑜𝑠(𝜃)
𝜕𝑟
𝜕𝑧
= −𝑟𝑠𝑖𝑛(𝜃)
𝜕𝜃

6
𝜕𝑧
=0
𝜕𝜑
Next, the following terms were defined by implementing the chain rule:

𝜕 𝜕𝑥 𝜕 𝜕𝑦 𝜕 𝜕𝑧 𝜕
= + +
𝜕𝑟 𝜕𝑟 𝜕𝑥 𝜕𝑟 𝜕𝑦 𝜕𝑟 𝜕𝑧
𝜕 𝜕𝑥 𝜕 𝜕𝑦 𝜕 𝜕𝑧 𝜕
= + +
𝜕𝜃 𝜕𝜃 𝜕𝑥 𝜕𝜃 𝜕𝑦 𝜕𝜃 𝜕𝑧
𝜕 𝜕𝑥 𝜕 𝜕𝑦 𝜕 𝜕𝑧 𝜕
= + +
𝜕𝜑 𝜕𝜑 𝜕𝑥 𝜕𝜑 𝜕𝑦 𝜕𝜑 𝜕𝑧

Plugging in the derivatives that were obtained earlier (on the previous page) in the equations
above yields:
𝜕 𝜕 𝜕
= 𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃) + 𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃) +
𝜕𝑟 𝜕𝑥 𝜕𝑦
𝜕
𝑐𝑜𝑠(𝜃)
𝜕𝑧
𝜕 𝜕 𝜕
= 𝑟𝑐𝑜𝑠(𝜑)𝑐𝑜𝑠(𝜃) + 𝑟𝑠𝑖𝑛(𝜑)𝑐𝑜𝑠(𝜃)
𝜕𝜃 𝜕𝑥 𝜕𝑦
𝜕
− 𝑟𝑠𝑖𝑛(𝜃)
𝜕𝑧
𝜕 𝜕 𝜕
= −𝑟𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃) + 𝑟𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃) + 0
𝜕𝜑 𝜕𝑥 𝜕𝑦
𝜕

𝜕𝑧

𝜕 𝜕 𝜕
The above systems of equations provide explicit definitions for , , and .
𝜕𝑟 𝜕𝜃 𝜕𝜑
𝜕 𝜕 𝜕
To solve in terms of , , and , a substitution method was used to solve one
𝜕𝑥 𝜕𝑦 𝜕𝑧
of the equations for one of the variables, in terms of all of the rest, then substituting the result into
the following equations and repeating the process for all three variables (i.e. solving the system

7
of equations). For this, Wolfram Alpha was used to obtain the following equations in terms of
𝜕 𝜕 𝜕
, , and :
𝜕𝑥 𝜕𝑦 𝜕𝑧

𝜕 𝜕 𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃) 𝜕
= 𝑐𝑜𝑠(𝜑)𝑠𝑖𝑛(𝜃) +
𝜕𝑥 𝜕𝑟 𝑟 𝜕𝜃
𝑠𝑖𝑛(𝜑)𝜕
+
𝑟𝑠𝑖𝑛(𝜃)𝜕𝜑
𝜕 𝜕 𝑠𝑖𝑛(𝜑)𝑐𝑜𝑠(𝜃) 𝜕
= 𝑠𝑖𝑛(𝜑)𝑠𝑖𝑛(𝜃) +
𝜕𝑦 𝜕𝑟 𝑟 𝜕𝜃
𝑐𝑜𝑠(𝜑) 𝜕
+
𝑟𝑠𝑖𝑛(𝜃) 𝜕𝜑
𝜕 𝜕 𝑠𝑖𝑛(𝜃) 𝜕
= 𝑐𝑜𝑠(𝜃) −
𝜕𝑧 𝜕𝑟 𝑟 𝜕𝜃

To solve for the x, y, and z unit vectors, the orthogonality of the unit vectors in the Cartesian
coordinates will be used in order to simplify the equations. Keeping in mind that the dot product
between two orthogonal vectors (90° apart) is zero, the following derivations can be made:

𝑥 • 𝒓̂ = 𝑠𝑖𝑛(𝜃)𝑠𝑖𝑛(𝜑)𝑥 • 𝑥 = 𝑠𝑖𝑛(𝜃)𝑐𝑜𝑠(𝜑) 𝑥 • 𝜽̂
= 𝑐𝑜𝑠(𝜃)𝑐𝑜𝑠(𝜑)𝑥 • 𝑥 = 𝑐𝑜𝑠(𝜃)𝑐𝑜𝑠(𝜑) 𝑥 • 𝜑
= 𝑐𝑜𝑠(𝜃)𝑐𝑜𝑠(𝜑)𝑥 • 𝑥 = −𝑠𝑖𝑛(𝜑)

The above equations can be rearranged to:

𝑥 = 𝑠𝑖𝑛(𝜃)𝑐𝑜𝑠(𝜑)𝒓̂ + 𝑐𝑜𝑠(𝜃)𝑐𝑜𝑠(𝜑)𝜽̂ − 𝑠𝑖𝑛(𝜑)𝜑

The same procedure was followed for the dot product of 𝑦 :

𝑦 • 𝒓̂ = 𝑠𝑖𝑛(𝜃)𝑠𝑖𝑛(𝜑)𝑦 • 𝑦 = 𝑠𝑖𝑛(𝜃)𝑠𝑖𝑛(𝜑) 𝑦 • 𝜽̂ =
𝑐𝑜𝑠(𝜃)𝑠𝑖𝑛(𝜑)𝑦 • 𝑦 = 𝑐𝑜𝑠(𝜃)𝑠𝑖𝑛(𝜑) 𝑦 • 𝜑 = 𝑐𝑜𝑠(𝜃)𝑦 • 𝑦 =
𝑐𝑜𝑠(𝜃)

8
The equations above can be expressed as:

𝑦 = 𝑠𝑖𝑛(𝜃)𝑠𝑖𝑛(𝜑)𝒓̂ + 𝑐𝑜𝑠(𝜃)𝑠𝑖𝑛(𝜑)𝜽̂ + 𝒄𝒐𝒔(𝜑)𝜑

Lastly, for the dot product of ẑ:

𝑧 • 𝒓̂ = 𝑐𝑜𝑠(𝜃)𝑧 • 𝑧 = 𝑐𝑜𝑠(𝜃)
𝑧 • 𝜽̂ = −𝒔𝒊𝒏(𝜃)𝑧 • 𝑧 = −𝒔𝒊𝒏(𝜃)
𝑧 • 𝜑 = 0

The above equations simplify for a final unit vector of 𝑧:

𝑧 = 𝑐𝑜𝑠(𝜃)𝒓̂ − 𝑠𝑖𝑛(𝜃)𝜽̂
Taking the gradient operator in Cartesian coordinates, where the gradient is equal to:

𝜕 𝜕 𝜕
𝛻 = 𝑥 + 𝒚̂ + 𝒛̂
𝜕𝑥 𝜕𝑦 𝜕𝑧
Now, solving for 𝑥 , 𝒚̂, and 𝒛̂:
𝜕
𝑥 = [𝒔𝒊𝒏(𝜃)𝒄𝒐𝒔(𝜑)𝒓̂ + 𝒄𝒐𝒔(𝜃)𝒄𝒐𝒔(𝜑) 𝜽̂ − 𝒔𝒊𝒏(𝜃)𝜑] ∗
𝜕𝑥
𝜕 𝒄𝒐𝒔(𝜑)𝒄𝒐𝒔(𝜃) 𝜕 𝑠𝑖𝑛(𝜑) 𝜕
∗ [𝒄𝒐𝒔(𝜑)𝒔𝒊𝒏(𝜃) + + ]
𝜕𝑟 𝑟 𝜕𝜃 𝑟𝒔𝒊𝒏(𝜃) 𝜕𝜑

𝜕
𝒚̂ = [𝒔𝒊𝒏(𝜃)𝒔𝒊𝒏(𝜑)𝒓̂ + 𝒄𝒐𝒔(𝜃)𝒔𝒊𝒏(𝜑) 𝜽̂ + 𝒄𝒐𝒔(𝜑)𝜑] ∗
𝜕𝑦
𝜕 𝒔𝒊𝒏(𝜑)𝒄𝒐𝒔(𝜃) 𝜕 𝑐𝑜𝑠(𝜑) 𝜕
∗ [𝒔𝒊𝒏(𝜑)𝒔𝒊𝒏(𝜃) + + ]
𝜕𝑟 𝑟 𝜕𝜃 𝑟𝒔𝒊𝒏(𝜃) 𝜕𝜑

𝜕 𝜕
𝒛̂ = [𝒄𝒐𝒔(𝜃)𝒓̂ − 𝒔𝒊𝒏(𝜃) 𝜽̂] ∗ [𝒄𝒐𝒔(𝜃)
𝜕𝑧 𝜕𝑟
𝒔𝒊𝒏(𝜃) 𝜕
− ]
𝑟 𝜕𝜃

9
Using Wolfram Alpha again to simplify the algebra results in:

𝜕𝑇 1 𝜕𝑇 1 𝜕𝑇
𝛻𝑇 = 𝒓̂ + 𝜃 + 𝜑
𝜕𝑟 𝑟 𝜕𝜃 𝑟𝑠𝑖𝑛(𝜃) 𝜕𝜑
Equation 5: The gradient of the temperature in spherical coordinates.

Knowing that:

𝒓̂ • 𝒓̂ = 𝟏
𝒓̂ • 𝜽̂ = 𝟎
𝒓̂ • 𝜑 = 𝟎

The following can be derived:

𝜕𝑇
𝒓̂ • 𝜵𝑻̂ =
𝜕𝑟
Therefore, we are able to derive Equation 4 for the heat flux in Problem 3:

𝜕𝑇
𝑞 = 𝒓̂ (−𝑲 𝜵𝑻̂) = −𝑲
𝜕𝑟

Using the above equation, it is now possible to derive an equation for the temperature as a
function of the radial distance:

1 𝜕 𝜕𝑇
𝜵𝟐 𝑻̂ = (𝑟 ) =0
𝑟 𝜕𝑟 𝜕𝑟
Where:

𝜕𝑇 𝜕𝑇
= 0 & =0
𝜕𝑡 𝜕𝑟

This means that:

10
𝜕 𝜕𝑇 𝜕𝑇
(𝑟 ) = 0, 𝑤ℎ𝑒𝑟𝑒 (𝑟 ) =𝜻
𝜕𝑟 𝜕𝑟 𝜕𝑟

This allows us to get an equation of 𝜻 in terms of r:

𝜕𝑇 𝜻
=
𝜕𝑟 𝑟

After integrating the above equation with Tinner as the integral boundary on one left and rinner as
the integral boundary on the right side, and rearranging the equation to be in terms of T(r), the
following equation is obtained:

𝑟
𝑇(𝑟) = 𝑻̂𝒊𝒏𝒏𝒆𝒓̂ + 𝜻 ∗ 𝒍𝒏( )
𝑟𝑖𝑛𝑛𝑒𝑟

At the outer position, router, the following is true:

𝜕𝑇 −𝑞 𝜕 𝑟
|𝑟𝑜𝑢𝑡𝑒𝑟 = = [𝑇𝑖𝑛𝑛𝑒𝑟 + 𝜻 ∗ 𝒍𝒏( )]
𝜕𝑟 𝐾 𝜕𝑟 𝑟𝑖𝑛𝑛𝑒𝑟

Simplifying the above equation produces:

−𝑞 𝜕𝑇𝑖𝑛𝑛𝑒𝑟 𝜕 𝑟
= + [𝜻 𝒍𝒏( )]
𝐾 𝜕𝑟 𝜕𝑟 𝑟𝑖𝑛𝑛𝑒𝑟

𝜕𝑇𝑖𝑛𝑛𝑒𝑟
Since is equal to zero, the equation further simplifies to:
𝜕𝑟

−𝑞 1 𝜻
= 𝜻 |𝑟𝑜𝑢𝑡𝑒𝑟 =
𝐾 𝑟 𝑟𝑜𝑢𝑡𝑒𝑟
𝒒
𝜻 = ∗ 𝑟𝑜𝑢𝑡𝑒𝑟
𝐾

Therefore, when we plug in the result we got for 𝜻 in the temperature equation, we obtain the
following equation for the gradient of the temperature in spherical coordinates:

𝑞 𝑟
𝑇(𝑟) = 𝑻̂𝒊𝒏𝒏𝒆𝒓̂ − 𝒓̂𝒐𝒖𝒕𝒆𝒓̂ ∗ 𝒍𝒏( )
𝐾 𝑟𝑖𝑛𝑛𝑒𝑟

11
Equation 6: the gradient of the temperature in spherical coordinates.

12
COMSOL Results

Problem 1:

Figure 1: Full 3-D Body Temperature Simulation

Figure 1 shows the 3 dimensional solution for the rectilinear parallelepiped when the initial
temperature conditions of T(z = Lz) = 450°K and T(z = 0) = 200°K are established and all other
surfaces are insulated.

13
Figure 2: Temperature Slice Plot

Figure 2 demonstrates that a temperature gradient only exists in the z-direction under the
insulated conditions. The takeaway from this plot is that the temperature gradient in the x direction
is zero while the thermal energy flows strictly from the top face of the block downward.

14
Figure 3: Temperature Line Plot

The line plot above further demonstrates that the gradient in the z direction is a constant
value, with the temperature increasing linearly from z = 0 to z = 𝐿𝑧 . This plot alone is arguably
sufficient to describe the solution of the problem considering that any line location along 𝐿𝑥 or 𝐿𝑦
would yield the same results.

15
Figure 4: Error Calculation

Although the error calculation seems to be unstable, the scale of the “Analytical vs
Experimental” chart tells us that the error is virtually zero. In comparing the experimental and
analytical data, the yielded error is between −2 𝑥 10−13 and 3 𝑥 10−13 , an incredibly small scale
and range.

16
Problem 2:

Figure 5: Full 3-D Body Temperature Simulation

Similar to Figure 1, Figure 5 shows the 3 dimensional solution for the rectilinear
parallelepiped. Main difference here is the band of width 0.1m and temperature 500°K placed
halfway up each side of the parallelepiped.

17
Figure 6: Temperature Slice Plot in XZ plane

Figure 6 shows the temperature distribution across the XZ plane length of the slice. As
we can see it gets hotter as we move further up the Z-plane, but stays constant as you move
along the X-plane.

18
Figure 7: Temperature Slice Plot in YZ plane

Figure 7 shows the temperature distribution across the YZ plane length of the slice. As
we established above in Figure 6, it gets hotter as we move further up the Z-plane. Similar to
the X-plane in Figure 6, the Y-plane follows suit and the temperature stays constant as you
move along the Y-plane.

19
Figure 8: Temperature Line Plot

The Line plot above demonstrates that the gradient in the z direction is not a constant
value like how we have seen in Problem 1. The slope is not a straight line and the relationship is
not linear anymore. In Problem 1, the increase in temperature was very uniform and constant
relative to its distance and followed a very exact path without any deviations. In this scenario, as
we have established from previous figures, the temperature in the Z-axis will increase as we go
up in the Z-plane, that is exactly what happened in Problem 1 too, but the temperature increased
much quicker, for example at x = 0.5m in Problem 1 corresponded to a temperature of 325°K,
looking at x = 0.5m in Figure 8 corresponds to a temperature of 430°K. However they both have
similar peaks, it's just that the peak is reached quicker in this problem. After the maximum
temperature is reached the slope becomes approximately zero until the end of the arc is reached
(x = 1m).

20
Problem 3:

Figure 9: Temperature Slice for Problem 3 Sphere

The temperature gradient presented through this chart is large in close proximity to the
center where temperature is a constant 500°K. Temperature change is relatively small once the
value becomes light blue.

21
Figure 10: Graphed 1D temperature data from Problem 3 Sphere

The graphed data demonstrates a non-linear relationship with temperature change and
distance, r. The temperature drops rapidly in close proximity to the heat source where temperature
is 500°K. The temperature continues to decrease between half the radius (~.055m) and the full
extent of the radius, however at a much smaller rate. This observation falls in line with the visuals
presented in the temperature slice, Figure 9.

22
Figure 10: Temperature Error along radius

Discussion

23
Problem 1:
Upon solving the equations it can be seen that the thermal conductivity falls out from the
solution. This occurs from the steady state assumption that the time derivative of the temperature
is zero. The term holding the thermal conductivity is a constant indicating the positional derivative
of the temperature must be zero. This assumption means that the thermal conductivity does not
matter in the steady state situation because it gets negated by the zero value of the derivative of
temperature.
COMSOL was used after being given values for all the parameters in the problem to go
and confirm what we have derived in the analytical solution. The results from COMSOL as we
have explained above mostly matched what we had found out from the analytical solution. There
is a very very small discrepancy between the simulation and analytical solution on the order of
10-13. This is a very minimal error that does not significantly impact the simulation. COMSOL
served as a great tool to check our answers and assumptions and confirm we have the right
approach to the problem.

Problem 2:
Problem 2 is set up similarly to Problem 1, with just one main difference. There is a band
of width w and temperature T3 placed halfway up each side of the parallelepiped. We were also
asked to do the same task and derive an equation for the temperature T. The addition of the band
in this problem leads to issues when trying to solve this problem analytically as the differential
equation is now a partial differential equation that is much harder to solve. Now we are left with:

𝜕
2𝑇 𝜕2 𝑇 𝜕2 𝑇
+ + = 0
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2

Instead of:

𝜕
2𝑇
= 0
𝜕𝑧 2

Similarly to Problem 1, which was far easier to solve. We cannot predict how the
temperature will behave in the X or Y direction.
We then employed COMSOL to help us understand more about how the temperature will
behave and set up the problem there after we were given values for the different parameters. We
added the band to the problem and followed the same analysis and approach as Problem 1.
To our surprise, the temperature along the X and Y axis also did not change similarly to
Problem 1. On the other hand, along the Z-axis, the temperature increased as we moved along
the axis, similar to Problem 1. However, that temperature change was significantly different than
the temperature change exhibited in Problem 1. In Problem 1, there was a linear relationship
between the arc length and the temperature, which resulted in a constant slope throughout. In
Problem 2, the temperature along the Z-axis seemed to increase much quicker than in Problem
1, reaching the maximum temperature of 450°K at about x = 0.6m. By comparison, in Problem 1,
at x = 0.6m, the temperature was only 350°K.

24
In this application, COMSOL helped us understand a situation with a very complicated
equation which was difficult to solve analytically. By solving these analytical partial differential
equations on our behalf, software such as COMSOL can save us valuable time and yield accurate
results.

Problem 3:

Problem 3 demonstrates heat conduction through a hollow copper sphere with a thickness
of 0.99m. The cavity is small relative to the overall size of the object. In this problem, we set an
initial condition of 500°K at the center of the sphere and defined the outward heat flux, q, as
10,000 𝑊/𝑚2 on the outer surface of the sphere. This heat flux describes a rate of heat transfer
passing through a boundary per unit area. An inner heat flux indicates a rate of heat transfer into
a system or boundary, while an outer heat flux indicates a rate of heat transfer out of a system
per unit area.
In this case, we are given a positive q, which indicates an outer heat flux. In other words,
heat is leaving the sphere at the outer boundary. This outer heat flux will cause the sphere to
release heat to its surroundings, cooling down in the process. We expected to see the temperature
decrease as the radius increases from the center. This study yields results when an equilibrium
state is reached.
Figure 9 shows that the temperature drops rapidly immediately outside the temperature
source of 500K. The large temperature gradient can be explained through the fact that this is a
3D sphere. A 3D sphere has a volume that increases exponentially with radius. In other words, if
the radius doubles, the volume will increase by a factor of 8. This exponential increase in volume
causes the heat to dissipate rapidly with r. Eventually, the gradient will decrease as the difference
in temperature with the outside surface decreases (ΔT). That is what we observe with the large
blue area in Figure 9.
Figure 10 numerically supports the conclusions drawn from Figure 9. The line chart
rapidly dips down at first. It continues to decrease however, at an increasingly smaller rate as the
radius increases.

25
Conclusion
After finishing this assignment, we believe that COMSOL is a great tool to help us
understand the behavior of different thermal problems. In Problem 1, we were able to solve the
differential equation without any issues, we then went to COMSOL to verify our solutions to make
sure we did not make any errors in the derivation. It's always a good practice to confirm the results
we get by hand calculations on COMSOL to get the full picture of what is happening.
Another example of why COMSOL is such a great tool was Problem 2. We ended up with
a very complicated partial differential equation that is very difficult to solve by hand. We then
headed to COMSOL to help us understand the behavior of the temperature as we were unable to
make any predictions or solve the equation analytically. We inserted the parameters from the
problem and received a full analysis of how the temperature varies in the problem.

26
Appendix

Assignment Description

27
28
29
30

You might also like