You are on page 1of 7

ONE DIMENSIONAL TRANSIENT HEAT CONDUCTION MODEL USING FINITE DIFFERENCE METHOD

Richard Jess Chan, Gromyko Geraldino Jr., and Joseph Mayormita Department of Mechanical Engineering University of San Carlos Cebu City, Philippines 6000
Abstract This paper presents a mathematical solution to a one dimensional transient heat conduction model representing the proposed cylindrical instrument used for measuring thermo physical properties such as thermal conductivity, specific heat, and thermal diffusivity. The model is composed of two cylindrical shell one being inside the other in a concentric manner while the test material being put between them. Constant heat flux is applied in the inner shell which causes heat conduction in outward radial direction to the outer insulated shell. Physical laws that govern the conduction heat transfer process were employed with the boundary conditions given. Using Finite Difference Method namely Explicit, Implicit, and Crank-Nicolson methods the model was able to give mathematical solutions. Implicit and Crank-Nicholson methods are found to be reliable solutions having an error of only 9.89% compared to 71.27% of explicit method. With the aid of computer programming the model was able to simulate and predict the characteristics of different test materials. It was found out that the lesser the difference between the inner and outer shell radius the faster the results can be achieved. The obtained mathematical model in this paper serves as a guide in the design of the proposed instrument.

1. Introduction Thermo-physical properties of any material such as thermal conductivity, specific heat, and thermal diffusivity are very important to be known. It can either be determined using steady state method or transient state method. Transient state method is often used for it can measure such properties at a shorter period of time, higher precision, and broader measuring range in contrast to the steady state method. One simple way of doing this method is to use a flat rectangular test specimen be subjected to a heat flux in one side thereby producing heat conduction. The temperature increase of the specimen with respect to time then determines such properties. However, due to its geometry such simple transient method causes less accurate measurement since the heat applied cannot be treated as one dimensional. Modifying its geometry into a hollow cylindrical shell with heat flux applied radially outward makes it one dimensional assuming it has infinite length. In this manner a more reliable and accurate data can be generated. This study aims to develop a mathematical model representing onedimensional transient heat conduction across a cylindrical layer of sample material. This method was originally proposed by AbdelWahed and co-workers which measures thermal properties of material sample contained in the annular space between two concentric, metallic cylinders with the inner cylinder being heated uniformly. This allows simultaneous measurements of specific heat, thermal conductivity, and thermal diffusivity.[1] A better solution to a mathematical modeling involves numerical approach which applies Implicit, Explicit, and Crank-Nicolson methods. Further simulation of the model using computer software may able to provide a faster, reliable, and graphical solution to the problem. 2. Modeling Mathematical models are used to predict behavior of physical systems, such as heat flow through a material, fluid flow, vibration and other engineering problems. Such models are useful in many engineering applications where cost estimates and safety considerations need be predicted to ensure the success of the project.

Mathematical modeling is classified into analytical approach and numerical approach. This study utilized the numerical approach namely Implicit, Explicit and Crank-Nicolson methods. Implicit method makes use of the time derivatives evaluated only at the old time in contrary to explicit method which uses only the time derivatives evaluated at new time. On the other hand, crank-nicolson method uses the average of the time derivatives at the old and new times. The model is governed by the laws of transient heat conduction through a cylindrical wall thereby following the energy equation with the given boundary conditions. Assuming the sample material simulates an infinitely long hollow cylinder with heat fluxes in radial direction only. The model can be treated as one dimensional which behaves in the equation; ( )

difference between inside and outside surfaces of the cylindrical sample becomes, *( ) [( ) ,( ) ] -+

which is independent of time. From this equation the thermal conductivity k of the material is derived as *( ) [( ) ] ,( ) -+

With the following boundary conditions: At :

At

:

By measuring q and ΔT during quasisteady portion of the heating process, thermal conductivity k can be evaluated since the second factor in equation (7) is only a function of the geometry of the test sample material. The time variation of the temperature during the quasi-steady heating process is derived from equation as,

The solution of equation (1) with specified boundary conditions is given by Eckert and co-workers: * ( ) [( ) ( ) ] + or The first term in this equation represents the quasi steady part of the solution, while the second term, F(r,t), represents the transient part of the solution which vanishes for t→∞. Practically, this term vanishes after a finite time t1 establishing a final quasi-steady temperature profile. After this time interval the temperature

[( )

]

As expected is independent of r. the thermal diffusivity follows from equation (5) as, *( ) +

From the known value of thermal conductivity (k), the measured heat flux q and the slope of the temperature-time history,

thermal diffusivity can be evaluated by the equation (10) since the second factor in this equation is only a function of geometry. 3. Methodology (Derivation of Numerical Solution) The numerical model was derived using control volume method. In this method 10 nodes were assigned radially to certain parts of the cylinder in this study. The arrangement of the nodes is shown in Fig. 1. Each node was treated as a control volume, where heat flux can flow in and out, and heat energy can be stored.

With the assumptions above the formulas for the numerical solution were derived from basic heat balance,

or

where:

From Fourier’s law of heat conduction:

where: -

⁄ ⁄ In this study three methods of solving the PDEs mentioned above were used and compared. These methods are Explicit method, Implicit method, and Crank-Nicolson method. The amount of heat entering node 1, , is equal to the amount of heat leaving node 1, , plus the energy store in that node, . For node 1 is the heat source from the inner surface of the cylinder, which is a numerical value in the equation. See Fig. 2. 3.1. Explicit Method For the explicit method, applying finite difference the equation at node 1 was derived as

Fig. 1: Cut-Away View of Cylinder and Placement of Nodes. The solution for the model was based on the following assumptions: 1. Heat flux flows from the inner surface of the cylinder to the outer surface along the radial direction only; no heat transfer along the axial direction. 2. The outer surface is adiabatic. 3. Cylinder thermo-physical properties are constant.

(

)( ( ) (

)

where: ( ( ) In explicit method the future values of temperature at each node, are computed based on the previous values of the temperature at the same node and its neighbor nodes, and . The approach employed in this study for solving the aforementioned equations is Gauss-Seidel iteration. 3.2. Implicit Method For the implicit method, similar analysis was used in deriving the equations. The only difference is that the spatial derivative ⁄ is approximated at an advance time level . That is, ) )

Fig. 2: Heat Balance at node 1. Solving for ⁄

where: ( ( ) )

Derivations of equations for nodes 2 to 10 are similar to the equation for node 1; however, is not a numerical value. Equation for nodes 2 to 10 are summarized below. For node 2 to 9:

Summary of the equations for each node is given below. For node 1:

where: ( where: ( For nodes 2 to 9: ) )

[ For node 10:

]

where:

[

]

For node 10:

[ For node 10:

]

where: ( ( 3.3. Crank-Nicolson In Crank-Nicolson method the approximation of the spatial derivative is the average of that in the explicit and implicit method, ( ) ( ) ) ) where:

Both the implicit and crank-nicolson methods formed a tridiagonal matrix and were solved using Gaussian elimination. Microsoft Excel and Visual Basic were used to solve the equations mentioned above because of the repetitive nature of the solution. 4. Results and Observations The three methods used in this study – explicit, implicit, and crank-nicolson, had varied results in temperature distribution and quasi-steady state temperature difference between inside and outside surfaces of the cylinder, which we will call ΔT. The temperature distribution and time required to reach quasi-steady state varied with different cylinder geometry, heat flux per unit length, and thermophysical properties.

Summary of the equations for each node is given below. For node 1:

where:

( For node 2 to 9:

)

4.1. Comparison of Explicit, Implicit, and Crank-Nicolson Methods The results of the three methods mentioned above were compared using the same test conditions: heat flux, thermophysical properties and geometry of the cylinder. The three methods showed different results in temperature distribution at any time t > 0. The temperature distribution for the implicit and crank-nicolson method did not have much difference compared to the explicit method. For the test condition of (properties of iron): ⁄

where:

⁄ ⁄ ⁄ ⁄

the time required to reach quasi-steady state for the explicit, implicit, and crank-nicolson methods are 10.08-sec, 11.42-sec, 11.12-sec, respectively. The temperature distribution in °C for the three methods at t = 1 second is shown in Table 1 for comparison. Table 1: Sample Temperature Distribution node Explicit Implicit CrankNicolson 1 30.0347 30.1090 30.1092 2 30.0283 30.0886 30.0888 3 30.0227 30.0712 30.0714 4 30.0180 30.0566 30.0567 5 30.0142 30.0447 30.0447 6 30.0112 30.0353 30.0352 7 30.0088 30.0282 30.0281 8 30.0072 30.0232 30.0231 9 30.0063 30.0203 30.0202 10 30.0060 30.0194 30.0192 However, the difference in temperature distribution between the implicit and cranknicolson methods disappear after the quasisteady state has been reached. The ΔT’s are 0.0308°C for the explicit method, and 0.0966°C for both implicit and crank-nicoloson methods. The exact solution for ΔT from Eq. (5) is 0.1072°C. The error is 9.89% for the implicit and crank-nicolson methods and 71.27% for the explicit method. 4.2. Thermophysical Properties and Heat Flux Different materials have different thermophysical properties which affect the conduction of heat within the material. An important property to consider is the thermal diffusivity which is a function of thermal

conductivity, specific heat, and density. While maintaining all other parameters constant, e.g. , etc., the materials with higher thermal diffusivity exhibit a lower temperature difference between the inside and outside surfaces of the cylinder. This is because heat flows through the material faster, so it takes less time for the temperature to change, and it also takes less time to reach the quasi-steady state. The time at quasi-steady state is denoted as . For the same elapsed time, the material with the higher thermal diffusivity will have a lower temperature at the inner surface because most of the heat is conducted away towards the outer surface; hence less heat is stored at the inner surface. When the input heat flux per unit area is changed by a factor x, ΔT also changes uniformly by the same factor, considering all other parameters are kept constant. That is, when the input heat is increased, ΔT also increases. also changes with input heat flux. When the input heat flux is increased, also increases and vice versa. 4.3. Cylinder Geometry The time required to reach quasi-steady state is also a function of the cylinder’s inner and outer radii, and , respectively. When the difference between the outer and inner radii is greater, is longer; when the difference in radii is smaller, is shorter. ΔT also increases with increasing difference in radii. The above statement is true when all other parameters are kept constant. 5. Conclusion The results verify that the implicit and crank-nicolson methods are more accurate than the explicit method. In estimating the ΔT, the implicit and crank-nicolson methods have an error of only 9.89% compared to 71.27% for explicit method. The equations from Chapter 2 shows that the thermal conductivity is a function of heat flux, and , and ΔT, which support the findings of the study.

The geometry of the proposed apparatus must have a small difference between and to allow for a speedy ( ) determination of the thermal properties of the test material. A low heat source will also speed up very slightly the determination of thermal properties since the quasi-steady state will be reached faster – low value of . However, since lower heat source also decreases the value of ΔT, this might present difficulties in reading the ΔT depending on the accuracy of the thermocouple used for determining the inner and outer surface temperatures. In fabricating the said apparatus, the researchers recommend that a high heat source be used because it would allow an accurate reading of the ΔT; and the difference in is only very minimal. Reference 1. J. P. Holman, Heat Transfer, 8th Ed. 2. Incropera, et. al., Fundamentals of Heat and Mass Transfer, 6th Ed. 3. Chapra, Canale, Numerical Methods for Engineers. 4. Eckert and co-workers 5.