Finite element methods are used to solve boundary value problems (BVP) or initial boundary value problems (IBVP) in engineering. BVPs are mathematical models of real-life situations. Such situations can be physical, biological, economic, and so on.
We will explore the problem of heat conduction and see how we build a finite element model and solve this problem. The first step will be to build a model. The model is the IBVP in the form of a partial differential equation or a variational problem. We will discuss whether the problem is well-posed. Then we will try to construct an approximate solution to the problem. Finally, we will discuss how good the solution is.[1]
Figure 1 shows a region through which heat is flowing. Points in the medium are represented by with
components with respect to the basis
and the origin. The heat flux into (or out of) the medium is where is the time. The heat flux takes place across part of the boundary of the medium (). The temperature on the remainder of the boundary () has a known value
. Heat sources inside the medium (for example, chemical reactions and plastic deformation) are given as a function .
The goal is to find the temperature distribution in the medium () as time evolves.
A realistic model for this problems needs two components:
Let us assume that there are no sources of energy other than thermal
energy. Then, the balance of energy states that:
Rate of change of thermal energy = heat generated by internal sources + heat flow into the body.
We have to convert this statement into mathematical form. To do this, consider an arbitrary part of the body () with
boundary . Let be the outward unit normal to the boundary.
The total thermal energy in at a particular time is given by the heat capacity of the body. The heat capacity is the amount of energy needed to raise the temperature (of a unit mass of the body) by one unit.
Let the mass density be . Then, the total thermal energy in at time is
where we have assumed that the components are with respect to the
Cartesian basis (). In the following, whenever we use index notation, the components are assumed to be with respect to that Cartesian basis.
There are two unknowns in equation (8). These are and and only one equation. Therefore, we need another equation that characterizes the material.
One possibility is the Fourier law which states that:
the heat flux is linearly related to the temperature gradient.
In mathematical form,
The quantity is a second-order tensor called the thermal conductivity tensor. The minus sign shows that heat flows
from hot to cold. Recall that a second-order tensor takes a vector to another vector (in this case a temperature gradient to a heat flux).
In index notation, we can write equation (9) as
If the region is homogeneous then is constant.
If the region is isotropic, the thermal conductivity tensor takes the form
Boundary conditions (BCs) are needed to make sure that we get a unique
solution to equation (12).
The temperature is prescribed on . Prescribed boundary conditions are also called Dirichlet BCs or essential BCs. In this case
The heat flux is given on the remainder of the boundary (). Such flux boundary conditions are also known as Neumann BCs or natural BCs. The flux boundary condition is
In index notation, the essential boundary condition is
Plug equation (9) into (14). We get
If the region is isotropic with thermal conductivity , we can define and
(also called the normal derivative). Then the flux BC simplifies to
↑ This writeup is based on the introductory chapter in Introductory Functional Analysis with Applications to Boundary Value Problems and Finite Elements by B. Daya Reddy, Springer, 1998, and the chapter on parabolic and hyperbolic problems in The Finite Element Method: Linear Static and Dynamic Finite Element Analysis by T.J.R Hughes.