Scribd
Upload
9 views3 pages

Weighted Residual

The lecture introduces the Finite Element Method (FEM) as a superior approach to solving partial differential equations (PDEs) compared to the Finite Difference Method (FDM), particularly for irregular domains and situations with additional physical properties. It discusses the equilibrium equation for a 1D structural problem, boundary conditions, and the concept of minimizing residuals using the weighted residual method, also known as the Galerkin approach. The lecture concludes with the formulation of the weak-form of the PDE, which relaxes the differentiability requirements of the solution.

Uploaded by

raghunandan gowda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views3 pages

Weighted Residual

The lecture introduces the Finite Element Method (FEM) as a superior approach to solving partial differential equations (PDEs) compared to the Finite Difference Method (FDM), particularly for irregular domains and situations with additional physical properties. It discusses the equilibrium equation for a 1D structural problem, boundary conditions, and the concept of minimizing residuals using the weighted residual method, also known as the Galerkin approach. The lecture concludes with the formulation of the weak-form of the PDE, which relaxes the differentiability requirements of the solution.

Uploaded by

raghunandan gowda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

CS601, Lecture 18/10/2022 – Finite Element Method

This lecture begins with an introduction to an approach to solving PDES. This approach is called
Finite Element Method (FEM).
Introduction: FDM is not practical/accurate approach to solve for all domains and all situations.
Few reasons why FDM is not practical are as follows:

• Domain structure: FDM cannot handle accurately domains with irregular


geometry/shapes. E.g., a domain with a curved boundary, domain with holes (with
irregular boundaries).
• Truncation error: what is the guarantee that successive derivatives of the unknown
𝑑2𝑢
function (e.g., 𝑢 in 𝐸𝐴 + 𝐹) decrease in magnitude when we write the Taylor series
𝑑𝑥 2
approximation? for e.g., 𝑓(𝑥) = 𝑒 𝑥 , the magnitude remains the same. As a result, when
we ignore the higher order terms of the Taylor series and truncate the series, the numerical
error / truncation error might be significant.
• Other physical properties: when there are additional mechanical engineering properties
that need to be satisfied, FDM doesn't consider these.
FEM is a method of solving PDEs that overcomes the shortcomings of FDM to solve practical
engineering problems in mechanical, electrical engineering, physics, aerospace engineering,
among others.
Example: 1D structural problem

𝑑2𝑢 (1)
𝐸𝐴 2 + 𝐹 = 0
𝑑𝑥
The above equation is also called as equilibrium equation. Typically, F is given as 𝑏𝐴, where 𝑏 is
the body force per unit area and 𝐴 is the area. In this problem, the goal is to find stresses and
strains at different points on the rod, which is fixed at one end and is subjected to a force at the
other end. An illustration is shown below:

We have: Stress (𝜎) / Strain (𝜖) = E (Young’s modulus)


𝜕𝑢
Strain (𝜖) = change in length / original length =
𝑑𝑥
𝜕𝑢
Substituting: 𝐸
𝑑𝑥
=𝜎
To solve the PDE shown in (1), we need boundary conditions. Typical boundary conditions look
like:
At 𝑥 = 0, 𝑢 = 0
(2)
At 𝑥 = 𝐿, 𝑢 = 𝐸𝐴𝜕𝑢/𝜕𝑥 (the value is specified)
The PDE of equation (1) together with the boundary conditions in (2) is called as strong-form of
the PDE equation. Suppose, 𝑢̃ is an approximate solution. Being approximate in nature, 𝑢̃ will not
satisfy the equation (1) at all points x i.e.
𝑑 2 𝑢̃ (3)
𝐸𝐴 +𝐹 =𝑅
𝑑𝑥 2
We say that 𝑢̃ has an associated error or Residual 𝑅. The goal is to minimize the residual. The
method to minimize the residual is called as weighted residual method. This method simply
multiplies the residual with a weighted function, takes the integral and equates to 0.

∫𝜔 𝑅 = 0

This approach is also called as Galerkin approach. 𝜔 are called as weight functions or shape
functions. Substituting for R from (3) in the above equation,

𝑑 2 𝑢̃
∫ 𝜔 (𝐸𝐴 + 𝐹) = 0
𝑑𝑥 2

𝑑2 𝑢
̃
=∫Ω 𝜔 𝐸𝐴 𝑑Ω + ∫Ω 𝜔 𝐹 𝑑Ω = 0, where ∫Ω denotes integral over the domain Ω.
𝑑𝑥 2
For the 1D rod problem, this is integral over 0 to 𝐿.

𝑑2 𝑢
̃
We can substitute for 𝜔 in the first term above with the following information from chain
𝑑𝑥 2
rule of differentiation:
𝑑 ̃
𝑑𝑢 ̃
𝑑𝜔 𝑑𝑢 𝑑2 𝑢
̃
[𝜔 ]= +𝜔
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥 2
Substituting:
𝑑 𝑑𝑢̃ 𝑑𝜔 𝑑𝑢̃
∫ (𝜔𝐸𝐴 ) 𝑑Ω − ∫ 𝐸𝐴 𝑑Ω + ∫ 𝜔 𝐹 𝑑Ω = 0
Ω 𝑑𝑥 𝑑𝑥 Ω 𝑑𝑥 𝑑𝑥
Rearranging terms:
𝑑𝑢̃ 𝐿 𝑑𝜔 𝑑𝑢̃
[𝜔𝐸𝐴 ] + ∫ 𝜔 𝐹 𝑑Ω = ∫ 𝐸𝐴 𝑑Ω (4)
𝑑𝑥 0 Ω Ω 𝑑𝑥 𝑑𝑥

̃ 𝐿
𝑑𝑢
In the above equation, [𝜔𝐸𝐴 ] is called the boundary term. The coefficient of the weight
𝑑𝑥 0
function in the boundary term represents the Neumann BC. This coefficient is also called as
̃
𝑑𝑢
secondary variable. 𝐸𝐴 is the coefficient in the boundary term and is the secondary variable
𝑑𝑥
above. 𝜔 represents Dirichlet BC and is also called as primary variable. 𝜔 denotes the unknown
function 𝑢, at boundaries. Whenever u=0 (from (2) 𝑢 = 0 at 𝑥 = 0), set 𝜔 = 0 . Applying this in
(4) we get:

𝐿 𝐿
𝑑𝑢̃ 𝑑𝜔 𝑑𝑢̃ (4)
[𝜔𝐸𝐴 ] + ∫ 𝜔 𝐹 𝑑𝑥 = ∫ 𝐸𝐴 𝑑𝑥
𝑑𝑥 𝐿 0 0 𝑑𝑥 𝑑𝑥

Equation (4) is called as weak-form of the PDE. The equation in (1) contains a second-order
derivative term. This means that there is a requirement that 𝑢 must be differentiable twice so
that 𝑢 is continuous (at all points x, 𝜕 2 𝑢/𝜕𝑥 2 is real). In equation (3), this requirement is weakened
in the sense that there are only first-order derivative terms. So, it is sufficient if 𝑢 is differentiable
once.

You might also like