You are on page 1of 18

An Introduction to 

the Finite Element 
Method Part 1: Introduction
by
Sergio Preidikman Orientation 
Solving a Problem by Finite Element Analysis
Especialización en Estructuras
Facultad de Ingeniería Civil Mathematical Models in One Dimension
Universidad del Cauca Two‐point boundary‐value problems
Popayán, Cauca, Colombia A Model Problem
22 de enero al 9 de febrero de 2018 01/25/2018 2
UNICAUCA 2018 by Sergio Preidikman

Orientation Orientation
• The finite element method is a general technique for • The finite element method involves four basic steps:
constructing approximate solutions to boundary-value
1. The weak or variational formulation of the problem;
problems (BVPs).
2. The construction of piecewise polynomial trial
• The method involves dividing the domain of the solution
functions;
into a finite number of simple subdomains, the finite
elements, and using variational concepts to construct an 3. The computation of the stiffness matrix and the
approximation of the solution over the collection of solution of the discrete system; and
finite elements.
4. The estimation of accuracy in the final Ritz
• Because of the generality and richness of the ideas approximation.
underlying the method, it has been used with remarkable
success in solving a wide range of problems in virtually
all areas of engineering and mathematical physics.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 3 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 4
Orientation Orientation
• To explain the basic approach of the FEM, consider a • It is straightforward to write a heat balance equation for
plate with a hole, as shown in the next figure, for which each point in the plate.
we wish to find the temperature distribution.
• However, the solution of the resulting partial differential
equation (PDE) for a complicated geometry, such as an
engine block, is impossible by classical methods like
separation of variables.
• Numerical methods such as finite difference methods
Plate with a hole. (FDM) are also quite awkward for arbitrary shapes;
software developers have not marketed finite difference
programs that can deal with the complicated geometries
that are commonplace in engineering.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 5 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 6

Orientation Orientation
• Similarly, stress analysis requires the solution of partial
differential equations that are very difficult to solve by Triangular Finite Element.
analytical methods except for very simple shapes, such
as rectangles, and engineering problems seldom have
such simple shapes.

• The basic idea of FEM is to divide the body into finite


elements, often just called elements, connected by nodes,
and obtain an approximate solution. This is called the
finite element mesh and the process of making the mesh
is called mesh generation.

Finite Element Model. Refined Finite Element Model.


01/25/2018 UNICAUCA 2018 by Sergio Preidikman 7 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 8
Orientation

Part 1.1: Solving a Problem by 
Finite Element Analysis (FEA)

Finite Element Model. Refined Finite Element Model.


01/25/2018 UNICAUCA 2018 by Sergio Preidikman 9

Solving a Problem by FEA Solving a Problem by FEA


• Solving a practical problem by FEA involves: • Most often, more than one cycle through these steps is
required.
1) learning about the problem,
• Time spent by the computer is a small fraction of time
2) preparing a mathematical model,
spent by the analyst, but the analyst must have an
3) discretizing it, understanding of what the computer is doing.
4) having the computer do calculations, and
5) checking results.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 11 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 12


Problem Classification Problem Classification
• The analyst must understand the nature of the problem. • Software has limitations and almost certainly
contains errors, yet the engineer, not the software
• Without this step a proper model cannot be devised,
provider, is legally responsible for results obtained.
nor can FEA software be told what to do.
• At present, software does not automatically decide that
non-linear analysis is to be undertaken if stresses are
high enough to produce yielding, that buckling is to be
considered if thin sections carry compressive load, and
so on.
• Although the trend is for software to be given more
decision-making capability, the analyst should not
abdicate control.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 13 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 14

Mathematical Model Mathematical Model


• Before undertaking FE discretization and a numerical • Material may be idealized as linear and isotropic.
solution, we devise a model problem for analysis.
• Depending on the dimensions, loading, and boundary
• This step involves deciding: conditions of this idealization, we may decide that
behavior is described by beam theory, by plate-bending
1) what features are important to the purpose at hand,
theory, by equations of plane elasticity, or by some other
so that unnecessary detail can be omitted, and
analysis theory.
2) what theory or mathematical formulation describes
• The simplified problem, with the analysis theory to
behavior.
be applied in solving it, constitutes the mathematical
• Thus we may ignore geometric irregularities, regard model.
some loads as concentrated, and say that some
“supports” are fixed.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 15 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 16


Mathematical Model Mathematical Model
• Because subsequent FEA is approximate and pertains • Also, initial modeling decisions are provisional.
only to the mathematical model, FEA is two or three
• It is likely that results of the first FEA will suggest
steps removed from reality.
refinements, in geometry (perhaps by restoring
• Modeling decisions are influenced by: geometric irregularities previously omitted), in
applicable theory (perhaps by adding in-plane stretching
1) what information is sought,
terms to plate-bending theory), and so on.
2) what accuracy is required,
3) the anticipated expense of FEA, and
4) its capabilities and limitations.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 17 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 18

Mathematical Model Mathematical Model


• As an example of modeling, consider the very simple Gravity
problem depicted in Fig. 1(a): F
tR
a) The ring thickness, t, (measured normal to the
figure) is uniform and is considerably less than the t
ring diameter. A
b) The material is considered linearly elastic, B E
R (mean)
homogeneous, and isotropic. C
We ask for stresses and deflections due to the ring’s own
weight as it rests on the ground.
D
Ground
(a) Figure 1. (b)
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 19 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 20
Mathematical Model Mathematical Model
• It is easy to arrive at the plane model in Fig. 1(b), in
which symmetry about the vertical centerline has been
exploited.
Details of pressure applied by the ground have been
(a) Ring in the vertical plane that rests on the ground,
discarded, replaced by a point support.
loaded by its own weight.
If mean radius R is perhaps 5t or more, the largest
(b) Mathematical model. stresses in the actual problem are circumferential flexural
stresses; then the theoretically infinite stresses at D
associated with a point support are not important (and
could only be calculated as high stresses by conventional
finite elements).

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 21 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 22

Mathematical Model Mathematical Model


• A two-dimensional model is adequate. • The foregoing conceptual models become complete
mathematical models when we decide on the appropriate
• Instead, if the physical structure is not a ring but a long,
analysis theory:
thin-walled pipe, should the model be three-
dimensional? Probably not. a) For a slender ring, it can be beam theory.
Stress analysts recognize that deflections and stresses are b) For a not-so-slender ring, plane elasticity theory is
essentially constant along the length, that they vary only appropriate.
near ends, and that the variation has only a small effect
c) For a thin-walled pipe in which end effects are to
on the largest magnitudes of deflection and stress.
be represented, thin-shell theory is appropriate.
However, in a long pipe the situation is more nearly
Elements based on the respective theories would be used
plane strain than plane stress; thus the model is changed,
for FEA of the respective mathematical models.
and appropriate data must be supplied to software and
appropriate analysis options chosen.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 23 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 24
Mathematical Model Preliminary Analysis
• Why not use a three-dimensional model? • Before going from a mathematical model to PEA, at
least one preliminary solution should be obtained, using
After all, reality is always three dimensional, and
whatever means are conveniently available:
elements for three-dimensional FEA are available. The
reason is cost. a) simple analytical calculations,
Demands on the analyst’s time and computer resources b) handbook formulas,
are likely to increase by a factor of 10 or more in going
c) trusted previous solutions, or
from two dimensions to three.
d) experiment.
Some of this effort may lead to a better mathematical
model.
Subsequently it will be used to check computed results.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 25 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 26

Preliminary Analysis Finite Element Analysis


• If we do this work before FEA rather than after, we • Use of general-purpose FEA software involves the
reduce a natural tendency to find answers that support following steps.
whatever FEA results have already been obtained,
 Preprocessing: Input data describes
especially if it took considerable effort to get them.
a) geometry,
It is easy to make mistakes in supplying data to software,
and even a crude preliminary solution may detect a result b) material properties,
that errs greatly due to a mistake in data input.
c) loads, and
Preliminary analysis for the ring problem of Figure 1 is
easy if R is considerably greater than t. Formulas for d) boundary conditions.
deflection and stress in a slender ring are available in
hand-books. If R is comparable to t, these formulas are
approximate but still useful for checking.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 27 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 28
Finite Element Analysis Preprocessing
 Preprocessing:
Software can automatically prepare much of the FE
mesh, but must be given direction as to the type of
element and the mesh density desired.
That is, the analyst must choose one or more element
formulations that suit the mathematical model, and state
how large or how small elements should be in selected
portions of the FE model.
All data should be reviewed for correctness before
proceeding.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 29 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 30

Finite Element Analysis Finite Element Analysis


 Numerical analysis: Software automatically:  Postprocessing: The FEA solution and quantities
derived from it are listed or graphically displayed.
a) generates matrices that describe the behavior of
each element, This step is also automatic, except that the analyst must
tell the software what lists or displays to prepare.
b) combines these matrices into a large matrix
equation that represents the FE structure, and In stress analysis, typical displays include the deformed
shape, with deformations exaggerated and probably
c) solves this equation to determine values of field
animated, and stresses of various types on various
quantities at nodes.
planes.
Substantial additional calculations are performed if
behavior is nonlinear or time-dependent.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 31 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 32


Postprocessing Postprocessing

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 33 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 34

Postprocessing Postprocessing

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 35 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 36


Check the Results Check the Results
• First, we examine results qualitatively and ask if they • If answers to such questions are satisfactory, FEA results
“look right”–that is: are compared with solutions from preliminary analysis,
and with any other useful information that may be
a) are there obvious errors?
available.
b) have we solved the problem we intended to solve, or
• One way to judge the accuracy of a discretization is to
some other problem?
look at plots of stress; software can plot either stress
• Boundary conditions are often misrepresented: contours or “stress bands,” which a zones of color.
Different colors are used for different levels of stress:
c) does the deformed FE structure show displacements stress bands, as will be shown subsequently, are
where there should not be any? discontinuous across interelement boundaries; strong
d) are expected symmetries present in the results? discontinuities indicate too coarse a discretization,
whereas practically continuous bands suggest
unnecessary fine discretization
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 37 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 38

Expect to revise Expect to revise


• Rarely is the first FE analysis satisfactory. • After another analysis cycle, the discretization may by
judge inadequate, perhaps being too coarse in some
Obvious blunders must be corrected.
places.
“Uncomfortably” large discrepancies between what is
Then mesh revision is required, followed by another
expected and what is computed demands explanation.
analysis.
Either physical understanding or the FE model, or both,
In analyzing a new problem, it is almost always
may be at fault.
appropriate to begin with a simple FE model, to which
Disagreements must be satisfactorily resolved by repair detail is added as the analyst learns more.
of the mathematical model or/and the FE model.
Each revision is an expected step on the way to an
adequate solution, not a penalty for failure in the
preceding attempt.
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 39 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 40
Outline of a FEA project
START
FE Software
Consider the physical  Obtain approximate  Plan a finite element 
actions involved.  results for subsequent  discretization of the  Preprocess: 
Devise (or improve) a  comparison with FEA  mathematical model.  Build the, finite 
mathematical model. results.  element model in 
the computer. 
Finite Element Analysis (FEA)
Physics

What is at fault? 
Are error 
estimates 
small? 
No Revise the 
finite element  Generate and 
Part 1.2: Two‐point boundary‐
Physical understanding  discretization. 
or FEA modeling? 
Does mesh 
revision do 
little to alter  Yes
solve equations of 
the finite element 
model. 
value problems
FEA results?  STOP
No
Yes
Postprocess: 
Are FEA results free of obvious errors, such as disagreement with boundary conditions  Display computed 
intended?  results for 
Are FEA results physically reasonable? examination. 
Do FEA results agree well enough with predictions and approximations obtained by 
other means? 

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 41

Two-point BVP 2nd Order DE Model Boundary Value Problem


Primary Secondary
• Consider the following two-point boundary-value Field variable
Problem data
variable

problem: of study
Heat transfer
u
Temperature
a
Thermal
c
Surface
f
Heat
Q
Heat
conductance convection generation
d d
  a  x  u  x    c  x  u  x   f  x  , 0  x  l
T – T∞ kA APβ f Q
Flow through Fluid head Permeability 0 Infiltration Point source
dx  dx  porous medium ϕ μ f Q
Flow through Pressure Pipe resistance 0 Point source

u  0   u0 , pipes
Flow of
P
Velocity
1/R
Viscosity 0
0
Pressure gradient
Q
Shear stress

u  l   ul .
viscous fluids vx μ –dP/dx σxz
Elastic cables Displacement Tension 0 Transverse force Point force
u T f P

here u0 and ul are given constants, and f(x) (on the right- Elastic bars Displacement
u
Axial stiffness
EA
0 Axial force
f
Point force
P

hand side), a(x) and c(x) are given “smooth” functions. Torsion of bars Angle of twist
θ
Shear stiffness
GJ
0 0 Torque
T
Electrostatics Electric potential Dielectric 0 Charge density Electric flux
constant
ϕ ε ρ E

d  d 
  a  x  u  x    c  x  u  x   f  x   0 for 0  x  l
dx  dx 
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 43 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 44
Model Boundary Value Problem Two-point BVP 4th order DE
k  thermal conductance;
d2  d2 
  convective film conductance; 2 
b  x  2  
u x   f  x, 0  x  l
dx  dx 
p  perimeter;
u  0   u0 ,
P  pressure or force;
d
T  ambient temperature of the surrounding fluid medium; u  x   0 ,
128 h
dx x 0
R with  being the viscosity, h the length, and d the diameter of the pipe;
d4 d2
E  Young's modulus;
b x u  x  Ml ,
dx 2 x l
A  area of the cross section;
d  d2 
J  polar moment of inertia.  b  x  2  
u x  Fl .
dx  dx  x l
d  d 
  a  x  u  x    c  x  u  x   f  x   0 for 0  x  l
dx  dx 
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 45 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 46

A Model Problem
• We begin by considering the problem of finding a
function u  x  , 0  x  1, which satisfies the following
differential equation (DE) and boundary conditions
(BCs):
d2
u  x   2, 0  x  1,
Part 1.3: A Model Problem dx 2
(1)
u  0   0,
u 1  0.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 48
A Model Problem A Model Problem
• The data of the problem consist of all the information • The data in our model problem are “smooth”; for
given in advance: example, the right-hand side = 2 and the coefficients are
– the domain of the solution (in this case, the domain is differentiable infinitely many times.
simply the unit interval 0  x  1), • As a consequence of this smoothness, there exists a
– the “nonhomogeneous part” of the differential unique function u  x  which satisfies the differential
equation (represented by the given constant (= 2) on equation at every point in the domain as well as the
the right-hand side), boundary conditions.

– the coefficient of the second derivative of u  x  (in • In this particular example, it is a rather simple task to
this case this is the constant +1), determine the exact solution to (1):
x x
– and the boundary values we demand the solution d2 d d
attain (in this case, zero at x  0 and at x  1). 0 d 2 u   d  0 2d  dx u  x   dx u  x  x0  2 x
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 49 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 50

A Model Problem A Model Problem


x
d2 Moreover,
0 d 2 u   d 
x x x
d d
0 d u   d  0 dx u  x  x0 d  0 2 d
x
d d 
 d  d u   d 
0
 x x
d d  2
u    u  x   u  0   u  x   x  2
d  0  dx x 0  2  0
x
d d
u  x   u  x    2d d 
dx dx x 0 0
u  x   u  0   u  x   x  x2
 x
 dx x 0 
 2  0
 2x
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 51 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 52
A Model Problem A Model Problem
And taking into account the boundary conditions,
u  x   x2  x
u  0   0 and u 1  0
we obtain,
u  x
u  0   0,
0 12 1
d  d x
u 1  0   u  x  1  12  0  u  x   1
 dx x 0  dx x 0

1 4 x2  x

 u  x   x2  x (2)

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 53 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 54

A Model Problem A Model Problem


• However, in most technical applications, one or both of • As an example of the first kind of difficulty, suppose that
these happy features of the problem are missing—either instead of f  x   2 being given as part of the data (the
there is no solution to the classical statement of the right-hand side of (1), we have the problem:
problem because:
d2 1
– some of the data are not smooth, or u x    x   , 0  x  1,
2  
dx  2
– if a smooth solution exists, it cannot be found in (3)
u  0   0,
closed form due to the complexity of the domain,
coefficients, and boundary conditions. u 1  0.

Where   x  1 2  is the Dirac delta: the unit “impulse”


or “point source” concentrated at x  1 2.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 55 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 56


A Model Problem A Model Problem
• The fact is that   x  1 2  is not even a function but is
rather a symbolic way of describing operations on
smooth functions defined by
 Something appears to be amiss!
   x  1 2   x  dx   1 2 
0
(4)
How can a function u satisfy (3) everywhere in the
for any smooth function   x  satisfying the boundary
 x  1 when its second derivative cannot
interval 0____
conditions.
exist at x  1 2 because of the very irregular data given
• We can convince ourselves that if any function u is to in the problem?
satisfy (3), then it must have a discontinuity in its first
derivative du dx at x  1 2; its second derivative d 2u dx 2
does not exist (in the traditional sense) at x  1 2.

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 57 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 58

A Model Problem A Model Problem


• The difficulty is that our requirement that a solution u to • Whenever a smooth “classical” solution to a problem
(3) satisfy the differential equation at every point x, 0 < x exists, it is also the solution of the weak problem.
< 1, is too strong.
• Thus, we lose nothing by reformulating a problem in a
• To overcome this difficulty, we shall reformulate the weaker way and we gain the significant advantage of
BVP problem in a way that will admit weaker conditions being able to consider problems with quite irregular
on the solution and its derivatives. solutions.
• Such reformulations are called weak or variational • More important, weak or variational boundary-value
formulations of the problem and are designed to problems are precisely the formulations we use to
accommodate irregular data and irregular solutions, such construct finite element approximations of the solutions.
as those in problem (3), as well as very smooth solutions,
such as that of our model problem (1).

01/25/2018 UNICAUCA 2018 by Sergio Preidikman 59 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 60


A Model Problem A Model Problem
• Examples of problems for which exact solutions cannot x x
d2
be found explicitly (even though they are known to exist) 0 d 2 u   d  0   x  1 2  d (5)
are found commonly in BVP in two or three dimensions.
• It is in the solution of such problems that the true power d d 0 0  x  1 2
u  x  u  x  
of the finite element method has made itself felt. dx dx x 0 1 1 2 x (6)
d2  U  x  1 2
u  x     x  1 2  , 0  x  1,
dx 2 (3) U  x 1 2
u  0   0, u 1  0.
1

1 1
Maple: u  x   U  x 1 2 x  U  x 1 2  x
2 2 12 x
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 61 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 62

A Model Problem A Model Problem


• 0  x 1 2 : • 1 2  x 1 :
x x x x x x
d d d d
0 d y   d  0 dx y  x  x0 d  0 U   1 2  d 0 d u   d  0 dx u  x  x0 d  0 U   1 2  d
x x x x x 12 x
d d d d
0 d y   d  0 dx y  x  x0 d  0 0d (7) 0 d u   d  0 dx u  x  x0 d   0d   1d
0 12
(9)

d  d 
y ( x)  y (0)   y  x   x  0 u ( x)  u (0)   u  x   x  x  1 2
 dx x 0   dx x 0 

d  d 
y (0)  0  y ( x )   y  x   x, 0  x  1 2 (8) u (0)  0  u ( x)   u  x   x  x  1 2, 1 2  x  1
 dx x 0 
 dx x 0 
(10)
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 63 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 64
A Model Problem A Model Problem
• 1 2  x  1 cont.: • Hence, finally we get

@ x  1 u ( x)  0  1
u  x    x, 0  x 1 2
2
d  (12)
u (1)   u  x  1  1  1 2 1
u  x  x  ,
1
1 2  x 1
 dx x 0  2 2
d  d
0   u  x   1 2  u  x   1 2 (11)
 dx x 0  dx x 0 0 12 1
x

1 1 1
 x x
1 4 2 2 2
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 65 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 66

A Model Problem A Model Problem


• or, • cont.
1
u  x    x, 0  x 1 2 1 1 1 1
2 u  x    x  xU  x  1 2   xU  x  1 2   U  x  1 2 
1 1 (12) 2 2 2 2
u  x  x  , 1 2  x 1
2 2
1 1 1 1
  x  x U  x  1 2   U  x  1 2   x
2 2  2 2
1 1 1
u  x     x  1  U  x  1 2    x  U  x  1 2 
 2  2 2 1 1
 xU  x  1 2   U  x  1 2   x (14)
2 2
1 1 1 1
  x  xU  x  1 2   xU  x  1 2   U  x  1 2 
2 2 2 2
(13)
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 67 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 68
Mathematical Models in One Dimension Integration by parts
First Mean Value Theorem for Integration
d du dv
If G :  a, b    is a continuous function and ϕ is an uv   v  u  product rule 
dx dx dx
integrable function that does not change sign on the
interval [a, b], then there exists a number x  (a, b) such b b b
d du dv
that b b  uv  dx   vdx   u dx
dx dx dx
  
a
G t  t dt  G      t  dt
x
a
a a a

b b
du dv
In particular, if   t   1 for all t  [a, b] then there exists  uv a   vdx   u dx  fundamental theorem 
b

x  (a, b) such that a


dx a
dx
b

 G  t  dt  G  x  b  a 
b b
dv du
 u dx  uv a   vdx
b
a
a
dx a
dx
The point G(x) is called the mean value of G(t) on [a, b].
01/25/2018 UNICAUCA 2018 by Sergio Preidikman 69 01/25/2018 UNICAUCA 2018 by Sergio Preidikman 70

You might also like