Basic Matrix Concepts - Devdas Menon

Advanced Structural Analysis
Prof. Devdas Menon

Department of Civil Engineering
Indian Institute of Technology Madras
Module - 3.1
Lecture - 17
Basic Matrix Concepts
(Refer Slide Time: 00:09)
Good morning. We are now starting lecture 17. This is our first session on module 3,
Basic Matrix Methods.
We have completed the review of Basic Structural Analysis. We took 16 hours to finish
16 lectures. The reason why we spent so much time is - you really cannot do anything
advanced unless you are strong in the basics; that is the continuous problem we face with
structural engineers. I think it is a very good idea to always go back to the basics and
strengthen them.
This is module 3, Basic Matrix Concepts. We will first cover simple mathematics,
principles of linear algebra which you have probably already studied. So, it is a quick
review and then will get into how you can apply concepts of matrices to do structural
analysis.
The book we will be following is the book I have written called Advanced Structural
Analysis published by Narosa in India and Alpha Science International; it is a hardbound
edition abroad. The topics covered in this course after the review of basic structural
concepts, will be matrix concepts and methods. Then, we look at how to apply these
concepts to structures with axial elements like trusses; next, how to apply them to beams
and grids; next, plane and space frames. Finally, in the seventh module, we will be
covering a bit of non-linear analysis as applied to elastic instability and second order
analysis.
In this particular session where we cover the mathematics underlying matrix concepts,
we will review what we mean by a matrix, a vector, the elementary matrix operations
including matrix multiplication and transposition. We look at what it means to know the
rank of a matrix; how it is useful in solving linear simultaneous equations, which is what
we need to do in structural analysis and how it can be done using matrix inversion.
Finally, we look at Eigen values and Eigen vectors because they have a role to play -
especially when you do elastic instability analysis, and later, in dynamics.

Before we do all that, we should see the role that matrices play in structural analysis.
Why do we need to study this topic which you had studied in mathematics? How does it
have a role in structural analysis? Could someone give an answer to that question?
Structural analysis is all about finding the response, the force response and displacement
response of a given structure when it is subject to given set of loads. The loads could be
direct loads, direct actions or indirect loads including support movements, environmental
changes and construction errors. So, where do matrices come here?
We can build up matrices for a one load to the other points at the location and we can
build the matrix and then for different position, different forces, what would be the
reaction for these forces?
Now, we are actually stepping back from calculation and trying to look at the big picture
because if you want a digital computer to do structure analysis, you have to see the big
picture and you realize that basically you are dealing with numbers.
If you want to find unknown forces, internal forces, bending moments, shear forces and
axial forces at different locations, you are actually dealing with numbers which can be
put nicely in arrays. You are dealing with vectors and you are dealing with properties of
the structure like the stiffness matrix and the flexibility matrix. That is how you can deal
with it in a matrix configuration. For example, from the input side, you definitely have
the load vector which are loads that are applied at different joint locations. You may
have known support displacements and so you have a support displacement vector; you
may have some initial deformations or temperature changes.
What you need to find is the displacement at all the degrees of freedom that you have
identified and the active coordinates; you need to find the support reactions; you need to
find the internal forces; you may also want to find the member deformations like the
elongations in truss members. In short, all the variables, all the quantities that you need
to find in the force field and displacement field can be expressed in a numerical manner,
but you need to conveniently do it and you need to do it in a systematic way.
If you want the computer to do it, you also need to first model the structure itself, which
is not that easy. That means you and I can see a truss and visualize it, but how does a
computer do it? Well, the computer does not think. You should be able to even define the
geometry of the structure using coordinates - Cartesian coordinates and you need to also
check what the computer interprets is the structure that you have. Please note you are
dealing with a mathematical model of an idealized structure. In modern software
packages you have a too -l visual ability, where you get back the picture of the structure
that you input. If you find some members are out of place or oriented the wrong way,
you can make corrections. You can use visual basic for example, to do such things; that
is a separate field.
Basically, you have some unknown quantities and some known quantities in the
displacement field and the force field. You need to know certain laws, which are hidden
in the structure and which enable transformations to take place.
So, basically, we want to transform the loads to the responses and this becomes easy,
relatively easy, when you are dealing with a linear analysis because it is a linear
transformation. You are actually converting one vector to another vector with the help of
a matrix. (Refer Slide Time: 03:35) For example, one of those sets of relationships are,
as I have shown there, relating the force field to the displacement field with the help of
the stiffness matrix which you can partition.
You have a large number of analysis software packages available today. Unfortunately
today, we are in a stage where structural engineers are no longer doing any manual
analysis and they are learning how to use the design or the analysis manual that comes
along with the package. You input some numbers and you get some output, but if you
really do not know what is going on, it is garbage in and garbage out. In fact, if you read
the preface to this book on Advanced Structural Analysis, I have raised some of these
concerns. In the olden days, when there were no computers, there was no choice and you
had to understand structural behavior. You have to know whether the bending moment
that you calculated makes sense or not. You have to know whether it should be sagging
or hogging, but now you do not know because you are getting some numbers thrown up
by the computer. The need to do analysis faster and quicker is so strong now that there is
no time to pause and reflect and, that old breed of structural engineers who really
understood what was going on is a vanishing species. It is important to always step back
to have some rough checks, which is where you need to have a good foundation of basic
structural analysis.
Here is the first sentence in the preface. There is an increasing tendency among modern
structural engineers to lean heavily on software packages for everything. In fact, even
design is done without thinking that you have computer aided design. At the press of a
button you get big reports which the client receives and thinks you have done a
phenomenal job. Actually, all you did was to press a button. In fact, today we have
reached a stage where quantity surveying, planning everything is programmed. The only
thing the computer has not yet been able to do is actually construct buildings at the press
of a button; that may also happen and hopefully the building will stand.
This induces a false sense of knowledge, security and power. The computer is indeed a
powerful tool and an asset for any structural engineer, but it is dangerous to make the
tool ones master and to make it a convenient substitute for human knowledge
experience and creative thinking. I hope you appreciate this. I hope you enjoyed what we
did in basic structure analysis. You must not lose that touch and always go back to it and
have simple ways of checking your answer. So, that is where matrices come in. Let us
quickly cover the mathematical concepts.
What is a matrix? Now you should think general because matrix concepts can be applied
to a wide variety of not just engineering issues, but even issues in biology, in social
sciences, economics and so on.
What is a matrix?
Set of things.
Set of things. There are no things. Computer does not see things.
Set of numbers.
Right. By definition: a matrix is a rectangular array of elements arranged in horizontal

rows and vertical columns. So, it should look neat and every number has a place - aij , ith
row, jth column. So, it is a rectangular array. I wish you had said rectangular, you did not
say it. It is not a circular array. It is a rectangular array.
The entries of a matrix are called elements. They are scalar quantities. They are not
things; they are commonly numbers. They can also be functions; they can be operators;
they can also be matrices which are called sub matrices. They could be algebraic
functions, but they should make sense. You have mathematics and you have physics. Our
job is to link the physics to the mathematics. Every number has a place and a meaning.
Typically, there are many ways you can write or designate a matrix. In text books, you
will find usually with the bold capital letter A, as shown bold, but when you do it on
paper you cannot make it bold unless you have the bold ball point pens. What we
normally do is we put braces outside it, box brackets commonly, as shown here.
There are alternative ways you can do it, but actually that one letter stands for a huge set
of numbers and variables which can occupy numerical values as shown here. Now, what
is m into n, m rows and n columns? That is called the order of the matrix. You know
very well, it is called the order of the matrix. Let us see.
When m is equal to n, what do you call that type of matrix?
Square matrix, very good.
When all the elements are 0, it is a null matrix. It has a symbol capital O. When the off
diagonal elements are zero, and all the diagonal elements are unity, it is an identity
matrix.
We can go fast because you seem to know everything, but at least you should know the
notations used. For example, a three by three identity matrix looks like that. (Refer Slide
Time: 13:50) What are these elements called? You have two diagonals in that matrix a
you have a diagonal in the other way. That is the principal diagonal. A matrix which has
non-zero elements only in the diagonal is called a diagonal matrix.
Is this a diagonal matrix? No. What are the different adjectives you can give for this
matrix?
Symmetric matrix.
By the way, a stiffness matrix looks like this. Typically, a stiffness matrix looks like this.
That is why I chose this. It is symmetric, very good. What else? It is square. Can a non-
square matrix be symmetric? No.
Upper triangle.
Is it an upper triangle? All the diagonal elements are positive. That is true. What is the
name of the matrix? I am not asking you to describe the contents of the matrix. This can
be any matrix. In structural engineering, it could be a stiffness matrix. It is a banded
matrix.
Why is it a banded matrix? Because you have only the diagonal elements and you have
elements next to the diagonal. So, it is forms a band. That is good. It is a banded matrix
and many of the elements are zero. There is an advantage in terms of storing the matrix if
many are zero, why? You can store only the banded portion. So, storage is easier. That is
called as sparse matrix. So, square matrix, symmetric matrix, banded matrix and sparse
matrix. Sparse means not filled, empty. This class is relatively full, but on days before
you have other exams, you have sparse matrix here. Some of you do not show up. Now,
it is not sparse.
What are these matrices called?
[Noise] (Refer Slide 16:08)
Very good. Lower triangle typically shown with capital L and upper triangle shown with
U.
You are familiar with partitioning of matrices. When do you need to partition matrices?
When you want to make life simpler and when you deal with a complex system, you
want to operate with sub systems. You do not want look at the big picture all the time.
You pick out little pieces from your big picture, but you should know where it is, how it
is positioned and so on. You can draw these partition lines and sub divide. Different sub
matrices will have different orders, but they should all add up to the order of the full
matrix. Quickly, you seem to know everything about matrices.
What about a vector? Is it a type of matrix?
Yes sir, it is a matrix.

Which is bigger? Matrix or vector, which is more general?
Matrix.
So, vector is one type of matrix. What is a vector?
Column matrix.
It is a matrix which has only one column. A vector is a simple array of scalar quantities,
typically arranged in a vertical column. Hence, the vector can be visualized as a matrix
of order m into one. You do not use a word order anymore; you use the word dimension.
'm' is the dimension and you can say elements, but the right word to use is components.
You have components in a vector and elements in a matrix. You have dimension in a
vector because one of the other order items is equal to unity. You have an order in a
matrix.
It is typically arranged like this (Refer Slide Time: 18:10) and again you have a bold
letter to describe the vector. Here the brace we use, just for convenience, is a curly brace,
but if you put a square brace you are not really making a mistake because it is after all a
type of a matrix. Do not put a curly brace on a matrix of size m into n where, n is not
equal to 1, but you are allowed to put a square brace even for a vector. When you have
something like this, some people call this also a vector, but actually the transpose of this
that is the vector. This is sometimes referred to as the row vector, but strictly not a
vector.
Vectors are important you know we are dealing with displacement vectors and force
vectors. So, they have some meaning. It is good to see the background. We can visualize
a multi-dimensional linear vector space Rm whose, dimension m is given by the
minimum number of linearly independent vectors with real components required to span
the space. These are beautiful words. There is a word called spanning, a word called
linearly independent vectors and there is another word called basis.
What is basis? Sorry, one at a time.
Linearly independent vectors, which can span the space.
Why do you need to get a basis?
[Noise] (Refer Slide Time: 19:37)
Do you see it has a role not just in physics, but in other fields of knowledge? Let us take
social sciences. In fact, it is more difficult there because if you really want to get
complete picture of say human behavior, you need to be able to construct a frame work
where you have these linearly independent vectors. It is very difficult because you have
to establish that they are independent and that provides a frame work for you to describe
behavior. So, these are not just mathematical terms; they have wide applications.
So, vectors are said to span a vector space, if the space consists of all possible linear
combinations of those vectors. You should be able to capture any vector in that space
with the help of your linearly independent vectors. Any set of vectors that are linearly
independent and also span the vector space is called a basis of the vector space. What is a
simplest basis that we the use in physics? i, j, k. In a three dimensional space; you are
familiar with this. You have i, j, k and you have unit vectors which form a good basis.
So, a 1 0 0, 0 1 0 and 0 0 1.
Any vector in that space, for example, 2 minus 1 3 you can write it in a column, but it
actually has a sense of direction V equal to 2i minus j plus 3k and it has a length which is
obtained by the resultant of the scalar quantities. So, you are familiar with it. It is also
called the magnitude. So, vectors are meaningful and a set of vectors V1, V2 to Vn having
the same dimension m is said to be linearly independent, if no linear combination of
them other than the zero combination results in a zero vector. That is the proof whether
you really landed up with linearly independent vectors. Otherwise, you have a problem
because there is a correlation between vectors.
Elementary matrix operations: I will go very fast you are familiar with it. You have
scalar multiplication. The scalar is never shown in bold; so, lamda is a scalar. You have
matrix addition and A plus B will be equal to B plus A - that is a property. What is the
property called? Commutative property, if you put a negative sign. So, we only say
addition, but addition is generic. It includes subtraction because you are multiplying with
minus 1, which is a scalar.
Matrix Multiplication is, when you studied it first time, it looked very funny because it
did not look like the multiplication of scalars, but now you are used to it. You have a
distributive property also. A plus B plus C is the same as A plus B which is a matrix plus
C. AB equal to C; you cannot just multiply any two matrices there. You have to satisfy
some requirement, what are those requirements?
Basically, you know what to do. If you want to find Cij, you have to take the ith row of
the A matrix and the jth column of the B matrix and you have to multiply each term with
its corresponding term so, they must match in terms of dimension. That is why m into n
order multiplied by n into p order will give you an m into p order. You are familiar with
this.
Let us look at it more closely. You know how to do it, but let us see if we can make some
physical sense out of it. I will give you a clue. You can do that multiplication in this
manner. What does this manner suggest? Now, this is something you may not have
studied when you did Linear Algebra. What does this signify? I can get the same answer
by doing the multiplication in this way. What does this suggest? It suggests a certain
principle. I took the first column. I took the first element of the first row. I took the
second column. I took the second element of the first row and I get the first column in
the resulting matrix so, what does this principle show? Every column vector of C is a
linear combination of the column vectors of the pre-multiplying matrix A.
Incidentally in terms of direction, if you multiply a vector with a scalar you are getting
the same direction; you are only changing its length. You can do another operation in the
other direction and you will prove that every row vector of C is a linear combination of
the row vectors of the post-multiplying matrix B. That is just a physical meaning. The
moral of the story is - you can blindly do matrix operation but, you can also do them with
your eyes open and figure out everything makes sense. You are playing with vectors.
And you have these properties. What are these properties? You are familiar with all of
them?
Associative.
Distributive.
Associative and distributive. Then, if you multiply with a null vector. Anything
multiplied with zero is zero. You get identity only replicates. The commutative property
is not there; you are familiar with it. AB is not equal to BA. You may not even be able to
do the multiplication sometimes.
Transpose - you are familiar with it. Transposition is an operation in which the
rectangular array of the matrix is rearranged or transposed such that the order of the
matrix changes from m into n to n into m with the rows change into columns preserving
the order. If the original matrix A is aij and the transpose will be aij transpose which is aji
and the orders changes from m into n to n into m. You are familiar with this and you
have these properties. If you take the transpose of the transpose, you get back the original
matrix. If you pre- multiply it with a scalar, you can take it out side and this is
interesting.
The third one is also simple. A plus B the whole transpose is A transpose plus B
transpose, but AB the whole transpose is B transpose A transpose. Here, you have to do a
switch; then only it works. Supposing B and A are the same, then what is the kind of
matrix that you get? If you pre-multiply a matrix with its own transpose, what is the type
of matrix you will get? You will get a square matrix plus it will be symmetric. How do
you know it is symmetric? You take the transpose of this matrix and you will get back
the original matrix. So, that is very interesting. You can demonstrate with examples. We
will not spend time. Sometimes, you see matrices where the transpose gives you a
negative sign. So, this is called a skew symmetric matrix.
Now, next, we will talk about the products of two vectors. You cannot multiply two
vectors; you know that. Why? Because you have to take the transpose of one of them.
You commonly meet this situation, when you want to find work. Take in structural
analysis, you have the force vector and you have the displacement vector. You cannot
multiply F into D; either you take the transpose of F and multiply it with D or you take
the transpose of D and multiply it with F. Then, you will end up with one by one
quantity. What is that? That is a scalar. So, what kind of product is this? Well, we say dot
product in physics when you deal with two and three dimensional spaces, but it is called
an inner product. In general, it is called an inner product. You can also show it in this
fashion. This is symbolic way of showing the inner product.
I will read this: The product, F transpose D equal to lamda results in a matrix of order
one in to one, which is nothing but a scalar. Such a product which is sometimes denoted
as (F,D) with those triangular braces is called inner product or dot product in vector
algebra involving two and three-dimensional vectors. This product essentially reflects the
projected length of one vector along the direction of the other vector; we are familiar
with this in physics. The magnitude of any vector V can also be viewed as an inner
product of V with itself; you are familiar with that. Vectors are said to be orthonormal, if
you do their inner product you either get 0 or 1. What is this symbol called? This delta is
called a Kronecker delta.
The commutative property of the inner product is what I explained a while back. You
can calculate work either by converting the force vector to a row vector or the
displacement vector to a row vector, you get the same answer. If you introduce the
relation between force and displacement through the stiffness matrix, you can arrive at
an interesting conclusion. Can you tell me what that conclusion is?
See, the first is a commutative property, the second is a stiffness relationship, and it is a
law. If you plug in the second property to the first property, what can you end up
proving?
K is symmetric.
Yes, that is really good. So, you take the inner product in this manner and you get lamda.
Now you substitute. Take the transpose of this; you still get lamda. Then, when you
compare these two expressions, D transpose k D is the same in both but, in one you have
k transpose. So, you end up proving k is equal to k transpose. It is a very elegant
mathematical proof, requirement of why the stiffness matrix must be symmetric.
We will study symmetry from the structural analysis point of view later. Now, let us get
to the real need for bringing in matrix. We need to solve equations and we want the
computer to do it; we do not want to do it ourselves because it is a pain sometimes.
Linear simultaneous equations can be written as AX equal to C. A is called the

coefficient matrix, as we saw in this case of the stiffness matrix, it is a property of the
structure. X is a vector like the displacement vector and C could be a load vector. C is
some set of constants.
If C is 0, well you should not say 0, you should say O because 0 is a scalar, O is a vector
or a matrix, then those set of equations are called homogeneous equations. You are
familiar with them. When do you need to look at the solution for homogeneous
equations? In solving linear simultaneous equations, even if you have a non-zero C, a
constant vector you may still need to inspect this solution. That is called the null space of
A, all possible solutions. When you need to do this? It is very interesting, though we do
not directly need it in structural analysis.
You need to do it because it is related to what is called the rank of the matrix.
What is the rank of the matrix A? It is a beautiful concept.
Number of independent we can find out.
You kind of vaguely know what it means. Let us explore it - rank of a matrix. The rank
of the matrix A is equal to the number of linearly independent column vectors of the
matrix and this number is identical to the number of linearly independent row vectors.
So, the maximum value of the rank r of any matrix of order m into n is given by either m
or n, whichever is lower and the minimum value is 1.
Let us look more closely at this: the subspace in the vector space Rm containing all linear
combinations of the independent column or row vectors is called the column space or
row space of A and this subspace has a dimension equal to the rank r. You have to finally
get the number of linearly independent vectors. Take a look at this matrix. It is a four by
three matrix. Obviously, the rank cannot exceed 3.
Is the rank 1, 2 or 3?
Three.
Two.
Two. Why do you say 2? That is good, that is quick, smart looking, but in general it is
good to keep quiet and do an analysis, before you pop out with the answer. Three, you
should never say because you cannot spot it so easily. Sometimes it is cleverly disguised,
you cannot catch it.
How do you capture the rank of a matrix? There is a beautiful technique. You have
studied this? Matlab will do this easily. It is called the row reduced echelon form.
Relatively easy way of determining the rank of a matrix is by reducing the matrix to its
form R through a process of elimination. You are familiar with this.
Yes or no? Yes, sir.
Now you need not assert yes, you should have the care and concern for other members of
your group. This is a matrix. Do not be a dominant diagonal. You should care for those
who are saying they do not have an exposure. You cannot insist they have an exposure.
All you can insist is, I know, that too we will check shortly whether how well you know
it. It is always good to say I do not know. I want to learn or I want to relearn.
So, the idea is, though it is a refresher for many of you, incidentally, we are just covering
the theory. I would say it is more for fun because nobody is going to do at this stage in
your life. You are going to use press buttons and let the computer do it, but there is a nice
story it is telling and I think it is worth understanding this. How do we do that? Any
matrix R you can finally, through doing elimination. Many of you, those of you who said
no, are you familiar with Gauss elimination? You basically play around with rows. You
multiply and add different rows. You are allowed to do that, but you have to do the all
the way including the column vector. You can play with them and that is allowed. You
do not lose anything in doing so. You can multiply row with 2 and subtract 3 from the
next row and so on. Those are row operations that you can do, but the idea is you do
them so that you can get an identity matrix in the left hand upper most corner.
You get an I matrix there and the idea is to get 0 below. You can always reduce R to this
form where you have null matrices at the bottom, identity matrix at the top left corner
and you I will get some matrix F. What is F called? It has a name, those of you who said
you know everything, please answer. F is called the free variable coefficient matrix. Why
is it called that? Let us see, it is very interesting.
Take a look at this matrix. Let us do it is step by step. What is the first thing we should
do?
Make it one.
Make what, one?
Top left.
Ok, that variable, so that is called the pivot.
You divide that row with 2 and then do an operation and get zeros in the other one. This
can be done. We will just go through it fast. You have done it before. In gauss
elimination you would have done this. That is our pivot, you will get that. You have got
lot of zeroes already. The top one also should be 0. This is ultimate. You got the identity
matrix.
Actually the last step is optional when you do simultaneous equation because you can do
back substitution and get it. You got this. Then you look at the partition. You captured
that I, F, O, O, beautiful. What is a rank of this matrix?
Two.
It is the same matrix we looked at earlier. Rank is clearly 2. It is the size of the identity
matrix. So, it is R into r. That is clear. Now, this is how you can do it. In fact you can
write a program and it can do on its own. Rank is 2. Let us look at a set of equations like
this.
Can this set of equations have a solution? Yes or no.
Infinite number of solutions.
Infinite solutions. Will they always have infinite solutions?
Yes.
It may not have because first the constant vector must satisfy conditions of consistency.
It is very interesting, but we do not need too much of it. I want you to read on your own.
I will just rush through it.
Let us look at this. Finally, you have to solve equations like this. Do you agree to this? If
you take out those two rows, they should look like this. That is D0, should be O and this
must be satisfied by C for a feasible solution set. If it is not satisfied, you cannot solve
this equation. Still, you might get infinite solution. That is a different matter, but you
cannot touch this, if you have a violation of consistency.
And whenever you have a problem like this, when you have free variables, you have to
look at the infinite solutions which come from the null space. So, you get one particular
solution. You must have done this in differential equation. Get one particular solution,
then you have to add the null space solution from the homogenous equation. Let us take
a look at this. You have X1, X2, X3 and you have this situation. If you take the row
reduced echelon form, it looks like that. Well, the last two rows, the constants should be
zero. You run the operations right through they will look like that. So, you cannot do this
for any arbitrary C. Do you agree? Yes. First that has to be satisfied and then only you
talk of a solution. Then also you may not get a unique solution because the rank is only
2, it is not 3. So, first you have to satisfy this. That means you can express c3 and c4 in
terms of c1 and c2.
Let us look at a problem like that. Let us say c1 is A, c2 is B, and then c3 and c4 must
necessarily satisfy this requirement. Then you can talk of a solution, but even then you
will not get a unique solution because the rank is not full. Let us take an example like
this and you do the row reduced echelon form, you will get something like that. How do
you solve this problem? Well, partition it; you can throw away the third and fourth rows
because they already satisfy consistency. They are not going to help you. You have only
two rows, but you have a problem there. You take the first set of equations. You can
write X1 and X2 in terms of X3. Clearly, you do not have a unique solution. So, you take
a particular solution; you can choose any value of X3 including minus 342.625, who
cares, any particular solution.
For convenience, we took X3 equal to 3 then you got a unique solution for X1 and X2, but
this is not a complete solution. You have to take the null space solution, the easiest null
space solution; let us take some scalar like lamda. Let X3 be lamda, plug them into those
equations and you will find that if X3 is plus lamda X1 and X2 must turn out to be minus
lamda, minus lamda. So, what is your complete solution? This is how give the complete
solution. It is for you to learn for a general knowledge. We do not fortunately have such
problems except, when your structure is unstable. In which case, you better worry about
this structure falling down rather than solving these equations. You have to make the
structure stable.
Now the other case we have is when R is equal to n; n is the number of columns which is
less than the number of rows. So, coefficient matrix A has the full column rank, but there
are linearly independent rows, which also implies that the number of rows exceeds the
number of columns. When do you have such a situation? When you have more equations
than unknowns, then you need to satisfy consistency. Again this comes in unstable
structure; so, we will just rush through this. You do not need it really. Now it takes this
form. There are no free variables here and the constant still must satisfy, because it
belongs to the same type. You run this exercise.

Let us just rush through this. Once you satisfied it, you get a unique solution. You do not
need to have the null space here. This is a second type. We are not dealing with it in
What do we deal with in structure analysis? Case three and case four.
What is case three? R equal to m less than n. When do you get such a situation? When do
you get more unknowns and less equations?
Statically indeterminate structure.
So, coefficient matrix A has a full row rank, but there are linearly dependent columns
which also implies that the number of columns exceeds the number of rows. We have
more unknowns than equations. It is a situation we encounter in statically indeterminate
structures. Now, you have only I and F, those Os vanish. You have a full row rank.
Your row reduced echelon form looks like that, I and F. You see, you have infinite
solutions, but you do not have to worry about consistency. Those equations are not there.
Since there are free variables present, you have infinite possibilities.
Let us taken an example like this: AX equal to C. You run this operation till you end up
with the row reduced echelon form. They are consistent. You do not have to check on
consistency but, now to make it look like in identity matrix, sometimes you have to
interchange rows. You know that operation. You have to interchange rows. In this case
you have to do that. They look like this and if your constants have numbers like 3, 6 and
2, after your elimination, they look like 3, 7 and minus 3.5.
You solve this; find the particular solution; find the null space solution and you get it. It
is similar to the fist type of problem except consistency is not there. We do not even do
this. We do it. When did we do it? We wanted to find displacements in statically
indeterminate structures after having analyzed it, remember? There we chose lamda to be
0. We did not waste time doing all this because you need any statically admissible
solutions to apply the principle of virtual work.

This case is straight forward. Full rank, if you are lucky, but if your structure is stable,
you are likely to get this situation. If it is statically determinate, this is straight forward.
In this case, the coefficient matrix A has a full column rank as well as full row rank. This
also implies that this matrix is the square matrix. Such a matrix is said to be invertible or
non-singular. If you have a matrix like this and all our stiffness matrices for stable
structure will look like this, then you do not have a problem; you also have a symmetric
matrix in our case.
You have a unique solution. You can use the Gauss elimination method or you can use a
row reduced echelon form and you can get the answer. In earlier days, we used to make
students actually painfully do all this, but today we do not do that. We say, you use the
computer; certainly be familiar with software like Matlab where you just press a button;
enter the matrix. All of you are familiar with Matlab? There are some free ware called
Skylab and so on, but I think we have many versions of Matlab and Mathcad. So, it is
very easy today to find the inverse of a matrix.
What is a problem with some matrices regarding finding inverse?
They are non-singular.
They should not be singular.
That is true. So, in structural analysis, where do you see a problem? We will see that.

Matrix inversion: When the matrix A is square and of full rank an alternative approach to
solving the equations is by the operation of inversion. X is equal to A inverse C. A
multiplied by A inverse is A inverse A is identity matrix. You are familiar with these. A
inverse inverse gives you back A and so on. The last property is very interesting. It is
like what you got in transpose. A B the whole inverse is B inverse A inverse. Some
mathematician gave a beautiful physical feel to this. He gave an analogy. Do you
remember? I think I mentioned this earlier. He said how do you put on shoes? You first
put on socks that is A matrix, then you put on the shoe. How do you reverse the process?
You have to first remove the shoe, and then you remove your socks. That is why A B
inverse is B inverse A inverse; very easy to remember, very nice analogy.
What is a determinant of a matrix? Why do you need to know it?
Whether it is invertible or not.
Whether it is invertible or not, yes. So in general it can be proved that the inverse exists
if a scalar property called the determinant of the square matrix denoted A with those
vertical lines or det A or delta A is not equal to zero. For diagonal matrix, the
determinant is given by the product of all the diagonal elements. You know that. That is
an easy way to find out.

I think you are familiar with finding determinants. We will not spend too much time on
this.
You know about co-factors. You really you do not need to do in this course in matrix
methods, but it is good to refresh these concepts. You know about the adjoint method of
finding the inverse, yes or no? You know about Cramers rule?
Yes.
Those who said no and if you are interested please read good book in matrix algebra or
just read the second chapter in this book on advance structure analysis. It is good to
browse through these and we can spend hours discussing these different methods of
finding the inverse like the Gauss-Jordan Elimination method, Cholesky Decomposition
method. You know these are very good for stiffness matrices, but we will not do that.

Cramers rule you are familiar with. It is very easy to do when your order is up to 3.
Beyond 3, manually, it is not worth it and we use elimination methods to do so. We will
stop here.
Maybe we will finish this.
Condition of a matrix: A matrix if it is ill-conditioned, you are not sure about the
accuracy of the inverse that you get. How do you know whether inverse is right or not?
You again multiply with original matrix, you should get the identity matrix. You will not
get, not necessarily. You will have lot of rounding off errors while you do elimination.
Unless your matrix is healthy, it is in good condition and there are tests to check the
condition of a matrix, you cannot be sure of finding its inverse. Therefore, you are not
show about the answers you get.
Now, fortunately for us, our stiffness matrices are all healthy matrices. They are well
conditioned. You know about the property of positive definiteness. You should read
these. I will just finish this. A square matrix of order n is said to be positive definite, if
for any arbitrary choice of n dimensional vector X, the product X transpose AX yields a
scalar quantity that is positive. If it is negative, it is ill-conditioned; if it is zero it is called
semi definite. You can use Eigen values to decide. Basically if your diagonal elements
are heavy and they are positive, you got a well conditioned matrix. That is the thumb rule
and stiffness matrix will be like that. Flexibility matrix, there is no guarantee. That is
another reason why flexibility matrix is not preferred for. It is too flexible. You can run
into problems of finding the inverse.
Thank you. We will stop here.
Keywords: Basic Matrix Analysis, Matrix Algebra, Eigen values, Eigen vectors, Gauss
elimination
Prof. Devdas Menon
Indian Institute of Technology, Madras
Module - 3.2
Lecture - 18
Good morning to you. This is lecture 18, the second lecture dealing with the third
module on Basic Matrix Concepts in this video course on Advanced Structural Analysis.
So, we are doing the third module, which we started yesterday. There was one topic
remaining in review of matrix algebra that was dealing with Eigen values and Eigen
vectors, we will quickly cover it. Then, we will start looking at how matrix concepts can
be applied in structural analysis.
So, this is covered in the second and third chapters of this book on Advanced Structural
Analysis.
So, Eigen values and Eigen vectors are something you have all studied while learning
mathematics. What are the applications of these from your own experience, so far in
mechanics, for example, have you ever applied it? You learn some technique in Maths.
Have you ever applied it? Never?
Solving simultaneous.
Solving simultaneous that is again in mathematics, but you do not have any applications
in engineering, in mechanics? Well, think about stress analysis. Locating, finding
principle stresses, locating principle planes, you have an application there. Think about
buckling; you have studied buckling in mechanics of materials; you deal with an Eigen
function there; the buckled mode shape is an Eigen function. So, it really has a wide
variety of applications. And one of the interesting applications is in solving simultaneous
equations; you can end up diagonalizing the coefficient matrix A. How does that help
making the coefficient matrix diagonal?
Other elements multiplied by
Sir, directly we can get it.
You get directly; the solutions become uncoupled. They no longer need to be solved
simultaneously; you can solve one at a time. So, that is the ultimate way of solving these
equations. If somehow, you can reduce the coefficient matrix to its diagonal form, which
you can do with the help of Eigen values. So, I will quickly go through this. If you recall,
we said that any given matrix A can be visualized as a device that is capable of
transforming a vector X, whose components are unknown to another known vector C,
which is your constant vector through a linear transformation and that is the meaning of
this equation AX equal to C; it is a transformation. You are transforming one vector X to
another vector C.
Now, where do Eigen values come here, an Eigen vectors come here? From whatever
you have learnt so far. What does an Eigen value transformation do? What does it do in a
physical sense? What will be the end result after you do the transformation? You will not
get the identical vector. You get a?
Scalar multiple.
You will get a vector, which is a scalar product of the original vector X. What is it
physically signify?
It is a vector, which is parallel.
That is right. You get a parallel vector. So, if A is a square matrix, then there exists
certain special types of vectors X called eigenvectors such that the transformation results
in vectors that are parallel to X; very simple. This means that the product AX is a scalar
multiple of X, and that multiple is often designated as a lambda is called as scalar
multiplier. And it is a unique property of the matrix A, and the word Eigen comes from
German, which really means one, unique.
For a non trivial solution, you see one solution is when X is null vector. The physical
example that you can appreciate is the case of buckling. How do you get the buckling
load, the critical buckling load? Well, actually if you remember you must have done the
Euler buckling analysis, you write down an equilibrium equation, assuming it to a
buckled, and you find if try to find P critical, which would facilitate that buckled shape.
So, you will end up writing an equation, where two solutions are possible; one is axis,
the displacement vector is a null vector that means, there is no buckling, the second is
there is buckling. When you have a null vector, it is called a trivial solution. In the case
of the column, it means that the column remains straight remains unbuckled. And at a
certain critical load, at what is known as a bifurcation point, two alternative equilibrium
positions are possible; and it will prefer to buckle; so it cannot take any further load after
that. So, that is how important it is.
Now, to find a non trivial solution, it is important that the matrix A must be non-singular.
In problems of structural stability and dynamics the square matrix A is usually
symmetric, which makes our life easier; and this implies that all the Eigen values will be
real, not complex; there are situations where you have imaginary numbers coming in, but
not in our real life problems in structures. And luckily for us, the Eigen vectors are going
to be distinct, and they will be linearly independent, and also mutually orthogonal. That
is a property of the eigenvector.
So, in stability problems, the Eigen values correspond to the buckling loads; and the
corresponding Eigen vectors and they referred to the Eigen functions if you are I dealing
with the continuum. They represent the mode shapes of buckling, right? So, you have the
first mode called the fundamental mode; then you have this next second mode, third
mode and so on. In practice everything buckles on the first mode. We will study this in
greater detail in the seventh module, but for now we cover the maths part of it.
And in dynamics, what do you think the Eigen values represent? In vibrations of strings,
you may have studied in school; you have the first mode, the second mode of the
vibration and so on. Is it not? So there, the mode it is called a natural mode, the mode
represents the shape, and it is similar to the shape you get in buckling; say the flexural
vibrations in a beam. We have the first mode, the second mode, third mode so on. It is a
shape that you get; and what is the Eigen value represent corresponding to any particular
natural mode? They are called natural frequencies, and you probably heard of the
phenomenon of resonance in the excitation coincides with the natural frequency; then
you have large amplification in the response. So, that is how important it is to study
Eigen values.
There is some properties at I think you know. The sum of all the n Eigen values of a
square non-singular matrix A of order n into n that means, you have n Eigen values is
equal to the trace, which is the sum of the diagonal elements of A, and the product
directly gives you the determinant. So, the properties of the matrix A are linked to the
Eigen values and the Eigen vectors.
So, you can take a simple example. You can prove this that these two properties are real.
You have probably come across this it is called characteristic equation, which is what
you get, when you want a non trivial solution. The determinant of this matrix will have
to be 0, which means the elements get shifted by lambda. You remember you must have
done this exercise, and you need to solve, you shifted the lambda X for the left side, so
now you have a homogeneous solutions.
Clearly for a nontrivial solution, A minus lambda I has to be singular, which means the
determinant must be equal to 0. So, this is how you have studied; and you end up with a
polynomial equation of order n in terms of lambda, the roots are Eigen values and the
corresponding X vectors, which will not have unique components you know since a rank
of this matrix is less than full, you will have linear dependence in the components of the
Eigen vector.
So, there are many ways of doing this and you have probably studied some of them. One
convenient way is to reduce this A minus lambda I to its row reduced form. And then
you can easily find out using the technique that we discussed yesterday. You can identify
the Eigen vector. Take an example, and you have done lots of these problems in think in
your earlier semesters. This is what you do. You get a cubic equation in this case, which
you can solve and identify the three Eigen vectors; you put them in ascending order. So,
the first Eigen value is a lowest, this gives the 3 6 and 9.

If you take one of them lambda 1 equal to 3, if you locate its row reduced form, it is
easy, then to find the Eigen vector. This is not the only way you can do it; you have
probably learnt it in some other way. But finally, you will find that it is c by 2, c and c;
or c can take any value. So, for example, in a pinned pinned column subjected to an axial
load; the Eigen shape is a sinusoidal shape. The sinusoidal shape only relates every
points displacement to every other points displacement, but it does not fix the
amplitude that is what is reflected here; c can take any value, c is like the amplitude of
that wave.
All that it guaranties is that all points on that wave have a definite relative amplitude. So,
that is the beauty of the mode shape; the mode shape does not dictate the absolute value
of the deflection. Similarly you can do it for all of them; and you can normalize these
Eigen vectors, you have studied them. Commonly what we do is we take one of them,
the top one you make it to unity and scale the others. But there are other ways of scaling
and once you normalized it you call it an orthonormal Eigen vector.
So, a vector like this, you can put together, and you can get a matrix; and that is called
the orthogonal Eigen vector matrix. And it has certain beautiful properties; one of the
properties of this matrix, which you see here, is that the transpose will be the same as the
inverse. So, finding the inverse is very easy, just take the transpose of it. For example,
once you have identified Q for this matrix A, if you take the transpose you will get it,
and you can also put the three Eigen values together into a diagonal matrix, and that is
called an Eigen value matrix, it will look like that.
And then if you do a little trick; Q that symbol is called capital lambda. Q capital lambda
Q transpose will give you back the original matrix A that is a fantastic property. And
what is the advantage of this property? This property helps you diagonalize the matrix A,
because your A is your original coefficient matrix. So, if you do that, you get back A,
this matrix A can be factorized or decompose into that form, and that procedure is called
Eigen decomposition. These are the techniques used for solving simultaneous equations
essentially.
And if we do not choose the eigenvector, we choose any other I mean we do not use the
orthonormal, we take any eigenvector. You can choose any eigenvector, then you will
end up with the symmetric matrix, and then also factorization is possible. Commonly,
when we apply to real problems of finding buckling load and so on, we have slight shift
in the formulation. Instead of saying A X equal to lambda X, we sometime encounter the
form A X equal to lambda B X.
We will see this how this happens later, in which case you get A minus lambda B into X
equal to 0, and you need to solve this equation. And typically X corresponds to the
displacement vector A corresponds to the primary stiffness matrix. So, there is a strong
relationship between the stiffness matrix, and the mode shapes, and the Eigen values and
lambda B corresponds to the geometric stiffness matrix which is something we will study
in module 7. Just want to show you that mathematics is a very powerful tool in structural
analysis, and it is good know at least conceptually what is going on.
You need not do any Eigen value long hand calculations, the computer will do it for you.
You can use those tools, but at least broadly you should never forget the background
what is going on under the black box. So, there are many iterative methods in our times
we have to really manually solve all these, and learn all the techniques and find out
which is the most efficient.
So, there is power method Jacobis method, Stodola-Vianello method, and so on. And
commonly we had to do this whenever we studied buckling instability or dynamics,
where you have multiple modes to be consider.
So, with this we conclude the maths part it, which is just a refresher. Now, we get into
how to apply matrix concepts to analyze the structures.
Now, remember we have 7 modules 2 modules was Basic Structural Analysis which
you do manually. This third module is basically introductory, it is the fourth, fifth, and
sixth where we directly apply matrix methods to problems. In module 4 we apply it to
trusses and axial systems. In module 5 we applied it to beams and grids. Grids are three-
dimensional.
In module 6 we apply it to plane frames and space frames, in module 7 we do instability
and second order analysis. You will find that, these concepts really start making sense
only when you start applying them. So, what we are going to do from now is only a kind
of introductory, we kind of planting seeds which will take a little time gestation period
for them to germinate. So, once you start applying, it becomes very easy. So, if you dont
fully understand at this stage, do not worry it will all fall into place as we go on. So, the
topics we are going to cover is: introduction coordinate system, and these look simple,
but you will realize that the first difficulty you encounter is how to convert all that you
have been doing on paper drawing a truss or a frame how to pass it on to the computer.
So, you need to have a coordinate system. You need to have a framework in which you
can put that structure, and identify everything that is going on in the structure. So, that is
how important coordinate systems is then you need transformation matrices. What do
you think these mean? Essentially, you will find you need two coordinate systems, when
you run a country like India, you have a central government the federal government, and
you have state governments and you have to have some control between the center, and
the states.
Likewise in a structure - the structure is one like the country the nation state India. So,
the structure must have a federal coordinate system, we called them global coordinate
systems. So, the entire structure is located in a global framework with one x, y, z, but
then when you get to the element level; the element may be inclined and located anyway.
It is you better to go in for a local coordinate system, the state government; and
transformation is all about switching from the local to the global or the global back to the
local. So, that is very beautiful. It is a linear transformation from one system of Cartesian
coordinates to another system.
Then we need to look at stiffness matrix. Again here, you can define the stiffness matrix
at the element level. And you can define the system matrix at the structure level. What
do you think is a difference?
That is for one element and this for the entire structure.
Like what is going on in your hostel. You have these groups the Andhra group,
Karnataka group and the Kerala group, and then you know the Maharashtrian group and
so. They are all element level groups, but we need national integration. So, we need to
connect the groups together, and make us live harmoniously, if you do not do it
harmoniously you will have separation. You have lack of compatibility which is the
cause of much strife in the world today, right?
So, similarly in a structure they can have their independence, but unless they operate
together harmoniously in the interest of the overall structure, you will have problems. So,
similarly you have to understand stiffness matrices, flexibility matrices both at the
element level, and how it all fits in at the structure level. And you have already studied
this whenever you encounter intermediate loads, because you want to limit degrees of
freedom, you have to pass them effectively to the joints. So, you need equivalent joints
loads, and you need to find fixed-end forces.
Now, we will also look at three different types of methods, you have stiffness methods
which again branch out into two which we are going to call conventional stiffness
method, and reduced element stiffness formulation, and flexibility methods; we will see
this shortly.
But first let us review, what we have learnt already I have shown this picture earlier, if
you take a truss, identify it is degrees of freedom, you will find that they are D1, D2, etc.
And at the corresponding locations in the same direction, you can identify forces we call
them F1, F2, F3 so on. These are. So, called external forces and joint displacements, they
effect the structure as a whole that is the central government. But at the element level,
you pull out one of those truss members that element has its own internal force which we
will locate at the ends of the element.
We have two forces at the two ends, and somehow these forces are link to F1 and F2. So,
actually you have a transformation there itself; how do you convert joint forces on the
structure to element level forces in different elements? You see you can see the potential,
and the basis for that must be equilibrium. Similarly, every member in this truss can
elongate, the two joints can move. Compatibility demands, that whatever moments you
get at the element end levels must match the moments you get in the structure.
So 2, 3 elements joining at one point will have the same movements in the global
coordinate system, but in the local you have to resolve them into their relative
components; that is compatibility. And structural analysis is all about getting a good
grasp of these knowing what is happening in your force field, knowing what is
happening in your displacement field, and defining the laws that relate these fields. Once
you have got a good big picture understanding, you can program everything, the
computer will handle everything for you. We are lucky we are dealing with linear
analysis so, the transformation are all linear, there is absolutely no problem.
So, let us begin with the basics of coordinate system, let us say I have a structure like
this. I have made it a little complicated in the sense you have inclined columns, which no
other more complicated types of problems you can get. And I have a structure here, it is a
plane frame. I first need the computer to kind of get hold of the structure, and be able to
draw it back to me and so, that I can confirm that this is exactly what I want.
So, I need to have a coordinate system first to identify the joints in the structure, right?
So, I first choose an origin in this case, I could have chosen usually you choose the
extreme lower most left end, but you can choose anywhere you in that plane. So, I have
put it somewhere there, and that is my origin and I have x and y; conventionally x is
pointing to the right, and y is pointing upwards, and that automatically defines z or as the
American says zee.
You have to follow the rules of vector algebra, which means i cross j must be equal to k,
and if you follow the right hand screw rule, you will find that the z will be coming
outside towards me from that plane of x, y. These arrow marks hold not only for forces
and translations. The double arrow in the same direction holds good for moments, and
rotations. So, you will discover the z positive direction will demand that I should
consider clockwise or anticlockwise rotations to be positive? Anticlockwise. So, this is
the first change we are going to make from what we have studied till now. It is a big
change, it is an abrupt change, because till now in displacement method we assumed,
clockwise rotational moments should be positive, because that is the way traditionally it
was done. You know, slope deflection method evolve like that.
But today we have a big picture, proper grasp of it, we want to apply matrix methods, it
should be consistent with our idea what is positive. So, we make this change, and we
should do it little effortlessly. And it is like getting an international driving license, you
know that in some countries including India, you have to stay to the left side of the lane,
but when you cross over to another country, there you have to be on the right side and
you are ok with it. So, then only you get the license to drive, then only you get the
license to apply matrix methods after having studied Basic Structural Analysis.
So having done that, you should first write down the coordinates for A, B, C, and D
which is easy to do. If someone has given you the geometry of the structure, you can do
that x, y, z of A, B, C. Well, some people do not like to use any letters at all. So, they
will call those joints 1, 2, 3, 4. It is ok, but you have too many 1s and 2s so, I thought in
the early stage it is better to do this. And then you have 3 elements so, you should
identify the 3 elements, the 3 states.
This is a country with 3 states: element 1, element 2, and element 3; it is clear; and I put
a circle around it so, that I do not mix up with some other numbers which I am going to
show up soon. What is a next thing I need to do by the way there is something more to it.
To define the element you also have to define the orientation of the element. What is a
start of the element? What is the start node, and what is the end node?
So, you have a choice there, but if you move consistently from left to right, and from
below to above then, A to B is called the incidence of my defining element A - element
1, B to C is element 2, and C to D is element 3. I need to do all these. I need to first
locate A, B, C and D. I need to identify elements, and their start node and finish node,
and then I have got that. What more I need to do. So, those are called nodes; nodes are
locations where you need to locate either a displacement or a force. So, you can have
nodes inside an element for example, if you have a concentrated load and you do not
want to convert an equivalent joint load, where you want to find the deflection in the
middle of a member, you can a create a node there, nothing prevents you. But the
moment you create node you have got two elements joining that node.
But we should not make life more complicated than is necessary. So, we normally limit
the nodes to the joints, but nothing prevents you from bringing in additional nodes. So,
you define elements and nodes, and next you need to define degrees of freedom, right?
We have primarily doing a stiffness formulation, later we look at the flexibility
formulation. So, the degrees of freedom here are in this plane frame you know at B, and
C, you have 3 degrees of freedom. And let us align those along our chosen Cartesian
coordinate system. So, along the x direction, we say the translation is one; along the y
direction positive upwards is 2, and anticlockwise rotation at the joint B is 3, does it
make sense?
So, we know at every joint, you potentially have 3 degrees of freedom - 2 translations, 1
rotation D1, D2, D3 would mean that. But then you move on to the next joint from B to
C, because you have only active degrees of freedom at B and C, at C what are your
coordinate numbers if you finish 1, 2, and 3; you need to do 4, 5, and 6; right? And that
is all you have or do you have more?
Do you have more degrees of freedom? Yes, at D also you have some more. That is a
roller so, it can move. You have an active degree of freedom there. So, it can move to the
right, and it can also rotate, right? Now, you can choose 7 and 8 there, is it clear? What
else do you have? If you want to take care of support reactions you need to have or you
have support moments. So, to complete the picture you should also look at what is called
as restrained degrees of freedom.
So, you do that. So, you can bring in 9, 10, 11, 12 got it. These are the restrained degrees
of freedom 9, 10, 11, 12, right? I have changed the color here, but when do it by hand
probably using the same black or blue pen. So, there is something nice, you can do to
identify restrained coordinates.
You can just introduce a slash, introduce a slash in the arrow marks; the slash tells you
that that degree of freedom is not active it is restrained. When you use software
packages, you have to feed it is a binary code. You have to show whether, it is restrained
or it is free to move. So, normally the numbers are 0 and 1.
So, here we are doing this. So, with this we have captured all the displacements at the
different joints D1 to D12, some of them are active, some are restrained; is it clear? And
by the way restrained does not prevent you from including support settlements, but then
you say that it is still a known quantity; usually it is 0, but it can be non-zero. So, you can
define a displacement vector DA and DR which I can partition like that. I am putting all
the restraints at the end of the list, is it clear?
The D vector - the displacement vector have displacements at active degrees of freedom,
active coordinate locations, and R stands for restrained coordinate locations. Now,
luckily A and R also other words to describe it from a force point of view. So, you can
also have FA which is conjugate with DA, and FR which is conjugate with DR.
It is exactly where you have active degrees of freedom that you can apply loads, isnt it?
So, A stands for actions which means loads. So, you have FA corresponding to DA, and it
is exactly where you restrained displacement you get reactions, and luckily for us R also
stands for reactions
So, FA, FR constitute our first vector in the global picture, and DA, DR constitute our
displacement vector in the global coordinate system. Does this makes sense? Now, the
word coordinate has two meanings here, we are referring to 1, 2, 3, 4 all the way to 12
also its coordinates, and the Cartesian coordinate system that we use are also coordinates.
So, you have to be careful in interpreting the word coordinate. Does it make sense to
you?
Sir, at the joints we have to provide the constraints like both can rotate freely about. Both
has to rotate at same theta.
No, if you do not say anything we assume we are dealing with the rigid joint. For
example, member 1, and member 2 are rigidly connected at B if you do not make any
statement. If you dealing with the ends - boundaries, there you to have to define the
supports, is it free or is it restrained? But you can have more complications which we
will introduce later not in the beginning, supposing I have an internal hinge, then it is
little difficult we will study that later.
We should begin simple then we will get it into complication, but it is a point. You can
bring in releases, but at the as beginners let us not introduce those complications. Let us
take simple problems - straightforward problems do you understand this? Is it clear?
Let us demonstrate how it is done. Let us take the same frame, and the physical meaning
of DA and DR is shown in this vector, you will find that the first element in this vector D1
actually points to delta B x. That means, at this joint B, if this moves to the right by some
quantity that component of the deflection to the right is called delta B x. D2 will be the
amount it moves upward right at B, D3 would be the anticlockwise rotation here; if you
end up with a clockwise rotation you should put a negative sign. Likewise you can cover
the all the way from D1 to D8, then you are left with D9, D10, D11, D12. Those are
restrained degrees of freedom, and unless you include support movements they are
tacitly taken to be 0, is it clear? No doubt about that.
Now, in this problem you have 8 active degrees of freedom, and 4 restrained degrees of
freedom totally 12 degrees of freedom, fine? Very clear no problem. Let us look at a
problem where you have some loads applied, how will you write down the force vector?
So, I have in this problem 3 loads, so what would you call 50 kilo Newton. What is plus
50 kilo Newton? Yes, you should talk in the language of the coordinates. F1 is 50 kilo
Newton, what is 100 kilo Newton?
Minus F2
Yeah. So, F 2 is minus 100, and what corresponds to 75 kilo Newton meter?
Minus F6.
Yeah F6 is?
Minus 75.
And what about the others?
0.
They are all 0 on the action side. FA is exactly what we said, you correctly said plus 50
minus 100, and minus 75 the rest are all 0 because though those joints have the potential
to take loads for this particular load case those values are 0. And the reactions are not 0,
but they belong to the unknown category; that is part of the response we need to find
them about, right? Does this make sense? Very simple. We are going in baby steps, this
is very straightforward.
Now, we also recognize that there is something called as stiffness matrix for the entire
structure, which means if you tell me that I have, you know on account of that loading I
have the displacements vector D, then the force vector that cause that displacement
vector must be F, and they are related by this equation F equal to k D.
If the displacements are unknown then I can easily get the displacements by inverting the
force. But it is not so simple, because you will find that in the vector F, you have FA and
FR. FA is known usually, and FR is not known. Similarly, in the displacement vector D,
you have DA and DR; DA is not known DR is known.
So, you cannot apply this directly, you have to play with it and we will see how to play
with it. Now we will switch from central government to state government. So, how do
we do that, what do you suggest?
Write the equilibrium equations.
No, we are talking about coordinate system.
(( ))
See, capital of India is New Delhi. Do you want the capital of Tamilnadu also to be New
Delhi?
(( )).
Where do you want it to be?
You want it to be here at Chennai, right? You want to be the boss. This is your origin.
This is your capital. So, this becomes your reference point, New Delhi talks to Chennai
and the whole of Tamilnadu is governed by Chennai. So, your coordinates of your origin
should be? So, you have the choice to choose. Usually it is put in the middle of the state,
but we do not usually put the origin in the middle of the member, right? Where would
you like to put the origin?
(( ))
Which one do you prefer?
A.
So, that is what you do. You start the origin at A, then you do not align your x, and y in
the Why should you follow what Delhi says? You choose your own local system,
right? But if I write same x and y there will be confusion. So, when I am dealing with
local coordinate system I put a star, I put an asterisk. So, whenever I put a star what does
it tell me, I am dealing with local coordinate not global coordinate, is it clear?
So, I have X star Y star for element 1; and I have to finally, link it to the global, because
the X and Y global I have shown below here. This X and that X have to be linked; how
do I link it? By the relative rotation. If they are parallel theta is 0 right? And I have to
choose a direction, so I consistently choose my anticlockwise direction. So, once I define
theta x 1, everything is defined; I know, how Y is oriented with Y star, Is it clear? Now,
having done that, I can do it for other states, right? So, I go to member 2, I do the same
thing, but now I say theta x 2 and theta x 3. So, the elements are L1, L2, L3 as their links.
Assuming they are prismatic, if they are not prismatic; let us see you have a stepped to
element, then it is worth creating a node wherever the step is taking place and treat that
as a separate state.
Is this clear to you? Now, what is the next step, next thing we need to do? We need to
again bring degrees of freedom. How many degrees of freedom do you have at any joint,
any node?
3. So, can we choose it this way? 1 star, 2 star, 3 star, 4 star, 5 star, 6 star. You agree that
these are independent so, I have 6 degrees of freedom for a plane frame element. What
do I label for the element 2? Should I do 7, 8 and 9? No. I will stick to 1, 2, 3. I am only
saying state number 2.
Does it make sense to you? Clear? Central government, state government, global, local;
good; now what? Now for each element, for the i th element let us say; this is simple
structure with 3 elements. I can have a structure with 525 elements. So, i is a generic
index that I will use. Now, this is a definition I use. If I say D without any star, it goes to
the central government automatically. So, I have to put the star, I can put it above or
below, I am putting it below, because I need to have other indices. And I am putting my i
here; i can be 1, 2, 3 to distinguish between the three states. So, does this make sense to
you?
So, I have a displacement vector 6 by 1; and I have a corresponding conjugate force

vector. Does this make sense? So now, I have creating a framework with which to play,
start playing my game, but without creating a framework which all of us share and
understand and agree to we will have problems. So, is this clear? Is the notation clear?
By the way, different books give different notations; should not get confused, the
concepts are the same. So, you have to define the local axis, you have to define the start
node, you have to define the end node, and everything falls into place. The theta is
crucial here.
Now, you can also combine all the elements together, and put them all in one group; for
example, I call a combined element displacement vector D star without the i, which
means it includes D1 star, D2 star, D3 star each of these is 6 by 1. So, what is the size of
this one element?
18 by one
18 by 1, if I want to. If I have 525, I would not waste time doing this, because it is too
complicated. The three elements maybe it is nice to put them all together, put all the guys
in one line. Similarly I can do for F star. Is it clear? So, this we will call the combined
element displacement and force vectors. It is optional, you may never need to do this, but
sometimes for small structures, it is good to put them in one line.
Now, the element has its own governance. So, at the element level, you have a
relationship between the element force vector and the element displacement vector, and
you can talk of an element stiffness matrix in the local coordinate frame work. Does it
make sense? So, you have an element stiffness matrix; you need for each element in that
structure, and you have a common structure stiffness matrix. And obviously, the element
stiffness matrices contribute to the structure stiffness matrix. Does it make sense? How
do you put it all together, is what we are going to study.
Take a look at this. I am showing you three examples from a very good book by
Kassimali, where he shows how we can do the numbering in an optimal way; numbering
of a global coordinate system. Here, I have shown you a continuous beam, a 7 span
continuous beam - 1, 2, 3, 4, 5, 6, 7. How many degrees of freedom do we have? What is
the degree of kinematic indeterminacy?
Six
It is a beam. We have six rotational degrees of freedom. I did not write A, B, C, D, but
they are there. How will you number them? Well, it is straight forward here; you number
them sequentially like this. Remember, anticlockwise positive; earlier we did clockwise
positive; now we are doing anticlockwise; 1, 2, 3, this is fairly obvious. Supposing I do
in a funny way, I put 1 here, then I choose to put 2 here, then I put 3 here. What do you
think will happen to the stiffness matrix? It is a good question.
It will just shift the columns.
Let me ask you. You are temporarily not in the class. What do you think will happen to
the stiffness matrix? Just to wake you up. It is an 8 o'clock class and half of you are half
asleep. Tell me.
Any one.
It will just shift the rows or a column.
What will happen to the type of matrix you get?
Same matrix, but same numbers, but elements will
No, some beauty in the matrix gets corrupted. What gets corrupted?
Symmetry.
The stiffness matrix has to be to symmetric we proved it; it will not be banded. It will be
all over the place, it does not looks nice; you cannot put it nicely. So, you want to make
it banded, it may still be banded, but the bands stretch to the extremes, you do not want
that. So, it is important to keep your band small; and there is a measure for it is called
bandwidth or half bandwidth. And these are your restrained degrees of freedom. So, let
us just look at the active degrees of freedom; you forget about the restrained degrees of
freedom.
Incidentally, you can partition your K matrix into active and restrained components using
these notations. Does it make sense to you? If I have F as FA and FR, and D as DA and
DR, I can write K as KAA, KAR, KRA, KRR, which means FA is KAA into DA plus KAR into
DR. Now, if DR is 0, then the equation is simply FA equal to KAA into DA which is the
most common way of dealing with the stiffness matrix. Is it clear?
So, let us look at KAA, and you can find DA from that. If you look at KAA, if you do this
numbering system, you get a beautiful band like that. And the half bandwidth as it is
called, turns out to be one that means, you have the principal diagonal, you have to take
only one more cell to the right and to the left of that principle diagonal; that is fantastic.
It cuts down your computational work, and your storage to a high degree of efficiency. I
am not saying, you need to do all this in your class assignments but when you actually
doing structures, when you are doing programming, it is important to figure out, how
best to do this?
So, let us take a more difficult structure like a truss or a frame. Now, you will find your
21 degrees of free active degrees of freedom in the truss, and 24 here; we will figure out
it. Now, let us take the frame; every joint has three; let us include axial deformation; how
many free joints you have? 8; 8 into 3 - 24. Clear, similarly for that truss. So, is there an
optimal numbering scheme? What do you think would be the best way to do the
numbering, so that you minimize your half bandwidth because that helps a lot?
Nearby places.
Nearby places. So, the number of arithmetic operation involved in solving a set of
equations increases the linearly with the square of the half bandwidth. And to minimize
the bandwidth of the stiffness matrix, the numbering sequence should be done along the
shorted direction of the structure. This is what we did in the previous example.
So, this really means if these are your active degrees of freedom number them one you
got 1 here 2, 3 then do not go here do not go along the longer direction go along the
shorter direction 4 5 then you can either go here 6, 7, 8, 9 and so on. Does it make sense?
If you do it this way you will get A Banded matrix which looks like this. It is a
symmetric matrix. The half bandwidth is roughly one third the degree of indeterminacy.
That is very good it is a neat arrangement.
And if you take the other example, you should number them this way along the shorter
direction 1, 2, 3, 4, 5, 6, 7, 8, 9, all the way here you got 24, and when you look at the
matrix it looks like that. You might still get 0 cells in between, but you cannot get it less
than this. So, half bandwidth is one-third and k turns out to be 8. Any questions?
After 1, 2, 3, we are not going up, but we are going (( ))
You should go along the shorter direction always.
Column is shorter than beam.
Not that way. We look at a bay, overall size of the structure. It is not height, the lengths
of the members do not matter at all; lengths do not matter, because they are just
properties. So, this is an important lesson to learn. We will stop here. Thank you.
Keywords: Basic Matrix Analysis, Matrix Algebra, Eigen values, Eigen vectors,
Stiffness Matrix
Prof. Devdas Menon
Module - 3.3
Lecture - 19

Good afternoon. This is lecture 19, module 3, Basic Matrix Concepts. We are doing the
third lecture in this third module. If you recall, we had just started the second topic of
Introduction to Matrix Structural Analysis; this is covered in chapters 2 and 3 in the book
on Advanced Structural Analysis.
We had covered the introduction, basic coordinate systems. And in this session, we will
look at the transformation matrices and stiffness matrix.
This is to refresh your memory. Please see what the real task for us is - we want to be
able to make the digital computer do structural analysis. We want, first of all, to have a
framework in which we show the structure geometrically, which means, we have to
identify the elements, the nodes with reference to some Cartesian coordinates.
We have to identify the active degrees of freedom, restrained degrees of freedom and
that is how we defined the vectors D and F - D is the displacement vector, F is the force
vector; the subscript A stands for actions if you are referring to forces and it also stands
for active degree of freedom if you are referring to displacements.
So, FA the sub matrix FA really is a load vector and FR is reaction vector - R stands for
reactions. R also stands for the restrained degrees of freedom. DR is the restrained
displacements; the displacements are usually zero, unless you have some known
specified support settlement.
Next, you will recall, we defined the - local axes. In this particular plain frame, for
example, you have three elements and we have the local coordinates - the local axes.
You have 6 degrees of freedom per element, so you have the element displacement
vector, the element force vector and we use some superscripts and subscripts to clarify.
You remember the star is what we use when we want to refer to the local system and the
superscript we put i referring to the i th element, so here i can be 1, 2 or 3. And you have
six different displacements and six different forces. We can assemble all the elements
together into one overall combined element displacement vector D star. In this case, what
will be the size of D star? 18 by 1, because D1 star itself is 6 by 1, D2 star is 6 by 1 and
D3 star is 6 by 1. So, this is convenient to do when you are dealing with small frames;
when it is a large frame we do not do this. Similarly, you have F star.
Now, let us look at transformation matrices. You need to know how to switch between
the structure and the element. So you have some relationships. This is a linear
transformation, so you can transform from D i to D i star.
This is a standard transformation. T i is called the transformation matrix - the element

transformation matrix - and the same T i holds good not only for displacements but also
for forces.
This is the conventional stiffness method. The interesting thing about the transformation
matrix is that, it is an orthogonal matrix which has this beautiful property of the inverse
being equal to the transpose; so, it is very easy to do the reverse switch, that is, you can
move from the local coordinates to the global coordinates. We will see how this can be
applied; so, this is one major type of transformation we do. You would have studied in
school coordinate geometry, so you know about what happens when you shift the axes,
you rotate the axes and so on, so it is something similar.
There is another set of transformation we will see later, where you can do what is called
a displacement transformation matrix. This is used especially in the reduced element
stiffness matrix - we will see this later. This can also be applied to the conventional
stiffness matrix.
When you do the flexibility method, you have the force transformation matrix; you can
do the switch from the global to the local coordinates. So, we will see these as we go
along. You will find that with increasing familiarity, you will understand everything, but
it is good to plant the seeds early. We will keep coming back to these slides and your
understanding will get strengthened.
So, you can work both from global to local and from local to global. Just to give you a
big picture view of the different methods we are going to study, the so called
conventional stiffness method is the way softwares work in computers. This is how it is
done by the computer; this is the master method very generic dealing with slightly big
matrices. I am not sure if this is what is taught in many Indian universities in matrix
methods, because what is taught is the simplified version of it, which is more convenient
to do manually.
So, we are going to refer to that as a reduced element stiffness method. It is equally
accurate, but here you have less flexibility; it has some limitations. And here, you deal
with smaller size stiffness matrices; the number of degrees of freedom are less. We will
discuss this in the next class. The stiffness matrix you get here has an inverse and that
turns out to be the flexibility matrix.
So, you have the third method that we need to study; it is called the flexibility method. In
software packages, you do not have the reduced element stiffness method and you do not
have the flexibility method.
So, in many advanced courses; they do not waste time with these methods. They just
cover the conventional stiffness method. In fact, many modern text books do only that.
Traditional text books cover everything, but sometimes in many universities, I think in
your gate exam, for example, they will cover the reduced elements stiffness methods
without calling it so. They will still call it a stiffness matrix method, but it is actually the
reduced version of the method and you learn flexibility method.
What we are trying to do in this course is to cover everything. But for manual use, to
demonstrate in the class and so on, it is probably easier to use the reduced elements
stiffness method but we will do both, because we like to know how the black box in the
computer works; so you should see the big picture.
So, we begin with the right royal conventional stiffness method. So, let us take a fresh
look at this conventional stiffness method. You can express the degrees of freedom both
along the local axes what we saw earlier, but you can also do it along the global axes.
So, for example - 1, 2, 3 instead of pointing towards the local longitudinal axis 1, we can
point towards the global x axis - can you see that? So, this is a big change - that means
we do the transformation here itself, so I am not putting the star anymore. Remember - 1
star, 2 star, 3 star - there the 1 was pointing along x star - this way - 1 was along that and
2 was the normal to it. Now, 1 is parallel to the global one, and 2 is parallel to the global
y axis and you will find the 3 does not change.
Why does not the 3 change? Because the vector pointing outward from the x, y plane is
the same, so that does not change and so there is no transformation needed there. So, it is
like you do the correction at this stage itself - at the element level. If you can define 1, 2,
3 like this, you can also define 4, 5, 6 and similarly, for all the 3 elements. Got it? So,
this is another way of doing it and this is the transformation that we are trying to do.
And so I am now defining element displacement vectors and element force vectors - they
are still element, because the i comes here; i refers to the i th element. This is for i equal
to 1, 2 and 3. The big difference between this and the previous element displacement and
force vectors is in designation how do I make the difference? That star is missing; the
asterisk is missing - have you understood?
Now, all we need to know is how to switch from the local axes system to the global axes
system and then the job is done. How do we do that? Let us take a look.
Let me explain once more. Please listen carefully and pay attention. This is what we did
in the last class. Here, you have 1 star, 2 star, 3 star and so on - 1 star, 2 star here for
sure. If this joint has moved here, the displacement along the local x axis is D1 star with
the i - i stands for the i th element. And this moment normal to it is D2 star. And this
change in angle anti clockwise is D3 star. Does this make sense? And Similarly, I can
define D4 star, D5 star, D6 star. Is this clear to all of us? Mind you, this theta i that I am
talking about here is not referring to a displacement; it is referring to the original
inclination of the member prior to loading with reference to the global x axes. Is that
clear? So, this is what we mean by the local axes system.
What do we mean by the global axes system? This is a similar representation of the
element end forces. Do you understand this now? Now, if I have a force acting this way,
I call it Fi2 star, along the longitudinal x star axis I call it Fi1 star, a moment acting this
way I call a Fi3 star. These three are conjugate with these three - can you see that? The
directions are the same - one are a set of forces, the other are a set of displacements. The
same here at the other end - this is the end node, this is the start node. Does it make sense
to you? So, this kind of description of element level forces is what we call the local axes
description.
We have another description - the global axes. Let us see how that looks. Now, I am
taking the same element and I am taking the same displaced configuration; but I am
choosing to define the displacement D i D i1 as parallel to the x axis - you see, D i D i2 as
parallel to the y axes and D3 i does not change - it is same as D3 i star; these two do not
change, you can see there is no change. But, these two change. Likewise here, this moves
horizontally - D4 i - D5 - and this D6 does not. Do you get the hang of it?
Similarly, here the component of the force along the x axis - global x axis - if F1 along
the y axis, global y is F2 and the moment is F3, F4, F5, F6 - is this make clear?
With this, graphically, I am trying to show you what we mean by defining the element
vector and displacement vector - force vector. You have two choices - one with the star
which tells you we have to align yourself along the longitudinal x star axis, the other one
is you align yourself along with the global axis. Is it clear?
Now, there is a relationship between these displacements. They are defined this way - the
boss here is central government that means the global axes is your foundation and from
that you switch to the local axes. You do it through a - transformation - linear
transformation using a standard matrix called the T i matrix - it is a transformation
matrix. Can you work out a relationship of this matrix? What is the size of this matrix?
2 by 2
3 by 3 sir
2 by 2
Wait, you have to understand the question first. What is the size of this - for the i th
element?
3 by 1
6 by 1
There are six degrees of freedom. You are right - three comes from one end, but the
element has two ends. You have a head and a tail and what we are saying is this you get
by transforming - this is a transformation matrix. You are transforming something else -
what are you transforming?
This is also 6 by 1 - this matrix is the transformation matrix which I call T i, obviously,
this has to be 6 by 6. But there is some nice property about this matrix. What is the nice
property? Apart from that, this matrix will look like this. Even this you can partition, this
you can partition, this belongs to the start node, and this belongs to the end node.
Because you are following stiffness formulation, will there be any relationship between
this and this? They are independent. Actually, this is independent with this and this. And
this whole thing is independent with these, so you will get some zeros somewhere here.
So, which will be zero? So, the off diagonal matrices will be null matrices and whatever
relationship you get here - let us say this is A, this will also be A - does it make sense?
Let me explain. You got 3 by 3 here. Now, this is related to this through this
transformation; this will be related to this through a similar transformation, isnt it? Does
not make sense?
Yes, but once you have a transformation, they are not independent. These two are
dependent, these two are dependent, what is their dependence? This transformation. You
are just rotating the axes, obviously, there is dependence there. Can you tell me what A
will look like? It is a functional direction cosine and it is not difficult to work out, so it
will look like this.
In Di star, first three elements will be identical to the next three elements. I mean, they
are pointing in the same direction.
I dont get you. What you said?
Sir, x star is in same direction along the entire element.

Along the longitudinal axis of the element, that is the axes x star, the displacement at any
point along the element will vary. The displacement at the extreme left end is our
unknown; we are calling that D1 star. At the end node, we are calling it as D4 star. D1 star
and D4 star are independent that is why we call them degrees of freedom. Any pending
doubts? You will get the hang of it soon, this is quite straight forward. Have you got it?
So, this is your A matrix and it is easy to establish.
Now, take a look at these relationships. This is exactly the relationship we wrote there.
This is the fundamental transformation matrix used in the conventional stiffness method.
It is very easy to remember. You have a cosine and you have a sine; everything is
positive, except, the second element for obvious reasons. You have cosine theta and
minus sine theta. Why is it minus sine theta? Because it goes to the negative side. I hope
you can work this out, this is straight forward. We will do it in great detailed when we
get into the actual problems. Do we get this? This is our A matrix; this is for the start
node and this is for the end node. Everything is same except that 1, 2, 3 gets replaced by
4, 5, 6. Does it make sense? So, if you want to put it all together it will look like this.
Clear?
Standard transformation - the property about this is - it is an orthogonal matrix. Now,

why did we align 1 star, 2 star and 3 star along the global x, y and z axis? Why did we
make it 1, 2, 3? What do we achieve by that?
We can span the entire displacements
Think in terms of displacements.
Any displacement or the entire vector space will (( )).
That you can span even with the local coordinates.
Let me give you a clue. First, we began with the structure. In that particular frame, how
many displacements we had? We had 12 - DA and DR. We had D 1, D 2, D 3 all the way
to D 8 - they were active degrees of freedom; D 9, D 10, D 11, D 12 were restrained
degrees of freedom. How are those linked with these? So, you have think in the stiffness
language. In the displacement methods what is fundamental? Displacements are
fundamental. And how do you link the state government with the central government?
How do you link the element displacement with the structure displacement?
This is very important; it is very simple. Remember, this was the first element, this was
the second element and this was the third element (Refer Slide Time 22:39). The
structure was kind of fixed here and roller here. Let us look at the active degrees of
freedom. If you remember - why do not I join them? This was element 1, 2 and 3 - what
were the degrees of freedom?
We said this joint, we start the active degree numbering here, we said this is 1, this is 2,
this is 3 remember? This is for the structure. Now, what is D 1 equal to? From the
element point of view, how will I write? See, you now know all the definitions. It is
exact equal to D first element, this element and fourth - Needs little thinking. Plus?
This element moves here, this joint also moves there - where is the plus coming? You
have to get out of statics where you do plus, to kinematics where you do not do plus. A
conjugal relationship, say between husband and wife, works as long as there is no
separation; so, do not separate out. This is compatibility - do you get it? In defining these
displacements, itself you are defining compatibility. What is this equal to? What is the
second element?
D21
Now, you got the hang of it; so that is why we are playing this game. See, there is a big
difference between trying to figure out what somebody else worked out and trying to
figuring it out yourself from first principles. Let us say you have to find a way to make
the computer do it; it is wonderful.
Go ahead. Let us continue. What is D 2? Somebody from this side. What is D 2 equal to?
D 1 5 equal to D 2 2. What is D 3 equal to? D 1 6 equal to D 2 3. You are getting the
hang of it now; so on and so forth, we will not waste our times. So, this is 4, 5, 6 etc. Got
it? That is the idea; the idea is you do the correction early stage, and then you are
satisfying compatibility straight away.
These are the fundamental transformations. The element transformation matrix T i is a

square, orthogonal matrix, which enables member end displacements and forces,
expressed in a global axes framework, to be transformed to the local axes system.
Transformation from the local axes system to the global axes is enabled by the transpose
of this matrix, which is equal to the inverse of the matrix; you can switch back and forth.
Let us give an example. This is a pin-jointed truss - pin-jointed frame. How many active
degrees of freedom, do you think there are? It is a plane truss (Refer Slide Time 26:44).
Three?
Two sir.
Where can you apply loads? Only at O; A, B, C, D are all supports. So, two active
degrees of freedom; these are coordinates 1 and 2 - one along the global x and one along
the global y. Then, how many restrained degrees of freedom do you have? (Refer Slide
Time 26:44).
Four. Eight
Four sir.
Why do say 4? Because it is along the member force acts, so for each member only one
at A 1, B 1, C 1, D 1.
He is strictly right, in the sense; you know the directions of your support reaction in this
case, because they have to be aligned along the members in this particular problem. But
if they say, you let go a little bit; it does not matter, you will finally get the same answer.
So, do not be too rigid, let us accept what they say. Let us say 3, 4, 5, 6, 7, 8, 9, 10, but
we know that the reactions are not independent but then, you are bringing equilibrium. I
am doing stiffness method, I know only compatibility. Until I bring in equilibrium, I
cannot say that the reactions are related; so, let us leave it like that. This is fine.
What about our element? All four elements I reduce to one element; why should I draw
four pictures - I like drawing one picture. All four, I cover with that angle - theta x i. If
the length you know this height of that truss is L - I can write an expression for the
length of all four members as L divided by sin theta. Got it? I need four values of theta,
they come from that truss; these are those four values I am measuring with respect to the
positive x axis. Got it? Ultimately, I want to generate the stiffness matrix, but let us begin
with the transformation matrices.
So, my local coordinates are 1 star, 2 star, 3 star and 4 star, got it? And 1, 2, 3, 4 is like
this - without the star; 1, 2, 3, 4 is along the global axes system and 1 star, 2 star, 3 star, 4
star is along the local axes system - by now you know.
Now, how do I write the transformation matrix? Is it clear till now? I have taken a very
simple example. How do I do it? Well, I also have the stiffness, let us say. Does this
make sense? It is the same A matrix we got without the rotation component, which is 1;
so, you have cosine theta, minus sine theta, sine theta, cosine theta - does this make
sense? This transformation is straight forward. Again, off diagonal is null - sub matrix -
is it clear?
Let us apply it. No doubts? We have done an example of a plane frame transformation
and we are doing it in a plane truss. You can do a beam later. How do we proceed? Now,
this is important. The size of this matrix is 4 by 4 - there are four elements for each of
them; so, there are four transformation matrices and for each of them it is 4 by 4.
So, if you take, for example, this element 2, we want to fill here and here; but we can
straight away mark the linking linking what? Global coordinates. See, we have
identified this element as 1, 2 and it is matching with this 1, 2. This is 3, 4 but, for the
global coordinates this is not 3, 4 - this is 5, 6 - are you getting what I am saying? Let us
write to help you understand what we are doing. For that particular problem - this is 5, 6
- this is 1, 2 - this is global coordinates.
This is 1, 2, 3, 4 - what is this? This is local coordinates along the global axes that means
I have already aligned it in this direction. Now, D 5 and D 6 in the structure must be
equal to - how do I write that D 4? Like the same way, we wrote here. For this second
element 2, how should I write that? In that second element D 2 4.
I do not write so much. For convenience, I do it in brackets here - the parenthesis here -
you get it? 1, 2, 3, 4 is local - 1, 2, 3, 4 this way - so, it is 1, 2, 5, 6 global coordinate -
this is helpful for me. This is a clue for me that the displacements that I get at the
element level are the same as the displacements I get at the global - no plus here - be
careful; it is exactly equal.

So, it helps me assemble my structure stiffness matrix from my elements stiffness - is
this picture clear to you? Like this, you have to play a game. You can similarly take for T
1, T 2, T 3, T 4 and I need to plug-in the empty boxes - how do I fill in those empty
boxes? Sine, cosine that is easy to do in a tabular format; I can do it effortlessly - is this
clear?
Let us go through this once more. For element 1 - 1, 2, 3, 4 are the connecting start and
end nodes - 1, 2, 3, 4 - 1, 2, 3, 4. For element 2, we have already finished. For element 3,
it is 1, 2, 7, 8 - 1, 2, 7, 8 which match with the local coordinates 1, 2, 3, 4 and for element
4, it is 1, 2, 9, 10 - does this make sense?
Now, you can do this neatly in a tabular fashion, mark your start node, mark your end
node, mark the theta value, calculate cosine theta and sine theta which are also called
direction cosines and get the length - in this case it is also function of sine. And just plug-
in those values - cosine theta, sine theta and all that. So, you understand - all these are
very easy to do - it is mechanical, the computer will do it effortlessly; Matlab you can do
it easily. So that is how you generate - generate what? Transformation matrices.
Now, we will look at stiffness matrix. We already know the definition K ij; it is called
the stiffness coefficient. It is the force F i generated at the coordinate i on account of a
unit displacement D j equal to 1 at the coordinate j, with all other degrees of freedom
restrained; you are very familiar with this definition.
In matrix forms, it looks like this in the local coordinate system, for each element - for
the ith element it looks like this and for the entire structure, it looks like this - got it? For
each element, it will look like this and for the entire structure it will look like this. Here,
you do not have a star, you do not have i, because all the elements contribute to this
matrix; this is for an individual element - is this clear?
Now, let us look at typical elements that you may get. Let us keep it simple, so we look
at a plane frame element with six degrees of freedom - we have already done that; plane
truss element - how many degrees of freedom? Four degrees - 1, 2, 3, 4, because you can
have four different displacements; and beam element, we assume that it axially does not
move, so you have only a translation, a deflection and the rotation, four degrees of
freedom, got it? 1, 2, 3, 4 - we will keep looking at these three elements again and again.
Obviously, the plane frame is the most generic, because it is a sum - it is a combination
of the beam and the truss.
The stiffness matrix for a plane frame element will look like this (Refer Slide Time
36:55). Let us not fill in those, but you know the meaning of this. It is a 6 by 6 matrix
relating F i 1 star to 6 star, D i 1 star to 6 star, so it is a 6 by 6 matrix. It is going to be a
symmetric matrix, it is a square matrix and how do you fill these? You already know a
physical approach - you can give one displacement and arrest the others; you can fill
these up, but you can also do it using many other techniques including the energy
method. We will see all this later. I am just introducing notations and am giving you a
framework on how to do matrix analysis.

Let us take one example, something you have never seen till now. I want you to generate
right here, right now in the midst of your yawning at 5.30 after doing a lab; I warn you to
wake up and generate the 4 by 4 stiffness matrix for a truss element - all of you.
How to do this? It is going to look like this - it is going to symmetric, generate it. By the
physical approach that means you do not need any calculations. So, what is the first thing
you will do? You have to draw a sketch where you apply only one unit displacement -
which one will you apply? So, D 1 star equal to 1 - can you draw a sketch? It will look
like this - you just have to push it horizontally.
You have an element. D 1s and all the others are restrained. What is the force you need
to push it? E A by L - axial stiffness. What is the reaction you will get at the other end?
With a negative sign - why? Because it has to be opposite, we have chosen 3 star
positive.
Now, what is k 2 star 1 star? You see this symbol - does this make sense to you? We
have got this value that is E A by L; we have got this value which is minus E A by L -
what is this value? And what is that? Why is it 0?
Because there is no reaction.
Because there is no reaction, there is no resistance, there is no reaction, there is no

problem, so those are 0.
Next sketch is going to be tricky. Now, you apply D 2 equal to 1 draw it. Give me the
four values. In other words, help me fill up this column, because we have already filled
up the first column. In the second column what are the values? First, draw the sketch and
then tell me the values. This is interesting; it looks tough, but actually it is extremely
easy. Can someone tell me the answers? 6 E I by - where is I coming? This is a truss
element. All are zeros - all four elements in that second column are zero? No, it cannot
be all zero - which is non-zero? 3 2 and 4 3 are zero - all of you get zero.
You see, all you have to do is to lift it up, like a rigid body. That member has stiffness
only along its own axis - he gave the correct answer. All you do is to lift it up - it will
move like a rigid body, because you are allowing that moment and it is pin jointed. See,
the first thing you learnt was, in a truss member you will have a force only in that
member and there is no shear force - is there no shear force in a truss member? So, how
can you have a non-zero?
I have a member like this - please look carefully - I have a member like this - for you this
is left end and I want this to go up. No problem, this is a rigid body motion, got it?
This is something new. You got a matrix with the second row zero. And this is very easy
to do the third one; fourth one is another set of zeros - so, what does that matrix actually
look like? Fill it up; it has got lot of zeros in it.
Sometimes your stiffness matrix looks like this, which is why, the reduced element
stiffness method said - why should I put lot of zeros? I do not need those zeros - so, it
becomes smaller. Have you got the hang of this? Is this clear? But, then the question why
should we have such a matrix? How is it helping us?
We are saying D i star into K i star is equal to F i star. I can have an element in a truss
like this - please listen - I can have an element in a truss like that which takes a shape
like that. Certainly, I have four values for these displacements, so the components of this
4 by 1 can be independent. No question about it. This is a valid vector and the
components of these are independent. But, in this vector you know that it can have forces
only along this direction. So, what does that mean?
By the way this will be one value. So, this will be D i 1 star, this will be D i 2 star - etc.
So, these four values can be independent that is why we call them degrees of freedom.
But, these are not independent; in the sense - of the four values, the second will be zero
and the fourth will be zero. And even worse, these two are not truly independent,
because the resultant of them must be aligned along this direction - is it clear?
We will see it is related to the rank of the matrix. What is the rank of this stiffness
matrix?
One.
Two.
1 - It is 1. You can see that this can be reduced to 1.

All that we learnt in concepts of matrix algebra are very deeply meaningful and they
have a physical meaning as well. So, we need this matrix, it is helpful and we need this;
we want the big picture no matter what we do, we should get correct results. We know
that even after you multiply, you should get zeros; obviously, if the second row and
fourth rows are zeros, you will end up with zeros - you have to; so, this has a physical
meaning.
For a beam element, without much ado, the stiffness matrix will look like this (Refer
Slide Time 45:45). We will study it in depth later. You already have a clue why it will
look like that. You can break it up into four parts - we will study it in detail later; do not
worry too much about it.
You can fill up this matrix. Here, it is a function of E I - remember 4 E I by L, 2 E I by L,

6 E I by L square, 12 E I by L cube - that is all that you can get. If you have a plane
frame element, you will have a combination of the truss and the beam; we will see this
later.
This is 6 by 6; right now it is an introduction, you should know that you have these
elements stiffness matrices. Now the challenge for us is we have got the transformation
matrix, we have got the elements stiffness matrix, they are properties of the dimensions
of the element, length of the element and the E and A or E and I values, how do I get the
elements stiffness matrix in the global axes system? Not for the structure, I am still
sticking to the element.
In the local axes system, this is what I have got - I have got K i star; what I want is - this.
You see how I am getting to do it. This D i can be written as T i D i - got it? This star.
Now, if I do that I am able to get into the global axes system and F i is also T i transpose
F star i - remember? You plug this in there, you get F star i as K star D i star from first
principles. When you substitute here, finally you will get it in this form - this is a
standard form - this is a standard transformation we do.
So, this is how you are able to derive the element stiffness matrix without the star, that is,
the element stiffness matrix aligned in the global axes system. Once, you have the
transformation matrix T i, which you got - it is just sine and cosine and maybe 1, and you
have the local element stiffness matrix - these are fundamental principle in matrices - is
this clear?
Now, look at this transformation. This transformation tells you that this is definitely
symmetric; so, when you pre multiply and post multiply with the transpose of a matrix,
you will end up with another symmetric matrix. This is something that can be done
effortlessly in the computer. So, you generate the T i for all the elements and generate K
i star for all the elements; do this exercise you got straightaway and you aligned all the
elements in the direction of the global axes system.
To give you a parallel, you have these national parties in the central government -
different political parties. I do not want to give any names. They will all have their little
groups in the state government also; they are all aligned - already aligned. They will only
vote for them, so it is something like that where you got the groups at the state level.
Let us take this one picture which is a beautiful way to understand it. I want you to draw
it. I am going to draw a playground with four corners. At the bottom line - You see how I
have done; it is very easy to do it. I am putting the forces on the left side and the
displacement on the right side; so, the first transformation that I have is the element
displacement vector in the local coordinate with the star. If I have this and if I pre
multiply with the element stiffness matrix, I get the element force vector in the local
coordinates - is this clear?
This is the meaning of the arrow. I am not writing any equations; I am symbolically
showing you this first transformation - it is a linear transformation. In other words, if you
give me this and I have this, I pre multiply this with this - I get this. That is why the
arrows are pointing this way, so this transformation I know clear?
What is the other transformation you know?
Global to
No multiplying, we are playing a game; we are not deriving any equation. This is the
other one - this is aligned along the global axes system; so, I have D i without the star, I
have K i without the star and I have F i without the star. This is in global axes system;
this is in local axes system - is it clear? So, both the arrows point this way; this is the
element stiffness matrix in the global axes system and this is the element stiffness matrix
in the local axes system - is it clear?
Then, what is the other relationship I have? I have this relationship. From global axes
system, using my element transformation matrix, I get the element displacement vector.
Now, I have got this nice play ground with four corners; so, let say I have this vector -
can I straight away go to this vector? How? I do not walk along this corridor; I just cut
across straight here. How do I move from here to here?
T transpose
Even before that - is there a relationship between this and this?
T i transpose
Yes, there you are - that is the inverse of the transpose, so your playground - your
corridors are now ready.
Now, let us do the diagonal. Does it make sense to you? Is this easy to remember? Those
matrices are not easy to remember; play ground is easy to remember. This is a nice
sketch; I want you keep drawing it till you master it. Four corners are clear? And the four
transformations are clear? They are all symmetric matrices; they are all square matrices,
so are these terms clear to you?
Now, let us take the diagonal - how will it look? It will look like this - this is equal to this
times this - this is equal to this times this; therefore, this is equal to this into this times
that - that is a shortcut - got it?
What is this in terms of this? Just take this and pre multiply by that. Instead of going
around along the corridors to reach here, I take a straight shortcut from there to there and
obviously, it will look like that - is it clear?
This is the first of many diagrams like this we are going to draw in this course, but they
are the easiest way of remembering the relationships. Let us play this game once more.
We began with this, then we did this, then we did this, then we did this (Refer Slide Time
53:31) - are all four crystal clear?
Then, we said take the shortcut from D i to F i star like that and then, we said my
objective is to get this in terms of all the other jokers; so, how do I do this? Straight away
like that - got it? That is all; so, you have got all the relationships to do the direct
stiffness method. Let us just demonstrate with one example; then, only you will
understand.
You have got this problem. You have got 12 degrees of freedom - 8 are active and 4 are
restrained. These are all element level degrees of freedom aligned in the global axes
system and these are the compatibility relationships; you yourself spelt it out very nicely.
We need to do this transformation - how do we do it?
I want you to see how - you have got the 3 T i values, you got the 3 K i star values; you
do the T i transpose K i star T i, you got the 3 K i values - what are those K i values?
You got the element stiffness - the computer does it effortlessly - for all the three
elements.
Now, the big question is how do you assemble the structure stiffness matrix? You have
done that - you did that and you got this. Take the first element in the first element, the
start node is matching 10, 11, 12 of the global coordinates and end node is matching 1, 2,
3 of the global coordinates. So, I am writing 10, 11, 12, 1, 2, 3 and I am saying I have
already got this - I have got these - is it clear? Do you understand the meaning of this?
I have assembled this - I have assembled this K 1 - how did I assemble this K 1? I did the
transformation that we saw in the last step; I am now moving from state government to
central government. I have got three states; all the three states have already been aligned.
The stiffness matrices are ready - K 1, K 2, K 3 are ready - how do I put up the big
matrix? What is the size of the big matrix? No, for the whole structure; no, there are 12
degrees of freedom - 12 by 12.
It is like this (Refer Slide Time 54:31); please listen carefully. I got my three elements -
in this, I got a 6 by 6 matrix; in this, I got a 6 by 6 matrix; in this, I got 6 by 6 matrix -
three elements. All of them, somehow, must fit into my 12 by 12 matrix of the whole
structure. You have to collect boxes from here and put them into the main 12 by 12 box -
the right box goes to the right slot. If you do not do that, you will miss out; so, there must
be a clever way of doing it - how do I do it? That is what I marked here.
I have a 12 by 12 matrix and you can see nicely they come in packages of 3 by 3; so, the
big 12 by 12 - I have 3 by 3; 3 by 3; 3 by 3; 3 by 3; 3 by 3; 3 by 3; 3 by 3; 3 by 3 and like
that - I have got all these slots.
Now, you just need to match these linking numbers you got 10, 11, 12, 1, 2, 3, 10, 11,
12, 1, 2, 3; so, this matrix I can make it look like this - got it? This is 10, 11, 12 global;
this is 10, 11, 12 global, so this should go into the main structure 10, 11, 12 global; it
goes straight there, because the stiffness has add on.
You have studied this in displacement method. If you have two elements and the
contribution of this to the structures stiffness matrix is something and this they add on to
the joints; we have done that understanding earlier. You have got an element stiffness
matrix contributing - like a donation coming from the state party to the central party, so it
should go to the right slot - are you clear? So, this will go to 10 K 10 to 12, K 10 to 12;
this will go to - this part is 1, 2, 3, 1, 2, 3 - so, this will go to the upper left corner. Please
listen carefully; this is interesting, because once you get it - you get it.
See, look at this matrix - this matrix is 1, 2, 3, 4, 5, 6. Unfortunately, my start node is 10,
11, 12 and my end node is 1, 2, 3; do not blame me for that - that is the way I did it. I did
it, because we decided that active degrees of freedom should come to the top of the list
and restrained degrees to the bottom. For my element level, the upper left corner became
10, 11, 12 - I mean it was not my choice, it happened and the lower left became 1, 2, 3;
but, I have to make these boxes and keep it ready. Does it make sense? Then, there is a
cross from 1 to 3 to 10 to 12 and 10 to 12 to 1 to 3 - is it clear? Does this make sense to
you? If it does not; go back and read up from the book.
Take the next slot. The next element is - this is easy 1, 2, 3, 4, 5, 6 - nothing to worry, it
will go to the right slot effortlessly; no need to play around.
And the last one is 4 to 6 which you need to modify, because you have got a mix up of 8,
9, 7. Sometimes, this happens in real problems, so you have to just rearrange the rows
and columns and you get 7, 8, 9.

At the end of the day, this is what you need to do; this is your 12 by 12 matrix. From the
first K 1, you get this, you get this, this, this (Refer Slide Time 60:16); you have to
appropriately put the boxes - does this make sense to you?
There will be some zeros and this is my K A A; this is my K A R, this is my K R A and

this is K R R. If you have understood - very good; if you have not understood - not to
worry, we will pick up this later.
This is just an introduction; we are not actually solving any problem, but we are trying to
see how you can make by proper planning the computer do everything for you.
We will carry on from this point in the next class.
Thank you.
Keywords: Basic Matrix Analysis, Transformation Matrix, Stiffness Matrix,

Conventional Stiffness Method, Stiffness Matrix, Flexibility Matrix
Prof. Devdas Menon
Module - 3.4
Lecture 20
Good morning. This is lecture number 20 in our course on Advanced Structural Analysis.
We are continuing with this third module on basic matrix concepts, as applied to
This is covered in the chapters 2 and 3 of this book on advanced structural analysis.
You will recall in the last class, we gave a big picture view. This picture really shows
how you can do various transformations to finally arrive at the? Arrive at what? What is
the end result of this?
Global Stiffness Matrix.

You will get the element stiffness matrix of every element aligned along the global axis.
So the end result of these transformation is this K i is equal to T i transpose, K star i T i,
after this, your next job is to assemble the matrix and get the structure stiffness matrix.
So, that assembly is where some of you had a difficulty. Remember we had some
difficulty understanding so, let me tell you a story.
There was a housewife who wanted to buy one of those gadgets or mixer, come grinder,
come juicer, come so many other things so, she bought it online and this big packet
reached her house. She was very excited because, she paid through credit card and all
that and when she open the box it looks beautiful there was a brochure but, it all came
like this matrices in small pieces and she was supposed to assemble them because
assembled is very big, the unassembled is very small so, for packaging purpose is came
in small pieces.
She was postgraduate student so, she thought she could easily put it altogether. So, she
opened the manual and she spent one hour trying to put it altogether and did not work,
just like you did not get the hang of this in the last class.
So, she kind of gave up, and in frustration she left the house and went out. When she
came back few hours later, she was amazed to see that, the matrix was well assembled
already and it has been connected to the socket, plug and she just put on the switch and
she heard a beautiful whirling sound.
Oh my god, how did this happen and she looked around and she realize there was a maid
servant in the house, who is absolutely illiterate. She was wondering, did she fix up this
wonderful gadget and she called and, she said did you do this? She said, she smiled
hesitantly and said yes. But you do not know how to read or write and she replied yes
mam, when you do not know how to read or write you have to use your brain.
So, your real challenge is you are literate, you know how to read or write, and you can
look at solved examples in text books. If they say step 1, step 2 you know what all that
means but, really putting things together on your own, you need to be little illiterate and
start using your brain.
So, here we have the challenge of both knowing how to read and write and also trying to
fit it all together. So, let us do this one more time, looking through the eyes of the maid
servant.
So, you have three elements there, for each element in the last class; we generated the T i
matrix, first we generated along the local coordinates, then we transformed it and put it
along the global axes but, we put some linking coordinates.
So, we have got those readymade stiffness matrix boxes and we just have to put it at the
right place in the big matrix. The structures stiffness matrix, what is the size of that? Its
12 by 12 and we have these small three 6 by 6 matrices, which we can again partition
into 3 by 3, 3 by 3, 3 by 3, 3 by 3 and we need to put the right 3 by 3 box in the right slot.
So, all you have to look at is, connectivity. Let me help you, take the first element, the
first element is connected to nodes 10, 11, 12 and 1, 2, 3 so all you have to do is, to pick
up the appropriate boxes and put them into the slots as shown here, does this make
sense? It fills up only those four slots.
The contribution of element one is only to these four slots 3 by 3 slots and it is direct
contribution. You just have to show it there, you do not have to play with it, you do not
have to multiply with the change sign, and nothing just put it there. Can you do it? That
is all you have to do, do it intelligently, right?
Element one has no contribution elsewhere in that big picture in that global, take element
two, where do you think it will fit in? It will fit in that simple, the top corner because 1,
2, 3, 4, 5, 6 element coordinates match with, global structure coordinates 1, 2 so that is
simple and the last one? Last one will connect 4, 5, 6 and 7, 8, 9. Does it make sense?
That is all, and in some places like 1, 2, 3 you will have two contributions, you just
algebraically add them up, because stiffnesses get added up. Does it make sense? That is
all you have to do and when you do an assignment of course, we will make it a small
problem you will get a firsthand feel of how to do it yourself. So, there is no way of
learning this except doing it by yourself, like the maid servant did it in our little story.
Finally, after you have got this matrix and we will see that there are many other ways of
getting the same matrix. There is a reduce element stiffness method, there is another
method, which we will study soon of getting this matrix because, this has the big
advantage that you are playing with little boxes, no matrix is more than 3 by 3.
So, your storage is very elegant, economical, only thing you must program it well, so that
it goes to the right slot. There is a crude way of doing it; we will look at that, for
beginners, that you will be more comfortable with that. But, when you do programming,
when you want to make, a matrix solve a very large problem, you cannot do that; that is
too unwieldy. This further you can partition to the active coordinates, you have eight
active degrees of freedom and the restrained coordinates. Does this make sense? And the
alternative is to use the, what we call the displacement transformation matrix T D which
is what I want to show you next.
But, before that let us say, you have got the stiffness matrix, you can write relationships
like this, basically F is equal to K D at the structure level but, D itself is subdivided into
active degrees of freedom, which you do not know. You do not know D A displacements
but, you know D R displacement, right? They are usually zero. On the left hand side, the
force vector F A is known to you, that is a load vector acting at the, F R is the reactions
you do not know. So, you can expand this into two equations, because a K matrix can be
subdivided and they will look like this. Does it make sense?
So, you need to solve the first equation, find out the unknown displacements D A, plug it
in those second equation, you get the support reactions here, from the global picture.
Then, the unknown displacements D A you should use, using equation similar to your
slope deflection equations and get the member level forces, does it make sense? This is
the big picture.
(Refer Slide Time: 07:38) Now, that is what you get in a truss, because there are no
intermediate loads but, in a beam, or in a frame, or even in an axial system; like a chain
where you have intermediate loads, you have to add the fixed end force effects, which
you have some introduction to already, so you have to modify that equations slightly. We
will look at that in detail later.
Then let us go back, let us take another example, let us take this frame which we looked
at yesterday. How do you go about doing it, first you find T i and K i star, we have done
this, what is a next step? Finding K i by doing this transformation which, in Matlab you
can do effortlessly for each element. Then, what is a next one? No, no this is at the
element level. Next is assembly, this is where you had difficulty. We assemble it to this
form and then, you can solve for whatever loading you get.
So, mind you up to this stage, you are dealing with the property of the structure and you
say, now give me the loads I can handle it but, the computer does this in a jiffy, the
moment you feed in the geometry of the structure in the material properties. It is ready to
handle any loading because; it is already going to work out this properties effortlessly if
your programming is good.
For example: in this structure you do it this way, let us look at K A A, it will be a 2 by 2
matrix, because you have only two active degrees of freedom and K A R for example,
will look like this but, in this method you have to always keep track of the linking global
coordinates which I have shown here. 1, 2 are the active degrees of freedom and 3, 4, 5
all the way to 10 are the restrain degrees of freedom and K R A will be the transpose of
K A R and you can calculate K R R.
You will never need K R R unless, you have support settlements because that is D R is
usually zero, the alternative is this using a displacement transformation matrix, this is for
people who are not very comfortable you know putting it in slots and all that. If
somehow you could have a robust method, where you do not have to keep track of this
linking coordinates then, you have got a blind way of doing it.
Not suitable for programming large structure but, ok at your level, ok to do manually and
I think that is the way it is taught in most universities. But, that is not how algorithms are
built in software packages so, let us look at that.
Remember, I told you about three transformations, this one we have finished, you
studied this, we know how to do the transformation using T i.
The second one is using the displacement transformation matrix, this is (( )) when you do
the reduce element stiffness method, which we are going to study next but, you can use
the same idea in conventional stiffness method and in the next class, we will look at
flexibility method where, instead of a displacement transformation matrix you have the
force transformation matrix, it will all fall into place.
Now, the inverse of those matrices helps you switch from the element level the local
level to the global level and there is a principle, you do not need to do the inverse there is
a principle called the contra gradient principle, a beautiful principle which will look at
shortly.
So, let us look at these transformations. Now, I am talking about structure level
coordinates here and element level coordinates here. I have combined them, so F star
really means F 1 star, F 2 star. I can put them all together neatly.
So, does this make sense? And I am saying, if you give me the global displacements they
are linked all in one go to the element displacements through a matrix, which I called the
T D matrix. T D matrix is the displacement transformation matrix and obviously, it must
satisfy compatibility. Likewise, in the flexibility method there is something linking this
to this, that must satisfy equilibrium and what would be the good name for that
transformation matrix, force transformation matrix, we will study that later. For the time
being, we will look at the displacement transformation matrix and see how it works.
So, since we know that the displacement vector is really made up of D A and D R then, I
can again partition my T D into T D A and T D R, does it make sense? T D A into T D R
ok and if you expand it, it will look like this. Do not get worried about those size of those
matrices, it is just how it works out, how it falls into place. Now, I can do something here
this D star is actually including all the elements in the structure. It is too messy
sometimes, so I just look at one element. When I look at one element the i th element, it
will look like this D i star and this T D now, becomes unique to that element so, it is T i
D which can be partitioned as, T i D A and T i D R. Does it make sense? This also you
can do, it is actually all the D i stars put together that actually make up the D star.
Now, let us take a look at an example: but, before that a definition, the displacement
transformation matrix T D is a unique matrix satisfying compatibility requirements for
any given structure and a typical element D i T D, T i D at L star j of this matrix may be
visualized as the displacement being transferred on account of a unit displacement at the
j th coordinate to the local L star coordinate of the i th element in the structure with all
other degrees of freedom restrained, that is quite a mouthful but, basically you get what I
am trying to say. Do you remember? When you have two subscript, the second subscript
goes to the cause and the first one goes to the effect.

Let us just try to understand what that means, take this same structure, let us see what it
means for the i th element. What should we do? Well really, we are trying to first arrest
all the degrees of freedom. So, you got the kinematic structure; all degrees of freedom
arrested then, one at a time let us say D 1 equal to 1, you get a shape like that. We are
trying to see the effect of this in each of the element stiffness matrices, each of the
element displacement vector that is what we are trying to see. In other words, if I apply
D 1 equal to 1, which elements will it affect? It will affect first and second, the third will
remain unaffected, if I apply D 2 equal to 1, likewise something like that will happen.
If I apply D 3 equal to 1 that will happen. So you already have an exposure of doing this,
you can actually do it from first principles, using the physical approach because, we did a
few examples earlier. But, the computer has no feel for any physical approach, it has to
do it mechanically, blindly and that too is possible.
Before that, let me talk of this matrix K star, which is sometimes refer to as the
unassembled stiffness matrix. Let me give you a simple example: let us take a truss.
Let us say the truss has four elements, so I have F first element. There is only 1 because,
1 star, 2 star, let us take the first element F 1 3 star, by now you know that the truss
element has four degrees of freedom. We worked out this stiffness matrix in the last
class. So, this is F four star got it. This is related to D 1 1 star, D 1 2 star, D 1 3 star, D 1
4 star through a relationship, which we are going to refer to as, K 1 star, right? This is the
element stiffness matrix 4 by 4 and you remember, some rows are zero. Which are zero?
Second and fourth. Because they refer to shear forces in a truss element, the shear force
is zero. So, we know that these are anyway going to be zero and so, the columns also will
be zero, right? And the non zero values what are they? They are very easy to remember.
So, the principle diagonal will have E A by L of that element 1, minus comes to the off
diagonal elements. Now, if I take the next one, which I will put as F 2 star ok I get
another 1 like that is equal to K 2 star into D 2 star, does this make sense to you?
Similarly, I get let us say there are m elements, I can write these matrices.
Now, if I want to relate all of them to a single matrix. So, this is really meaning F 1 star,
F 2 star etc. This will be right? How will this look like? Let us say, this is dimension m
by 1, m members this will be m by 1, and each of them will be 4 by 1.
So, the actual size will be 4 m by 1, right? Let us look at, the sub matrices and not worry
about this so, we do not worry about this. What will be the size of this matrix m by m?
How will you make it? How will you arrange it? You have got all the individual
elements matrices, how will you arrange this?
No.
Sir k one star k two star.
Wonderful, see the element stiffness matrix of one element has no relationship with the
element stiffness matrix for the other element. So, if you really want to put in a matrix
form, there is only one way you can put it. This will be a diagonal matrix where, you
have say K 1 star here, K 2 star there, all the way here and you have zeros everywhere
else, is it clear? Because, then only I get this is equal to this times that, which is this, this
is equal to this times that and nothing else. So, this is sometimes referred to as the
unassembled, because you really not assembled the structure you just put together all the
element fellows together but, it is not the structure stiffness matrix. So, this is called the
unassembled element stiffness matrix. Does this make sense?
That is what shown in the slide here because it is always the diagonal matrix and if you
do the transformation here and see it is a simple transformation. I need to explain this
transformation, this depends on the contra gradient principle which I showed you but, I
have not proved it, i will prove it little later. Let us accept this principle that, the reverse
works this way and so you can prove that it takes a same form, this form is familiar to
you, because we did T i transpose K i T i, you have a similar transformation here. So you
see, you get the same structure stiffness matrix from the displacement transformation
matrix this manner.
So, let us say you get T D, let us say you got K star, you got K by this method and if you
do not want to do it for the whole structure, you want do it element by element that is
also fine but, then you have to sum up for all the elements. I want you to understand that,
unlike the previous case where we dealt with K i, which was always limited to the size, it
is a square matrix size 6 by 6. This K is a structure stiffness matrix size is as big as the
global degrees of freedom, you will understand when we do an example and then, you
can have these relationships, after you find the displacements, you can get the member
end forces.
So, let me put it in a different way. If I want to generate the structure stiffness matrix
using this approach, the displacement transformation matrix. First, I create this
relationship at the element level putting all of them together, so it is unassembled. Is it
clear? Exactly, what I have shown on the board.
Then, in the real structure I know this is what I want. Now, my playground is different,
now my playground is on one side, the goal poles will have the structure level
coordinates, on the other side, I have all the element level coordinates put together in an
unassembled manner but, all of them are there. Then, my T D will take this path because,
D star is T D into D, I have not partitioned D A, D R, but you can do that. Is this clear?
Let me ask you, is this step clear from here to there? Is this step clear from here to here?
This your conventional stiffness matrix. Is this step clear from here to here? Ok. That is
how to get this T D is what we will see shortly but, once you got T D you have the same
game going on, that will be T D transpose. We will prove that it is so. Right now you
accept it as it is and then, you have this short cut along the diagonal and so you will get
this.
So, you know this also establish the same relationships. Let us see this problem which
we did in the last class, I am just showing you to remind you that we did this using T i.
How do we do the same thing using T D? So let us see, remember in T i those where our
non zero elements. The transformations were very simple, you had a cosine theta and a
sine theta and a negative in one of those.
Let us, look at the same problem and here we are using the T D in the conventional
stiffness method. What do you need to do? Well, I am not going to the global axis, I am
sticking to my original 1 star 2 star 3 star 4 star, I am not doing any switch at this stage.
Then, I need to develop this matrix, right? It will look something like this. First of all, in
this truss element; I have four degrees of freedom so, do you understand D i 1 star all the
way to D i 4 star. This is by T D matrix and on the right side, I put all the elements that
are there in this all the coordinates in the structure, global coordinates. There are two
active coordinates and there are eight restrained coordinates, right? So, the size of my T
D matrix for any element i, there are four elements is 4 by 10. Are you getting it? It is
simple, it is 4 by 10.
Now, most of the elements are zero, let us prove it. Now how do I fill up these column?
If I want to fill up this column, what should I apply? D 1 equal to 1 and everything else
0. If I want to fill up this, what should I do? D 2 equal to 1 everything is 0, you want to
fill up this D 3 equal to 1 everything else 0, right?
(Refer Slide Time: 24:50) Now, do not you agree this box will be the same as the box we
have got earlier, because it is a same transformation? Cosine theta, sine theta, minus sine
theta, cosine theta because we are aligning to the global axes because, the global axes are
pointing horizontal x and vertical y.
(Refer Slide Time: 25:16) So, I have taken a short form. I have said let C i be cosine
theta i, because I need not write so many letters and my matrix looks smaller and sine
theta, I write as S i. So, do you agree this is so but, there is one more box of C i, S i
where should that come, well it depends on which of the four elements you are dealing
with, right? Let us take the first element. This set of C i, S i minus S i, C i will come
some way here, where will it come? Will it come here? Will it come here? Will it come
here? Or will it come there? Well check the coordinates, the element number one is
linking 1, 2 and 3, 4 so this is 1 is not it, this is 2, this is 3, this is 4 because that is what
the global coordinates point to.
So, where will it come? It will come there, for i equal to 1, so that top corner is common
for all of them, because 1, 2 is common for all the four elements, in this particular
example. What about for, i equal to 2, the second element; where will it go? You see,
you are just kicking around a football. What about i equal to 3? i equal to 4? That is it?
So, you have got four T D matrices, for all the four elements K star is very easy, what is
it? It is the same, it is this one, right? You can play the game and you know, you can
actually work it out. It is very easy to do it
And find T D and K i as we pointed out, generate this, sum it up and you will get to the
exactly the same form, right? Sure? Ya? No. Sorry that should be sine that is a same box
which is going. That minus should not have been there agreed.
It is a same box well pointed out, thank you. Good. Students should have sharp eyes to
catch errors.
So, let me sum up, because we are just showing the big frame work, we are not solving
any problems. What do you do in this stiffness method? Well, your first job is to make
sure that, the computer is able to deal with the same structure that you have in mind.
So you have to join, you have to give information regarding the joint labels, the global
axis coordinates identify the active and the restrained degrees of freedom, put them at the
end of the list, give them labels and identify the incidence which is your start node,
which is your end node and give the element material and cross sectional properties. In
fact, if you take any software that is the first thing you need to do. You input it and once
you get the coordinate, it should draw that structure for you to see and confirm and say
yes. In fact, in the big software packages, you not only see the line you can do what is
called an extrusion.
Let us say, you have given an I-section, you have to make sure that the I points in the
right way. I mean it does not get flipped over to all that come, you can see clearly. So,
sometime we make errors in the input but, you should be able to see it using some visual
C plus plus. I am not asking you do any of that but, then this is the first step.
Next step; what do you think is the next step? After you have given the input.
Give the loads.
That comes much later. I mean I am not saying what you have to give. You are right.
You have to give the loads, what this going on inside? The moment you give this it is not
waiting for loads. It is trying to be ready to handle your loads know so, by the time you
keen the loads it would have done its homework, what is the homework? It will generate
the element properties, right? So, we have done that the element stiffness matrix, the
transformation matrix, either you can do T i software packages do only T i but, let us say
humans will do T i D, because now you are not comfortable putting those brackets and
serving them into box.
So, T D is the blind way. You know you are really covering all the global coordinates.
So, what if most of the elements of that T D matrix is 0? Then, you do the product (Refer
Slide Time: 29:51) so that, you are able to generate the structure stiffness matrix and
there are two ways you can do it and finally, you find that its K A A that you need to
invert, because only that forms the matrix, which will help you get the unknown D A.
So, you do not have to invert the whole K matrix for example: that truss problem only K
1 only 1 and 2 those coordinate. So, K A is really small and when you learn in the
university, they teach only that K A A, so you do not see K A R, you do not see K R A,
you do not see K R R, so you do not get confused. Now you get confused but, then you
are trying to see the big picture and trying to see what is really going on, you are trying
to capture the complete displacement field, the complete force field and you decide, what
do you want to do.
Next, you put in the load data and this we will see later you generate the fixed end force
vector, you find the unknown displacements by using the inverted K A matrix, compute
the element level forces, compute the support reactions, you got everything and all this
happens in a jiffy at the press of a button but, inside the black box, we are trying to see
what is inside the black box and how is it working.
So remember, we have got some idea of what the conventional stiffness method is. We
are now, going to look at the reduced elements stiffness method and we will also look at
flexibility method in the next class.
But, something is common to these and I want you to understand. In the last class, we
talked of three element stiffness matrices, one was for a truss element same as the one I
showed on the board, second was for a beam element but, we said there is a problem
with the rank of these matrices, what is a rank of the first matrix? 1. So the question is,
can you find the inverse of the element stiffness matrix? Is it possible?
No.
So, is that a flexibility matrix for the element stiffness matrix? Obviously, because it is
singular, it is rank is not full so, you cannot invert it, does it mean there is no element
stiffness matrix? I mean no element flexibility matrix? Because we have studied that the
flexibility matrix is the inverse of the stiffness matrix, but clearly inverse is not possible
so, what is wrong? Or what is right? Physically, why is it not possible? That is the
answer to a different question. Why are you not able to find a flexibility matrix for these
elements? The answer is actually ridiculously simple but, I want you to say it and
mathematically it turns out so. So, your intuitive hunch is justified by mathematics. Now,
you do not have the intuitive hunch. You try to see the mathematics hence, and give me
the hunch behind it. Why cannot you have a flexible matrix? Why cannot you invert it?
What is a reason? I will give you a clue.
The clue is, take the flexibility matrix, what does it tell you? If you give some forces,
you can find the displacements. If I have a structure and I push it, the flexibility tell me,
how much the deflection is? Now, here it will not be able to tell me, what the
displacements are? Why?
Because, there is something when you look at those pictures, which should make you
nervous, when you built such structures. Why are you not nervous? You should get
nervous when you look at those elements, see they are just hanging, floating in the air,
how can you build structures like that? If you push them they will just this is not outer
space, right? So, you do not get any unique results. These are unstable structures, the
way we have drawn them. That is the simple reason, they are unstable structures.
You have rigid body movements so, you have independent displacement vectors which is
fine but, you do not have independent force components in the force vector. That is a
simple answer but, there is more to it. So rank of this is one. What is the rank of the
second one?
Two.
Two. Is it not, two. And what about this one, the plane frame element?
Two. Three. Three.
One is for the truss. It is 3. Why it is 3? You should figure out by now, you can multiply
you know they are all dependent, they are not independent. So, here is the summary of it.
The element stiffness matrix is singular, non invertible. Physically, this is associated with
instability, inadequate restraints, rigid body moments. The components of D i are
independent but, clearly the components of F i star are linearly dependent. The number
of independent force components is limited by the number of independent equations of
static equilibrium so, take the truss element.
Theoretically, you have four forces but, only one equation of equilibrium. So, that is why
the rank is one. It makes lot of sense. So, I am using the notation q for the conventional
stiffness method, where q is four for the truss, plane truss element but, it reduces to 1 so,
I have put a q tilde that sign is called tilde. q tilde is 2 for beam element. q tilde is 3 for a
plane frame element. Now, I want you to fill in the blanks. What about a space truss
element? 6 by 6 is what you get, if you follow the conventional stiffness method. What is
a rank of that matrix?
Four. One. Six.
Plane truss 4 became 1, Space truss 6 will become?
Two.
Wrong.
Three.
Wrong. 1. Argument is simple plane truss, what is the equation of equilibrium? 1. I tilt it
and put it in space, it is still 1. So you did not get it. It is beautiful. Who would manually
ever dream of doing the conventional stiffness method to do a space truss, when the
reduced element stiffness method offers you such a dramatic, drastic reduction in the
quantum of work involved? So that is why, it is good fun manually to study the reduced
element stiffness method.
So, for a space truss it remains 1, for a grid element it is 3, for a space frame it is 6, 12
become 6. I bring a new notation here. The tilde is my notation for telling you that this
element stiffness matrix is different from the stiffness matrix, we talked about till now.
This is dealing with the reduced size of the matrix, this has an inverse because the rank is
full here. This has an inverse and that inverse is called the flexibility matrix of the same
size so, it is beautiful, that is why it is good to look at the reduced element and flexibility
in one go because, the size of all the matrices are identical. Does it make sense? Ok.
Next
Let us look at this problem, we did this earlier, right? Now, I have given three supports. I
can give any three independent supports, you see these things and I have actually made it
simply supported. These two rigid link say that, this joint cannot move, this rigid link
says, it can move horizontal, it is like a roller support but, it cannot move vertically.
Now, this element is stable. If I apply loads on it, it will resist those loads. So, that is a
kind of element you are going to deal with from now on and actually you should be more
comfortable with these visually because, they are familiar to you and you know you
cannot play around with them, they are not mechanisms, they are good solid elements.
But, I have chosen simply supported, you can choose a cantilever also, you will get the
same results but, we will choose it simply supported. Is it clear?
Now, how do I mark my degrees of freedom? Earlier I did, 1, 2, 3, 4, 5, 6 for each, how
do I do it now? How many do I get for each element?
There is in one force for the first element.
I have an axial degree of freedom because, I am not assuming inextensible. If I assume

inextensible that goes, so that is my 1. Then, I have two rotations so, 1 star, 2 star, 3 star
much better than 1 star, 2 star, 3 star, 4 star, 5 star, 6 star, right? This is the reduction; it
is a reduced element stiffness matrix we are dealing with. Is it clear? Would you like to
do this or the other one?
This.
This, when you are doing problems for yourself, that when you want to sell your
software and make money.
So, let us proceed. So, you have 1, 2, 3 everywhere, right? Ok. These are my local
coordinates, now I can put it all together and then, I get my combined element stiffness.
So, for each of them it is like this got it, that 6 became 3 that is all much easier to handle
and there are three elements and I can put them all together and that is my combined
element stiffness matrix. Does it make sense? Everything that we did till now, except the
size is dramatically reduced and you put a tilde to make sure that, you are clear. So, I
have both, I have displacement vector, force vector combined and when I put the
relationship between them, it will be that diagonal unassembled stiffness matrix or I put
it the other way I get it in this fashion. Now, each of them at the element level will look
like this.
The reduced element stiffness formulation has the advantage of enabling us to deal with
transformation and stiffness matrices of small sizes. It is especially suitable when we
wish to make simplifying assumptions in structural analysis, such as ignoring axial
deformations in frame elements. However, it should be used with caution, when dealing
with sway type problems and those stertorous problems; we have seen where the
columns are sloping. There you have to use extra brain work not worth it. might will use
the software way of doing it.

So, you must use the right instrument, for the right occasion. You do not use an elephant
to drive a small nail but, we end up doing that. So, let us look at the different types of
reduced element stiffness matrices; take this, it is a 1 by 1 matrix only one degree of
freedom. What do you think the degree of freedom is? Well, I have shown this is simply
supported but; I can pull it only in one direction. What is the displacement D 1 star for
this, physically what does it mean?
Extension
Elongation
It is just the elongation. So, that is easy for us, it is what we called E I earlier. Now, we
are comfortable, we are connecting with, what we have studied in structural, Structural
analysis is actually very simple and you should not lose that simplicity. So now, we
realize that is not it much better to deal with this, than with that but, you have to deal
with this, when you dealing with large problem. So, you have to know both. Ok. That is
your truss element and we call that axial stiffness, what about your beam element?
E I by L
Two. You are familiar with this, the value will be 4 E I by L, 2 E I by L. What about the
plane frame, three? It is nothing but, the truss element and the beam element put together
and we assume that the axial stiffness is independent.
We ignore any axial-flexure interaction that is what we do in first order analysis, we will
study later in the seventh module there is an interaction, because it behave like a beam
column. Right now, we pretend there is no interaction, not only we, all the have software
packages also happily pretend that, there is no interaction but, actually you have
something called a P-delta effect which we will see later, that is called second order
analysis. Does this make sense? Ok.
What about the flexibility matrix? That is very easy, I put it this way, I flip it over and I
call this axial flexibility and look at this. This makes sense, you remember if I apply a
moment here, what is a rotation I get there? L by 3 E I and what do I get here? L by 6 E I
with an opposite sign that is what this matrix tells you. So, it is not difficult and by the
way, it is an inverse of the previous one and similarly, this works out.

So, if you want to put it all together. These are the three values for stiffnesses and these
are three values for flexibility. Very easy and you already have learnt this and manually,
you can do this but, you should also know the other one. How do I generate the
components of this matrix for a plane truss element? I will end with this definition.
So, let us take a truss element like that. How do I get the T D matrix for this truss
element? Well, now I realize there is only one degree of freedom though in my global
structure, I may have 1, 2, 3, 4, now this 1, 2, 3, 4 I have shown for convenience, it may
actually be 5, 6, 21, 22, in the actual structure. So, those are happening there but, the
elongation is given by, how much these move, so this 1, 2, 3, 4 is arbitrary. Now,
basically I want to know, if these joints move by some known amounts, how much
elongation do I get? That is a meaning of this T D i, T D i is just how much elongation
the i th element, got it?
So, the elongation due to D i 1. Actually, this i can be removed. It can be just D 1
because, it is a global thing. How much is that? So, you did not put i here. This i can be
removed but, it is linked to the other i, in the conventional stiffness method.
What are these values? How do you work them out? Well, let us do it from first principle
you put D 1 equal to 1. How much will be the elongation? The component of that is cos
theta, will it be plus or minus elongation? It will be minus. You put D 2 equal to 1. It will
be? Cannot you see the length of the element is reducing? It will be sine. It will be minus
sine theta. If you take the other end and you pull it, horizontally and you lift it up, that is
your transformation matrices. Does it make sense?
That is the answer to these question marks. We will stop here. Have you understood
todays class? A little better understanding of yesterdays class, right? So, you will find
that, with subsequent classes this knowledge will get solidify but, you have to do some
reading on your own. Thank you.
Keywords: Basic Matrix Analysis, Displacement Transformation Matrix, Conventional

Stiffness Method, Reduced Element Stiffness Method,
Prof. Devdas Menon
Indian Institute of Technology, Madras\
Module - 3.5
Lecture - 21


Good morning. This is lecture 21. I hope to cover basics matrix concepts with this lecture so this
will be the fifth lecture on this topic. This is covered in the second and third chapters of the book
on advanced structural analysis. There was one topic which we did not cover in the last session
that is equivalent joints loads. You are already familiar with the need for this, you need this
whenever you have intermediate loads. We have used it when we did slope deflection method
and moment distribution method.
So when intermediate loads act in between the joints of a structure, they can be converted to
equivalent joint loads to facilitate formulating the load vector in matrix analysis that is because
we want to limit the degrees of freedom.
Ah so you have two types of problems, one where you have direct actions which are distributed
and you can find the fixed end forces you have done it for beams but, it is possible you might get
them in axial elements e.g. a column with self-weight that is a distributed load. We will see how
to handle that.
So this notation f i star f what does it represent in the stiffness method? In the i th element the
fixed end forces that small subscript f stands for fix end forces along the local axis, ok? so it is a
local axis system, right? You can get fixed end forces from another problem indirect loading.
Take a truss when you let us say you have a lack of fit problem in the truss or let us say you have
a temperature change problem that is indirect loading.
Ah how will you handle that? How will you convert into fixed end forces?
First find tension in that bar
How will you find that tension?
Elongation (( )).
Where is elongation? Let us say it is statically determinate then why are you taking that as a
fixed end force? Elongation will not give you the fixed end force it will give you elongation. Say
l alpha delta t is a free elongation you get it in a bar from that where is the tension there is no
tension in the bar yeah.
Then there is an elongation (( ))
Please understand the physics. I have a statically determinate truss let us say I am doing by the
stiffness method. I heat all the bars let them all get heated. Where do I get tension?
No fixed force no tension.
No there is you are right but, you are not able to give the physics behind the problem what you
said is right you might get tension or you might get compression that is a different issue but, you
will get an axial force, when? In which structure will you get an axial force?
Sir if it is determinate we can move the structure at a.
We are doing by the stiffness method that is what you learnt till now, right? In the stiffness
method how do you handle environmental change in bars in a truss for example? It is a little
tricky we will study it in detail in the next module but, I will give you a clue. All structures are
kinematically indeterminate even if they are statically determinate.
So we are always working as our base the primary structure. So in the primary structure no
movements are allowed at the joint (remove so) you have to arrest all the joints. Now you try
heating those bars you will get compression in those members or you cool those bars you get
tension in those members. So in the primary structure you get them and they get passed on to the
joints that is how you handle it in stiffness method it is very interesting.
So do you agree I can write down a set of initial displacements either contraction or elongation in
a truss and if I just multiply it with the elements stiffness of that I get the fixed end forces we
will see this in detail later. So you got the local fixed end force vectors for all the elements either
caused by direct actions that is intermediate loads or by indirect loading.
What do you do next in this conventional stiffness method which we studied? You have to
convert these to the global axis system, how do you do that?
Transformation matrix
What is the transformation matrix you will use?
(( ))
Will you use t i will you pre multiply by t i? not t i then what? (remove t i)
Tf
Dont mix up problems. We are doing conventional stiffness method t f comes in flexibility
method so it is easy to get confused. You remember the t i enables you a transformation from the
global axis system to the local axis system both for forces and displacements.
Now we are trying to switch from the local to the global so, what is the transformation you have
need to do?
T i transpose
There you are. t i transpose that is all you have to do. So add up all your elements fixed end
forces, apply the transformation in the reverse direction so that you would go back to the fixed
end forces and then you can generate the fixed end force vector for the structure by assembling it
in some appropriate way and you will realize you will get some fixed end forces coming to the
active degrees of freedom they are going to be your loads and some will pile on to the restrained
degrees of freedom they will add on to your reactions
So you pick up your fix equivalent joint loads from f f a. a is at the active degrees of freedom and
then you have to apply it in the opposite direction you remember in the classical stiffness method
and that is your equivalent joint loads. Im just brushing up what we have done earlier but, we
are doing it in a matrix format.

So the basis for equivalence is that the equivalent joint loads should generate the same conjugate
displacements at the joints on which they act as the original intermediate loads at these locations.
Remember we... I have shown you this earlier you have a original structure and you have first a
restrained structure where you arrest all the degrees of freedom and then you have a structure in
which you apply nodal loads.
The joint displacements here are zero because you arrested everything. The joint displacement in
the original structure are non zero so obviously if you do super position the equivalent joint load
should act in such a way that you end up getting the same unknown displacements in the original
structure at the joints not in between and that is the basic principle of equivalent joint loads.
If the displacements at these joints are arrested in the original structure then the resultant fixed
end forces generated at the artificial restraints when applied in an opposite direction correspond
to the equivalent joint loads. You have to add up these equivalent joint loads with whatever
additional joint loads that you may get, (remove ok)
Ok. Now we move on. We had a quick look at the conventional stiffness method there is no way
you fully understand it unless you apply to problems but, the idea here is to just lay road map
now we are going to look at the reduced element stiffness method we will see the flexibility
method after that they are similar so it is easy to see them.
If you recall, we know why we can take advantage of these reduced element stiffness method
because the rank of the element the conventional stiffness matrix at the element level is less than
full so you can reduce it to the minimum to full rank and weve already discussed three types of
elements we will see other types later but, this is elementary so we looked at the truss element
you have only one degree of freedom reduced.
We looked at the beam element two degrees of freedom. We have chosen just for convenience a
simply supported arrangement you can very well choose a cantilever but, let us not get confused
we will take one at a time and the plane frame element is a combination of the truss element and
the beam element so you have got three degrees of freedom and three displacements correspond
to one translation and two end rotations in terms of forces one actual force and two end
movements
Now you may ask what about the shear forces and all that? All that is statically determinate.
Once you got the end movements you got the shear forces, right? So we also looked at this recall
we talked about the displacement transformation matrix and we derived this from first principles
using what we call a kinematic approach. Ill show you, you can also get the same matrix the t d i
matrix from a static approach using this principle called the contra gradient principle which we
are now going to look at.
But before that let us just do an application of the t d i matrix. Let us take this frame which we
have been playing along all this while you have got this three element frame the primary
structures are arrested we want to if you push d one equal to one what and if you have to
generate the t two d a matrix which are put it with a circle here.
How will you get this matrix? What will be the size of this matrix? Im just taking the second
element which I have shown here. Second element has three degrees of freedom one two and
three. When I apply a unit displacement to the structure here arresting all the other displacements
this will effect elements one and two. Just see the effect in element two. how will you convert it
to whatWhat effect will it cause?
It will cause a compression in that member, agreed? And remember you have only one degree of
freedom in that element in the reduce format and that displacement is what? We said it is an
elongation, right? so we want to get that quantity what will it be? It will be. it is a component
of one you know would not it be cos theta? Will it be plus cos theta or minus cos theta? minus
cos theta, right? And will you get so we have got this value d i one star is going to be minus
cos theta. What about d i two star for element two? Will you get any moments? Will you get any
rotations? No because you are arrested that here. You wont get anything. There will you get any
rotation here? No. ok(remove ok)
Ok. Youll understand when I show you the matrix but, at least tell me the size of the matrix.
What is the size of t two d a matrix? You remember there are how many degrees of freedom here
in this problem? There were eight active degrees of freedom and four restrained degrees of
freedom, right?
So this is (remove this is) the matrix will look like this. Ok I am just pulling out the t two d a
matrix here with the circle. For this element if I have to generate the matrix the first column this
is a three by eight matrix first column is what I write down when I apply d one equal to one to
this structure. Do you agree? When I apply d one equal to one to the structure I get just a
contraction in that element and so corresponding to this first degree of freedom I get minus cos
theta of that element two. I do not get a rotation at the left end. I do not get a rotation at the right
end.
So do you understand the physical meaning of first row, second row, third row. The first row
refers to what I get here when I keep applying a unit displacement to the structure. The second
row refers to the rotation here. The third row refers to the rotation here. Likewise I can move
ahead and go to apply d two equal to one which means I now lift up this joint and keep
everything else restrained when I lift up this joint what do you think i will get? I will get minus
sin theta.
Do you understand or you do not? You understand. Do not worry we will come back to this later.
Why the minus sign?
Because you will get a contraction, the length of the material well that question you should
have asked here. Why did you get a minus sin here? See when I push this horizontally here I got
a contraction so this was a minus sin. When I lifted it up then also the length reduced, I got a
minus sin. But when I pull the right end here right I get a plus sign because you can see the red
line is elongated and when I lift it up at the right end I get elongation, right?
So what I have written here, the displacement transformation matrix gives me the element
elongations caused by a unit movement, unit translation at the left end and the right end keeping
all the other degrees of freedom, clear? So I get minus whenever I push the left end and I get
plus when I push the right. Is that clear now? Ok? I do not seem to. Yes or no?
yes.
Give me a resounding yes if you understood. But do not worry with time you will understand.
So broadly you get the picture hang of this matrix. now this (remove now this) When you do
reduced element stiffness method and you would do flexibility method you cannot program it as
blindly as you could the conventional stiffness method because you are trying to take a shortcut
you have to apply your brain a little bit especially in generating the fixed end force vector which
we will see later, so you can do this.
Ah now let us look at the contra gradient principle. I have shown this earlier but, here at the
lower level I put the element level forces, the combined element force vector f star which
actually refers in a case of a truss to all the axial forces in the truss. The combined element
displacement vector d star which are the bar elongations in the truss. The structure level
displacement vector which are the joint displacement in the truss and the structure level force
vector f, ok?
F I can include both the active degrees of freedom and the restrained degrees of freedom or I can
limit to the active degrees of freedom. Now the transformation we did was, if I know the joint
displacements then I get the individual member deformations in this case the bar elongation by
applying the t d matrix as a transformation I just demonstrated that. And this actually gives a
compatibility relationship between the structure joint displacements and the member bar
elongations, right?
Likewise there is a set of equilibrium relationships which relate f star with f. and.. Do you think
this relationship is linked with the t d matrix? Can we prove that relationship? How do you prove
that relationship?
This is a very important concept. How do you relate the element level forces to the structure
level forces in the way we related the element level displacements to the structure level
displacements? What principle will you use? Ah?
Principle of super position like
But whenever you are doing transformations of matrices you are using principle of super position
whether you like it or not. that is how the. It is a linear combination of vectors. So that is
implicit. What is the answer you gave?
(( ))
But you have to talk matrix language. You are just saying what I wrote here. I am asking you.
How do I build a relationship between this and this? There you are it has come out.
Using this. So you use the principle of virtual work your real work is that principle which ties
together the force field with the displacement field. So do you realize, the principle virtual work
says the total external virtual work product is equal to the total internal virtual work product?
Now, here all the forces are real, all the displacements are real but, it is the product is still virtual
because I am not applying any factor half. I am just taking the forces. So do you agree to this
equation that I have written here? That if I take the structure level forces and find the total work
done, it is simply given by either d transpose f or f transpose d similarly, if I take the internal
virtual work, I can write either d star transpose t f star or f star transpose d.
Now we will use the other one in the flexibility method. Now, how to proceed after this because
you will end up with a beautiful equation. What can we now substitute in this equation?
Instead of d with ah
Instead of.?
D star.
No, that is not what you need. You have to bring in t d into the picture, man.
(( )).
Equal to t d.
Ah
T d into d star.
You replace d star with?
T d into d.
T d into d that s that is how you bring into picture and you have to take the transpose of it.
Now a b the whole transpose is b transpose a transpose, right? So do not you get this nice
relationship? Then what? What is this establish?
(( ))
Then you keep your eyes open and compare the left hand side with the right hand side and then
you arrive at a simple equation.
F is equal to.
F is equal to t d (( ))
So what does this mean? This means that is your relationship. That is a powerful relationship.
That means once you (remove would) establish the t d matrix, you are actually tying you
(remove one you) are also linking simultaneously a relationship between the force vector at the
element level to the joint force vector. In terms of a truss what we are saying, if you know the bar
forces in the truss then the joint loads that could have caused those forces are related in this way
but, it is not a very useful relationship in the sense of normal problem is, you have unknown bar
forces when you have known loads. Ok?
The real problem is that but, this is useful in building up the force transformation matrix. Now
because you are doing it in the reverse direction you know this was the normal direction for t d,
you are (remove one you are) moving in that contra direction, the slope is moving in the
reverse direction hence the word contra gradient. Gradient is the slope.
You know normally things flow downhill but, here it is going to flow uphill when you are
dealing with the force field. Ok? and yeah.
(( )) x star into t d into f star
Sorry.
Similarly you can write f star equal to t d into
You cannot do all that.

But t d is not (( ))
You cannot do all that. See in the stiffness method t d is unique. From first principle we derived
it. What you are saying is something we will do in the flexibility method. There we will bring the
t f matrix.
Right now, do not jump ahead because you will make mistakes. Do it slowly. Is this proof crystal
clear? It is a simple proof beautiful proof. Let us apply this proof but, before that a statement.
The contra gradient principle as applied in the stiffness method of analysis states that if in any
structure displacement transformation matrix t d can be established based on compatibility
conditions to transform the nodal displacements in the structure to member displacements d star
equal to t d d then the transpose of t d serves to transform the member forces to the nodal forces
acting in the form of loads and reactions on the structure. f is equal to td transpose f star. Ok.
Remember this problem? We wanted to find the t d matrix. Well, if you use statics you can find
the t d transpose matrix then you take the transpose of the transpose you will get the t d matrix.
right? So now if I look at this relationship, f star in this you have four (remove a four)
components here, right? f i one, f i two, f i three, f i four this force internal force here is actually
n i you know that is what I called f that is what that is what(remove one that is what) the
element force there is, right?
Actually this is written the wrong way, yeah the f i is four by one, right? Then what you get here
what you get here (remove one what you get here) these are the four vector joint that I need.
That means if I take this structure and imagine I have a bar and have a unit axial force in that bar.
What are that load components that I get at the two ends of the bar which will which will
(remove one which will) satisfy equilibrium, that is a static approach. Can you tell me what
these four components what will be the what should this force be equal to?
(( ))
Plus or minus?
Minus.
Because there is axial tension.
Minus. Plus. Plus. Minus.
Then how will you satisfy sigma f x equal to zero?
Minus.
What about here? This one? It will be plus or minus?
Plus sir.
And what will this be?
Plus.
Plus, what?
Sine.
Sine theta. What will this be?
Minus.
There. I hope it does not take too long.
So you you (remove one you) see you have got this. That is your t d transpose matrix, you take
the transpose of that you get t d you see, very clever way of getting the t d matrix in the reverse
direction. This is statics, previously we did kinematics. You have a choice. Normally, students
are weak in kinematics. They love doing sigma f x equal zero sigma f y equal to zero. You can
get it this way and take the transpose you get the t d matrix but, if you are good in geometry you
can do that. I want you to be good in both and you can do either way you are. Is this clear?
We will see examples of this in trusses later. Now you got you got (remove one you got) the
stiffness matrix at the element level. You got the t d matrix which relates the element level
displacements to the structure level displacements. Our job is to find the structure stiffness
matrix.
Remember we were not very comfortable with the conventional stiffness method where, you
took put the linking coordinates and you have to put it in the slot and so on. Here, you will have
no problems because you got this relationship. This is t d. Now, you got the same playground.
Everything is familiar to you. You got a new beautiful relationship from the contra gradient
principle which is what? Which is this.
Does this make sense to you? The structure level on top, the element level below. t d flowing
downward for displacements, t d transpose flowing upwards for forces, then take the diagonal.
The diagonal just says that, if you tell me what the displacements are in a truss joint
displacement immediately I will tell you what the bar forces are by doing this transpose. This
matrix.
Remember I put a k tilde to to (remove one to) remind myself that I am dealing with the
reduced stiffness matrix, so agree? So this is the unknown. Bar forces are nothing but, this times
this times that. Is it clear? Then if I want to take this shortcut directly from d to f what do you
think k will be? k will be t d transpose k tilde star t d.
So you will find repeatedly this kind of pattern emerging you have to pre multiply the stiffness
matrix whether it is a regular stiffness matrix or the conventional stiffness matrix by the
transpose of some transformation matrix in this case t d and also post multiply with t d. So that is
that is (remove on that is) what you do. When you do this you will get the full stiffness matrix
without any effect. So that is to sum up the reduced element stiffness method. We will go
through the details later. It similar to the conventional stiffness method. The steps are similar but,
you will really understand this only when we do some problems.

So we are now set to move on to the last topic which is the flexibility method. The first problem
you will notice the main [dis/difference] (remove this bracket) difference between flexibility
and stiffness method is that this (remove this) you have even at the global coordinates level
you have to be careful. Why? because here you have a choice of redundance. Now this particular
structure, is it statically indeterminate or determinate?
Indeterminate, sir.
It is determinate?
Indeterminate.
Indeterminate. What is the degree of indeterminacy?
One.
One. Which would you like to choose as a redundant?
The roller sir (( ))
The support is not redundant is a force. Ok? you will so corresponding to that redundant
support you want the primarily structure to be cantilever. Say that. That means, you want to
get rid of that roller support, right?
So in terms of statics you have to replace it with a force, right? And that is your single redundant
here. So as far as your external loads are concerned there is an one to one similarity between
flexibility method and stiffness method. You have eight degrees of freedom active degrees of
freedom I have put the same numbering one, two, three, four, five, six, seven, eight. I am not
showing the reactions yet. Actually flexibility method is more suited for manual analysis, so you
can always get the reactions once you got the free body.
So you can always get the reaction once you got the free body (remove so.body) So for the
present let us not worry about reactions but, I will show you can find reactions if really want to
also. We will skip that. So what do you do, remove it. This is your primary structure and to
complete the picture introduce introduce (remove on introduce) your redundant as the next
coordinate. So you put the f x the redundant forces. Here there is only one supposing at three so
youll put nine ten eleven as your degrees of freedom in addition to the active degrees of
freedom we will call them redundant degrees of freedom, ok?
So for example, if you had this problem which i showed you earlier, then v d is your redundant x
one and this is how you would write the force vector, right? The first part the f a part we have
seen earlier. You are just putting f nine now. f nine is your additional unknown redundant, ok?
You need to pay attention right (remove youright) So f vector is f a f x. d vector is d a d x .so
this is the major departure from the stiffness method because the stiffness method we have d a d
r, f a f r. in flexibility method for statically indeterminate structure we have f a f x and d a d x as
a force vector displacement vector. does it make sense to you? Right? Ok.
In these vectors what is known and what is unknown? Well, fa is known these loads are usually
given to you f x is unknown this is the redundant you need to find. d a is not known and d x is
known because d x gives you youe (remove youe) compatibility conditions right? So what is
your compatibility condition? Well, in this particular problem d x is having only one element d
nine and it is zero but, suppose you add a support movement settlement then it will not be zero,
youll put the correct value. We are familiar with this we did method of consistent deformation
so it falls into place so far so good? Ok.
Next point if you want to find support reactions you can do it. We will see it later. In the
flexibility method of matrix analysis we can replace supports by appropriate rigid links so the
force in a rigid link becomes your you convert the external reaction to an internal force in an
imaginary rigid link. It is a little trick you can do but, it adds to your size of your matrix. You
might not find it convenient to do in in practice. Ok.

Similarly we can assign support displacement if required by way of initial deformations in the
rigid links. Ok we will see that later. So now we are going to look at what one of you asked
know? Can we do this transformations? So this is our t f transformation. Remember we did t d
transformation we looked at t d transformation in the stiffness method both the conventional
stiffness method and the reduce stiffness method. Now you do not worry about t d you look only
at t f and we have based (remove webased) obviously t f must be based on based on (remove
one based on) equilibrium alone.
So you can uniquely define t f only for a statically determinate structure because you have a
unique statically admissible solution. For a statically indeterminate structure you cannot uniquely
define t f please note that. Demonstrate with an example, very familiar example. This is plane
truss. You identify the active degrees of freedom. There only five of them, right? One, two,
three, four, five. Ah (remove ah)
You have one element. Remember you done the reduced element stiffness method. The
flexibility method follows the same as same (remove as same) system as the reduce element
stiffness method. So what is what is (remove on what is) your unknown displacement in that
element? It is a bar elongation. What is the unknown force in the element? It is a bar force.
Tension assume positive. Elongation assume positive. Ok (remove ok)
Ok. So that is your degree of freedom. We want to write down the t f matrix. We want to write
down the t f matrix, which is f star is equal to t f into f a. How do you write the t f matrix?
(( ))
What do you need to do to write the to generate the t f matrix?
Apply external
Apply?
Unit load.
Unit load. How many of them? One at a time. How many of them? You answer the question that
I have raised.
(( ))
Whats the size of t f matrix?
Five by five.
Five by five. by five (remove by five) so far so good. How do I get the first column in the t f
matrix?
(( ))
Apply unit load. Where?
(( ))
At one. So it is called apply f one equal to one. Then how do I get the second column?
F two equal to one.
F two equal to one. So I apply so I have to apply five times and one at a time. is it clear?
Lets do one of them. So this is my first transformation matrix. Five by five, it turns out to be a
square matrix. It is a unique matrix, it is interesting. So f a has five components, d a has five
components they are referred to the joint movements, right? You know what f one is, you know
what d one is. f two, d two etc. This is external. What about internal? f star has also five because
there are five bars, right?
Now I have written here a f one star, f two star, f three star but, you can write if you want f one
one star because really your degree of freedom is one star. So whether I write just star here or
one star here in the case of a truss it does not matter you can choose either. Ill show you
alternatively how you can do it.
So this is nothing but, the bar force and the internal d one star is nothing but, the bar elongation,
Do you understand what this means? Everything make sense? Now we need to generate the t f
matrix so let us do that. So let us take that matrix apply f one equal to one. Can we quickly
analyze this? Are there any zero bar forces here? By inspection tell me which bar force ah.
One, four.
One and four. It is quite fast. well look at the support d roller support from there you can make
out one of those two bar forces zero which is zero? Four is zero, right. Then you go up to the
joint b, which is zero? One is zero. Then what is the force in bar five? One. Is it tension or
compression?
Compression.
Its compression so minus one. Then you go to joint c, what is the force in bar three?
(())
The component the horizontal component of the axial force in bar three must be equal to one.
So how much would plus or minus?
Plus. Plus.
Plus. Very good. Then the vertical component of this must be the force in bar two. How much is
that?
Four by (( )) minus four by three.
Solved. Like this you do five times you can get that matrix and some of them are really easy. that
is what it looks like so do not get scared t f matrix is very comfortable because these we were
raised on statically determinate structures that is of orders so it is very easy to generate the
especially the second one f two equal to one has only one element. ok.
So you understand what t f matrix is. Good. Now let us take any element in that. That element
we used to call n three one remember small n for unit loads. n three one refers to the force in bar
number three caused by f one equal to one. Remember, we have used this notation earlier. It is
called the first transfer coefficient. Why it why it (remove one why it) is called transfer
coefficient?
(( )) it transfers that much load to the (( ))
Yeah it is like all these five bars form on organization a partnership and whatever money comes
or goes out to that you know it is an immediate transaction online bank account some money
come in some fraction of it goes. So that is exactly what is happening, you apply a load f one
equal to one on the structure. Can you see what I have done? You apply f one equal to one
immediately you got a transfer n three one to bar number three in this case five by three but, do
not take that analogy too far because you are actually transferring more money than that came in.
Right?
So this is the meaning of force transfer coefficient. Right. the force transformation matrix is a
unique matrix satisfying static equilibrium enabling structure level forces or loads to be
transferred in inverted commas to element level forces in a given statically determinate structure
in the case of a statically indeterminate structure remember it is not possible to define such a
unique matrix because multiple statical admissible solutions are possible.
So how do you handle an indeterminate structure? Primary structure.
(( ))
Primary structure. Primary structure is statically determinate. And apply?
(( )).
Apply super position. Right, that is what we did in consistent deformation. So that is what we are
going to do. So here again you have a contra gradient principle and actually have (remove
have) a principle which is the one that proves why you do not need to satisfy explicitly
compatibility in statically determinate structure. you remember, all structures have to satisfy
three requirements. What are the three requirements? Equilibrium, compatibility and force
displacement relationship which include the stress-strain relationships, constitute relationships.
Statically indeterminate structures equilibrium alone cannot help you get the unknown forces, we
have force to look into the other two. Statically determinate, you are getting away we just
managing looking at equilibrium. The question arises a good student should ask this question,
does it mean somehow automatically compatibility satisfied? Yes, prove it.
So the contra gradient principle actually proves it. How do you prove it? We we (remove one
we) did a similar proof earlier we did principle of virtual work. Can we do the same principle?
Then we will get the same old result. No, there we proved t d transpose here we have to prove t f
transpose.
So how do you use the same principle? You flip what?
(( )) f star n a
Ah you (remove you) there we did d transpose f now we do f transpose d. Now because we are
keeping reactions out of this game anyway reactions do not do any work if the supports you are
move. We have f a transpose d a, agreed? is equal to f star transpose d star, agreed? Total
external virtual work product is equal to total internal virtual work product. What do you need to
substitute now?
F star equal to t f.
There. Now you are good. This is simple logic. What do you end up proving by comparing?
(( ))
D a is stable.
(( ))
Fantastic. So without your knowledge when you are satisfying equilibrium in a statically
determinate structure you are also simultaneously satisfying compatibility, provided your
structure is statically determine. Ok.

The contra gradient principle, as applied in the flexibility method of analysis, states that if in a
statically determinate structure, a force transformation matrix t f can be established, based on
equilibrium conditions, to transform the loads on the structure to the member forces f star is
equal to t f into f a, then the transpose of t f star servers to transform the member deformations to
the active structure displacements. d a is equal to t f transpose t star.
This implies that the same equilibrium conditions serve as compatibility conditions in a just
rigid structure, with a difference that t f pertains to a transformation from the global coordinates
to the local coordinate, while t f transform transpose pertains to a transformation in the opposite
direction in the contra direction in the contra gradient direction that is from local to global when
you are relating displacement. That is it.
So you see you will never forget the contra gradient principle. It has two two (remove one two)
parts to it, one part when you find t d transpose that is in stiffness method, the other part is when
you find t f transpose. You will you will (remove one you will) be very comfortable with
flexibility method. Unfortunately, our focus is not the flexibility method, because the computers
do not to do that method for reasons we will see shortly.
Lets see the implication of this t f transpose. We have done this, right? Now what are the
implications? Let us take a bar in that truss, the third bar. For the fun of it let us heat that bar.
How does this t f transpose matrix help us find find what?
(( ))
Ah
(( ))
Displacement at c? If I say e three initial that is the initial displacement. It could be a lack of fit,
it could be heating that bar by say one mm. What do I end up finding? Let us take n three one.
Five by three. We did a bank transfer recently, right? for the forces. What is five by three going
to be? Displacement, it must be a displacement. Where? At b or c?
(( )) at c
Well you (remove you) what you need to do is to see. Where all the bars will move? I have
heated you know, the compatibility will satisfy. Remember this five by three is the force you got
in bar three when you applied f one equal to one. So when you change the length of the bar three
you will get the displacement at one. Where is one where is f one? D. This much. So it is
fantastic. Where (remove where) You derived from statics of matrix you can get that matrix to
get joint displacements. and let us give a (remove and...a) So d a is t t f transpose d star let us
give another demonstration of it. Same truss. Now I have a lack of fit. I arbitrarily have different,
you know? It it (remove one it) came to the site the bars came with different lengths with
(remove with) that is how bad manufacturing is.
So bars one and two are found be two long by one point five mm, bars three and four two long
by three mm, bar five too short by one mm, find the various joint displacements. Well, most
people will have to necessarily either do this by geometry or actually do the experiment and then
measure how much it moved.
But if you are intelligent, if you are brilliant, you can invoke this beautiful contra gradient
principle and get it in a jiffy. How do you get it in a jiffy? Well, first you have to convert these
into the language of matrices. How will you write that? You have an initial displacement element
level matrix which (remove matrix which) vector which you can write like that. i n i short for
initial. Do you agree? I put plus when I have a larger length put minus when I have a shorter
length. Does it make sense? Right?
Then I invoke this powerful principle. these are my (remove thesemy) This is my t f matrix
and I take the transpose of that matrix I have got all the answers in one shot. If I am using
matlab. Correct (remove correct) You you (remove one you) have to agree matrix method is
very powerful, if you know how to use it, when to use it and when not to use it. When not to
abuse it, when not to misuse it and in practice the misuse and abuse is wide spread. Ok.
So if you compare now the two transformations, you get another interesting result for a statically
determinate structure. t f transpose inverses t d a and t d a inverse is t f transpose. So you can get
a potential question in the exam where I give you one of them and ask you to figure out the other
you have to bring the same.
The contra displacement matrix that is a name I have given but, they remove (they) it is not a
very common name. You can say t f transpose matrix which is based on a flexibility formulation
for a statically determinate just rigid structure and the active displacement matrix t d a which is
based on a reduce elements stiffness formulation for the same structure are both square matrices
describing compatibility relationship between the local and global coordinates one being the
inverse of the other.
Flexibility matrix definition you know. Just like the stiffness matrix, each element is called the
flexibility coefficient. f i j is defined as a displacement d i generated at the coordinate i on
account of a unit load at the coordinate j that is f j is equal to one with no other loads acting on
the structure. In stiffness method, we said the same thing as it with, all other displacement
arrested then only the definition is complete. You can write it this way. You can put all the
elements together this way. Ok?

Its called the unassembled, With (remove with) the diagonal form and you need to get the
structures flexibility method. Now it is simple. We have played this game so many times, it is
not difficult. You know the element flexibility stiffness is is (remove one is) just the inverse of
what we did for the reduce element stiffness matrix and so this we know if you expand it will
look (remove itlook) it will have a diagonal form like this. Remember, we did it recently.
And then you can either write equations, if you have fond of writing equations or you can play
this game.
I i(remove one I) suggest you write this game. It is very easy, right? First at the element level f
star is now you see the direction is moving to the right because flexibility is always opposed to
stiffness, they are two different paths. So if you know the forces, element level forces you
multiply by the flexibility matrix, you get the element level displacement, right? In in (remove
one in) a bar you multiply the bar force in a truss, multiply the bar force with its flexible
(remove flexible) axial flexibility you get the bar elongation, right.
At the structure level,you need to find f, relating f a with d a. how do you get f from f star? small
f. I have put small f because capital f goes to force. Well, let us do it the right royal . We will
play the game by its rules so first this this (remove one this) transformation. It flows this way
and on the left side t f transpose and contra gradient says this flows this way. You take the
diagonal it will flow this way and you take the shortcut it goes that way. That is your proof t f
transpose f star t f is your structure flexibility matrix.
If you got these matrices with you, you do not multiply those matrices. put(remove put) Give
kill (remove kill) that damn computer the job to do it. Press the button get your f matrix. So
you got statically indeterminate structures. Here the problem is d f cannot be uniquely defined.
So what you need to do? Primary structure. For a chosen primary structure d f is unique, what
does it mean? Each student in this class can arrive at the different t f matrix so it is difficult for
the examiner to check the results. So he has to wait for the final results to come out.
But whoever does it has to follow the same path. Now I say t f a and t f r. I you know (remove
you know) have partitioned it. t f a refers to that part which is dealing with the loads on the
primary structure and t f r. Sorry its t f x is that which deals with the unknown redundant and a
(remove a) you can do it at the individual element level or you can take the combine level.
Right?

So you can write down these equations. They are compatibility equations. We will study them at
(remove at) in great depth. I just want to rush through it because this is just a road map and you
can write down the procedure just as we wrote the procedure earlier for the flexibility method. I
want to conclude with this last slide summing up all that weve learnt till now.
Broadly, stiffness and flexibility method differ in some some (remove one some) terms. What
are the basic unknowns in stiffness method?
(( ))
In Stiffness methods, they are forces?
Forces (( ))
In stiffness method, the unknown are?
Displacements.
In flexibility method, the unknowns are? This is childs play. Ok.
Next one. What is the type of indeterminacy in stiffness method?
Kinematic
Kinematic verses.
Static. static. (remove one static)
Brilliant. what is a unique (remove whatunique) Do you have a unique primary structure in
the stiffness method.
No.
Do you have a unique primary structure, that means will all the students in a class...
Yes.
Yes, what about the flexibility method?
No.
Well depends. if all of you (remove ifyou) If all fools think alike then it will be unique but,
otherwise potentially it will be different, right.
What about the force displacement relation format? What is the format you have in stiffness
method?
f is equal to kd
Yeah but, if you want to write it in word, what is a format?
f is equal to kd
Its a stiffness format and flexibility format, come on. Is the matrix that you get the stiffness or
flexibility always well-conditioned? Yes, in the stiffness method.
No
Well, sometimes yes though (remove though) but, you cannot say always, right?
Governing equations?
Ah compatibility.
Compatibility in which one?
(( )) flexibility method compatibility.
Equilibrium in stiffness. Compatibility in flexibility. Lastly how do you deal with the
intermediate loads?
Equivalent loads
In stiffness method, equivalent joint loads. In flexibility method?
(( ))
No no (remove one no) on the computer. Your number of degrees of this is the real this is
the achilles heel of flexibility method. This is what destroys the beauty of the method you have
to borrow the same damn concept from the? It is a tragedy because there is no other way you
can convert to that means you have a you have a (remove one you have a) statically
determinate portal frame with the uniformly distributed load load (remove one load) on top you
have to pretend that, that beam is fix-fix. Make it statically indeterminate and borrow from
stiffness method how they calculate the equivalent and then put it here. That is a problem. So
flexibility method is fantastic for trusses and not particularly great when you deal with
intermediate loads.
Thank you.
Keywords: Basic Matrix Analysis, Displacement Transformation Matrix, Conventional Stiffness

Method, Reduced Element Stiffness Method, Equivalent Joint Loads

Basic Matrix Concepts - Devdas Menon

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Basic Matrix Concepts - Devdas Menon

Caricato da

Copyright:

Formati disponibili

Advanced Structural Analysis

Prof. Devdas Menon

(Refer Slide Time: 00:50)

(Refer Slide Time: 01:15)

(Refer Slide Time: 03:04)

(Refer Slide Time: 03:04)

(Refer Slide Time: 09:15)

(Refer Slide Time: 11:06)

Right. By definition: a matrix is a rectangular array of elements arranged in horizontal

When m is equal to n, what do you call that type of matrix?

Square matrix, very good.

What are these matrices called?

[Noise] (Refer Slide 16:08)

(Refer Slide Time: 16:58)

What about a vector? Is it a type of matrix?

Yes sir, it is a matrix.

So, vector is one type of matrix. What is a vector?

What is basis? Sorry, one at a time.

Linearly independent vectors, which can span the space.

Why do you need to get a basis?

[Noise] (Refer Slide Time: 19:37)

(Refer Slide Time: 22:01)

(Refer Slide Time: 23:39)

(Refer Slide Time: 23:39)

Linear simultaneous equations can be written as AX equal to C. A is called the

What is the rank of the matrix A? It is a beautiful concept.

Number of independent we can find out.

(Refer Slide Time: 31:35)

Yes or no? Yes, sir.

(Refer Slide Time: 34:29)

Make what, one?

Ok, that variable, so that is called the pivot.

Can this set of equations have a solution? Yes or no.

Infinite number of solutions.

Infinite solutions. Will they always have infinite solutions?

(Refer Slide Time: 43:34)

(Refer Slide Time: 43:47)

Statically indeterminate structure.

(Refer Slide Time: 46:06)

What is a problem with some matrices regarding finding inverse?

They are non-singular.

They should not be singular.

(Refer Slide Time: 47:54)

What is a determinant of a matrix? Why do you need to know it?

Whether it is invertible or not.

(Refer Slide Time: 49:26)

(Refer Slide Time: 49:37)

(Refer Slide Time: 50:32)

(Refer Slide Time: 50:42)

Maybe we will finish this.

Thank you. We will stop here.

(Refer Slide Time: 00:35)

(Refer Slide Time: 01:06)

Other elements multiplied by

Sir, directly we can get it.

It is a vector, which is parallel.

(Refer Slide Time: 08:37)

(Refer Slide Time: 11:10)

(Refer Slide Time: 12:48)

(Refer Slide Time: 14:14)

(Refer Slide Time: 16:22)

(Refer Slide Time: 21:27)

(Refer Slide Time: 23:34)

(Refer Slide Time: 35:18)