Dynamics of Machines

Prof. Amitabha Ghosh

Department of Mechanical Engineering
Indian Institute of Technology Kanpur

Module-13 Lecture-5
Approximate Method of Vibration Analysis: Rayleigh - Ritz Method

It will be better explained with the help of an example.

(Refer Slide Time: 00:31)

So, to compare the values what we get by this approach, that is Rayleigh's method by
Grammel’s modification, let us take up again a simple problem of a cantilever beam
uniform for which we know the answer. Our objective is to compare the Rayleigh's
method and Rayleigh's method using the Grammel’s modification. Just to demonstrate
that, we will take a simple [example] and we can also compare the exact value of the
answer.
(Refer Slide Time: 01:03)

Let us take a cantilever beam of uniform cross-section and of length L and they are all
uniform. Now, we assume the deflection satisfying the condition, at least the geometric
condition will have to be satisfied to know that, so we just take it like this: Xa is equal to
cx square. So, the slope is this and x we do not need that anymore. So, let us see… of
course, we can do that. At x is equal to 0, Xa is equal to 0 that satisfies the condition. The
slope is also 0, that is another geometric condition that is also being satisfied. At the free
end, we are supposed to have this as 0, this is 2C, this is 0; but, it should have been 0,
because, there is no momentum acting here, but it is not satisfying that condition, but
anyhow it is satisfying the geometric boundary condition, which are very important and
they must be always satisfied.
(Refer Slide Time: 03:30)

Now, with this what will be the inertial loading? Inertial loading is x square. It will be
like parabola and we consider an element of length d zeta, at a distance zeta. Now, let us
find out the moment of this inertia force because of this, produces how much bending
moment at distance x from the origin? So, what is the inertia force of this element? That
is equal to the mass of the element that is A into d zeta into rho. Then, omega square into
Xa square; now Xa is this part, so it will be C x square. So that is the inertia force.
Moment of that at x is equal to X, due to this element will be omega square A rho C x
square d zeta into zeta minus x.

So, this is a force, inertial load and this is the distance. So, this will be the moment and
since this moment or the positive direction of moment was like this (Refer Slide Time:
06:53) and for that the slope of the neutral axis, as x increases, gradually decreases. Here
it will be gradually increasing. So, they will be in negative. Therefore, total moment will
be the moment of all the inertia loading of all the elements from zeta is equal to x zeta is
equal to L. So, therefore, moment as a function of x at distance x will be: minus omega
square A rho C integral x to l x square zeta minus x d zeta. So, we have taken coordinate
[08:11 min] zeta is the variable not x, x is a parameter; at zeta is equal to x what is the
[08:31 min] if you want to find out [08:35 min]. So, this if we estimate, we will find this
is: minus omega square rho AC by 12 into x to the power 4 plus 3L to the power 4 minus
4xL cube. Now, if we use this: omega square is equal to M square by EI dx by x 0 to L,
again now x is the variable 0 to L rho AX square dx. If we now use this M here and x, of
course, is we have assumed like this then this will lead…. Now of course, if you
remember that omega will also come in this side, because, of this twice omega to the
power four. So, when you solve the whole thing, we will get omega square is equal to
12.46 1 by L to the power of 4 EI by rho A or omega is approximately square root of this
3.53 1 by L square root of EI by rho A. So, this is the result we get using the Grammel’s
modification that means we find out the moment at x due to the inertial loading which is
caused by an assumed deflection Xa and substitute it here in the original equation of
Rayleigh. Then solve the equation for omega square because, omega square will be on
both sides and we get this (Refer Slide Time: 10:54).

Now, if we try to find out with the same function or same displacement function Xa just
by following Rayleigh's method and if we take, then what we get?

(Refer Slide Time: 12:22)

Omega square is simply 0 to L EI d square Xa dx square, whole square dx by 0 to L rho A

Xa square dx. Now, we know dx square dXa is 2C; so this will be simply 4C square into
EI dx that is 1 by L - that is the numerator. What will the denominator be? The
denominator will be: rho A x square that is C square x to the power 4, if you integrate it
will be simply L to the power 5 by 5. So, that is 20 EI by L 4 rho A or omega is
approximately square root of 20 is 4.47. So, this is the value we get by using simple or
straight forward Rayleigh's method. For a cantilever beam, we have already solved by
exact method. The exact value we already calculated earlier it was equal to this (Refer
Slide Time: 13:18). Now, if you compare, this is the exact. I am using this simple straight
forward assumption about the displacement function dx square, the Rayleigh's principle
gives this. Even in the same assumption about this, if we use the modification proposed
by Grammel, we get this. Now, you can see that how close it goes to the exact value.
Therefore, the accuracy can be substantially improved in the Rayleigh's method if we use
the moment expression directly, which is determined by finding out the inertial loading
based on the assumed mode shape and that results in substantial improvement of the
value of the [14:30 min].

So, I think this will be enough for the time being, but we have to now keep in mind that
whether it is possible to find out the higher mode or higher frequency. All these cases,
this Rayleigh's principle is valid in case of fundamental frequency as we know.
Therefore, all the time we are assuming something for the first mode deflection, if we can
find out the second mode shape and put it there, obviously, we will get the second mode,
but most of the time you see that it is convenient to solve the problems satisfying the
boundary conditions, it will give us the first mode frequency. So, in the next presentation,
we will try to see if it is possible to get a methodology by which we can have some
approximate value of the higher frequency.

So, following the approach proposed by Rayleigh and even considering the modification
suggested by Grammel, it is always found that we get the fundamental natural frequency
or the first mode natural frequency. However, when it is required to find out higher
frequencies or higher mode frequencies, it is possible to devise a method of this
approximate nature following the same principle and employing one of the important
properties of Rayleigh's quotient.
(Refer Slide Time: 16:29)

We have seen that Rayleigh's coefficient is equal to this (Refer Slide Time: 16:40). So,
this is the maximum potential energy during normal mode oscillation and this is the
maximum kinetic energy divided by the natural frequency square. So, this is the
Rayleigh's coefficient and with this Rayleigh's coefficient as the natural frequency,
omega square, square of that. The maximum potential energy expression will depend
whether it is a transverse oscillation in that case it becomes this (Refer Slide Time:
17:44). If it is longitudinal oscillation, then again its expression is different, we have seen
that. So, in general you can keep it as the maximum potential energy during normal mode
oscillation divided by this quantity.

Now, there is one property of this Rayleigh's coefficient which says: R is stationary when
Xa becomes an exact mode. So, whatever we may assume for this displacement function
or the deformation pattern or mode shape whatever you may say, R of this quantity will
be always higher than a quantity which will be minimum or stationary as Xa becomes a
real mode, that is exactly it coincides with the natural mode. So, therefore the omega
square what we will get which we associate with this, this omega square will become
lower and lower as Xa becomes a better and better approximation and when Xa becomes
in-distinguishable from the exact mode, then the real natural frequency will be the lower.
So, this property of Rayleigh's coefficient is used and this proposal was first suggested by
Ritz. The whole process will be also called Rayleigh Ritz process of Rayleigh Ritz
method.

(Refer Slide Time: 19:58)

So, in Rayleigh Ritz method, this assumed deformation shape is expressed in the form of
a series or in terms of a number of functions. So, where, each Xi satisfies the boundary
conditions as much as possible. Here, better approximation means satisfying more
boundary conditions, better results will be obtained and at least the geometric boundary
conditions. Now, if we do that, then what will happen if we substitute, this is also a
function dependent on quantity depend on Xa and derivative; here also it is dependent on
Xa; so as a whole when this quantity is found out, so, the Rayleigh's coefficient will be
equal to this (Refer Slide Time: 22:43), for transverse oscillation in beam mode.

So, when you substitute this, this becomes a function of C1, C2 up to Cn. So, this Xi's are
definite functions E, I, rho, A these are all known quantities, L is known; so, only thing
what remains as a parameter on which R will depend will be the N number of
coefficients. So, now if R be a function of all these coefficients we have chosen, so what
will be the best way to select this coefficient? Here comes the utilization of this property.
(Refer Slide Time: 24:40)

We should select C1 C2 and so on up to Cn in such a way that we get a minimum value of

R. That means it will be nearest to the real value of the natural frequency square. Thus a
condition which satisfies that, they are so chosen that the R will be always higher than the
natural frequency square. So, the best possible choice of R, for the given values or the
given functions Xi's, can be those for which C1, C2, Cn are so that R is minimum. That
condition says: thus del R del Ci will be equal to 0, if R takes a stationary value at the
chosen position then, all these [25:55 min] N number of that will be 0. Therefore, this
represents N number of homogeneous equations in C and therefore a non-trivial solution
of the C will be possible, only if the determinant is 0. Therefore, when this is done, this
represents this N equation. Now the determinant of these equations if we differentiate
this, this is the numerator and this is the denominator it will be this (Refer Slide Time:
27:03). How do we get it? Let us consider (Refer Slide Time: 28:25) this as the
numerator and this is the denominator. So del R del Cr will be equal to this.
(Refer Slide Time: 28:35)

This will be: del N del Ci by D minus which will be N by D square del D and this equal
to 0. So what we can do now? This gives us below there will be D square, so there will be
DN DCi minus N. Since D square is not 0 [29:35 min] D and this is nothing but that. This
is the numerator and this is the denominator (Refer Slide Time: 29:55).

(Refer Slide Time: 30:36)

Now, we further note that omega square is equal to… this is the denominator, this is the
numerator. So we know that omega square is equal to numerator by denominator. So,
when you use this quantity here then, using this… now of course, you should remember
that. Using this in the above equation what we get finally is this (Refer Slide Time:
31:28). So, there will be N such equations, it is very easy to see if we use this. Therefore
we can always take this common and pick out this. Once we pick out this, it goes below
here and this becomes omega square. So, it becomes omega square and then we take this
(Refer Slide Time: 32:40). So, therefore it is straight forward. So, then what is done is
that these N equations in Ci - C1, C2, C3 up to Cn - and these homogeneous equations can
have a solution only when determinants becomes 0. We will try to explain the whole
process with the help of a specific example.

Let us take the same case of a paper beam, we have found out the fundamental frequency
by Rayleigh's method, by Rayleigh's method we did Gravel modifications; we also know
the exact value. So, it will be easier for us to assess or answer if we get from this at least
the first mode. So, what we will try to do, because, we have to keep our analysis, board
work and also objective is to demonstrate the use of the principle to find out higher mode;
we will take only a case where we will find out up to second mode. So, the example is the
same.

(Refer Slide Time: 34:20)

So, the same paper beam, the size of thickness is varying linearly and reaching the value
eight after a length L and it is symmetric. So, if you fix the coordinate here the origin and
the x-axis. Now, the question comes that if we were interested only up to two modes -
omega1 and omega2 only. Therefore we would need only two terms in the expression Xa
is equal to C1X1 plus C2X2 and X1 and X2 should satisfy the boundary conditions, at least
a geometric one. So what do we do? We select X1 as this (Refer Slide Time: 36:02). So,
these are the two functions of X; at X is equal to L, X1 is equal to 0.

If we differentiate once with respect to X, again this term 1 minus x by L remains and at
x is equal to L that will be 0. So, slope is also 0. So, these are the two boundary
conditions, geometric conditions, which are satisfied. This one also say at x is equal to 0
also it is becoming 0. So, it is not really satisfying the condition, but other end if you see
that x is equal to L it is satisfying and both the slope and the value of x deflection is 0. So
therefore, X1 and X2, X1 is satisfying the geometric conditions, X2 also is satisfying the
geometric boundary conditions, but neither X1 nor X2 will satisfy all conditions. Any how
our objective was that whether it satisfies the two conditions here onwards. So, we have
to keep that in mind only two conditions are being partly satisfied.

Now, with these two things our equations become this (Refer Slide Time: 37:35). So, we
substitute this quantity which becomes this after integration of this.
(Refer Slide Time: 39:11)

This after integrating becomes this (Refer Slide Time: 39:19). So, now this yields two
equations. These two equations will be now algebraic equations. So, here it is Ci;
therefore, if you differentiate with respect to C1, we get this equation. Now, if you
differentiate with respect to C2 partially, we get the other equation (Refer Slide Time:
40:22).

So, now these are our final equations - homogeneous equations - in C1 and C2 and as I
have been mentioning that non-trivial solutions will be possible only when the
determinants of the coefficients equal to 0. We can see the determinants of these
coefficients contain this term omega square. So this will ultimately give us an equation in
omega square. So, the determinants of this - we will find out.
(Refer Slide Time: 42:20)

So, for non-trivial solutions, this becomes like our characteristic equation or this
becomes… (Refer Slide Time: 44:11). So, this quadratic equation in omega square, we
ultimately obtained [45:00 min] while lambda is equal to this. Solving this equation, we
get two values of omega square and one value is omega1 which is equal to this (Refer
Slide Time: 45:19). So, the first mode frequency and this is omega2. We get the second
mode frequency also, which is approximately 5. If we try to compare this, the exact value
if we remember is this. Thus, we find that, here the difference is only in the third decimal
place and that so again you can see it is less than 1 percent maybe 0.1 percent. Therefore,
the first mode we get quite accurately; the accuracy of the second mode has to be again
found out by analysis, but it can be shown it is also reasonable accurate. We could find
out even higher frequencies, only then it will be necessary to take or add one more term
here like [47:11 min]. Obviously, computation will be more; there will be three equations
C1 C2 C3 and the determinants will give rise to a third degree equation omega square
which has to be solved for the [47:26 min].

So, this brings us to the end of our discussion on approximate methods for solving
continuous systems, which are otherwise difficult or otherwise not so easy for solution
using analytical exact method as we have discussed. I think normally what is done in a
real life system will be approximated by a model systems and in the model system, one
tries to keep as many paths idealistic as possible. Then, of course in some cases, it may
not be possible to have a theoretical solution or analytical proof or solution. Approximate
methods like this can be used to make a quick estimate. However, now-a-days with the
advent of computer and computers being so cheap, generally a complicated structure can
be always modeled and using a finite element technique we have complete [vibration]
analysis can be solved; standard software are available. In spite of that, the designers
sometimes may find this quite handy and convenient for getting approximate answers.