Smart Structures

Professor Mohammed Rabius Sunny

Department of Aerospace Engineering
Indian Institute of Technology, Kharagpur
Week 04
Lecture No: 22
Static Analysis of beam for Induced Strain Actuation using Energy Principles
(continued)
Part 06

In the last lecture, we started with the virtual work equation for this beam with piezo's and
from those equations, we saw how we can get a governing equation that we are familiar
with.

Now, we will see how by starting from the same virtual work equation, we can get a
governing equation in a different form. So, again we look at equilibrium equation, but we
will derive it in a different form. So, after we start with the virtual work equation, we
obtained this form.

(Refer Slide Time: 1:45)

So, if I don't shift the derivatives here, if I don't do the integration by parts, if we written it
as it was, we have it here and rest of it we know how to get. Now, what we will do is instead
of integrating by parts here and here, we will do the integration by parts here and here, and
that will give us a governing differential equation in a different form. So, if we look at the
first term which was our integral E A total, if I integrate by parts, it gives me: E A total,
del u0 by del x multiplied by integral of this and integral of this, is nothing but delta of u0
and it is evaluated at x equal to 0 and L and then we have minus of differentiation of this.
So, del del x of E A total, del u0 by del x, multiplied by integral of that. So, delta of u0 0 to
L dx. So, we can see it here, these terms. So, these terms come from this.
𝐿 𝐿
𝐿
𝜕𝑢0 𝜕𝑢0 𝜕𝑢0 𝜕 𝜕𝑢0
∫ 𝐸𝐴𝑡𝑜𝑡 𝛿( ) 𝑑𝑥 ⟹ 𝐸𝐴𝑡𝑜𝑡 𝛿𝑢𝑜 | − ∫ (𝐸𝐴𝑡𝑜𝑡 ) 𝛿𝑢0 𝑑𝑥
𝜕𝑥 𝜕𝑥 𝜕𝑥 0 𝜕𝑥 𝜕𝑥
0 0

Similarly, we can do the same thing for the next term where we have delta of del u0 by del
x multiplied by E S total, into del 2 w by del x 2. After we do the integration by parts, we
get this term. And then, we had Np multiplied by delta of del u0 by del x and that were
integrated from 0 to L. So, if I do integration by parts here, it gives me Np multiplied by
the integral of this, which is delta of u0 0 to L minus del Np by del x into variation of u0,
and which we can see here. So, this had a negative sign before. So, that is why it was
negative, it is positive, what we see here.
𝐿 𝐿
𝜕𝑢0 𝐿 𝜕𝑁𝑝
∫ 𝑁𝑝 𝛿 ( ) 𝑑𝑥 ⟹ 𝑁𝑝 𝛿𝑢𝑜 | − ∫ 𝛿𝑢0 𝑑𝑥
𝜕𝑥 0 𝜕𝑥
0 0

Then we had a term like E S total, multiplied by del u0 by del x, multiplied by variation of
the second order derivative of w and this was integrated from 0 to L. If I integrate this by
part, that gives me E S total, del u0 by del x variation of delta w by del x, which is evaluated
at the limit 0 and L. And then, minus I would have derivative of this and integral of this.
So, now, the other term that I get as minus, I am writing here, multiplied by variation of
del w by del x. But again, if I integrate this by parts, this would give me minus del by del
x, E S total, multiplied by variation of w and that gets evaluated at 0 to L. And then, plus
this differentiated twice, E S total. So, here this term is missing, it is E S total, del u0 by del
x multiplied by delta w. And here, we have the same thing, but differentiated once more
multiplied by delta w dx. So, that is how we get this term, this term and this term.
Accordingly, we can find out rest of the terms by doing the integration by parts and we get
this entire expression.
𝐿
𝜕𝑢0 𝜕 2𝑤
∫ 𝐸𝑆𝑡𝑜𝑡 𝛿 ( 2 ) 𝑑𝑥
𝜕𝑥 𝜕𝑥
0
𝜕𝑢0 𝜕𝑤 𝐿 𝜕 𝜕𝑢0 𝐿
⟹ 𝐸𝑆𝑡𝑜𝑡 𝛿 ( )| − [ (𝐸𝑆𝑡𝑜𝑡 ) 𝛿𝑤]|
𝜕𝑥 𝜕𝑥 0 𝜕𝑥 𝜕𝑥 0
𝐿
𝜕2 𝜕𝑢0
+∫ (𝐸𝑆𝑡𝑜𝑡 ) 𝛿𝑤𝑑𝑥
𝜕𝑥 2 𝜕𝑥
0
(Refer Slide Time: 7:29)

Now, if we look at this entire expression, we can see that the variations are in del u0 w, del
w by del x, and that is all. So, if we separate out the components, that are multiplied with
delta of u0, delta of w, and delta of del w by del x, we get this. In this expression, we have
these two terms that are integrated, whereas, these expressions are: these three expressions
are all defined at the boundaries. Now, if we look at these boundary terms, we can see that,
this is nothing but our N, normal force N.

So, this thing inside the bracket is normal force N. Why? Because, we have seen so far,
that E A total, plus E S total, multiplied E A total, del u0 by del x, plus E S total, del 2 w
by del x 2, minus Np, minus N, equal to 0, which means that this equal to N. So, this is
equal to N here.

𝜕𝑢0 𝜕 2𝑤
𝐸𝐴𝑡𝑜𝑡 + 𝐸𝑆𝑡𝑜𝑡 2 − 𝑁𝑝 − 𝑁 = 0
𝜕𝑥 𝜕𝑥
Similarly, using the same reason, this is my M. And this is nothing but del M by del x,
which gives me shear force. Now again, we can make similar argument and we can say
that these terms are 0 for our case because at x equal to 0, we have the axial displacement
specified which is u0, which means that delta of u0 is 0. At the other end the normal force
is 0.

Similarly, here we have the slope del w by del x specified, which means the variation of
the slope is 0 at this end. And at x equal to L, we do not have the bending moment. So, the
bending moment is 0 at x equal to L. And at x equal to 0, we have the displacement
specified. So, del w is 0 here and at x equal to L, we do not have any shear force. So, that
helps me get rid of all these terms and we are left with only this and this. And again, we
know that delta u0 and del w are independent and arbitrary variation, which tells me that
these integrands must be individually 0.

(Refer Slide Time: 10:43)

And from that same argument we can say that this is our one governing differential
equation, and this is our other governing differential equation. So, we get the governing
differential equations in a different form. Now, if I compare this form with our previous
form, the difference is that while getting the previous form, we did integration by parts in
these terms where Np and Mp were present, sorry, where Px and Pz were present whereas,
here we did integration by parts in the other terms. And there, it was written in terms of
variation of epsilon 0 and kappa. Here it is written in terms of u0 and w. And from there,
we get these two equations. Now, we will see using these two equations, how we can get
the solution to the problem.

(Refer Slide Time: 11:46)

Now, these two equations are now written in a compact form as this. So, where we have
this operator. So, u0 comes here, and it gets operated here, w comes here and it gets operated
here and we get the first equation. Similarly, u0 comes here and w comes here and we get
the second equation. We are going to solve it by using a technique named the Galerkin
technique.

(Refer Slide Time: 12:24)

So, let us apply the technique. Again, let us assume that we have u0 as this. So, u0 is written
as a combination of this product and within this product, we have phi and q. So, phi ui is
chosen function, chosen function to approximate u0 and similarly phi wi is chosen to
approximate w. If we can choose in such a way that they satisfy both essential and natural
boundary conditions, I mean the geometric and force boundary conditions, that is good,
but at the bare minimum they should satisfy the essential boundary conditions. So, these
phis are all known to us, because we are choosing them and they are being multiplied with
unknown constants q ui and this q wj. Now, q ui, q wj are unknowns to be found out. Now,
we have assumed that our u0 has this m number of terms and w has this n number of terms.

So, the vector q u ranges from 1 to m and the size of the vector q w is n. Now, we can
write a vector u which has u0 and w. And then, we can write it in terms of this phi's and
this q's as this. Here we have first m terms, which is populated with this q u's and then rest
of the n terms are 0. And here, we have first m term 0 and then rest of the n terms populated
with this q w's. So, this is of size 2 multiplied by m plus n. And this is of size m plus n
multiplied by 1. Now, we can call this entire vector as one q vector with size m plus n. So,
within this q vector the first m number of q's are q u’s and rest of the n number of q's are
q w’s and it is multiplied with the same thing. So, in a compact way we can call this as phi
matrix multiplied by q matrix. So, this is our phi matrix and this is our q vector.

(Refer Slide Time: 16:12)

So, again the governing equation rewritten here. So, if we put the approximations of u0
and w so, that becomes our phi multiplied by q. Now, because they are approximations,
they may not be exact with u0 and w. So, if I take the right-hand side to the left-hand side,
and if we substitute u0 w vector as q as phi q, then we can say that, this entire thing may
not be exactly 0 because phi q may not be exactly equal to u0 w. So, there is can be some
error which is remaining and that error is this. So, the close we are to the real value of u0
and w this is the error.

(Refer Slide Time: 17:11)

Then what we do is: we multiply this error function by this approximation phi’s and then,
so, if we go back to our phi vector. So, this is our phi 1, this is our phi 2, this is our phi 3
phi m and so on and we can go up to phi n. So, we may say that, we may like to call this
as our phi 1 vector and we can go on. So, this is our phi m vector and we may call each of
them as a vector and we can say that, this phi matrix consists of, this capital phi matrix
consists of all this small phi vectors.

(Refer Slide Time: 18:40)

Now, in this error function, if I multiply the - what we are doing here, if I multiply the i th
phi, the transpose of that with the error function and then integrate from 0 to L and equate
that to 0 and we can do it for all these phi's. So, that gives me total m plus n number of
equations. And those m plus n number of equations would have these q's as the unknowns
and that will give me. So, it will give me m plus n number of equations with m plus n
number of unknowns and that can be written in a matrix form as this. Now, this we call
stiffness matrix and this we call force vector. In the stiffness matrix, the generalized i th
component can be written as this. So, the so k ij is equal to phi i multiplied by this matrix
multiplied by phi j. So, we know that it is 1 by 2 because the i th column of phi is 2 by 1,
and similarly this is 2 by 1 and this is 2 by 2.

So, after this multiplication this gives me one value the ij th value of the Kth matrix. Let
us assume that our phi j, let us assume that, we are taking from our phi matrix, we are
taking this as i and this as j. So, let us assume that i is equal to 1and j is equal to 2. So, in
this case, it would look like this: phi u1 0 multiplied by this vector del by del x, E A total
del by del x, and then, we have this phi u1 and 0. So, when I multiply this column with
this rho, I have del del x of E A total, del phi 1 del x, plus 0. And here, I have del 2 del x 2
of E S total, del phi 1 del x as 0. So, as we said, let us assume i is equal to 1 j is equal to 2.
So, it is 2. So that gives me phi u1 0, and here, we have del del x of E A total, del phi ut by
del x. And here, we have del 2 by del x 2 E A total, del phi u2 by del x. And of course, this
has to be integrated from 0 to L. This also has to be integrated from 0 to L. Again, we
multiply this rho with this vector and after we multiply, we are remaining with only one
term. So, this term multiplied with by this term, that will give me just one value. This
multiplied by 0 will give me 0. So, finally, we will get 1 by 1 term which is our k ij. So, in
this particular case, which is k 12. So, accordingly we can just put any i here and any j here,
and accordingly, we will get the k ij and this entire matrix will be populated.

(Refer Slide Time: 23:53)

Similarly, when I find out phi I, this becomes the expression. So, phi i is multiplied with
del Np by del x, and del 2 Mp by del x 2, and then we have phi i transpose. So, here also
we have phi i transpose and here also we have phi i transpose multiplied with minus Px,
Pz vector. Now, generally while evaluating this, we don't directly evaluate this, we shift
the derivatives here also by integrating by parts and then we see that the boundary terms
go 0 and the rest of the integral evaluated. So, instead of the differentiation being here, we
get the differentiation here and we evaluate. So, later on when we do an example, we will
see it in more details, how the derivative is shifted and how this term is evaluated in a better
way.

So, with this we get the k matrix and the force vector and by solving we get our q. And
once we, and once we get our q. So, we have q u, we have q w's. So, we get q ui, q wj from
here, and then we put it back in our approximation that u0, as a function of u is summation
of phi ui into q ui, and w as a function of x is equal to phi wj into q wj. So, we just put it
here, and we put it here, and that gives us the desired solution. So, we get our solution. So,
that is one way to solve these differential equations.

(Refer Slide Time: 25:53)

So, here we started with the differential equation and then, we multiplied them by each of
these approximation functions and finally, this differential equation was converted to a set
of algebraic equations. And then by solving the algebraic equations, we get our solution.

So, with this I would conclude it here. We will look into some other similar techniques in
the next lecture.

Thank you.