Smart Structures

Professor Mohammed Rabius Sunny

Department of Aerospace Engineering
Indian Institute of Technology, Kharagpur
Week 05
Lecture No: 28
Energy Harvesting and Vibration Control (continued)
Part 04

In the last lecture, we looked into the mathematical formulation of a beam with
piezoelectric patches for energy harvesting applications. And we followed a paper for that.

Today we will look into the same formulation, but from a different perspective. We will
start with the governing differential equations and we do the formulation. So, in the
previous lecture, we did it by starting from the Hamilton's principle and there, we put the
approximations. Here we will start with from the governing differential equation and we
will do it. And after that, we look into the vibration control applications.

So, the governing equations in general for a 3D case as, we have seen that is this. Sigma i
j comma i plus Bj, which is a body force is equal to rho uj double dot and the other
differential equation for the electrical domain is the Gauss law.

𝜎𝑖𝑗,𝑖 + 𝐵𝑗 = 𝜌𝑢̈𝑗

𝐷𝑖,𝑖 = 0

Now, if we again have our beam under consideration and if we have x as the axial
dimension z as the vertical dimension and y is the other out of plane dimension. So, let us
consider this and with Euler Bernoulli assumption and piezoelectric effect, under Euler
Bernoulli assumption and considering piezoelectric effect, the linear momentum equation
becomes this.

The corresponding governing differential equation is m b w double dot, plus del 2 w by del
x 2 of EItotal, del 2 by del x 2 of EItotal, del 2 w by del x 2 minus pz minus del 2 Mp by del x
2.

𝜕2 𝜕 2𝑤 𝜕 2 𝑀𝑝
𝑚𝑏 𝑤̈ + (𝐸𝐼𝑡𝑜𝑡 ) − 𝑝𝑧 − =0
𝜕𝑥 2 𝜕𝑥 2 𝜕𝑥 2

So, we are very familiar with this equation. We already derived it. And for the other
equation, is just this, the Gauss law is just this.

𝐷𝑧,𝑧 = 0
Considering the fact that this piezoelectric patches are polarized in z direction, and so all
the electric field voltage is in the z direction. Electrical displacement in z direction. So,
these are the two equations now that we have to use to form the final energy harvesting
equation. And here we will follow the notations that we were using in the strain induced
actuation problems. So, in the previous lecture, we use the notations that was in the paper,
but here, we have come back to the notations that we are using in the induced strain
actuation problems. So, let us assume w as, as we have assumed phi wj q wj and the electric
field was assumed to be - let us assume the electric field. So, Ez is equal to V, the voltage,
which is a function of time, multiplied by psi. Psi is a function of z, and here q w j is also
function of time.

So, we will make these assumptions and then we will go to the derivation. So, our approach
is to multiply the first equation by phi wj. So, if we do it.
𝑁

𝑤 = ∑ 𝜙𝑤𝑗 𝑞𝑤𝑗 (𝑡)𝐸𝑧 = 𝑣(𝑡)𝜓(𝑧)

𝑗=1

So, let us multiplied by phi wi. Let us take i as the index, and then it is - we are just
multiplying phi w i with the first equation and in the first equation, we put these
approximations and we integrate it from 0 to L.
𝐿
𝜕2 𝜕 2𝑤 𝜕 2 𝑀𝑝
∫ 𝜙𝑤𝑖 (𝑚𝑏 𝑤̈ + 2 (𝐸𝐼𝑡𝑜𝑡 2 ) − 𝑝𝑧 − ) 𝑑𝑥
𝜕𝑥 𝜕𝑥 𝜕𝑥 2
0

And Mp is as we know as for the definition Mp is the area integral of minus z into elastic
modulus multiplied by the free strain. And free strain, as we know that, it is d31 multiplied
by electric field Ez or E3, whatever, Ez is equal to E3. And we have already assumed Ez as
this.

𝑀𝑝 = ∫ −𝑧𝐸𝜀𝑝 𝑑𝐴 = ∫ −𝑧𝐸𝑑31 𝐸𝑧 𝑑𝐴 and 𝐸𝑧 = 𝐸3

𝐴 𝐴

So, finally, after putting this - after putting these approximations and following the
procedure that we have followed; finally, this gives equation in this form. So, we have
already seen how to derive the set of ordinary differential equations. From this kind of
equations, it needs some shifting of derivatives also, which we have to do. And finally, if
we do all these things these equations come - is equal to F.

[𝑀]{𝑞̈ } + [𝐾]{𝑞} − [Θ]{𝑣} = {𝐹}

So, here, Mij is (can put qw), Mij is mb, phi wi, phi wj dx which is equal to Msij plus Mpij, as
per the paper discussed. So, as per the paper discussed.
 𝐿

𝑀𝑖𝑗 = ∫ 𝑚𝑏 𝜙𝑤𝑖 𝜙𝑤𝑗 𝑑𝑥 = 𝑀𝑠𝑖𝑗 + 𝑀𝑝𝑖𝑗

0

(Refer Slide Time: 9:24)

And then we have Kij is equal to EItotal, and which is equal to Ksij plus Kpij, as per the paper.
So, we are just showing the equivalence between the expressions using the two different
formulations and notations.
𝐿

𝐾𝑖𝑗 = ∫ 𝐸𝐼𝑡𝑜𝑡 𝜙𝑤𝑖,𝑥𝑥 𝜙𝑤𝑗,𝑥𝑥 𝑑𝑥 = 𝐾𝑠𝑖𝑗 + 𝐾𝑝𝑖𝑗

0

And we have theta i, which is minus 0 to L A z psi d31 Young's modulus multiplied by phi
i comma x x dA dx. Now, please understand d31 multiplied by E, gives us E31. So, this is
phi. So, we are using phi here itself and Fi is 0 to L phi wi pz dx.
𝐿

Θ𝑖 = − ∫ ∫ 𝑧𝜓𝑑31 𝐸𝜙𝑖,𝑥𝑥 𝑑𝑥
0 𝐴

F𝑖 = − ∫ 𝜙𝑤𝑖 𝑝𝑍 𝑑𝑥
0
Now, here, by assuming pz to be a distributed function of x, it is written in this way. While
discussing the paper, we saw that those forces were considered to be discrete. So, that is
why they were summed over.

So, that is about the first equation. In the second equation, which is D z z equal to 0, we
multiply psi, and then have V and then, from there we can show that this becomes q
multiplied by - So, q let us put capital Q as charge. So, capital Q is equal to charge because
we are using small q to denote the coefficient associated with the displacement
components. So, it is ds plus V Dz psi multiplied by v, and then dV.

So, now, this term can be written as - we can sum it over. So, we can write phi wj comma
x x q wj plus e 3. So, electric field. So, which is psi multiplied by v and that multiplied by
epsilon 3 3. So, here to multiply e here, e 3 1, and this is epsilon. This multiplied by psi
dV plus Q multiplied by psi ds.

So, this gives us the equation as Theta T, multiplied by qw plus Cp, Cp is just one value
here. So, we can - we may just write it as Cp. Cp v plus Q is equal to 0. Phi T, we have
already defined.

Cp comes to be volume integral of epsilon 3 3, epsilon sigma square dV. So, we have got
these two equations. From the first equation, which was a version of the linear momentum
equation, we got this. From the first equation which was a version of the linear momentum
equation, we got this equation and the from the second equation, which is gauss law, we
got this equation. And these two equations are equivalent to the equations that we got in
the last lecture while discussing the paper in a different type of formulation.

(Refer Slide Time: 15:10)

Now, we look into vibration control applications and the corresponding mathematical
analysis. For that, this paper by Gordon Gee and his co-authors can be referred to. Here the
idea is that we have a piezoelectric beam, sorry a beam which has piezoelectric patches
for sensing and actuation. So, it has two piezoelectric patches at the top and bottom. The
first top one is used as actuator and the bottom one is used as sensor.

Now, it is not needed that the actuator and sensor has to be placed back to back, they can
be placed at other locations also, depending on the control strategy. Now, from the
piezoelectric sensor - so this structure is under some dynamic load. So, in the experiments,
the dynamic load can be given by a shaker or anything. In real life, it is due to the external
factors. So, it is under some dynamic load and then - so we can just show some loads here.
We can denote this as pz. Now, from the sensor, the charge output comes, and then, there
are two modes as suggested by this paper. One is the charge amplifier mode; through which
we get voltage from the charge. So, voltage output from the charge. So, it can be a charge
amplifier, or there can be current to voltage converter.

So, we will talk about it, what it is? Then from this, the voltage output comes and then, it
is multiplied by a multiplied by a gain. We can call it controller and then finally, from there
a feedback voltage is given here, which we may call v voltage. Now, the from here, the
charge or current comes here. The output of this charge amplifier or current to voltage
converter is a voltage. This voltage when it goes to the controller, it is multiplied by a gain
and then the final voltage comes here which actuates the actuator.

Now, if you are interested in the details of this charge amplifier or current to voltage
converter, this paper can be referred. The circuit diagrams are given there.
Now, if we write this, write the dynamics of this, then from there, the charge can be written
as - charge output from sensor is Q which is equal to Dz integrated over the area. So, we
have the electrical displacement Dz here. Let us call this as x, this dimension as z and this,
let us call this as y. So, the same convention that we have been using. So, the Dz here, and
it is already normal to the - it is already oriented along the z axis which means it is already
normal to the xy plane. So, Dz integrated over the area that gives us the charge output Q.
Now, the charge output Q is e31 of the piezoelectric material multiplied by the strain. So,
the strain there - so, the strain here is nothing, but epsilon 0 minus z into kappa, as we
know, now, epsilon 0. So, strain is equal to epsilon 0 minus z into kappa. So, epsilon 0
comes here, z for this case is this and we have to multiply kappa here and kappa comes
here.

𝑡𝑏
𝑄 = ∫ 𝐷𝑧 𝑑𝐴 = ∫ 𝜀31 (𝜀0 + ( + 𝑡𝑐𝑠 ) 𝜅) 𝑑𝐴
2
𝐴 𝐴

𝜀 = 𝜀0 − 𝑧𝜅

Now, the thickness of the host beam is tb and we can call the thickness of the piezoelectric
patch which is used as actuator to be tA, and the thickness of the piezoelectric patch which
is used as sensor to be tcs, s stands for sensor, A stands for actuator. Now, with this, we
can rewrite this expression as – integral from x0 to x0 plus lc. And again, if you want, we
can call this as x0 as we have been calling and this dimension it is better to show in a
different color just to avoid confusion. This dimension is x0 plus lc. So, we are assuming
that the length of the sensor and actuator same. Again, that is not a requirement they can
be different as well. So, we are assuming lc, that lcs is equal to lcA is equal to lc, but that is
not a requirement, they can be different as well.

Now, we also know that in our analysis, we have been assumed u0 x t to be summation of
j is equal to 1 to M multiplied by phi uj into q uj, and w as N phi wj q wj.
𝑀

𝑢0 (𝑥, 𝑡) = ∑ 𝜙𝑢𝑗 𝑞𝑢𝑗

𝑗=1

𝑤(𝑥, 𝑡) = ∑ 𝜙𝑤𝑗 𝑞𝑤𝑗

𝑗=1

So, with that approximation, this expression can be written as - e31 multiplied by bs, bs is
the width of the sensor and then, we multiply this row vector phi u 1 x to phi u M x and
then we have tb plus tcs, q w 1 x double dot, sorry, w comma x x plus tb by 2 plus tcs. and
this entire row vector is multiplied by the column vector and dx.
 𝑄
𝑞𝑤1
𝑥0 +𝑙𝑐 ⋮
𝑡𝑏 𝑡𝑏 𝑡𝑐 𝑞𝑤𝑀
= ∫ 𝑒31 𝑏𝑠 {𝜙𝑢1 ,𝑥 ⋯ 𝜙𝑢𝑀 ,𝑥 ( + 𝑡𝑐𝑠 ) 𝜙𝑤1 ,𝑥𝑥 ⋯ ( + 𝑠 ) 𝜙𝑤𝑀 ,𝑥𝑥 } 𝑞𝑤 𝑑𝑥
2 2 2 1
𝑥0 ⋮
{ 𝑞𝑤𝑁 }

So, what we are doing is - if I multiply this q vector with this row vector, that gives me this
expression.

Now, we can write this entire expression as AT multiplied by q. q is this column vector and
AT is this vector multiplied by bs into e31 and integrated from x0 to x0 plus lc. So, here AT
is of size 1 multiplied by M plus N. And q is of size M plus N by 1.

𝑄 = {𝐴}1𝑇 x (𝑀+𝑁) {𝑞}(𝑀+𝑁) x 1

So, again just to summarize what we have done here is - we need the expression for charge
from the sensor. So, the charge is surface integral of the dz and dz is e31 multiplied by the
strain. So, e31 and the strain expression is here in terms of epsilon 0 and kappa. And epsilon
0 and kappa comes in terms of the derivatives of this phi u and phi w. Here, this is not q,
this is phi. And when this vector, after being multiplied with bs and e31 integrated along
the length of the sensor, it gives me a vector called A.

Now, we have to write the expression for the voltage that we get from the charge amplifier.

(Refer Slide Time: 25:52)

If you want the voltage from the charge amplifier, from the charge amplifier the voltage
that we get is VA is equal to 1 by Cf multiplied by Q, where Cf is capacitance of the or we
can call feedback capacitance of the charge amplifier. And then, this expression can be
written as 1 by Cf multiplied by the A vector multiplied by the q vector.

1 1
𝑉𝐴 = 𝑄 = {𝐴}𝑇 {𝑞}
𝐶𝑓 𝐶𝑓

So, the input voltage to actuator can be written as - v is equal to, we multiply by some
gain, GA with the VA. So, it is GA by Cf AT q.

𝐺𝐴
𝑣 = 𝐺𝐴 𝑉𝐴 = {𝐴}𝑇 {𝑞}
𝐶𝑓

And current to voltage converter, for that the voltage: let us write it as may be Vc is Rcv
multiplied by Q. So, R is a resistance. If we look at that structure diagram of the current to
voltage converter, the position of R is there. So, R is multiplied by Q dot. Q dot is the
current. So, that being multiplied with the resistance gives us the output voltage from the
current to voltage converter. So, that is equal to 1 by Cf multiplied by AT into q dot.

1
𝑉𝑐 = 𝑅𝑐𝑣 𝑄̇ = {𝐴}𝑇 {𝑞̇ }
𝐶𝑓

And again, input voltage to actuator is v is equal to, let me call it - there is another gain.
So, let us call it may be GV multiplied by Vc. So, we can call it Gc also. So, it becomes Gc
(it is not Rc, Rcv), so, Gc multiplied by Rcv AT q dot.

𝑣 = 𝐺𝑐 𝑉𝑐 = 𝐺𝑐 𝑅𝑐𝑣 {𝐴}𝑇 {𝑞̇ }

So, what happens is - the charge amplifier gives an output which is our VA or Vc and
accordingly the controller multiplies with GA or Gc, and the product of it is the feedback
voltage V that comes to the actuator.

(Refer Slide Time: 30:16)

Now, we know that for the actuation case, we have already seen that the dynamics can be
written as this. And that is equal to x0, x0 plus lc, BT, NP, MP. And here, BT is already
known to us and this is multiplied by NP. And then it is integrated over the domain of the
piezoelectric actuator, I mean, the domain along the x axis and that gives me the actuation
force.
𝑥0 +𝑙𝑐
𝑁𝑃
[𝑀]{𝑞̈ } + [𝐶]{𝑞̇ } + [𝐾]{𝑞} = ∫ [𝐵]𝑇 { } 𝑑𝑥 = [𝐵𝐼 ]{𝑆}𝑣
𝑀𝑃
𝑥0

And then we can write this entire thing as BI. So, we are calling it BI means, this B matrix
integrated over x0 to x0 plus lc. So, BI multiplied by a vector S, we are introducing a new
vector S into the actuation voltage v. So, S is a vector which is required to convert the
voltage v to NP and MP. So, NP and MP is a vector S which, on being multiplied with v
gives me the NP and MP. So, S can be written in terms of tcA, the width of the actuator bA
and the beam thickness tb and the elastic modulus of the piezoelectric patch and d31 or we
can write e31 also. So, in terms of those, this can be written - the S vector can be found.

So, we have now this and then, we can rewrite the expression as Mq double dot plus C
multiplied by q dot plus K q. And then we have, the BI, we have S, and then, we have the
gains GA divided by Cf multiplied by A transpose, multiplied by q.

𝐺𝐴
[𝑀]{𝑞̈ } + [𝐶]{𝑞̇ } + [𝐾]{𝑞} = [𝐵𝐼 ](𝑀+𝑁)𝑥2 {𝑆}2𝑥1 [𝐴]𝑇 1+(𝑀+𝑁) {𝑞}(𝑀+𝑁)𝑥1
𝐶𝑓
Now, this is for we will do that. Now again, this entire expression can be written as - LA
multiplied by q, because we know that, this is M plus N over q, S is 2 by 1, this is 1 by M
plus N, and this is M plus N by 1. So, finally, this comes to be M plus N over M plus N
and this we know it is M plus N over 1.

[𝑀]{𝑞̈ } + [𝐶]{𝑞̇ } + [𝐾]{𝑞} = [𝐿𝐴 ](𝑀+𝑁)𝑥(𝑀+𝑁) {𝑞}(𝑀+𝑁)𝑥1

Now, this is charge amplifier mode. Now, when for current to voltage mode, the same
expression for current to voltage mode it becomes - this becomes LC over q dot. And where
LC is BI over S, then we have Gc Rcv, we already defined - Gc Rcv, and we have A transpose.

[𝐿𝑐 ] = [𝐵𝐼 ]{𝑆}𝐺𝑐 𝑅𝑐𝑣 [𝐴]𝑇

So, for the charge amplifier mode, we have Q here and for current to voltage mode we have
q dot here. So, we can see depending on the control strategy whether we used charge
amplifier mode or current to voltage mode, we have q or q dot here. And the gains are
decided depending on - I mean, it is a matter of control system. So, we used various
techniques, various optimizations to find out the optimal gain. So that this entire system
can be controlled.

Now, in real life, there can be unwanted vibration due to several reasons, aero elastic
instability for aircrafts can be one of the reasons. So, if it goes to the unstable region,
dynamically unstable region, flutter can occur and the vibration may look like this. In such
cases if this kind of controls are applied, then the vibration can be stabilized.

(Refer Slide Time: 36:47)

So, we have discussed about the energy harvesting and the control systems based on the
formulations that we have been doing so far.

Now with that let me conclude this lecture.

Thank you.