2.7 HW6

  2.7.1 problem description
  2.7.2 problem 1
  2.7.3 problem 2
  2.7.4 problem 3
  2.7.5 Key solution for HW 6

2.7.1 problem description

PDF

PDF (letter size)
PDF (legal size)

2.7.2 problem 1

   2.7.2.1 Verification using Matlab ffteasy.m

3.41 in text: A periodic disturbance consists of a sequence of exponentially pulse repeated at intervals T, such that Q(t)=FeλtT for 0<t<T, and Q(t±T)=Q(t). The parameter λ is nondimensional. Determine the complex Fourier series representing the force. Evaluate the first 5 coefficients when λ=0.1,1,10. What does this reveal regarding the influence of λ on the frequency spectrum?

Let Q~(t) be the Fourier series approximation to Q(t) given byQ~(t)=12n=Fnein2πTt

WhereFn=2T0TQ(t)ein2πTtdt=2T0TFeλtTein2πTtdt=2FT0Tet(in2πTλT)dt=2FT(et(in2πTλT)in2πTλT)0T=2Fin2πλ(eT(in2πTλT)1)=2Fin2πλ(ein2πeλ1)

But ein2π=1, hence Fn=2Fin2πλ(eλ1)

Hence Eq ?? becomesQ~(t)=12n=2Fin2πλ(eλ1)ein2πTt=Fn=(eλ1)in2πλein2πTt=Fn=1eλλ+in2πein2πTt

For n=2,1,0,1,2 we obtainQ~(t)=Fn=221eλλ+in2πein2πTt=F(1eλλi4πei4πTt+1eλλi2πei2πTt+1eλλ+1eλλ+i2πei2πTt+1eλλ+i4πei4πTt)

For λ=0.1Q~(t)=F(1e0.10.1i4πei4πTt+1e0.10.1i2πei2πTt+1e0.10.1+1e0.10.1+i2πei2πTt+1e0.10.1+i4πei4πTt)=F{(6.026×105+7.572×103i)ei4πTt+(2.41×104+1.514×102i)ei2πTt+0.952+(2.4099×1041.5142×102i)ei2πTt+(6.026×1057.572×103i)ei4πTt}

For λ=1Q~(t)=F(1e11i4πei4πTt+1e11i2πei2πTt+1e11+1e11+i2πei2πTt+1e11+i4πei4πTt)=F{(0.00398+0.05i)ei4πTt+(0.016+0.098i)ei2πTt+0.632+(0.016+0.098i)ei2πTt+(0.00398+0.05i)ei4πTt}

For λ=10Q~(t)=F(1e1010i4πei4πTt+1e1010i2πei2πTt+1e1010+1e1010+i2πei2πTt+1e1010+i4πei4πTt)=F{(3.877×102+4.872×102i)ei4πTt+(7.169×102+4.505×102i)ei2πTt+0.1+(7.169×1024.505×102i)ei2πTt+(3.877×1024.872×102i)ei4πTt}

We notice that as λ became larger, the DC term became smaller. Since the DC term represents average value of the whole signal, then we can say that as λ gets larger, then the average becomes smaller. This means the energy of the signal becomes smaller as λ becomes larger.

2.7.2.1 Verification using Matlab ffteasy.m

From above, we found for λ=1Fn=2Fin2πλ(eλ1)=2Fin2π1(e11)

and the first 5 found to be



n Fn


2 0.00398+0.05i


1 0.016+0.098i


0 0.632


1 0.0160.098i


2 0.003980.05i


To verify the result with ffteasy.m using λ=1, Using F=1, and using T=1. This below shows the result for F0,F1,F2 and we see that the DC term F0 agrees, and that complex component of F1,F2 also agrees. The real parts are little larger than what I obtained using the above. This might be a scaling issue, and I was not able to determine the reason for it at this time.

EDU>> T=1; del=0.01; t=0:del:T; lambda=1; xt=exp(-lambda*t/T);
EDU>> (1/length(t))*fft_easy(xt,t)

ans =

   0.6326 + 0.0000i
   0.0190 - 0.0986i
   0.0072 - 0.0502i

2.7.3 problem 2

   2.7.3.1 Part(a)

pict

We are given that m=1200 kg, f=5 Hz, ζ=0.4 andz(x)={x5x20<x<0.200.2<x<4

A plot of z(x) for first 20 meters is

z[x_] := Piecewise[{{x - 5 x^2, 0 <= x < 0.2}, {0, 0.2 <= x <= 4}}]
z[x_] /; x > 4 := z[Mod[x, 4]];
Table[{x, z[x]}, {x, 0, 21, .1}];
ListLinePlot[%, PlotRange -> {All, {0, .07}}, Frame -> True,
 FrameLabel -> {{"z(x) hight or road (mm)", None}, {"meter",
    "bumps on road"}}]

pict

We need to be able to express z(t) as Re{Zei2πTt} where T is the period of the function z(t). Hence we need to represent z(x) as Fourier series approximation then replace x=vt and use the result.

The period T=4 meter. Let z~(x) be the Fourier series approximation to z(x), hencez~(x)=12F0+Re(n=1NFnein2πTx)

WhereFn=2T0Tz(x)ein2πTxdx=1202/10(x5x2)einπ2xdx=1202/10xeinπ2xdx5202/10x2einπ2xdx

Using integration by parts udv=uvvdu, letting u=x and dv=einπ2x then v=einπ2xdx=ieinπ2xnπ2hence02/10xeinπ2xdx=xieinπ2xnπ2|021002/10ieinπ2xnπ2dx=210ieinπ2210nπ22nπ02/10ieinπ2xdx=410ieinπ10nπi2nπ(einπ2xinπ2)0210=410ieinπ10nπ+4n2π2(einπ2x)0210=410ieinπ10nπ+4n2π2(einπ22101)=4i10nπeinπ10+4n2π2einπ104n2π2=einπ10(4n2π2+2i5nπ)4n2π2

Now we do the second integral 02/10x2einπ2xdx.

Integration by parts, udv=uvvdu, letting u=x2 and dv=einπ2x then v=ieinπ2xnπ2 hence02/10x2einπ2xdx=[x2ieinπ2xnπ2]021002/102xieinπ2xnπ2dx=8100ieinπ10nπ4inπ02/10xeinπ2xdx

But 02/10xeinπ2xdx was solved before and its results is Eq 2.1, hence02/10x2einπ2xdx=4100ieinπ10nπ24inπ(einπ10(4n2π2+2i5nπ)4n2π2)=8i100nπeinπ10einπ10(16in3π385n2π2)+16in3π3=einπ10(8i100nπ16in3π3+85n2π2)+16in3π3

Putting all the above together, we obtain Fn asFn=1202/10xeinπ2xdx5202/10x2einπ2xdx=12[einπ10(4n2π2+2i5nπ)4n2π2]52[einπ10(8i100nπ16in3π3+85n2π2)+16in3π3]=einπ10(2n2π2+i5nπ)2n2π2einπ10(20i100nπ40in3π3+205n2π2)40in3π3=einπ10[2n2π2+i5nπ20i100nπ+40in3π34n2π2]2n2π240in3π3=einπ10(40in3π32n2π2)2n2π240in3π3

Now F0=2T0Tz(x)dx=1202/10(x5x2)dx=1300 Hencez~(x)=12F0+Re(n=1NFnein2πTx)=1600+Re(n=1N(einπ10(40in3π32n2π2)2n2π240in3π3)einπ2x)=1600+Re(n=1Nei(nπ2xnπ10)(40in3π32n2π2)einπ2x(2n2π2+40in3π3))=1600+Re(n=1N40n3π31iei(nπ2xnπ10)2n2π2ei(nπ2xnπ10)2n2π2einπ2x+40n3π31ieinπ2x)

But x=vt, hencez~(t)=1600+Re(n=1N40n3π31iei(nπv2tnπ10)2n2π2ei(nπv2tnπ10)2n2π2einπv2t+40n3π31ieinπv2t)

Therefore the forcing frequency is nϖ1=nπv2 or from 2πf1=πv2, hence f1=v4Hz.The above can be written asz~(t)=1600+n=1NRe(40n3π31iei(nπv2tnπ10))n=1NRe(2n2π2ei(nπv2tnπ10))n=1NRe(2n2π2einπv2t)+n=1NRe(40n3π31ieinπv2t)=1600+n=1N40n3π3sin(nϖ1tnπ10)n=1N2n2π2cos(nϖ1tnπ10)n=1N2n2π2cos(nϖ1t)+n=1N40n3π3sin(nϖ1t)=160040π3n=1N1n3sin(nϖ1tnπ10)2π2n=1N1n2cos(nϖ1tnπ10)2π2n=1N1n2cos(nϖ1t)+40π3n=1N1n3sin(nϖ1t)

Where ϖ1=πv2

To verify the above, here is a plot for different number of fourier series terms showing that approximation improves as N increases. This was done for v=5m/s and for 5 seconds.

pict

2.7.3.1 Part(a)

The equation of motion ismy+c(yz)+k(yz)=0(2.1)my+cy+ky=cz+kz

From earlier, we found that fourier series approximation to z(t) isz(t)=1600+Re(n=140n3π31iei(nϖtnπ10)2n2π2ei(nϖtnπ10)2n2π2einϖt+40n3π31ieinϖt)=1600+Re(n=140n3π3einπ101ieinϖt2n2π2einπ10einϖt2n2π2einϖt+40n3π31ieinϖt)=1600+Re(n=1einϖt[40n3π3ei(nπ10+π2)2n2π2einπ102n2π2+40n3π3eiπ2])

Let Zn=40n3π3ei(nπ10+π2)2n2π2einπ102n2π2+40n3π3eiπ2 Then above can be simplified toz(t)=1600+Re(n=1einϖtZn)

Where ϖ=πv2, hencez(t)=Re(n=1inϖeinϖtZn)

Hence, let yss(t)=Ren=1Yneinϖt

Hence Eq 2.1 becomesn=1mn2ϖ2Yneinϖt+n=1icnϖYneinϖt+n=1kYneinϖt=n=1icnϖeinϖtZn+k600+n=1keinϖtZnn=1(mn2ϖ2+icnϖ+k)Yneinϖt=n=1(icnϖ+k)Zneinϖt+k600n=1(mn2ϖ2+icnϖ+k)Yneinϖt=k600+n=1(icnϖ+k)Zneinϖt

HenceYn=(icnϖ+k)m(nϖ)2+icnϖ+kZn

Let D(rn,ζ)=icnϖ+km(nϖ)2+icnϖ+k=i2ζmωnatnϖ+ωnat2mm(nϖ)2+i2ζmωnatnϖ+ωnat2m=i2ζnϖωnat+1(nϖωnat)2+i2ζnϖωnat+1=1+i2ζrn(1rn2)+i2ζrn

Where in the above rn=nϖωnat where ϖ is 2πT which means it is the fundamental frequency of the forcing function and ωnat is the natural frequency.

Then Eq ?? becomesYn=D(rn,ζ)Zn

And the steady state solution yss(t) becomesyss(t)=k600+Re(n=1D(rn,ζ)Zneinϖt)

Now we can answer the question. When c=0 then D(rn,ζ) reduces to km(nϖ)2+k=11(nϖωnat)2=11rn2, henceyss(t)=k600+Re(n=111rn2Zneinϖt)

So the displacement yss(t) will be resonant when rn=1 or nπv2ωnat=1 or v=2ωnatnπ

Hencev=2(2π5)nπ=20n

Hence v=20,10,5,2.5,1.25, meter/sec will each cause resonance. To verify, here is a plot of yss(t) with no damper for speed near resonance v=19.99 and comparing this for speeds away from resonance speed. This plot shows that when speed v is close to any of the above speeds, then the displacement yss(t) becomes very large. Once the speed is away from those values, then yss(t) quickly comes down to steady state F/k value.

pict

2.7.4 problem 3

pict

pict

The function is periodic with period T=2τf(t)=Pτt0<t<τ0τ<t<2τ

and f(t±T)=f(t). Let f~(t) be the Fourier series approximation to f(t), hencef~(t)=12F0+Re(n=1NFnein2πTx)

WhereFn=2T0Tf(t)ein2πTtdt=22τ0τPτteinπτtdt=Pτ20τteinπτtdt

Using integration by parts udv=uvvdu, letting u=t and dv=einπτt then v=einπτtdt=ieinπτtnπτhenceFn=Pτ2[(tieinπτtnπτ)0τinπτ0τeinπτtdt]=Pτ2[(τieinπττnπτ)inπτ(einπτtinπτ)0τ]=Pτ2[(τ2ieinπnπ)+τ2n2π2(einπτt)0τ]=Pτ2[(τ2ieinπnπ)+τ2n2π2(einπ1)]

einπ=cos(nπ)=(1)n, henceFn=Pτ2[(τ2i(1)nnπ)+τ2n2π2((1)n1)]

Hence for even nFn=Pτ2[(τ2inπ)]=Pinπ

and for odd nFn=Pτ2[(τ2inπ)2τ2n2π2]=Pnπ(2nπ+i)

F0=Pτ20τtdt=Pτ2(t22)0τ=Pτ2(τ22)=P2

Now Eq ?? becomesf~(t)=12F0+Re(n=1NFnein2πTx)=12F0+Re(even nFnein2πTt+odd nFnein2πTt)=p4+Re(even nPinπein2πTt+odd nPnπ(2nπ+i)ein2πTt)=p4+Re(Pπeven ninein2πTtPπodd n1n(2nπ+i)ein2πTt)=P4+Re(Pπeven ninein2πTtPπodd n(2n2π+in)ein2πTt)

To verify, here is a plot of the above, using P=1 and τ=0.5 sec for t=02 seconds. This shows as more terms are added, the approximation becomes very close to the function. At N=40 the approximation appears very good.

pict

Now we need to write f(t) as sum of exponential to answer the question. f~(t)=12F0+Re(n=1NFnein2πTx)

where ϖ is the fundamental frequency of the force given by 2πT=2π2τ=πτ

Hence, let yss=n=Yneinϖt, thenRe(mn=(nϖ)2Yneinϖt+cn=inϖYneinϖt+kn=Yneinϖt)=12F0+Re(n=1NFnein2πTx)n=(m(nϖ)2+icnϖ+k)Yneinϖt=12F0+Re(n=1NFnein2πTx)

HenceYn=Fnk1(1(nϖωnat)2)+i2ζnϖωnat=Fnk1(1(nr)2)+i2ζnr

Henceyss=12F0+Re(n=1Yneinϖt)

Finding Yn for τ=π3ωnat

where r=ϖωnat. When ζ=0.04 and τ=π3ωnat, hence now r=2π(2τ)ωnat=2π(2π3ωnat)ωnat=3, thereforeYn=Fnk1(1(3n)2)+i6(0.04)n=Fnk1(19n2)+i0.24n

The largest Yn will occur when the denominator of the above is smallest. Plotting the modulus of the denominator (19n2)2+(0.24n)2 for different n values shows that n=1 is the values which makes it minimum.

This happens since for any n>1 the denominator will become larger due to n2 and hence Yn will become smaller. So n=1 will be used.

pict

For n=1, we obtainY1=F1k1(19)+i6(0.04)

But F1=Pπ(2π+i), henceY1=Pπ(2π+i)k1(19)+i6(0.04)=Pπk(2π+i)8+i0.24=Pπk2π+i8i0.24=Pπk(2π+i)(8+i0.24)(8i0.24)(8+i0.24)=Pπk(0.075759+0.12727i)

ThereforeY1=Pk(0.024115+0.0405i)

Here is a list of Yn for n=110 with the phase and magnitude of each (this was done for pk=1)

pict

From the above we see that most of the energy in the response will be contained in Y1 and adding more terms will not have large effect on the response shape. This is confirmed by the plot that follows.

Plot for the steady state

Since yss=12F0+Re(n=1Yneinϖt)

Where now r=ϖωnat. When ζ=0.04 and τ=π3ωnat, hence now r=2π(2τ)ωnat=2π(2π3ωnat)ωnat therefore r=3yss=p4+Re(n=1,3,5Yneinϖt+n=2,4,6Yneinϖt)=p4+Re(n=1,3,5Fnoddk1(1(nr)2)+i2ζnreinϖt+n=2,4,6Fnevenk1(1(nr)2)+i2ζnreinϖt)=p4+Re(n=1,3,5Pnπ(2nπ+i)k1(1(nr)2)+i2ζnreinϖt+n=2,4,6Pinπk1(1(nr)2)+i2ζnreinϖt)=p4+pkRe(n=1,3,51nπ(2nπ+i)(1(nr)2)+i2ζnreinϖt+n=2,4,6inπ(1(nr)2)+i2ζnreinϖt)

Now let r=3, ζ=0.04. Normalizing the equation for ϖ=1 which implies τ=π and k=1 and p=1, then the above becomesyss=14+Re(n=1,3,51nπ(2nπ+i)(1(3n)2)+i2(0.04)3neint+n=2,4,6inπ(1(3n)2)+i2(0.04)3neint)

Here is a plot of the above for t=020 seconds for different values of n

pict

We see from the above plot, that yss(t) does not change too much as more terms are added, since when r=3, then Yn for n=1 contains most of the energy, hence adding more terms did not have an effect.

Repeating the calculations for τ=3πωnat

r=ϖωnat. When ζ=0.04 and τ=3πωnat, hence now r=2π(2τ)ωnat=2π(23πωnat)ωnat=13, thereforeYn=Fnk1(1(nr)2)+i2ζnr=Fnk1(1(13n)2)+i23(0.04)n=Fnk1(1n29)+i0.0267n

The largest Yn will occur when the denominator of the above is smallest. Similar to above, we can either find n which minimizes the denominator (by taking derivative and setting it to zero and solve for n) or we can make a plot and see how the function behaves. Making a plot shows this

pict

From the above we see that the smallest value of the denominator happens when n=3.

so using n=3 we findY3=F3k1(1(3r)2)+i2ζ3r=F3k1(1(313)2)+i2(0.04)313=F3k1i0.08

But Fn=Pnπ(2nπ+i), hence F3=P3π(23π+i)

ThereforeY3=P3π(23π+i)k1i0.08

HenceY3=pk(1.3263+0.28145i)

Here is a list of Yn for n=110 with the phase and magnitude of each (this was done for pk=1)

pict

We see from the above that |Y3| is the largest harmonic.

Plot for the steady state

Since yss=12F0+Re(n=1Yneinϖt)

Where now r=ϖωnat. When ζ=0.04 and τ=3πωnat, hence now r=2π(2τ)ωnat=2π(23πωnat)ωnat=13,  therefore from aboveyss=p4+pkRe(n=1,3,51nπ(2nπ+i)1(1(nr)2)+i2ζnreinϖt+n=2,4,6inπ1(1(nr)2)+i2ζnreinϖt)

Now let r=13, ζ=0.04, and assuming τ=0.5 then ϖ=2π2τ=π0.5, and assuming k=1, then the above becomesyss=14+1kRe(n=1,3,51nπ(2nπ+i)1(1(n13)2)+i2(0.04)13neinπ0.5t)+1kRe(n=2,4,6inπ1(1(n13)2)+i2(0.04)13neinπ0.5t)

Here is a plot of the above for t=020 seconds for different values of n

pict

We see now that after n=3 that the response did not change much by adding more terms, this is because more of the energy are contained in the first 3 harmonics with Yn being the the largest.

2.7.5 Key solution for HW 6

PDF