Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 10

(1) case(d) A System of linear second order ordinary differential equations is given below:

     m1  ! E1 ( x1  x0 )  E 2 ( x2  x1 )  k1 ( x1  x0 )  k2 ( x2  x1 ) x1

(1) (2) (3)

     m2 2 ! E 2 ( x2  x1 )  E 3 ( x3  x2 )  k2 ( x2  x1 )  k3 ( x3  x2 ) x    m3  ! E 3 ( x3  x2 ) x3  k3 ( x3  x2 )

Equations (1) (2) and (3) are rearranged as:

    m1  ! (E1  E 2 ) x1  E 2 x2  E1 x0 )  (k1  k2 ) x1  k2 x2  kx0 ) x1    m2 2 ! E 2 x1 )  (E 3  E 2 ) x2  E 3 x3  k2 x1  (k2  k3 ) x2  kx2 ) x

   m3  ! E 3 x2  E 3 x3  k3 x2  k3 x3 x3
Then these equations are converted into matrix/vector notation as:

  M ! Ax  Kx  f x
Where x( t ) ! ( x1, x 2, x3)T M is a diagonal matrix involving the masses mi A is a symmetric tridiagonal matrix involving the damping coefficients i K is a symmetric tridiagonal matrix involving the stiffness coefficients ki  f is a vector involving the ground displacement xo and velocity x0 then we got:

By neglecting damping mechanism A and dynamics of the unforced system f e.g. (A=0, f =0), then: =-Bx where B ! M 1 K  x Where

x(t ) ! e iwt (i2= -1; v is a constant vector)

1 0 0

0 2

0 0 3

  x1 2 !  x 3  x

 E2 E1  E 2 E 2  E3 E2 0  E3

 0 x1 k1  k 2   E 3 x2   k 2 x 0 E3  3

 k2 k 2  k3  k3

0 x1  k 3 x 2 k 3 x 3

m2 m3 adj M ! 0 0
M ! m1m2 m3
M 1

0 m1m3 0

0 0 m1m2

adjM det M

1 / m1 M
1

! 0

1 / m2 0 0 1 / m3
must satisfy:

We need to show that vector v and natural frequencies

( B  w 2 I )v ! 0
By substituting x(t ) ! eiwt ,

 x(t ) ! iwe iwt ,

  (t ) !  w 2 eiwt to the X !  Bx  x

We have obtained  w 2 eiwt v !  Be iwt v ( B  w2 I )eiwt v ! 0 ( B  w2 I )v ! 0 where


(frequency) is the eigenvalues of characteristic equation B  w2 I ! 0

k1  k2 m1 k B!  2 m2 0

k2 m1 k 2  k3 m2 k  3 m3


0 4 2 0 k3  !  2 3  1 Where mi ! 3, k1 ! 6, k2 ! 6, k3 ! 3 , m2 k3 0  1 1 m3

Power method is used to obtain an approximation to the dominant eigenvalue, Pm , and corresponding eigenvector, vm , of the matrix B.

y1

1.0000

2.0000

1.0000

4.0000

1.0000

5.000

1.0000

27/5

1.0000

y2

1.0000

-2.0000 -1/2

-7/2

-7/10

-42/10

-21/27

y3

1.0000

1/2

1/10

8/10

4/27

2.00

4.00

5.00

27/5=5.4=w2

=5.4;

= 2.32 1.4142; 2.3761]

m =[0.5952;

= 2.3761 exact solution using MATLAB


% natural frequencies clear m=3; k1=6; k2=6; k3=3; M=zeros(3,3); K=zeros(3,3); for n=1:3 M(n,n)=m; end K(1,1)=k1+k2; K(1,2)=-k2; K(2,1)=-k2; K(2,2)=k2+k3; K(2,3)=-k3; K(3,2)=-k3; K(3,3)=k3; B=inv(M)*K; disp('natural frequencies:' ) f=sqrt(eig(B)) disp('max frequency:' ) fmax=max(f) return

% natural frequencies clear m1=1; m2=3; m3=3; k1=6; k2=6; k3=3; M=zeros(3,3); K=zeros(3,3); for n=1:3 M(1,1)=m1; M(2,2)=m2; M(3,3)=m3; end K(1,1)=k1+k2; K(1,2)=-k2; K(2,1)=-k2; K(2,2)=k2+k3; K(2,3)=-k3; K(3,2)=-k3; K(3,3)=k3; B=inv(M)*K; disp('natural frequencies:' ) f=sqrt(eig(B)) disp('max frequency:' ) fmax=max(f) return

With varying the value of mi it is observed the behaviour mostly can be affected by m1 rather than m2 and m3 since increasing the value of m1 decreased frequency more than by increasing m2 and m3 as the investigation below may confirm that: m1=1; m2=3; m3=3; fmaX =3.6315 m1=3; m2=1; m3=3; f max=3.3183 m1=3; m2=3; m3=1; fmax=2.4495

It has been observed that by either equal mass distribution or by just increasing the value of m1 less frequency is achieved but the equal mass distribution is more likely because it gives lesser frequency than the other case. m1=3; m2=3; m3=3; fmax =2.3613 m1=9; m2=3; m3=3; f max=1.9680

2) case(d)
L=100; D=3; c=0.02; As=25; Qt=10; H(0)=12; Q(0)=30

dH 1 Q  Qt ! dt As

H n1 ! H n  hf n
A dQ !  g H  c Q Q L dt Qn1 ! Qn  hf ( Qn ,tn )

tn

Hn

f(Hn)

Qn

f(Qn)

12

0.8

30

-20.7

0.5

12.4

0.38

19.65

-13.88

12.59

0.11

12.71

-10.91

1.5

12.64

-0.112

7.2

-9.43

12.58

-0.30

2.48

-9.04

At h=1.5 and Hn =12.64 water reaches its maximum value compared to exact solution, as shown below in graph 1.
% surge tank model % ode system integrated using ode45 function system=surge clear global g As A L c QT g=9.81; L=100.0; D=3; c=0.02; As=25.0; QT=10.0; HO=12; QO=30; % pipe area A=(pi*D^2)/4; zO=[HO,QO]; %initial values tspan=[0,100] %time interval [t,z]=ode45(@rhs,tspan,zO); plot(t,z) % ++++++++++++++++++++++++++++++ function f=rhs(t,z) global g As A L c Q T f=zeros(2,1); f(1)=(z(2) -QT)/As; f(2)=-((g*A)/L)*(z(1)+(c*z(2)*abs(z(2)))); return

Value of L, D and c are changed to 50, 1.25 and 0.01 respectively to investigate how the system will respond to changing values.

tn

Hn

f(Hn)

Qn

f(Qn)

12

0.8

30

-20.7

0.5

12.4

0.38

19.65

-13.88

12.59

0.11

12.71

-10.91

1.5

12.64

-0.112

7.2

-9.43

It is observed that lots of iterations are required to find maximum water level since decreasing L, D, c also it will reach its highest peak after many iterations with Eulers method. And its exact solution is shown in graph.

As t

, values of H and Q become constant about -2.5 and 10 respectively. But system behaviour

depends on values of c as it decrease system fluctuate and never become constant and it increase, fluctuation in the system will disappear. C in the system behaves like damping, so when system is damped, there will be no frequency on that, while system is undamped, frequency will go up and never stop. Clearly all investigations are in figure below.

1 when t

2 when c=0.001

3- when c=0.05

4- when c=0

You might also like