Advanced Numerical Analysis for Chemical Engineering

Final Examination -1 (3 hrs.)

1. Consider matrix 2 3
1 2 5
6 2 1 4 7
6 7
6
M =6 1 2 5 7
7
4 0 1 2 5
2 0 2

(a) Find a basis for column space and for null space of M . (3 marks)
T
(b) Project vector b = 1 1 1 1 1 onto the column space of matrix M . Also,
…nd component of b in left null space of M . (5 marks)
(c) Does the set of functions of the form
f (t) = a + bt
,where a and b are nonzero real constants, constitute a linear vector space? Here
set of scalars is real line and t 2 [0; 1] : (3 marks)
(d) Consider application of …nite di¤erence method to solving ODE-BVP with non-
equidistant grid point i.e.
zi = zi+1 zi ; i = 0; 1; 2; ::::
Derive expressions for approximating the second derivatives (3 marks)
(2) 2 yi+1 yi yi yi 1 1 (3)
yi = y ( zi zi 1 ) + ::::
zi + zi 1 zi zi 1 3 i
2. It is desired to …t the data given in the table below
r (rate) C T
1 0.8 4
7.5 0.8 5
0.7 0.4 4
5 0.4 5
a nonlinear reaction rate equation of the form
2C
r=K exp( A=T )
2+C
(Note that, to simplify numerical the calculations, scaled values of temperature variable
are given in the above table).

(a) Rearranging the model as

r(2 + C) A
ln = ln (K)
2C T
determine the parameters A and K using linear least square estimation. (5
marks)

1
 (b) Estimate the covariance matrix of the parameters estimated in part (a), i.e.,
b = ln(K) A . (4 marks)
(c) Suppose, instead of using linearizing transformation given above, it is desired to
estimate model parameters using Gauss-Newton method. Perform one iteration of
Gauss-Newton starting of the estimate of A and K generated in part (a). (5
marks)

3. It is desired to apply the method of …nite di¤erence to solve the following PDE

@C @ 2C
=
@t @t2

Boundary Conditions : C(0; t) = C(1; t) = 0

Initial Condition : C(z; 0) = 1

where t and z represent dimensionless time and dimensionless length, respectively.

Assuming ’n’equidistant grid points and de…ning vector
T
x = C1 C2 ::: Cn

we obtain the following set of ODE-IVP from the PDE

T
dx=dt = Ax ; x(0) = 1 1 ::: 1
2 3
2 1 0 :::: 0
6 1 2 1 :::: 0 7
1 66
7
7
A= :::: :::: :::: :::: ::::
( z)2 6
4 0 :::: 1 2 1
7
5
0 :::: 0 1 2

(a) Suppose that it is desired to solve the resulting linear algebraic equations analyt-
ically as x(t) = [ exp( t) 1 ] x(0) where A = 1
: Show that vector
T
v(k) = sin(k z) sin(2k z) ::::: sin(nk z)

is an eigenvector of matrix A with eigenvalue

2
k = [cos(k z) 1]
( z)2

where and k = 1; 2; :::n and z = 1=(n + 1): (Show calculations for 1st ; ith and the
last row). (5 marks).
(b) Comment upon the asymptotic behavior of the resulting solution as t ! 1:
(Justify your comments). (4 marks).

2
(c) Suppose, instead of solving the problem analytically, the set of ODE-IVP is to
be integrated using Crank-Nicholson method (i.e. trapezoidal rule). Find the
condition on the integration step size ’h’in terms of eigenvalues of matrix A so
that the approximation error will decay exponentially and approximate solution
will approach the true solution. (5 marks).
Note: Crank-Nicholson algorithm for the scalar case can be stated as
h
x(n + 1) = x(n) + [f (n) + f (n + 1)]
2

(d) It is desired to derive 3’rd order Gear’s implicit integration formula of the form

x(n + 1) = 0 x(n) + 1 x(n 1) + 2 x(n 2) + h 1 f (n + 1)

for numerically integrating an ODE-IVP of the form

dx=dt = f (x; t) ; I:c: : x(tn ) = x(n) (1)

from t = tn to t = tn + 1. Setup the necessary constraint equations and obtain

coe¢ cients f i g and 1 . (4 marks)
Note: The exactness constraints are given as
p
X
i = 1; (j = 0)
i=0
p p
X X
( i)j i +j ( i)j 1
i = 1 ; (j = 1; 2; :::::::; m)
i=0 i= 1

(e) A chemical reactor is modelled using the following set of ODE-IVP

dC 1 C
= 2C 2 (2)
dt V
dV
= 1 V (3)
dt
Linearize the above equations in the neighborhood of steady state C = 0:5 and
V = 1 and develop a linear perturbation model. Obtain the analytical solution for
the linearized system starting from initial condition C = 0:7 and V = 0:8: Also,
compute sti¤ness ratio and comment upon asymptotic stability of the solution.
(4 marks)