IIT BOMBAY, DEPT. CHEMICAL ENGG.
Intro. to Numerical Analysis
CL 238 /244 (Jan-April 2025)
1. For each question, marks are indicated beside the question in bold.
2. Total marks: 70. There are questions on both sides of the paper.
3. This is a closed notes exam. No formula sheets are permitted.
4. Most of the marks are for correct steps. No marks for correct answers
with incorrect steps.
END-SEM EXAM (24th-APRIL-2025)
1. IVP (12 marks): Consider the following IVP:
2y dy ay (x=1) =-1
dx2 =-4y(x) -5x dx' asxs b, y(x= 1) =1, dx
where a=land b= exp(2). Do the following:
Classify the equation as linear/nonlinear, homogenous/inhomogenous and constant/variable
coefficient. No marks unless al1 the answers are correct. If you answer
partly, zero credit. [1.5]
Choosinga step size of h=, obtain the value of y(1+ h) using the RK4 method
(see formula provided at the end of the paper). [7]
" Calculate the forward error at x =l+h,given that the exact solution of the differ
ential equation verifying the initial conditions are [1)
yelx) = x In(xe) Note: e exp(1.0)
" Using one iterationof the Newton-Raphson method, estimate the backward error
at x=1+ h. Use the initial guess as 1 + h for the NR method. (2.5]
2. BVP with finite difference (13 marks): The following boundary value problem is given
to you
d'y
dx? =-4y() + 4x, 0sxsn/2
with boundary conditions y(0) =0and (=nI2) =0. Do the following:
into 5 equi-spaced intervals h - draw a picture indi
Discretise the domain (x),
Using the central difference approximation for the
cating the discretised domain.
formulae involves points i -1, iand i+ 1) and the first
second derivative (this 1),
derivative in the boundary condition
above (involves the points i+1 and i -
condition. [3]
discretise the differential equation and the boundary
algebraic equations governing the values of yi = y(x;) X; =
" Obtain a set of equation taking into account
ih, i= 1,2,3, 4,5) from thediscretised governing
both boundary conditions. [5]
and report the values of
Solve the above obtained algebraic equations directly solve - do not use any
matrix and
y1 Y2...ys (use your calculator to invert the
fancy numerical algorithm). (2]
boundary
" Given that the exact solution to the differential equation satisfying the
error at x5.
conditions is ye(x) = x+ sin(2x), calculate the magnitude forward
[1]
Use one iterationof the Newton-Raphson method, toobtain the mnagnitude of the
backward error at x5. Use initial guess to be 1.6707. [2]
3. Finite difference formula (10 marks): We wish to evaluate the fourth-order deriative
(with Ax).atthe point xË using the values of y at xi+2, Xi+l) Mi, Xi-1 and xË-2 (equi-spaced
Derive a formulae for this using the Taylor series approach discussed in the
class and obtain the truncation error (i.e. the largest term which is neglected in this
approximation as Ax-0) [10]
4. Gauss quadrature -(10 marks) Consider a problem where you encounter the following
integral:
V21+*)
Dothe following:
Using symmetry of the integrand,express this as a standard Gaussian quadrature
integral. (1)
" Derive the equations governing the 3 point Gauss quadrature formula
Show that the roots of the equations derived agree with the above formulae. [6]
" Evaluate the integral above (continuing from part 1of this problem) using the
above 3-point formula. (2]
" Calculate the magnitude of the error in this numerical integration given that the
exact value of this integral is 0.108709. (1]
matrix notation (only), find the parabola that
5. Linear least squares - (10 marks) Using
following data points (invert the relevant matrix
best fits (in the least squares sense) the
directly using your calculator)
X] =-1, y1 =1/2
x) = 1, y2=-1
Xg = 2, y3 =-1/2
X4 =3, y4 = 2
the relevant matrix which ensures that the fitted parabola min
State the property of matrix
square error. What attribute of the relevant
imises (and not maximises) the least
[8+1+1]
ensures that this property is satisfied?
Linear operators- (6 marks) A linear operator ® needs to satisfy the following property
6.
nos. aand b:
operating on functions f(x) and g(x) for real
o (af() +bg(] = a6(f(x))+bo (g0)
Which of these are linear operators (provide justification):
" o dx (isa dummy variable)
" Adding the number 3 to any function say f(). [2+2+2]
7. Matrix norm - (4 marks) For any matrix B, is the following statement true or false.
Provide reasoning:
p(B) ||B||
where || ·|| is any matrixnorm, discussed in the class. [4)
8. Coupled ODEs- (5 marks) Analytically solve the following system using the method
of eigenfunction expansion discussed in class (dot indicates derivative with respect to
time). Express the final answer using suitable constants of integration.
ij =X1t x2, 2 = 4x] +X2
USEFUL FORMULAE
" The RK4 algorithm for solving the (vector) differential eqn. = f(x, Y(*) as an IVP is:
h
yn+) =yn)
+2k,+
6 2k +k
where
k =f(*,yn)
h
ko=f(x+ h, yn+hk))
with x, =1+ nh, yn) =Y(x).
the eigenval
" The spectral radius of amatrix B is defined asp (B) = max; |;l, ; being
ues.
" Matrix norm properties for matrices A and B:
||A||>0 if A#0
- llyA|| = Iyll|A||for any scalar y
- |A-B||s||A||·||B|
- ||A-x||s||A||||Z|| for any vector i.
" Remember that ||~|| of avector ~is measure of its length and is hence always 0.
