Sie sind auf Seite 1von 13

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 1
Problem 1.1 Compute the oating-point-representation of x = (76.2)10 and x = (0.0752)10 for p = 2 and t = 10 Problem 1.2 In order to compute (71.32)10 + (6.463)10 (58.91)10 (57.32)10 p = 8 and t = 6.

1. determine the binary representation (6 digits) of these numbers, 2. compute the binary result, 3. round and transform back. Problem 1.3 Consider the parabola dened as p(x) := x(w x) with an unknown parameter . By a measurement it is known that p(d) = q . 1. Derive a formula for h := p( w 2 ) as h = H (d, q, w ) 2. Assume the following values: d = 2 (exactly) , q = 1.96(1 + r), w = 100(1 + t)

with small relative errors r, t. Derive an exact formula for the resulting relative error in h. 3. Obtain a rst order estimate for the resulting relative error. Problem 1.4 In the context of friction and movement the function f : [0, 1] [0, /2[ R, f (s, t) := s cos(t) + s sin(t)

occurs. Compute a rst order estimate for the relative error of f for the data s = 0.6, |rs | 0.01 Problem 1.5 Consider the algebraic equation xn + ax 1 = 0, where a > 0 , n N> 1 . t = /3, |rt | 0.05.

1. Show htat tis equation has exactly one positive root z (a). 2. Obtain a formula for the condition number of z (a) and bounds for it. Problem 1.6 Derive safe algorithms to evaluate the following expressions for small |x|: 1x 1 1+x 1+x
3

1 + x2

1 x2

1 cos(x) . x

Give examples demonstrating the validity of your solutions. Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 2
Problem 2.1 A person at the top of a tower of height h observes a distant point which appears at the horizon. He measures the distance from the foot of the tower to this point as s = 25 1km and knows that the radius of the earth is r = 6366 10km. 1. Verify that h = h(s, t) = r
s ) 1 cos( r . s cos( r )

2. Obtain a rst order relative error estimate of h. 3. How can one simplify the computation of the error estimate by treating the expression for h a a composed formula? Problem 2.2 Consider the following two ways to evaluate the expression x2 y 2 : s1 =x2 s2 =y 2 s3 =s1 s2 t1 =x y t2 =x + y t3 =t1 t2

1. Evaluate this expression for x = 727 and y = 724 in both ways in 3-digit-decimal-oatingpointarithmetic. 2. Give an error analyis for both evalutations for machine numbers x, y and explain the dierence. Problem 2.3 The integrals
1

an :=
0

tn dt t+5

satisfy the recursion 1 5an1 , a0 = ln(1.2). n Show that these equations dene a stable recursion when they are used in the backward-direction: an = an1 = 0.2( What is a suitable starting value? Problem 2.4 Let the sequence (an ) be recursively dened by an+2 = 10.5 an+1 5 an . 1. Verify that the solution depending on the inital values is an = 2a1 a0 n 20a0 2a1 1 10 + . 19 19 2n 1 an ). n

2. Compute the an recursively in oating-point-arithmetic for the initial values a0 := 2, a1 := 1 and a0 := 2/3, a1 := 1/3.

Compare with the exact solution and explain. Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 3
Problem 3.1 Consider the iteration x0 := 1, dened with the function g : R R, g (x) := 1 x(45 10x2 + x4 ) 24 xn+1 := g (xn ), n N0

Show that the sequence (xn ) converges, determine its limit and the order of convergence. Problem 3.2 Consider the function f : R>0 R, 1. Sketch the graph of this function. 2. For each zero of f determine a suitable starting interval for the bisection method and compute approximations for the zeros. Problem 3.3 Consider the following algorithm for a given x ]0, 1[: z=0, d=1/2, u=x-1/4 for n = 2..N: if u<0: d=d/2, u=2u+z+d+d/2 else: z=z+d, d=d/2, u=2u-z-d/2 Verify that this is a sophisticated way to compute an approximation of x on the basis of bisection. How can this algorithm be extended to R>1 ? What is its advantage? Problem 3.4 The function f : R> 0 R, f (x) := ln(x) + x2 2 has two zeros (cf. problem 2). Try to rewrite the equation f (x) = 0 as simple xed point problem which satises the conditions of the contraction mapping principle. Remark: You will have to treat the two zeros dierently. Problem 3.5 Let I := [a, b] R be a closed interval and g : I I an increasing continous map. Dene the sequences (xn ), (yn ): x0 := a, xn+1 = g (xn ) and y0 := b, yn+1 = g (yn ). f (x) := ln(x) + x2 2

1. Prove the following statements (xn ) is increasing and (yn ) decreasing. (xn ), (yn ) converge and their (in general dierent) limits are xed points of g 2. Apply these results to g : [1, 4] R, g (x) := x4 + 20x . 2x3 + 10

Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 4
Problem 4.1 The equation 3x2 + tan(x) = 0 has two negative solutions. For each of them dene a suitable xed-point-iteration and verify the conditions of the contraction mapping principle. Compute an approximative solution. Problem 4.2 Let z R be a k-fold zero of a smooth function f : R R, which means: There exists such that f (x) = (x z )k g (x), |x z | < where g is a smooth function with g (z ) = 0. Prove that z is a simple zero of the function h(x) := f (x)/f (x). Problem 4.3 A typical situation for the geometrical interpretation of Newtons method is the following: Consider a twice continuously dierentiable function f : [a, b] R with: f (a) < 0, f (b) > 0, f (a) > 0, x [a, b] : f (x) 0. R> 0

Prove that Newtons method with initial value x0 = b converges decreasingly towards the unique zero of f. Problem 4.4 Sketch the graph of the function f : R R, f (x) := x2 + exp(x) 5

and compute approximations for its zeros by Newtons method. Try as starting values x0 = 5, x0 = 0, x0 = 5. Problem 4.5 Let f : [a, b] R be a continously dierentiable function satisfying the following conditions: f (a) < 0 < f (b) and x [a, b] : 0 < m f (x) M

with suitable values m, M R. Show that 1 f (x) M satises the conditions of the contraction mapping principle. g : [a, b] R, g (x) := x Apply this principle to compute the positive solution of f (x) := x 1.32 sin(x) Problem 4.6 Let z R be a zero of the polynomial
n

P (x) =
k=0

ak xk ,

an := 1.

Prove the estimates:


n1

|z | max{1,
k=0

|ak |}

and

|z | max{|a0 |, 1 + |a1 |, . . . , 1 + |an1 |}

Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 5
Problem 5.1 The polynomial P : R R, P (x) := x5 5.7x4 7.4x3 + 54.5x2 + 12x 130

has ve real zeros. Compute them. Start with x0 = 7. Problem 5.2 Solve the nonlinear system of equations 2x y = 0.2 exp(x) 2x + 2y = 0.2 exp(y )

by the 2-dimensional Newtons method. Start with (0, 0) and execute 2 steps. 4-digit-oatingpoint-arithmetic is sucient. Problem 5.3 Show that the nonlinear map G : R 2 R2 , G(x, y ) := 1 1 exp(x2 y 2 ), (x2 y 2 ) 6 5

satises the condition of the contraction mapping theorem on the set D := [1, 1] [1, 1] := {(x, y ) R2 | 1 x 1, 1 y 1} with respect to the . Problem 5.4 Compute all solutions of the complex equation 2 + 3i = (x + yi)3 . by applying Newtons method to the real and imaginary part of this equation. Use the initial approximations: (2, 1), (1, 1), (0.5, 1). Problem 5.5 Dene 2 1 A := 2 4 2 1 10 20 4 4 2 3 8 8 10 2 4 2 y := 16 2 14 6 y := 30 36
-Norm.

Obtain an approximate solution with error estimate.

1. Solve the system Ax = y by the elimination method of Gauss with column-pivot-search. 2. Gather the appropriate values to obtain the corresponding triangular decomposition. 3. Use this decomposition to solve Ax = y . Problem 5.6 Dene 1 1 A := 2 3 1 1 1 3 1 2 , 2 1 1 3 3 4 2 3 y := 5 5

Compute the triangular decomposition A = LR into a lower and and upper triangular matrix (Lii = 1) and use it to solve the system Ax = y . Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 6
Problem 6.1 Prove Lemma 3.1.5. Problem 6.2 Let A be a bandlimited n n - matrix, i.e. there exists m N such that aij = 0 if |i j | > m.

Prove: If A has a L-R-factorization then L and R are bandlimited too. Problem 6.3 Dene 1 p p p2 + 4 A := 3 3p 2 2p with a real parameter p. 1. Compute the Cholesky-decomposition of A (depending on p). 2. Use it to compute the solution of Ax = y for the case p = 1. Problem 6.4 Dene the matrix 8 A := 4 1 Problem 6.5 Solve the linear system of equations Ax = y , 3 11 14 A := 4 20 23 , 12 18 17 1 0 1 6 7 2 3 3p 10 6 2 2p 6 13 5 9 y := 16 19

Compute a QR-decomposition f ur A by the Gram-Schmidt-procedure.

6 y := 47 11

by Jacobi-rotations. Obtain and verify the Q-R-decomposition of A. Problem 6.6 Solve the linear system of equations Ax = y , 2 2 3 A := 4 14 3 , 4 15 3

15 y := 15 3

by Householder-transformations. Obtain and verify the Q-R-decomposition of A. Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 7
Problem 7.1 Write an algorithm to solve a linear system of equations Ax = b where A is a positiv denite tridiagonal matrix. Try to save storage space. Test your algorithm with the example: ai,i := 2 + h2 , ai,i1 := 1, bi := h2

with i = 1 . . . n 1, h := 1/n and n N variable. Problem 7.2 Let A Rnn be regular, u, w Rn . Derive a condition so that the matrix B := A + uwT is regular and determine a representation for B 1 making use fo A1 . Hint: Try to determine x from Bx = y . Problem 7.3 Examine the problem to compute HA where 2 H = I vv T , c with v Rn and c := v
2 2

is a Householder-matrix and A a given n n-matrix. Count the number of additions and multiplications for the following two methods: 1. Compute H explicitly and then HA by standard matrix multiplication. 2. Compute wT := v T A and then 2 HA = A vwT . c Problem 7.4 Solve the following linear system Ax = y 14 25 8 5 A := 0 18 8 22 Problem 7.5 Fit a circle through the following points by the least-squares-principle: (x, y ) = ((0, 0), (4, 0), (0, 4), (5, 6) Hint: If the circle is to be xed by its center (p, q ) and its radius r then (x, y ) satisfy: 2px 2qy + p2 + q 2 r2 = x2 y 2 . by Householder transformations: 45 4 1 9 10 7 . y := 81 18 9 27 17 29

Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 8
Problem 8.1 Determine the least-squares-solution for the overdetermined system equation: 2 1 1 2 1 1 1 2 b := A := 2 1 3 0 1 1 3 4 Problem 8.2 Consider the overdetermined system Ax = b where 3 3 6 3 3 6 A := 3 9 0 3 9 12 Ax = b by solving the normal .

6 3 b := 6 . 3

Transform it by Housholders method into a system with upper triangular matrix and read o its least-squares-solution. Problem 8.3 Determine the coecients a, b, c R which minimize the least-squares-sum
5

a + b 21+xi + c 24xi yi
i=1

for the data

x 1 0 1 y 19 12 12

2 19

4 72

Problem 8.4 Solve the linear system of equations 12 1 A := 2 1 1. by the Jacobi-iteration, 2. by the Gau-Seidel-iteration Will theses methods converge? Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities). Ax = y with 2 8 3 1 3 1 2 1 10 1 1 6 29 5 y := 42 24

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

Problemsheet No 9
Problem 9.1 Let A be an n n-matrix satisfying A

< 1.

Prove: The matrix I A is regular and

y Rn :

(I A)1 y =
j =0

Aj y.

Hint: Consider the iteration: xn+1 = Axn + y, Determine an estimate for (I A)


1 .

x0 = 0.

Deduce a sucient condition for the following conclusion for n n-matrices A, : If A is regular then A + is regular too. Problem 9.2 Determine exactly all eigenvalues and corresponding eigenvectors for the following matrices: 1 1 5 1 1 3 3 1 1 1 3 A = 3 5 3 B = 5 1 1 1 6 6 4 1 3 1 1 Does A or B admit a diagonal representation? Problem 9.3 Apply the theorem of Gershgorin to the following matrix 1 + i 0.2 0.2 0.1 1 i 0.2 A := 0.5 0.5 3 0.3 0.3 0.1 Problem 9.4 Dene the following matrices: 38 71 28 55 105 45 15 125 45 C := 115 A := 92 159 62 B := 75 143 241 93 75 105 25 235 25 5 205 85 425 185 A and to AT : 1 0.5 . 0.1 5

Try to determine approximately the dominant eigenvalue and a corresponding eigenvector. Use the initial vectors: (1, 1, 1) and (1, 2, 6). Problem 9.5 Let A = LLT whith 2 L := 1 1 0 2 0 0 0 2

1. Execute 2 steps of the inverse power method to compute an approximation for the smallest eigenvalue of A. Begin with x0 := (9728, 5120, 6144) 2. This problem has an integer solution. From your computation guess the exact smallest eigenvalue and a corresponding eigenvector. Verify your guess. Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

10

Problemsheet No 10
Problem 10.1 Let A be an n n Hermitian matrix. Dene the Rayleigh-quotient: R(x) := (Ax, x) (x, x) (x Cn \ {0}).

Show: All eigenvalues of A are real, if 1 (n ) denotes the smallest (largest) eigenvalue of A then x Cn \ {0} : 1 R(x) n Examine the convergence of (R(xn )) if the sequence (xn ) is computed by the power method (i.e. xn := An x0 ). Hint: An Hermitian matrix has an orthonormal basis of eigenvectors. Problem 10.2 Determine the largest eigenvalue of the matrix 5 4 A := 1 1

4 5 1 1

1 1 4 2

1 1 2 4

by the power method. Estimate the eigenvalue by the Rayleigh-quotient. Problem 10.3 Solve the following initial value problems 1. (t 2)x (t) = x(t) + 2(t 2)3 , x(3) = 2. 2. x (t) = 2x(t)/t + t/2, x(1) = 1. 3. x (t) = t3 /(1 + x(t))2 , x(0) = 2. 4. tx (t) x(t) t2 x(t)2 = 0, x(1) = sin(ln(2)). Hint: Derive a dierential equation for y (t) = x(t)/t. Problem 10.4 Consider the initial value problem x (t) = tx(t) , 1 + x(t) x(1) = 1.

1. Determine an exact equation F (x(t), t) = 0 for the solution x(t). 2. Compute an approximation for the value x(2) from this equation. Use Newtons method, iterate until 2 digits remain unchanged. Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

11

Problemsheet No 11
Problem 11.1 Consider the initial value problem: x (t) = 2tx(t)2 , x(2) = 1 . 5

1. Determine the exact solution. What is its range of denition? 2. Compute an approximation by Eulers method. Problem 11.2 Consider the initial value problem x (t) = [t + x(t)]2 , Solve it approximately by Eulers midpoint method. Problem 11.3 The method of Heun is dened by (t, x, h) := [f (t, x) + f (t + h, x + hf (t, x))]/2. 1. Determine the truncation error for this method. 2. Apply this method to compute an approximate solution for the initial value problem x (t) = 1.5t2 (x(t)2 1), Problem 11.4 Consider the one-step-method dened by (t, x, h) := f (t, x) + 0.5h(t f (t, x) + f (t, x)x f (t, x)) 1. Determine the truncation error for this method. 2. Compute an approximation of x(2) by this method where x (t) = t2 exp(2x(t)), Problem 11.5 solve the following initial value problem approximately by the Runge-Kutta-method: x (t) = Problem 11.6 Transform the initial value problem x (t) = (tx (t) x(t))3 , x(0) = 1, x (0) = 2 2t , 1 + exp(x(t)) x(1) = 0. x(1) = 0. x(1) = 0. x(1) = 0.

into a rst order system and solve it by Eulers midpoint method. You may compare with the exact solution x(t) = 2t 1 t2 Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

12

Problemsheet No 12
Problem 12.1 Compute the interpolating polynomial p in Lagrange form for the data x = 0, 1, 4, 9. (x, x), Problem 12.2 Compute the interpolating polynomial p in Newton form and the values p(3) and p(3) for the followint data x y Problem 12.3 Compute the interpolating polynomial in Newton form which aproximates the function: f : [0.5, 0.5 ] R, Use the values of f (x), f (x) := sin(x). -2 179 -1 54 0 1 2 159 5 -24

1 1 1 1 1 1 x = , , , 0, , , 2 3 4 4 3 2 How can one utilize the fact that f ist an odd function? Problem 12.4 Dene the function f : R> 0 R, f (s) := s + s3 . In order to determine an approximation for z R>0 with f (z ) = 1.5 1. determine the interpolating polynomial p2 in Newton form for the data (f (x), x), x = 0.5, 1, 2

2. then determine the interpolating polynomial p3 for the data (f (x), x), x = 0.5, 1, 2, p2 (1.5)

Determine p3 (1.5) and improve it by 1 step of Newtons method to solve f (s) = 1.5. Problem 12.5 Let s : [0, 12] R denote the interpolating natural cubic spline function for the data x y Compute s(5). Problem 12.6 Compute the natural cubic spline aproximation for the function: f : [0.5, 0.5 ] R, Use the values of f (x), f (x) := cos(x). 0 1 2 7 6 11 10 3 12 -1 .

1 1 1 1 1 1 x = , , , 0, , , 2 3 4 4 3 2 How can one utilize the symmetry of f ? Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

UDE, Mathematik: Numerical Mathematics for Engineers (ISE) (W 13/14)

13

Problemsheet No 13
Problem 13.1 Compute approximate values for the integral
2 1

sin(x) dx x

1. by Simpsons rule, with stepsize h = 2, 1, 1/2, . . . 2. by the Gaussian quadrature formula of 7.2 applied to the interval [1, 2] the intervals [1, 1.5] and [1.5, 2] Problem 13.2 Let f : [a, b] R be a function and S : [a, b] R the corresponding natural cubic spline interpolation with equidistant n + 1 points. Show that
b

S (t) dt = T (h)
a

h3 ba (z0 + 2z1 + . . . + 2zn1 + zn ), h := , 24 n

where T (h) denotes the result of the trapezoidal rule und zi are the moments of S . Apply this method to compute with n = 3 an approximation for
3 0

1 dt. 1 + t2

Problem 13.3 Determine an analytic expression for the solution of the initial value problem x (t) + 2tx(t) = t2 , x(1) = 2.

This involves an integral. Evaluate it with Simpsons rule on the intervals [1.0, 1.5], [1.5, 2.0] and obtain an approximation for x(2). Problem 13.4 The Gauss-quadrature-formula with 3 abszissas reads
1

f (x)dx
1

1 5f ( 9

3 ) + 8f (0) + 5f ( 5

3 ) . 5

In order to determine an approximation for


4 0

1 dx 1 + x4 + 3x2

apply this formula to 1. the interval [0, 4], 2. the interval [0, 2] and [2, 4].

Please note: In the exam you may use one sheet of size din A4 with informations only. For computations you may use a simple pocket calculator (no programming or computer algebra facilities).

Das könnte Ihnen auch gefallen