Runge-Kutta Methods: Absolute Stability and Time-Step Restriction, Assignments of Mathematical Methods for Numerical Analysis and Optimization

Solutions to problems related to the stability and time-step restriction of runge-kutta methods for numerical solutions of ordinary differential equations. It includes calculations for the two-stage runge-kutta method and the ode y' = -y/(1+y), as well as estimates and checks for the forward euler and 4th order runge-kutta methods.

Typology: Assignments

Pre 2010

Uploaded on 08/31/2009

koofers-user-1o9
koofers-user-1o9 🇺🇸

10 documents

1 / 3

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Jeffrey Hellrung
Friday, November 09, 2005
Math 269A, Assignment 06
1. Consider the two stage Runge-Kutta method
y=yn+αdtf(yn),
yn+1 =yn+dt (β1f(yn) + β2f(y)) .
Assume that the constants α,β1, and β2are chosen so that the method is second order (Assignment
3, Problem 1).
(a) Find the interval of absolute stability for the method.
(b) Does the interval of absolute stability depend upon the coefficients α,β1, and β2?
Solution
From Assignment 3, Problem 1, we have that
β1+β2= 1, αβ2=1
2.
(a) Setting f(y) = λy, we obtain
y=yn+αhf(yn) = yn+αλhyn,
yn+1 =yn+h(β1f(yn) + β2f(y))
=yn+h¡β1λyn+β2λyn+αβ22yn¢
=¡1 + (β1+β2)λh +αβ2(λh)2¢yn
=¡1 + λh + (λh)2/2¢yn
,
hence we require that
¯¯¯¯
1 + λh +(λh)2
2¯¯¯¯
<1λh (2,0).
(b) Clearly, the interval of absolute stability is independent of α,β1, and β2(assuming these constants
are chosen such that the method is second order).
2. Consider the ODE whose numerical solution you computed in Assignment 4, Computational Problem
3. dy
dt =µ0 1
βαy, y(0) = µ1
0.
(a) When α= 100.0 and β= 1.0 determine the eigenvalues of the Jacobian of this system of equations.
(b) Estimate the time-step restriction so that
i. Forward Euler
ii. 4th Order Runge-Kutta
yield a qualitatively correct solution. This is just an estimate, so you can do it graphically.
(Attached are the plots of the regions of absolute stability for Runge-Kutta methods of orders
1-4.)
(c) Use your program from Assignment 4 and check the validity of your estimate. For each method,
determine if your timestep estimate is close to the timestep required to get a “reasonable looking”
numerical solution.
1
pf3

Partial preview of the text

Download Runge-Kutta Methods: Absolute Stability and Time-Step Restriction and more Assignments Mathematical Methods for Numerical Analysis and Optimization in PDF only on Docsity!

Jeffrey Hellrung Friday, November 09, 2005 Math 269A, Assignment 06

  1. Consider the two stage Runge-Kutta method

y∗^ = yn + αdtf (yn),

yn+1 = yn + dt (β 1 f (yn) + β 2 f (y∗)). Assume that the constants α, β 1 , and β 2 are chosen so that the method is second order (Assignment 3, Problem 1).

(a) Find the interval of absolute stability for the method. (b) Does the interval of absolute stability depend upon the coefficients α, β 1 , and β 2?

Solution From Assignment 3, Problem 1, we have that

β 1 + β 2 = 1, αβ 2 =

(a) Setting f (y) = λy, we obtain

y∗^ = yn + αhf (yn) = yn + αλhyn,

yn+1 = yn + h (β 1 f (yn) + β 2 f (y∗)) = yn + h

β 1 λyn + β 2 λyn + αβ 2 hλ^2 yn

1 + (β 1 + β 2 )λh + αβ 2 (λh)^2

yn

1 + λh + (λh)^2 / 2

yn

hence we require that (^) ∣ ∣ ∣ ∣1 +^ λh^ +

(λh)^2 2

∣ <^1 ⇒^ λh^ ∈^ (−^2 ,^ 0).

(b) Clearly, the interval of absolute stability is independent of α, β 1 , and β 2 (assuming these constants are chosen such that the method is second order).

  1. Consider the ODE whose numerical solution you computed in Assignment 4, Computational Problem
    1. dy dt

−β −α

y, y(0) =

(a) When α = 100.0 and β = 1.0 determine the eigenvalues of the Jacobian of this system of equations. (b) Estimate the time-step restriction so that i. Forward Euler ii. 4th^ Order Runge-Kutta yield a qualitatively correct solution. This is just an estimate, so you can do it graphically. (Attached are the plots of the regions of absolute stability for Runge-Kutta methods of orders 1-4.) (c) Use your program from Assignment 4 and check the validity of your estimate. For each method, determine if your timestep estimate is close to the timestep required to get a “reasonable looking” numerical solution.

Solution

(a) The Jacobian of the system is simply ( 0 1 −β −α

which has characteristic polynomial

P (λ) = λ(λ + α) + β = λ^2 + αλ + β

whose roots are λ+ = α 2

4 β α^2

λ− = α 2

4 β α^2

Notice that for α^2 ≫ β, these can be approximated as

λ+ ≈ − β α

, λ− ≈ −α.

Thus, with α = 100.0 and β = 1.0,

λ+ ≈ − 0. 01 , λ− ≈ − 100.

(b) i. The interval of absolute stability for the Forward Euler method is (− 2 , 0), hence, since λ+, λ− < 0, λ+h, λ−h ∈ (− 2 , 0) ⇒ h ∈ (0, 0 .02). ii. The interval of absolute stability for the 4th^ Order Runge Kutta method is (− 2. 785 , 0), hence, since λ+, λ− < 0, λ+h, λ−h ∈ (− 2. 785 , 0) ⇒ h ∈ (0, 0 .02785).

(c) i. The following table gives the approximated solution yN to t = 10 using Forward Euler for the given timestep h: h yN

  1. 02083 (− 4 · 1012 , 4 · 1014 )
  2. 02040 (− 3 · 104 , 3 · 106 )
  3. 02020 (2. 71296 , − 180 .796)
  4. 02008 (0. 900043 , 0 .477619)
  5. 02004 (0. 905576 , − 0 .0756808)
  6. 02000 (0. 904819 , 1 · 10 −^16 )
  7. 01996 (0. 904922 , − 0 .0102694) ii. The following table gives the approximated solution yN to t = 10 using 4th^ Order Runge Kutta for the given timestep h: h yN
  8. 02857 (− 2 · 1012 , 2 · 1014 )
  9. 02817 (− 2 · 103 , 2 · 105 )
  10. 02801 (0. 477522 , 42 .7264)
  11. 02793 (0. 898467 , 0 .636066)
  12. 02786 (0. 904822 , 6 · 10 −^4 )
  13. 02778 (0. 904917 , − 0 .00890484)