APPLICATIONS OF FD APPROXIMATIONS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS

Size: px

Start display at page:

Download "APPLICATIONS OF FD APPROXIMATIONS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS"

Georgina Butler
5 years ago
Views:

1 LECTURE 10 APPLICATIONS OF FD APPROXIMATIONS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS Ordinary Differential Equations Initial Value Problems For Initial Value problems (IVP s), conditions are specified at only one value of the independent variable initial conditions (i.c. s) For example a simple harmonic oscillator is described by A d2 y B Cy gt y 0 y o V o y location dependent variable t time independent variable p. 10.1

2 Boundary Value Problems For Boundary Value Problems (BVP s) conditions are specified at two values of the independent variable (which represent the actual physical boundaries) Example General Initial Value Problems Any IVP can be represented as a set of one or more 1st order d.e. s each with an i.c. Example d 2 y dx 2 + D Ey hx y 0 dx A d2 y B C gt y 0 y o yl Let z and we can develop a system of 2 first order O.D.E. s which are coupled z y 0 y o dz ---- B Ā -- C gt z z 0 v A A o y o y l v o p. 10.2

3 Therefore the general IVP can be written as: f 1 y 1 y 2 y n t y 1 0 y 10. n f n y 1 y 2 y n t y n 0 y n0 Possible solution strategies for higher order o.d.e. s Solutions of 1st order d.e. s single equation with associated i.c. extension to coupled sets Solution of higher order equations without reduction to a 1st order system must develop a method for a specific equation not as general may be advantageous in certain cases p. 10.3

4 Solution to a 1st Order Single Equation IVP fyt with specified i.c. yt o y o Euler Method The Euler method is a 1st order method We evaluate the o.d.e. at node j and use a forward difference approximation for j j fy j,t j y j + 1 y j fy t j t j y j 1 + y j + t f y j t j p. 10.4

5 Simply march forward in time From time level j (time ) To time level j + 1 (time t j + 1 t j + t ) t j y j+1 y j } t f (t j, y j ) t j t j+1 We note that ft j y j equals the slope at t j. Therefore to obtain y j + 1 simply add t ft j y j to y j p. 10.5

6 Example Solve yy with the i.c. y 0 1 Apply the Euler approximation to solve this IVP equation. Apply a time step equal to t 0.5. Solve up to t 3 0 t 3 y t t i i.c. information available p. 10.6

7 Discretize the o.d.e. at a general node i i y i y i Approximate using a forward difference approximation i y i + 1 y i y t i y i y i 1 + y i t y i y i Next Value Previous Value + Run Slope Equation relates a known time level i to the new time level i + 1. This process is known as time stepping or time marching p. 10.7

8 The i.c. indicates that y o 1 Take 1st time step i 0 i Note that t i i t + t o i t where t o 0 in this case y 1 y o ty o y o y y at t Take next time step i 1 i y 2 y 1 t y 1 y 1 y y at t p. 10.8

9 Take next time step i 2 i y 3 y 2 t y 2 y 2 y y at t Take next time step i 3 i y 4 y 3 t y 3 y 3 y y at t p. 10.9

10 Take next time step y at t Take next time step y at t We can continue time marching t Numerical Solution using Numerical Solution using Exact t 0.5 and the Euler t 0.1 and the Euler Solution Method Method (i.c.) (i.c.) p

11 General Observations for Solving IVP s Solution to o.d.e. s can be very simple using finite difference approximations to represent differentiation Accuracy is dependent on the time step As t 0, the solution gets better t! We need to understand the error behavior IVP s are solved using a time marching process Begin at one end and march forward up to the desired point or indefinitely At each time step, we introduce a new unknown, and solving the discrete form of the IVP at node j. y j + 1, which is solved for by writing Solutions to Boundary Value Problems Boundary value problems must be 2nd order o.d.e. s or higher We apply FD approximations to the various terms in the differential equation to obtain discrete approximations to the differential equations at points in space. Unknown functional values at the nodes will be coupled and require the solution of a system of simultaneous equations matrix methods. p

12 Example Consider a stea state 1-D problem d 2 y dx 2 + Ay B with b.c. s specified as y 0 0 and yl 0 Consider the following discretization of the domain j j-1 j j+1 n-2 n-1 n jn+1 x0 x xl At each node, we introduce an unknown (total of n+2 unknowns) The O.D.E. is approximated at generic node j as y j y j + 1 2y j + y j x 2 + Ay j B We have used a central difference approximation of O x 2 p

13 Therefore we generate n + 2 equations (1 for each interior point and 1 for each boundary node) to solve for the n + 2 unknowns: y 0 0 y 2 2y 1 + y 0 + A x 2 y 1 B x 2 y 3 2y 2 + y 1 + A x 2 y 2 B x 2.. y j + 1 2y j + y j 1 + A x 2 y j B x 2. y n 2y n 1 + y n 2 + A x 2 y n 1 B x 2 y n + 1 2y n + y n 1 + A x 2 y n B x 2 y n p

14 Collect coefficients of unknowns and write in matrix form: y 0 y 1 y 2 y 3 y n 1 y n y n B x 2 B x 2 B x 2 B x 2 B x 2 0 where 2 + A x 2 p

15 Notes on solutions Tri-diagonal systems are very inexpensive to solve when a specialized compact storage tri-diagonal solver is used We can reduce the system of simultaneous equations from n + 2 n + 2 to n n by incorporating the b.c. s into the discrete form of the differential equation at nodes 1 and n If the o.d.e. is nonlinear the method no longer generates linear set of simultaneous algebraic equations but a nonlinear set! Example Solve the following nonlinear o.d.e. d 2 y dx 2 + Dy 2 E Using a central difference approximation yields nonlinear algebraic equations y j + 1 2y j + y j x 2 + Dy j E p

16 Solution strategies include: Iterative solution of algebraic equations puts the nonlinear term on the r.h.s. and iterates until convergence. There may be convergence problems. Linearization of the nonlinear terms. Use Taylor series to approximate the nonlinear terms. p

Homogeneous Equations with Constant Coefficients

Homogeneous Equations with Constant Coefficients MATH 365 Ordinary Differential Equations J. Robert Buchanan Department of Mathematics Spring 2018 General Second Order ODE Second order ODEs have the form