21.04.2015 Views

Lecture Notes in Differential Equations - Bruce E. Shapiro

Lecture Notes in Differential Equations - Bruce E. Shapiro

Lecture Notes in Differential Equations - Bruce E. Shapiro

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

365<br />

Figure 33.3: Illustration of the Backward’s Euler Method. Instead of construct<strong>in</strong>g<br />

a tangent l<strong>in</strong>e with slope f(t 0 , y 0 ) through (t 0 , y 0 ) a l<strong>in</strong>e with slope<br />

f(t 1 , y 1 ) is constructed. This necessitates know<strong>in</strong>g the solution at the t 1 <strong>in</strong><br />

order to determ<strong>in</strong>e y 1 (t 1 )<br />

y 2<br />

y 1<br />

y 0<br />

t 0 t 1 t 2<br />

to improve the calculation of y n before mov<strong>in</strong>g on to the next mesh po<strong>in</strong>t.<br />

A common method is to use fixed po<strong>in</strong>t iteration on the equation<br />

y = k + hf(t, y) (33.37)<br />

where k = y n−1 . The technique is summarized here:<br />

• Make a first guess at y n and use that <strong>in</strong> right hand side of 33.36. A<br />

common first guess that works reasonably well is<br />

y (0)<br />

n = y n−1 (33.38)<br />

• Use the better estimate of y n produced by 33.36 and then evaluate<br />

33.36 aga<strong>in</strong> to get a third guess, e.g.,<br />

y (ν+1)<br />

n<br />

= y n−1 + hf(t n , y (ν)<br />

n ) (33.39)<br />

• Repeat the process until the difference between two successive guesses<br />

is smaller than the desired tolerance.<br />

It turns out that Fixed Po<strong>in</strong>t iteration will only converge if there is some<br />

number K < 1 such that |∂g/∂y| < K where g(t, y) = k + hf(t, y). A more<br />

stable method technique is Newton’s method.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!