20.07.2013 Views

Молодой учёный

Молодой учёный

Молодой учёный

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

10 Математика<br />

«<strong>Молодой</strong> <strong>учёный</strong>» . № 3 (50) . Март, 2013 г.<br />

with a discrete function that satisfies the estimate<br />

h<br />

h<br />

2<br />

ξk -1 ≤ck ( - 1) h.<br />

(29)<br />

1<br />

Because of Taylor series in x of r ( tk, x)<br />

in the vicinity of point x i we get equality<br />

xi+<br />

12<br />

k<br />

r( tk, xi) = ∫ r( tk, x) dx h+ e<br />

x<br />

i<br />

i-12<br />

k<br />

where ei<br />

2<br />

≤ ch 1<br />

2<br />

1 ∂ r<br />

with c1 = max ( t , ) .<br />

[0,1]<br />

2 k x<br />

24 x∈<br />

∂x<br />

(30)<br />

Because of Theorem 1<br />

k-1<br />

xi+ 12 Ai+<br />

12<br />

r k = r k-1<br />

x<br />

k 1<br />

i 12 A<br />

-<br />

- i-12<br />

∫ ∫<br />

( t , x) dx h ( t , x) dx h.<br />

Instead of let use its piecewise linear periodical interpolant Then<br />

k-1 Ai+ 12<br />

r( t 1 1, ) k k x dx h -<br />

A<br />

- =<br />

i-12 k-1<br />

Ai+<br />

12<br />

k<br />

r 1 int ( t 1,<br />

)<br />

k<br />

A<br />

k x dx h h<br />

-<br />

- + i<br />

i-12<br />

k<br />

hi<br />

≤ 2<br />

k-1 i+ 12- k-1<br />

i-12 1<br />

2 =<br />

x∈[0,1]<br />

2<br />

∂ r<br />

2 k<br />

∫ ∫<br />

where chA ( A ) with c max ( t , x)<br />

.<br />

8 ∂x<br />

Thus, we get equality<br />

k-1<br />

Ai+<br />

12<br />

k k<br />

k i = k-1<br />

A<br />

int k -1<br />

+ i + i<br />

i-12<br />

r( t , x ) ∫ r ( t , x) dx h h e .<br />

(32)<br />

For we use (21) and (28):<br />

k-1 k-1<br />

Ai 12 A<br />

h + i+<br />

12 h<br />

k i = k-1 int k 1 k 1<br />

A<br />

- + - k 1<br />

i-12 A<br />

-<br />

i-12<br />

∫ ∫ (33)<br />

r ( t , x ) r ( t , x) dx h ξ ( x) dx h<br />

where values of are constructed by piecewise linear periodical interpolation.<br />

Now let subtract (33) from (32), multiply its modulus by h , and sum for all i = 0, 1,...,n–1:<br />

r( t ,) r ( t ,)<br />

⎛<br />

⎜h h<br />

⎝<br />

e h ξ ( x) dx<br />

⎞<br />

⎟.<br />

⎠<br />

Due to Theorem 3 last terms ( ,) ⋅- in brackets ( ,) ⋅ is ≤ evaluated ( + ) by + . Thus<br />

∑ ∫ (34)<br />

k ⋅-<br />

h<br />

k ⋅<br />

h<br />

≤<br />

1<br />

0≤≤ i n-1<br />

k<br />

i +<br />

k<br />

i +<br />

k-1<br />

Ai+<br />

12<br />

k-1<br />

Ai-12<br />

h<br />

k-1<br />

r tk h<br />

r tk h<br />

1<br />

c1 c2 2<br />

h<br />

h<br />

h<br />

ξk -1<br />

1<br />

h<br />

h<br />

2 h<br />

h<br />

k k<br />

1<br />

1 2 k 1<br />

1<br />

r( t ,) ⋅-r ( t ,) ⋅ ≤ ( c + c ) h + ξ - .<br />

(35)<br />

Let put c = c 1 + c 2 then this inequality is transformed with the help (29):<br />

h<br />

r( t ,) ⋅-r ( t ,) ⋅ ≤ ( c + c ) kh<br />

k k<br />

h<br />

1<br />

1 2<br />

that is equivalent to (27).<br />

We can see that at last level we get inequality<br />

h<br />

h<br />

2<br />

( tm,) ( tm,) cT h .<br />

1<br />

2<br />

r ⋅-r ⋅ ≤ t<br />

(36)<br />

In some sense we got a restriction on temporal meshsize t to get convergence. For example, to get first order of convergence,<br />

it is enough to take<br />

t = ch<br />

with any constant c independent of t and h. But this restriction is not such strong up to constant as Courant–Friedrichs–<br />

Lewy (CFL) condition:<br />

(31)

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

Saved successfully!

Ooh no, something went wrong!