26.12.2013 Views

Adaptivity with moving grids

Adaptivity with moving grids

Adaptivity with moving grids

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

102 C. J. Budd, W. Huang and R. D. Russell<br />

where the function v(y) is given by the Barenblatt–Pattle profile (Barenblatt<br />

1996). It is also well known that any positive initial data lead to a solution<br />

u(x, t), which converges towards to a self-similar solution in the sense that<br />

t 1/3 u(t 1/3 y, t) → v(y).<br />

Similarly, it is shown in Budd et al. (1999b) that if the monitor function<br />

is chosen to give a scale-invariant scheme (for example M = u), then this<br />

scheme admits a set of discrete self-similar solutions on a <strong>moving</strong> mesh<br />

which take the form<br />

U i (t) =(t + C) −1/3 V i , X i (t) =(t + C) 1/3 Y i .<br />

Note that the product<br />

W i ≡ U i X i = V i Y i<br />

is invariant in time. In Figure 5.1 we show the results of a computation presented<br />

in the computational domain, using a scale-invariant <strong>moving</strong> mesh<br />

method <strong>with</strong> M = u and a centred finite difference discretization. In this<br />

calculation the initial data at t = 0 were taken to be an irregular function,<br />

and results are plotted at times t =0andt = 10. We also show (dashed)<br />

two discrete self-similar solutions <strong>with</strong> values of C chosen so that they initially<br />

lie above and below the solution. Note that the solution calculated is<br />

sandwiched between these two functions. We also present the corresponding<br />

mesh X i (t) and the scaled mesh Y i (t) =X i (t)/t 1/3 . It is clear from<br />

these figures that the computed solution converges towards the discrete selfsimilar<br />

solutions (in correspondence <strong>with</strong> the continuous theory) and that<br />

the <strong>moving</strong> mesh tends towards the one for which Y i is constant in time so<br />

that X i (t) scales asymptotically as t 1/3 . Similar figures for solutions of the<br />

porous medium equation computed using a scale-invariant ALE method in<br />

two dimensions are given in Baines et al. (2006).<br />

5.2. Blow-up and related problems<br />

5.2.1. Parabolic blow-up<br />

As mentioned in Section 5.1, a significant success in the application of <strong>moving</strong><br />

mesh methods occurs in the study of parabolic partial differential equations<br />

(and also of systems of PDEs) which have solutions that blow up,<br />

so that the solution, or some derivative of it, becomes infinite in a finite<br />

time T . We now consider such problems in a little more detail. Blow-up<br />

in the solution often represents an important change in the properties of<br />

the model that the equation represents (such as the ignition of a heated gas<br />

mixture), and it is important that it is reproduced accurately in a numerical<br />

computation. A survey of many different types of blow-up problem is presented<br />

in Samarskii et al. (1973). Blow-up typically occurs on increasingly<br />

small time and length scales, and hence it is usually essential to use both

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

Saved successfully!

Ooh no, something went wrong!