July 3, 2004 23:17 WSPC/Book Trim Size for 9in x 6in wsdac04
24 Computational Methods for PDE in Mechanics
following initial and boundary conditions for equation (3.1.7)
∈
U(x,0) =ϕ(x),0≤x≤1,
U(0,t)=g
1(t),U(1,t)=g 2(t),0<t≤T,
(3.2.5)
which are discretized as
∈
u
i,0=ϕi,i=0, ..., m, m∆x=1,
u
0,j=(g
1)j,um,j=(g
2)j,j=1, ..., n, n∆t=T,
(3.2.6)
whereϕ
i=ϕ(x i), (g
1
)j=g1(tj), (g
2
)j=g2(tj). The starting values
u
i,j,i=0, ..., m, j= 0 are given by (3.2.6) 1. Next, for every fixedj,the
valuesu
i,j+1,i=1, ..., m−1, are computed by means of (3.2.3). Lastly,
the valuesu
0,j+1,um,j+1 are provided explicitly by (3.2.6)2.
However, reliable results are guaranteed only when the method isstable
with respect to data perturbations (errors) which inevitably arise during the
computational process. Furthermore, theconsistencyof the finite-difference
scheme,aswellastheconvergence, must be analyzed. All these questions
will be discussed in the next sections.
3.3 Matrix form
Formula (3.2.3) can be organized in a more compact form, which will be
used later. Indeed, by defining the column vectoru
j
u
T
j
=[u 1,j...um−1,j], (3.3.1)
and the (m−1)×(m−1) matrixA
A=
1−2r
r
r
1−2r
·
r
·
·
·
·
r
·
1−2r
r
r
1−2r
, (3.3.2)
the following matrix form can be given to (3.2.3)
u
j+1=Au j+bj, (3.3.3)