Integral equation methods
385
The
method
to
be
described starts from Green's formula (the third
identity) in the form
L
{«>'(q)ln
Iq-pl-
«>(q)ln'
Iq-pl}
dq
=
7J(p)«>(P),
(8.111)
where p, q are vectors specifying points of the plane and on L respec-
tively; the prime denotes differentiation at the point q along the inward
normal to the domain and dq
is
the differential increment of L at q.
If
P
is
a point in D then
7J(P)
=
2n,
but
if
p
is
a boundary point on
L,
7J(P)
= O(p), where O(p)
is
the internal angle at the point p, i.e. the angle
between the tangents to L
on
either side of p. In the latter case (8.111)
becomes (Jaswon and Symm, 1977)
L
{«>'(q)ln
Iq-pl-
«>(q)ln'
Iq-pl}
dq
-O(p)«>(P) = 0,
pEL.
(8.112)
This provides a linear relationship between the boundary values of a
solution
«>
of (8.110) and those of its normal derivative. Hence, given
values of either
«>
or
a«>/an
at each point of L
we
have a linear equation
for the other corresponding boundary values,
or
a pair of equations when
«>
is given on a part of L and
«>'
on the remainder. We note that provided
L has no cusps,
7J(p)
is never zero and so by substituting the solution of
(8.112), together with the original boundary data, into (8.111) the value of
«>
at any point p in D and on L can
be
obtained.
Exc~ptions
to the
statement that the integral equation arising from (8.112) has a unique
solution are discused by Jaswon and Symm (1977) and Baker (1979). The
general basis of a numerical procedure for solving the integral equations
is
to
divide the boundary L into smooth intervals such that any comers
or
changes in form of the boundary occur at the points of subdivision. The
functions
«>
and
«>'
are approximated by constant values within each
interval.
The
boundary L itself
is
approximated by a polygon by replacing
each interval by two chords which join its end points to
the
nodal point
within it. A system of simultaneous equations results, in which all the
coefficients of the discretized
«>
and
«>'
can
be
computed analytically.
8.7.1. Two-dimensional annular electrochemical machining problem
We now discuss in more detail the integral-equation method of solving
a quasi-steady-state model of the electrochemical machining process
outlined in §2.12.3 used by Christiansen and Rasmussen (1976). Their
system and terminology are illustrated in Fig. 8.27, where
f 0
is
the
cathode and
f 1 the anode and the problem
is
to
find a function
«>
satisfying (8.110) in
the
annular space D subject to boundary conditions
«>
=
«>0
on
fo,
«>
=
«>1
on
fl.
(8.113a)