L U7 = 2-00, Ug = 4-0, Ug =
13.6. ] ONE DIMENSIONAL HEAT EQUATION
Consider one dimension heat conduction equation
Ou _ 2 eu
ot ax?
Where c? ~ h/sp is the diffusivity of the substance in cm2/sec.The solution of (1) gives the
thin uniform rod after a time
We can solve (1) by the following three methods :
() Schmidt method.
(ii) Crank-Nicolson method.
(iii) Du Fort and Frankel method.
() Schmidt Method :
Consider a rectangular mesh in x/-plane with spacing h along x-axis and the
spacing & along t-axis
Let us denote the mesh point (x, ¢) by (ih, jh).
Then wehave M4 = Hit Ah {See equation (9) of § 13.4]
2, -21
i Pu _ Mina j Mist Miet) [See equation (5) oF 13-4)
ae
Now putting these values in (1), we get
Bi js tee Ou; 5+ Mig
he
3
ok
or (p41 9 =e ina) his Medd )
: wt
Let us take u. = ke”/h2, which is known as mesh ratio parameter. %
yj = te 1,7 Bid * Me LD
any 4, j + (1 ~ 20) oj + OL reap)
at (x) fj -1) mesh point in terms of
Wijed
. aj, j41 = : ‘
‘The equation (2) determines the value 0! :
the Priston values at the points % 1) xjand Sie at the time 7
Thus the equation (2)is a relation ble i values at the two time
levels j +1 and j. Hence it is known as teed fom
‘The schematic form of (2) is shown 1n Pig. 28.0%
_ >See ee ae ee _
ret NIT SEE \G) CAD De WIILIER Coe is Sn nn
respectively, assuming second and higher powers of h are negligible, and
13.4, | NOTATION FOR FUNCTIONS OF ‘SEVERAL VARIABLES
Assuming that 1 is a function of the independent variables x and ¢, Subdividg or
the xt-plane into set of equal rectangles of sides dx = h, &¢= hk as shown in Fig. 199
and the points of intersection of the dividing lines are called mesh points or nodal and
points or grid points or pivotal points.
ay or
with |
s
(1) ar
: (a)
OF Dee +
4 Ph, jk),
a
ale AD TD Torun
G-De — e
cD
with I
and
—+—_4__{ ~
4 Gh th +h ¥ or
Fig. 13.2
Let the co-ordinates (x, t) of the mesh point P be and
x=ih,t=jk
where i and j are integers, a
Denote the value of u at P by
uy,
w(x, 6) =ulih, jk) =u, ; @
In case of two independent variables x and t, the equations (4), (6), (7) and (8)
of § 13.3 may be written as
2,
Fu _ ue th, t)~Qu(x, )+ule—h,t) 09
7 = Me M(x, t) + ue — h, n? (2)
ae So ti + O(h?)
Ou _ ule +h, t) ~ u(x, t)
Ox in
_ du _ Ube, t)~ ube ~h, t)
ox h
4 +O(h) (8)
+ Oh), wl) iSand (2)
ibdivide
‘ig. 13.2
rnodal
of Partial Differential Equatio
ron ‘using the notation (1), these equations becomes
_ UME + WA jk) 20 th,
jk) + w=) hj)
Ri ee outa a
ne a
with a leading error of order h?
and uy = EAE + Dh, ik) = wlth, jh)
* h
Fe uy, = TSG
- h
Da y= elit, jk) — w= I) hj
= A
or
with leading error of (4) in both equations.
Similarly, the approximations for the derivatives with respect to ¢ using notion
(1) are given by
u lik G+ 1) h) — 2u (th, jk) +u lth, G= DA)
ul paw
ty ja 2a
or ye
with leading error of order h”
au _ ulih, +1 kl=utih, jh) =
Ot k
and w=
joa
or ut
(ih, jh) — wtih
and a= A
yj Madd
or un k
with leading error of O(#) in both equations.3. | FINITE-DIFFERENCE APPROXIMATIONS Ti
| ‘PARTIAL
When a function wv and its derivati
Ves al i 4
funetions of x, then by Taylor's theorem, valued finite and
= ni hn? 3
ue +h) = u(x) + hw +5 woe we) ry
2 3
and u(x — h) = u(x) ~ hu’(x) + = u(x) -# UH) +o ~A2)
Adding these equations, we get
ue +h) + ue — A) = Qu(x) + hue) + Oh) AB)
where O(h*) denotes the terms containing fourth and higher powers of h.
Assuming that these are negligible in comparison with lower power of h (£e.,
for small value of h), then equation (3) becomes
u(x +h) + u(x —h) = Que) +h? uw"),
ux) = 5g ee + hy — Bula) + ule) --A4)
with a leading error on R.H.S. of (4) of order h?,
Now subtracting (2) from (1), we get
Ge +h) — ute —h) = 2hul@) +3 ua) + OG) 8)
Again negligible the terms of order h® and higher order, then equation (6)
becomes ys
u(x +h) — u(x —h) = 2Qhu' (x).
KW) = a [ue +h) ~ ue h))
4 (6)
with an error of order h?
i imates
Equation (6) clearly approximal
the slope of the tangent at P by the slope
of the chord AB and is called a
central-difference approximation,
owith leading error of O(h) in both equations.
Similarly, the approximati
oe ‘ations for the derivatives with respect to
ig te a u lih (+1) kh—2u ih, jh) +u lik, G- DR)
we E -
Uy eau :
ae ign a uy jn1
with leading error of order k”
d du _wlih, G+) R)=wih, jb) OS
i. MoE hk ae
or paths
_ du _ wtih, jk) =u liyi= 4)
pn Me ot hk
or es St ee
ia) | with leading error of O(k) in both equations.
and (8) 73.5. | SOLUTION OF LAPLACE EQUATION
‘ Consider the Laplace equation in two dimensions
Pu Pu
(2) Pu , Fu _
ax?” ay?
(3) Let us take a reetangular region R for which u(s 9) is known at d
vf { ‘that an exact subdivision of Ris possible, #0
Assuming
ry a network of square mesh of side / as shown in Fig. 13.3.