FEATURE E
ADVANCED NODAL
FOR PEDESTRIANS
Yung-An Chao
ERFD Core Engineering
crneveail Mawlery Bert boop.
METHODS
The basic physics of advanced nodal theory is deseribed in simple terns,
Different nodal methods are reviewed with a winizal use of mathenatics,
PROLOGUE
Advanced nodal theory is easy to un-
derstand.
After more than a decade of deve
opment and applications, advanced
nodal methods have overcome the
skepticism thet classical nodai
methods received in the early years
and acquired a firm stetus and
general acceptance in the reactor
physics community. However, despite
the success of advanced nodal
methods in practical applications, the
Question for some people still
remains what, after all, really makes
them work? Is there anything magic
about them? Are we getting some-
thing for nothing? NO, edvanced
nodal theories are neither complicated
or magical. in fact, the physics in-
volved is quite simple, although the
resulting mathematical equations ere
more sophisticated than the good old
finite difference equations. in this
article, | will elaborate on this point
with minimal mathematics. 1 will elso
review different nodal methods from
the seme perspective.
NODAL METHODS EVOLUTION : THE
CLASSICAL TO THE ADVANCED
‘There have been two chronological
phases of nodal method development,
the classical and the advanced. The
mid-seventies was the dividing tine
of the two periods. The classical
nodal methods started with BWR
methods development in the early
sixties. They are based on more
heuristic arguments and rely to some
extent on empirical adjustments to
achieve adequate accurecy. The first
Hh oye
Advanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
advanced nodal method was intro-
duced by KWU, followed by several
others. They are besed on more rig-
erous enelysis, and are much more
accurate, FLARE and PALADON ere
examples of codes using classical
nodal methods. CUBBOX, QUANDRY,
ANC, and NOVA ere examples of
codes using advanced nodal methods.
Classica! nodal methods use coupling
coefficients,
Leakage Probabftity
Nodal methods ere introduced to save
computing time as compared to the
finite difference method. The latter
approximates derivatives by linear
difference ratios, end therefore re-
quires many small meshes to be ac-
curate, which is 1ime consuming, The
original idea behind @ classical nodel
method is to resolve the conflict
between accurecy and speed if
March 1988
Several small meshes sre combined
into a large “node” to speed up the
computation, the finite difference ap-
Proximation is no longer valid. Since
the finite difference approximation is
applied to the derivatives to evaluate
currents, which account for leakages,
the question is whether a different
method can be used to evaluate the
leakages between adjacent (large)
nodes. The central idea of a classi
nodal method is to relate the nodal
leakage currents to t
‘one
Aeutron source in a homogeneous
node to calculate anaiytically the
transit. probabilities of the source
Yo the surrounding regions.
ssary to make the simple
assumptions of fist source distrib-
ution, node homogeneity, and # spe-
cific surrounding environment in order
to make the problem solvable.
However it is also precisely these
assumptions that cause the classical
nodal methods to be heuristic and
empirical,
Two very important features about
classical nodal methods should be
2 Advanced Nodal Methods for Pedestrians— to be solved.
TECHNOLOGY TRENDS
March 1888
emphasized before we turn to the
discussion of advanced nodal
methods. The key elements of a
classical nodal method ere the coup!
ing coefficients, which are precalcu-
lated independently of the problem
For this reason we
may classify classical nodal methods
as “nodal coupling methods”. The
other point is that in order to be able
to calculate the coupling coefficients,
the nodes are assumed to be homo-
geneous.
Advanced nodel methods ‘sew up node
wise solutions.
Sewing up nodevise solutions
“oo™
Bee ee Taney
The assumption of homogeneous
nodes provides the key clue to the
development of advanced nodal
methods. If the nodes are already
assumed to be homogeneous, we
/ Should be able to do a lot better than
Phir ba pen
Vv
fn inher wate ne
Yi
peat
°y
nodal coupling methods provide. The
problem of neutron diffusion over @
homogeneous node is s0 much
simpler that we know a lot more
about the characteristics of its sol-
ution. If we could properly “sew up”
these individual solutions valid for
each of the homogeneous nodes, we
would be able to obtein the solution
for the whole core. Instead of pre-
uiating coupling coefficients by
juming flat neutron sources, ad-
venced nodal..methods «connect the
nodewise solutions via “coupling
equations”, which result from inter
_boundsry conditions. Beceuse
explicit representations of nodewise
solutions are made in advanced nodal
methods, we may classify them as
“nodal representation methods” in
contrast to classical nodal coupling
methods. More detailed discussions
‘on the basic ideas of the nodal rep-
resentation methods are given in the
next section.
WHAT MAKES ADVANCED NODAL
METHODS WORK ?
Homogenization is the key.
Let us start with @ simple one di-
mensions! problem, 8 homogeneous
medium of, say. 200 em length. To
solve this problem with the finite
difference method, we need to divide
it into at least 40 meshes, each 5 em
fong, to get # solution of acceptable
accuracy. Therefore, we need to in-
troduce at least 40 unknowns in a set
of 40 by 40 matrix equations, which
Advanced Nodal Methods for Pedestrians = 3
Balt aTECHNOLOGY TRENDS
result from the 40 interface boundary
conditions between adjacent meshes.
However, for such @ simple problem,
we can trivially obtain the exact an-
alytic solution, which is simply @
linear combination of @ sine and &
cosine function. The coefficients of
the two terms are the only two ine
dependent parameters to be deter-
mined by the core boundary
conditions. Thus, we can say thet for
this problem, we can use @ “nodal!”
method with only 8 single (extremely
coarse) node of 200 em to obtain &
solution even more accurate than that
of the finite difference method with)
aroivarly fine meshes. Is. tis|
statement trivial? Yes, nevertheless
the essence of the nodal method is
all contained herein, The very im-
portant point to recognize here is that
the medium being homogeneous does
the trick. If this 1D core were not
homogeneous, one would not be able
to obtain the complete solution with
a single node and two parameters,
and divi¢ing the core into meshes
would be necessary.
A nodal solution can be exact.
Now let us generalize the above
example to a slightly more interesting
fone. Suppose that the core contains
10 different sections, each 20 cm
tong and of a different homogeneous
material composition. Since for each
of the 1D homogeneous problems we
can obtain its analytic solution in
terms of two parameters, the com-
plete solution can be represented in
A
March 4988
terms of 20 parameters. Actually we
can do better than that, The two
parameters for each section can be
Getermined in terms of the boundary
conditions of that section, which can
be prescribed by the flux values at
the two ends of the section. There-
fore the complete “exact” solution
can be represented in terms of only
31 parameters, which are the fluxes
at the 9 interfaces and the 2 end
boundaries. So we conclude that for
this problem we can use # nodal
method with 10 (coarse) nodes and
11 parameters to obtain a solution
even better than that of a finite dif-
ference method with arbitrarily fine
meshes. Again the trick is thet each
of the 10 sections is 8 homogeneous
medium. Another very important
\point to be noted is that since the
1odal solution is analytic within each
section, it provides flux values con-
tinuously at every point. On the
contrary, the finite difference solution
provides flux Values only at discrete
points. :
refine the core model
further by introducing 15 fuel rod
subintervals in each of the 10 sec-
tions, each fuel rod with a different
burnup. The core now contains 300
intervals of different media. A finite
Gifference model for this core will
have typically 300 meshes with 300
unknowns. If we follow the same
analysis as that above, we need 300
nodes to obtain an exact nodal sol-
ution, There. is no advantage other
than that the nodal solution will still
Next, we
4 Advanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
be the exact solution while the finite
difference solution is not. But for
meshes so small, this advantage is
no longer significant. However, if
we know in practice that the solution
Variation across the 15 rod intervals
in each of the 10 sections is either
smail or very local, we may “homo-
genize” each section and obtain, as
described before, the 10 node and 11
Parameter exact solution to the hom-
ogenized problem. This method is,
of course, & lot faster than the finite
difference method using 300 meshes,
and its solution provides & continuous
and smooth global description of the
flux verietion over the core. This
solution should be very good for
calculating any node averaged quanti-
ties although it does miss local
Structures, which can be recovered,
however, with @ post-caiculation re-
construction or superposition method.
We see once again that the key here
is the homogenization of the nodes.
Thus far, we can reach two conclu-
sions through this 1D example. The
first is that for a core containing a
number of homogenous regions, there
is no point in using @ fine mesh finite
difference method. One ‘would be
much better off to use a nodal
method to obtain 8 whole core sol-
ution by “sewing up” analytic sol-
utions for each of the homogeneous
regions. The second conclusion is
that if @ heterogeneous core can be
decomposed to distinct regions,
which cen be epproximately homo-
genized, one would still be better off
to use a nodal method to obtain the
March 1986
Smooth continuous solution and then
reconstruct local variations.
When we generalize the above 1D
example to 2D or 3D, however, com-
Plications rise. For simplicity of
discussion, we will consider here only
the 2D case. For 8 2D homogeneous
region, an analytic solution to the
diffusion equation usually can not be
obtained. The geometry of the region
ang the boundary conditions can
complicate the solutions tremen-
Gously. Even for the simplest case
of & homogeneous rectangular region,
the closest that one can get to an
analytic solution is an integral repre-
sentation in terms of the boundary
values and the Green's function for
the rectangular region, which ean only
be expressed ss en infinite series
expension. Only for an extremely
imple boundary condition on this
rectangular region, can this integral
representation result in @ closed sol-
ution. The question now is whether
one can still take any advantage of
the homogeneity of the regions to
do better than using the fine mesh
finite difference method. Well, al-
though one can not obtain the exact
analytic solution for » homogeneous
rectangular node, one may be able to
obtain an approximate one. If the
approximate solution has acceptable
accuracy and contains fewer pereme-
ters than the number of meshes re-
quired for a finite difference methoo
solution, one would still be better
off. It is easy to see that this should
be possible. For a homogeneous
Advanced Nodal Methods for Pedestrians 5TECHNOLOGY TRENDS
region, the solution must be a smooth
function. Now consider fitting to e
smooth curve. One can choose dis-
crete points on the curve and join
them by straight lines, which is ane
ogous to the finite difference method,
Or one can fit it with functional
forms containing parameters, which
is enalogous to nodal methods.
Fitting with line segments
versus parameterized
functions
It seems obvious thet even if we do
not know the exact or correct fune-
tional form, we should expect to do
much better with reasonable func-
tional representations than straight
tine segments. On the other hand if
the region is not homogeneous, the
Solution may have zigzags or local
Structures. For that type curve, a
Smooth functional representation can
only describe its smooth “back-
ground” shepe, The local structures
then have to be reconstructed on top
March 1988
of the smooth background. How to
obtain such en approximate snelytic
solution and how many parameters
are needed for its representation are
the crucial questions to be answered.
Different nodal representation
methods differ in their answers to
these questions. This leads to the
discussion of the next section, »
review of different advenced nodal
methods.
REVIEW OF ADVANCED NODAL
METHODS
Ditferem methods use different sol-
ution parameterizations.
Since the analytic solution for @ ho-
Mogeneous node can not be obtained
in @ closed form except for the one
dimensional case, nodal represen-
tation methods try to construct the
single node solution representation
from the one dimensional solution.
This solution is usually obtained by
fegrating over the transverse vari-
ables, for example, integrating over
the y and z variables to reduce the
diffusion equation to # one dimen-
sional equation in x The resulting 1D
equation contains en extra term re-
Presenting the neutron leakage in
transverse directions, which ean be
treated as en additional source term.
However, this 1D equation can not be
solved until the transverse leakege
term can be explicitly specified. This
is where approximations are intro-
duced. A quadratic approximation to
the tensverse leakage is introduced,
6 Advanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
the parameters of which ere in turn
relaied to the solutions to the similar
1D equations along the transverse di:
rections, Thus, all the three 1D
equations are coupled, The so called
analytic nodal method (QUANDRY
code developed at MIT and used by
STUDSVIK) solves each of the 1D
equations analytically, while the nodal
expension method (used in the
CUBBOX code developed at KWU and
our ANC) represents the 1D solution
by a polynomial expansion consistent
with the transverse leakage approxi
mation. There are other methods
having mixed features from both
methods. There are also variants of
each method, for example, using
partial currents or net currents to
implement interface boundary condi-
tions. In any case, all the methods
essentially approximate the single
node solution with @ representation
involving the parameters appearing in
the 1D solution ang the transverse
leakage term. These parameters are
the unknown variables entering the
nodal equations resulting from con-
Recting the single node solutions
through the interface boundary con:
tions. Unlike the nodaf”" coupling
methods, where only the node
average flux is determined, the nodal
representation methods determine all
the representation parameters, from
which both tne average and the dis-
tribution of the node flux cen be
seduced.
March 1988
What is the most economical paremet-
erization?
ht is important to reatize thet ed-
vanced nodal methods actually solve
for the representation parameters
rather than just the assembly aver-
aged fluxes. This has two implix
tations. More parameters provide
more information, which is good. On
the other hand, to determine more
parameters requires more calculation
and computation time. Therefore it
is important to address the question,
whether the parameters are theore-
tically independent, end what is the
Most economical parameterization to
use. These questions opened up a
new perspective to the understanding
of advanced nodal methods, and mo-
tivated our work on the “interface
flux nodal method” discussed in the
following paragraph.
Let us consider a 2D homogeneous
rectangular node, and ask how the
solution of the diffusion equation
over the node can be approximated
with a minimum number of parame-
ters, First, the solution is uniquely
determined by the boundary condi-
tions, which can be @ specification
of the boundary flux value or the
boundary current value. In fact the
exact solution is an integral, over the
four sides of the node, of the product
‘of the boundary value and 2 known
function, the so-called Green's func-
tion, So the best approximation to
the solution with a minimum number
of parameters is the one that ap-
proximates the boundary value profile
Advanced Nodal Methods for Pedestrians 7TECHNOLOGY TRENDS
With a minimum number of perame-
ters. If one chooses to use current
85 the boundary value, then the in-
terface current profile needs to be
Parameterized. The same current
Profile is shared by the two adjacent
nodes, and the parameters are deter-
mined through the nodal equations
resulting from the condition of inter-
face flux continuity. On the other
hand if flux is chosen as the bound-
ary value, the interface flux profile
is approximated. The seme flux
profile is shered by two adjacent
Nodes, and the parameters are then
determined through the_condition of
interface current continuity. When the
former approach is taken, parameter-
izing the interface current profile with
8 parabolic function, it can be proved
that the resulting nodal equations are
very similar to those used in the
above cited analytic nodal method
{QUANDRY}, In fact, they can be re-
Produced with only slight modifica-
tions, The so called interface flux
nodal method chooses to paremeter-
ize the interface flux with @ parabolic
function, and the resulting - nodal
equations are the ones used in the
NOVA code. The advantage of par-
ameterizing an interface flux profile
over an interface current profile is
that the neutron flux function is the-
oretically continuous everywhere. On
the other hand, an interface current
is continuous in its direction, but its
Profile over the interface is not nec-
essarily 2 continuous function.
Merch 1988
All the different advanced nodal
methods have about the same eccu-
tacy and computing speed. With two
group end one node per homogenized
assembly, they give excellent pred-
ictions of node averaged powers. In
Addition, they can also provide very
good power gradient predictions
inside the nodes.
SUPERNOVA : A SUPERFAST
NEO-CLASSICAL NODAL COUPLING
METHOD.
The minimum number of parameters is
one.
The efficiency of & nodal represen-
tation method depends on how many
perameters per node are used to re-
Present a nodewise solution. The
minimum number is, of course, just
one parameter. But is that possible?
This question turns our discussion to
a very recent development in ac-
vanced nodal’ methods, the methogol-
ogy of the SUPERNOVA’ (SPNOVA)
code. In order to provide an ade-
quate background for the discussion,
we will first describe related aspects
of the NOVA code.
Although the above advanced nodal
Methods are much faster than the
finite difference method, in some
application areas, where many re-
Peated dimensional calculations are
required such as fuel management,
core monitoring and kinetics analysis,
there is still a very strong need for
even faster computing methods. For
8 — Agvanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
loading pattern search applicetions,
NOVA wes revised to 8 simpler for-
mulation to increase its speed. A
two group two dimensional albedo
Model was introduced to replace the
baffle reflector region. The “Effec-
tive Fast Group” mode! wes intro-
duced to collapse the two group
formulation to an effective one group
formulation. The assumption of the
FG model is that the thermal flux
leakage is small compared to the fast
flux leakage so thet the ratio of the
two can be: treated as ® small ex-
Pansion parameter. In keeping the
expansion to the first order of the
Parameter, an epproximate thermal
to-fast flux ratio can be obtained.
Using this flux ratio in the fission
source term of the fast group dif-
fusion equation leads to the following
EFG diffusion equation,
LtVPO, OL keg, oD
where L, is the fest neutron diffusion
length and k«# is the EFG multipli~
cation factor, This factor differs
from the conventional multiplication
fector ke in that it includes the
thermal leakage effect to the first
order. The revised NOVA assumes
that equation (1} can provide an ac-
curate approximation for the fast flux
The physical argument for this as-
sumption is that because of the large
fast neutron diffusion length, the fast
flux distribution is not very sensitive
to the detailed local variations of the
fission source term. The use of
equation (1) reduces the nurverical it-
March 1988
eration scheme from two group to
one group. After the fast fiux is
obtained, the thermal flux, however,
is not simply deduced from the fast
flux by using the essumed fast 10
thermal flux ratio. An additional
correction, accounting for higher
order thermal leakage effects, is
carried out on a node by node basis.
The EFG model is 2 forebeer of
SPNOVA. From equation (1) we see
that its solution depends on two pa~
rameters, k* and L,. An iterative
numerical solution to equation (1)
contains two loops, the outer iter-
ation loop and the inner iteration
loop. The outer iteration loop
updates the fission source term on
the right side of equation (1), and the
inner iteration loop solves the fixed
source problem for the flux by nu-
merically inverting the diffusion op-
erator on the left side of equation (1).
The outer iteretion calculation in-
volves the parameter ks, while the
inner iteration depends only on the
parameter L, and the core boundary
condition. It i
that_the same diffusion ¢ is
fepeatediy inverted after every outer
iteration. If the inverted operator
could be pre-calculated and stored,
inner iterations could then be elimi-
Rated and the resulting nodal equation
could contain only one parameter per
node. This would be worth doing
only if the seme inverted operator g/
could be used for. different problems.
Fortunately, for a PWR core, th
of Ly is nearly independent of the
Advanced Nodal Methods for Pedestrians 9TECHNOLOGY TRENDS
fuel enrichment and burnup, and hence
is independent of the core loading
condition, We name this property’ of
an epproximately constant L, "Dif-
fusive Homogeneity", which means
that a PWR is essentially homogene-
us with respect to the diffusion of
fest neutrons despite the heterogene-
ity in its reactivity distribution. This
has @ very important consequence,
That is, the diffusion operator in
equation (1) is essentially unique and
therefore needs to be inverted only
once.
Inner iterations are elimineted.
After the diffusion operator is in-
verted, equation (1) becomes,
= [G]k«9,) (2)
where [G] is # square matrix releting
each node to the other ones. This
“G-matrix” is the Green’s function of
the diffusion operator (essentially its
inverse) cast in a discrete matrix
form. This Green's function is valid
for the whole core, unlike the Green's
function discussed in previous sec-
tions which is valid only far each
individuat node. The — G-matrix
depends only on L,, the core geom-
etry and the core boundary condition.
SPNOVA is a renewed nodal coupling
method.
The physical meaning of the G-matrix,
5 it stands in equation (2), -can be
March 1968
interpreted as the diffusion probabil
ity of source neutrons from one node
to other nodes. This reminds us of
the coupling constants in the classical
nodal coupling methods. indeed
equation (2) is the simplest form of
2 nodal coupling equation. Are we
then meking progress or retrogress-
ing? How is this different trom 2
Classical nodal coupling method?
Well, recall how the classical coupling
coefficients are obtained. They are
calculated with over-simplitied as-
sumptions and approximations, The
G-matrix elements, however, stand
for more accurately calculated coupi-
ing coefficients. For this reason, we
may categorize SPNOVA as a “neo-
classical” nodal coupling method.
Another difference in the classical
Rodel coupling methods is that they
have more complicated coupling
schemes so that inner iterations and
Mterative updating on the coupling
coefficients are still required.
To sppreciate the simplicity of the
Methodology, one notes that equation
(2) is the only nodal equation to be
Solved iteratively, it involves only
one parameter per node, the coupling
coefficients need not be updated
during iterations, and there are no
inner iterations to perform. These
features combined with the use of
one node per assembly and the
removal of baffle and reflector
makes SPNOVA a factor of fifty to
® hundred times faster than conven-
tional nodal codes.
10 Advanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
NODE HOMOGENIZATION : A BASIC
PROBLEM FOR NODAL METHODS.
A discussion on advanced nodal
methods would not be complete
without some remarks on node hom-
ogenization. By now, it is not nec-
essary to repest that node
homogenization. is the central as-
sumption of ail the nodal methods.
But it may not be obvious thet hom-
ogenizetion can become a rather
subtle problem. There are two prob:
lems _with the homogenization of e
Node, The first is thet it depends on
the node boundary condition, which
one does not know unless the exact
Solution of the problem is already
available, Therefore, one always
Reeds to assume some “typical” or
“expected” nodal boundary condition.
The second probiem is more subtle
and less familiar to nonexperts. Even
if the exact flux solution for a het=
erogeneous node were given, one still
could no: perform a “rigorous” hom=
ogenization, To expiain this in
simple way, let us consider the fol
lowing simple 1D and 1G diffusion
equation over one node,
> (6/dx) [0 (x) (4/dx) 6 (x) J
+ Ee (x) 9G)
= Ir Gx) eG) @)
One would like to homogenize the
above equation to the following form,
Bie? /ex)$(x) + Ee 30)
= Er G(x) (a)
March 1988
Such that its solution preserves all
the reaction rates given by the sol-
ution of equation (3). We will see
that this is simply impossible.
Rigorous homogenization is impossible!
Assuming the flux values are given
on the two boundaries suffices to
uniquely determine the solutions, glx)
and $x). The solution gx) depends
on the three parameters, B, Ea , and
tr. The question is whether the
three group constants can be so
chosen that the node integrated flux
is preserved and the node integrated
value of each of the three terms in
equation (3) is also preserved. Since
the three terms in equation (3) are
related, we need to impose conser-
vation on only two of them.’ There-
fore, we have three independent
conditions for three unknowns,
uniquely determining the three group
constants. Now the function $x! is
totally fixed, there is no freedom to
adjust the currents at the “two bound-
aries, which depend on the slopes of
Bod there. So the currents at the two
boundaries can not be forced to agree
individually with the true currents,
falthough the sum of the two bound-
ary currents is the total node leakage,
namely the first term of equation (d),
and is indeed preserved.) Therefore
we conclude that the homogeneous
solution and the heterogeneous -sol-
ution have the same boundary fluxes
but not the same boundary currents,
Hed the boundary currents been spe-
cified es the boundary conditions
Advanced Nodal Methods for Pedestrians 11TECHNOLOGY TRENDS
instead, with the seme argument we
would conclude that the flux on each
of the boundary ends would not be
Preserved. in short, there are simply
Ret enough homogenized group con-
stants to be adjusted to meet all the
conditions on preserving reaction
tates, flux level, end the boundery
values of both the flux and current.
The only way to resolve the problem
is to introduce more parameters in
order ‘to bring in more degrees of
freedom. This is why the “mysteri-
‘ous discontinuity factors” are intro-
duced. One simply uses the ratios
of the boundary values-of the fluxes
{or currents) between the heteroge-
neous and homogeneous solutions to
alter the interface flux (or current)
continuity condition. The homogene-
ous solution is not required to be
continuous at en interface, instead it
is required to be discontinuous by 2
specific amount. Of course the
correct discontinuity factors can be
caiculatée only if the exact solution
is already available. In practice, one
could choose @ “typical” case to
calculate them and assume that they
apply approximately to other cases.
Obviously one could introduce pa-
Femeters other than the discontinuity
factors to solve the homogenization
Problem. For instance, one could
assume directional dependence in B,
SO that instead of one diffusion co-
efficient. per node, there would be
several diffusion coefficients apply-
ing to different directions.
March 1988
EPILOGUE
What next?
| have given » conceptual discussion
of advanced nodal theory with ali
details omitted. Although the basic
physical idea is indeed very simple,
88 any good physics should be, the
implementation may become rether
involved mathematically. Further dis-
cussion on the subject would get into
the details and would no longer be
light reading to the readers. There-
fore | will wind up this article with
2 few personal opinions, which un-
doubtedly reflect my biased views.
For years | have heard people say
‘that reactor physics is dead, there is
nothing new on the horizon, and all
that is needed are bigger computers.
Of course from the point of view of
fundamental theory, reactor physics
may be mature. However, from a
practical point of view, reactor
Physics methods are still complicated
and clumsy. Physics is a science of
approximation. it models a compli-
cated system or problem in a simple
wey to capture its major character-
istics of interest. As long as a
physics model is found complicated
for certain applications, there is a
need for model simplification and
there is “new” physics to be devel
oped. Beyond a certain point one
may mot call it pure physics any
more, and that is the beginning of
So-called engineering physics.
12° Advanced Nodal Methods for PedestriansTECHNOLOGY TRENDS
The technology is not mature
In my opinion, today’s technology for
core design and analysis is not yet
& mature technology. We ttle
capability for optimizing the design
end operation of & core. Conse-
quently, core designers may not find
design as challenging, creative and
enjoyable as it could be. However,
the recent developments in nodal
methods will allow us to do a lot
more things than we could ever think
of doing before. As a result, de-
signers with more powerful and flex-
ible tools will be able to do more
interesting designs. in particular, for
the first time it will be feasible to
develop useful optimizers. Despite
many unfruitful attempts in the past,
1 believe that optimization will
Fesurge es a main theme of the next
6T Téactor physics methods, along
With design tools on workstations
and Pts.
In conclusion. reactor physics may
be @ mature science but reactor
physics methods are not mature
technology. There are many interest-
ing problems to be solved and many
powerful tools to be developed. A
new era is yet to come. But... will
it? It would be very sad if we lack
the vision needed for this time to
arrive.
March 1988
Author
Yung-An Chao is 8 fellow engineer in
the Methods Development and inte-
gration group of CNFD Core Engi-
neering. He can be reached et WIN
293-2556.
Advanced Nodal Methods for Pedestrians 13