Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Ann. nucl. Energy, Vol. 15, No. 12, pp. 553 559, 1988 0306-4549/88 $3.00+0.

00
Printed in Great Britain. All rights reserved Copyright © 1988 Pergamon Press plc

NEUTRON SOURCE STRENGTH DETERMINATION FOR


O N - L I N E REACTIVITY M E A S U R E M E N T S

J. E. HOOGENBOOM a n d A. R. VAN DER SLUIJS


Delft University of Technology, Interfaculty Reactor Institute, Mekelweg 15, 2629 JB Delft,
The Netherlands

(Received 5 July 1988)

Abstract--A method is described to determine the effective neutron source strength in a nuclear reactor,
which must be known when calculating the time-varying reactivity from inverse reactor kinetics for a
reactor at low power. When for an initially subcritical reactor the reactivity is changed and kept constant
after the change, the effective source strength can be determined from a linear regression of reactor power
to a function proportional to the emission rate of delayed neutrons, which can be calculated from the
reactor power history. In view of the relatively strong noise present in the reactor power signal at low
power, a grouping method for the regression is preferred over the least-squares method.
Experiments with a reactor simulator with known source strength showed good agreement. Application
to actual reactor signals gave consistent and satisfactory results.

INTRODUCTION constants fli a n d 2~ are the fraction a n d decay c o n s t a n t


of group i a n d are k n o w n from literature (Keepin,
F o r reactivity m e a s u r e m e n t s d u r i n g the o p e r a t i o n of
1965).
a nuclear reactor inverse kinetics m e t h o d s are often
To express p(t) in terms o f n(t), e q u a t i o n (2) is
used based on the p o i n t - r e a c t o r kinetic model. F o r
solved using a n integrating factor exp (2it) which
o p e r a t i o n at low power, for instance at start-up, the
results in :
c o n t r i b u t i o n of the n e u t r o n source m u s t be explicitly
taken into account. In the p o i n t kinetics model, this
Ci(t ) =fli it e ~,u-")n(t')dt'. (3)
implies knowledge o f a q u a n t i t y p r o p o r t i o n a l to the
source strength, which we shall call the source term.
F o r the d e t e r m i n a t i o n of this source term a new W h e n e q u a t i o n (3) is substituted into e q u a t i o n (1), it
m e t h o d has been developed with good statistical prop- follows t h a t :
erties. This m e t h o d has been tested o n a reactor core
p(t) = fl-n(t)~i fii21
simulator a n d has been applied to the H O R reactor
at the Interfaculty R e a c t o r Institute o f the Delft Uni-
, f_ ~e a,u-C)n(t')dt'

versity of Technology, AS A dn
n(t) + n dt (4)
INVERSE KINETICS Usually, the term A/n dn/dt is so small t h a t it can be
neglected.
T h e s t a n d a r d point-kinetic equations can be written In general, n(t) is n o t m e a s u r e d directly, b u t a q u a n -
as (Keepin, 1965) : tity p r o p o r t i o n a l to n(t), usually via the response o f
dn p(t)-/~ . . a n e u t r o n detector. The m e a s u r e d q u a n t i t y will be
n(t) + ~ 2,C,(t) + S, denoted by P(t) which is equal to cn(t), with c a
dt S --7 (1)
p r o p o r t i o n a l i t y constant. A l t h o u g h the n o t a t i o n P
dCi /L suggests t h a t the reactor power is measured, P can in
- '~ n(t) - ).~ C,(t),
/k
(2) fact be any q u a n t i t y p r o p o r t i o n a l to the n u m b e r of
-d5
n e u t r o n s in the reactor. Now, the e q u a t i o n for p(t)
where n(t) is the total n u m b e r of n e u t r o n s in the core, becomes :
C~ the total n u m b e r of precursors of delayed n e u t r o n s
o f group i, S the total n e u t r o n source strength, A the 1 " I' cAS
p(t) = f l - i ( ~E',- fl'~' L~o e *," '" P(t')dt'- P(t-) "
n e u t r o n generation time a n d p the reactivity to be
determined. S is supposed to be c o n s t a n t in time. The (5)
553
554 J . E . HOOGENBOOMand A. R. VAN DER SLU1JS

Equation (5) is suitable for on-line calculation of the at constant power. Therefore, determination of the
reactivity provided that the quantity cAS is known. source term for reactivity measurements is necessary.
It is not necessary that A, S and c are known separ-
ately. The term cAS will be called the source term Sp.
Its dimension is the same as that of the measured D E T E R M I N A T I O N O F T H E S O U R C E T E R M Sp

quantity P. Suppose the following experiment is performed. A


The integral in equation (5) can be calculated recur- reactor is operated in a subcritical state with reactivity
sively from a sampled version of P(t), using linear Pl and a constant power P~ = - Sffp ~. At t = t, reac-
interpolation between two consecutive samples tivity is changed, for instance by displacement of a
P,, 1 = P(t, ~) and P, = P(t,j at a time interval At control rod, bringing the reactivity to P2 (smaller or
apart : greater than p~) at time t = t2, after which the reac-
tivity is kept constant. If p2 < 0, power will ultimately
In = e ~,(',, c)p(f) dt' stabilize at a level P~: = -Sp/p2. Figure 1 shows the
c~
possible behaviour of p(t) and P(t) for a positive
reactivity change. As reactivity is constant for t > t2
=e a'atl, i+
fi f
i
P~--(t,--t') we can write equation (8) as :
g(t) = ( f - p 2 ) P ( t ) - Sp. (10)
P,-P, I~e x,(,.-C)dt, Now, if we conceive g as a function g(P) of P (in a
× At 3 strict sense g is not a function of P, because it not

=e-a'A'l, 1+2i
t,.( 1 1-e~ ~A,~
}
only depends on the value of P at a certain time but
also on its history, so it looks more like a functional)
the relation between g and P is linear and constitutes
a line in the g - P plane with slope f - p 2 with intercept
P. 1 ( 1 --e-~A'~ -Sp at the g-axis. Figure 2 shows schematically the
2, e ~,A, ~ /. (6)
lines in the g - P plane for the cases p~ < P2 < O,
P2 < Pl and P2 > 0. The line g = f P in Fig. 2 rep-
An analogous expression has been given by Mogilner
resents the route for quasi-stationary changes, i.e.
et al. (1974). However, their formula contained a sign
infinitely small changes in reactivity, and thus power.
error. N o w the following function is introduced :
F r o m sample values Pi = P(t) we can calculate
gl = g(ti) according to equation (7) and by fittingg(P)
g(t) = ~ f , 2 s
i f'
,a - o o
e ~,('-")P(t') dt'. (7) to a straight line for t > t2 we obtain the source term
St, and at the same time the quantity f - P 2 , and hence
F r o m equation (3) it can be seen that g(t) is pro- the reactivity in the final state.
portional to Y2iCi, which is the total decay rate of F o r the determination of Sp and P2 it is not necess-
delayed neutron precursors or the emission rate of ary that P is initially kept constant ; it is only necessary
delayed neutrons. Equation (5) can be written as : that P(t) is known for a sufficiently long time to
evaluate g(t) and that the regression is carried out for
p(O = f g(O s, (8~ those points (gi, P3 for which p (or, equivalently, the
P(t) P(t) " control rod position) is constant.
Two methods for calculating the coefficients of
Note that for constant P(t) = P, we have g = fliP. F o r
equation (10) have been used, namely :
a subcritical reactor with constant power P due to the
presence of the neutron source S we have : • The standard least-squares method (Draper and
Smith, 1981). This implies minimization of:
Sp Sp
p=--fi or P- (9) Z { g i - ( a P , + b ) } z,
P i

For reactors operating at sufficiently high power (or with respect to a and b. Then Sp = - / ~ and
more precisely neutron flux) the effect of the source P2 = f - - a with 6 and/~ the estimates of a and b.
term on the calculated reactivity can be neglected, • The so-called grouping method (Bartlett, 1949).
as is often done for power reactors. However, for The procedure for a function of the form
operation at low power, the inclusion of the source y = a x + b is:
term in equation (5) is essential, because otherwise (1) arrange the observations of x in order of
zero reactivity will be obtained for a subcritical reactor magnitude ;
Neutron strength determination 555

tI t2
tim~
I P~
I J

J
l

c:t l o
o_
0J I
P2 PI
PI- V :,

tl t2 time
Fig. 1. Schematic time-behaviour of reactivity and power in an experiment to determine the source term
of a reactor.

(2) divide this sequence of observations into three g(t) contains far less noise than P(t), due to the fil-
equally large groups ; tering effect of the integration in equation (7), results
(3) calculate the averages of x and y for group 1 will improve if g is used as the independent variable
and 3 (denoted by 2~, f l and £3, f3, respec- and P as the dependent variable, which has been done
tively) and of the total number of obser- in the analyses of experiments with the least-squares
vations (£ and f ) ; method. However, this method will still give biased
(2 and f ) ; results, as the noise in g is not zero and, moreover,
(4) estimate a by ~ = 0 7 3 - f 0 / ( 2 3 - Y , ) will be correlated with the noise in P.
and b by/~ = 3~-~£. Also, the accuracy of both methods will be affected
by the time interval for which the power curve is
For application to equation (10) P was chosen
analyzed. As the point density of the (g, P) curve
for x~ and gi for y~ so that ~ and/~ estimate f l - p 2
increases where power stabilizes, that part of the line
and - Sp, respectively.
might thus become overemphasized.
The accuracy of these methods depends on the As already mentioned before, the standard least-
nature of the data. A necessity for the least-squares squares method is prone to be biased when applied to
method is that the independent variable has to be the (g, P) or (P,g) curve. The grouping method,
known exactly, which condition is clearly not fulfilled however, produces an asymptotically unbiased esti-
as the experiment will be performed at low power, so mate under general conditions, as will be shown. Sup-
the power signal may contain appreciable noise. As pose that the observations xi and yi contain errors :

X i = Pi~-J'li , E[~i] ~- O,

yi = g , + ¢ , , E H = 0,
with E denoting the expectation operator. Sub-
stitution of the form fk = 1IN E(k)y~ for .93, fl, £3 and
g ~into the expression for the estimator ~, with N the
number of observations in each of the three groups
and (k) denoting the group for which the summation
has to be done, results in :

//// ,o
~= ~ g~+~¢,-
1 1 1
///../ ,o.,?o "'(D "'(O
, ,/ / 7 - ' ~ / fin) ,02 <6 ": o |-- 1 1 1 . (11)
//// iv (3) (1) "'(1)
-% /,~ "' ~P
Usually, neither

2~i-E¢~ nor Eqi-Eth


(3) (I) (3) (I)
Fig. 2. Typical behaviour of the g-P curve for three different
reactivity changes. will be equal to zero, so the estimate differs from the
ANE 15:12-D
556 J, E. HOOGENBOOMand A. R, VAN DER SLUIJS

! I I I I I I I I I I I
Control rod position (orb scale)
o.14-
f (o)
!

0.12-

0.10~-
-

-- 0.08-

o~ -

n 0.06-

0.0q-

0.02- i

O. 00
0 0.2 0. u, 0.6 0.8 1.0 1.2 1.t&xlO~-
Time (s)
Fig. 3a. Reactor power of the simulator RESIDEL at a reactivity change.

1.2 I m m m n m m n m m n m

(b)

1,0

0.8

~ 0.6
E

(.9 0.LI
cj=t

0.2

~ flitted Line
0. O
0.02 0.0q 0.06 0.08 0. I0 0.12 0.1q.
Power (W)

-0.2
Fig. 3b. g P curve corresponding with Fig. 3a.
Neutron strength determination 557

Table 1. Results o f source-term determination applied to reactor core simulator R E S I D E L

G r o u p i n g method L S method
Time interval (s)
for analysis p~ x 103 P2 × 104 Sp Pl x 103 P2 × 104 So

Without 190-500 -5.584 --8.443 9.999 x 10 s 5.540 8.444 9.991 x 10 s


noise 190-800 --5.585 -8.444 9.990 × 10 s --5.585 --8.443 9.989 x 10 s
19(~1280 --5.600 -8.469 1.002x10 4 -5.015 -7.142 8.970x10 5
800-1280 5.475 8.274 9.792x10 s 0.140 21.29 --2.510x10 4

With 190 500 --5.576 --8.434 9.978x10 5 5.577 --8.434 9.978x10 5


noise 190 800 --5.591 --8.458 1.000x 10 4 --5.577 --8.434 9.997 x 10 5
190-1280 -5.581 --8.441 9.986x10 5 --5.992 8.807 1.072x10 4
800 1280 -6.461 --9.768 1.156x10 4 515.7 7.860 9.227x10 3

correct value of a. However, for increasing N the numbers. In the applicable power range, the relative
estimate will become more precise as : variance of the noise was chosen inversely pro-
portional to power. Calculations were done with a
1
lim ~ q l = E[~/~] = 0 PDP-11/84 computer.
In Fig. 3 P(t) and the ~ P diagram are illustrated
and for an experiment with noise. N o t e the difference in
noise level in g and P.
1
lim ~Y,~j~ = E[~i] = O. Exactly the same experiment was carried out with-
N~ 3c
out noise. Table 1 shows some results, where p i is the
The same conclusions apply to the estimator b. reactivity before rod displacement and P2 the reac-
As the statistical errors in g and P are correlated tivity after displacement. The experimental value for
and the variance in P is not constant (the relative P l is obtained from equation (9) for a constant power
noise in P normally decreases with increasing P at P1. The exact value of p~ was - 5 . 5 8 5 x 10 -3 and of
low power), it is very difficult to estimate the variance P2 - 8 . 4 4 4 x 10 4.
of Sp. Table 1 shows that the grouping method produces
good results under all circumstances, even in the time
interval from t = 800 to 1280 s where power stabilizes ;
EXPERIMENTS WITH A REACTOR SIMULATOR
the results of the least-squares method are less sat-
The method for determining Sp based on equation isfactory when the number of points where the power
(10) has been tested with the reactor core simulator stabilizes increases. This was already pointed out in
R E S I D E L of the Interfaculty Reactor Institute of the description of the two methods.
the Delft University of Technology. The simulator is
based on point kinetics and the values for 21 and /~i
are taken from Keepin (1965) for the thermal fission E X P E R I M E N T S W I T H AN A C T U A L R E A C T O R
of 235U, with the total effective fraction of delayed
neutrons equal to 0.0075. The simulator also includes The H O R reactor of the Interfaculty Reactor Insti-
a thermohydraulic model, but this does not effect our tute is a pool-type reserch reactor with highly enriched
results as they have been obtained at low power. (93%) 235U fuel. A 19 Ci C m Be neutron source is
The exact values of the reactivity and source term placed in the vicinity of the core.
are known: the latter is constant and equal to The power signal was sampled with a PDP-11/84
9.99 x 10- s W. Noise can optionally be added to the computer with a C A M A C interface; the sampling
power signal using Gaussian distributed random period was 100 ms. As the experiments were per-
formed at low power, the influence of fuel and mod-
erator temperature could be neglected. Table 2 shows
T a b l e 2. R e s u l t s o f s o u r c e - t e r m d e t e r m i n a t i o n o f t h e H O R
the results of the analysis with the grouping method
reactor by means of the grouping method
of three measurements.
pt p2 Sp Figure 4a shows P(t) and Fig. 4b shows g vs P for
the first measurement; the starting situation of the
pincrease - 1 . 9 8 x 1 0 3 _7.74x10-4 4.33x10 4 first measurement corresponds with the end situation
p decrease -7.48 × 1 0 - 4 - - 1.97 x 10 -3 4.28 x 10-4
pincrease - 2 . 2 2 x 1 0 3 _8.70x10-4 4.23x10-4 of the second. The third measurement is a different one
in the same power range.
558 J . E . HOOGENBOOM and A. R. VAN OE~ SLUIJS

0.~ I I I I I I I I I I I I l l l l

(a)
Control rod position (arb. scale)
,r ............ ,.,m I .... L,,

t
0.6
J

~.O.q

o
13_

0.2

I i I I I I I I I I I | I I I I I
0.0
100 200 300 tlO0 500 600 700 800 900
Time (s}

Fig. 4a. Digitally-filtered power of the H O R reactor at a reactivity change.

6 i I I I I I I I

I (b)

3
E

t9 2

/ / ~ Fitted Line

I I I I I I
0.2 0.tl 0.6 018
Power (W)

-I
Fig. 4b. g - P curve corresponding to Fig. 4a.
Neutron strength determination 559

DISCUSSION AND CONCLUSION c, = ASI[I and c2 = p/ft. This form is linear if p is


constant and in that case cl and c2 could be deter-
To investigate whether the source term found for
mined by a regression method. The drawback of this
the H O R reactor is of the right order of magnitude
method is that the regression would be based on the
the following simple calculation may provide insight.
two variables x(t) and y(t) which both may contain
If c is the proportionality constant between power
considerable (and strongly correlated) noise because
and the total number of neutrons in the core, then
they both contain a term n(t). Application of this
c = Qf" v" Zf with Qr the energy released in a fission,
method to the data of Fig. 3 using the grouping
v the average velocity of the thermal neutrons and Ef
method leads to results for S o and P2 close to
the macroscopic fission cross section. Furthermore,
the results from Table 1 (analysis time interval 190-
A = 1/(vvS~0 with v being the neutron yield per fission
1280 s). However, application of the least-squares
so that :
method resulted in values of Sp and p that differ a
S " Qf factor 3.7 with the data in Table 1, for which the least-
so = c A S = - -V . (12)
squares method still gave reasonable results.
It may be concluded that the grouping method
Knowing that the total neutron emission of the source
applied for regression On equation (10) is well-suited
is about 4.7 x 107 n s ~and that the source is (geome-
to estimate the source term (and reactivity) of a reac-
trically) directed half towards the core of the H O R ,
tor, especially when strong noise is present in the
the effective source strength might be around
measured reactor power signal which will normally
2.35 x 107 n s-~ which results in a value of the source
be the case at low power.
term of about 3.2 x 10- 4 W (v = 2.4 ; Qr = 200 MeV).
The intrinsic neutron source of the core increases this
value so the determined values are in reasonable agree- REFERENCES
ment with physical consideration.
Bartlett M. S. (1949) Biometrics 5, 207.
In the past (B/irs et al., 1977 ; Mogilner et al., 1974), B/irs B., Kgtll L. and Visuri P. (1977) Nucl. Instrum. Method.
methods for the off-line determination of the source 143, 199.
term and the reactivity have been proposed based on Draper N. R. and Smith H. ( 1981) Applied Regression Analy-
the form : sis. Wiley, New York.
Keepin G. R. (1965) Physics of Nuclear Kinetics. Addison-
y(t) = c~ +CzX(t), (13) Wesley, New York.
Mogilner A. I., Fokin G. N., Chaika Y. V. and Kuznetsov
with x(t) = n(t) and y(t) = n(t)-AE2iCi(t)/fl. Then F. H. (1974) Soy. Atom. Energy 36, 453.

You might also like