-

¥}

 

 

OAK RIDGE NATIONAL LABORATORY

operated by
UNION CARBIDE CORPORATION
for the
U.S. ATOMIC ENERGY COMMISSION

ORNL- TM-2315
copy no. -4 (0O

DATE - 8/27/68
Instrumentation and Controls Division

MEASUREMENT OF HELIUM VOID FRACTION IN THE MSRE
FUEL SALT USING NEUTRON-NOISE ANALYSIS

D. N. Fry  R. C. Kryter J. C. Robinson'

ABSTRACT

Investigations were made at the MSRE to determine if the amount of
helium gas in the fuel salt could be measured using neutron noise analysis. The
neutron power spectral density (NPSD) was measured at different reactor operating
conditions and compared with analytical model predictions of the NPSD for the

same conditions.
The results of experimental tests and analytical studies have shown that the

principal source of small neutron density fluctuation observed in the MSRE is helium
bubbles circulating in the fuel salt. The NPSD was most sensitive to changes in

. fuel temperature: the NPSD in the region of 1 cps increased by a factor of almost

50 when the average core outlet fuel temperature was decreased from 1225 to
1180°F. The measurements showed that NPSD in the frequency range from 0.5

to 2 cps varies as the square of helium void fraction as predicted by the model,

and that the minimum void fraction is more nearly zero than the previously accepted

value of 0. 1%,
It is concluded that changes in the circulating void fraction can be inferred

with good sensitivity directly from neutron noise measurements, and, consequently,
NPSD can complement and enhance the value of the MSRE reactivity balance

calculations.

 

IConsultant from the Nuclear Engineering Department, University of
Tennessee, Knoxville, Tennessee.

NOTICE This document contains information of o preliminary nature
and was prepared primarily for interncl use at the Oak Ridge National
Laboratory. It is subject to revision or correction and therefore does

not represent a final report.

OISTRIBUTION OF THIR AOCUMENT IS UNLIMITED
 

 

 

LEGAL NOTICE

This report was prepared as an account of Government sponsored work. Neither the United States,

nor the Commission, nor any person acting on behaif of the Commission:

A. Makes any warranty or reprssentation, expressed or impliad, with respact to the accuracy,
completeness, or usefulness of the information contained in this report, or that the vie of
ony information, apparatus, method, or process disclosed in this report moy not infringe
privately owned rights; or

8. Assumes any liabilities with respect to the use of, or for domages resulting from the use of
any information, apparatus, method, or process disclosed in this report.

As used in the above, “person octing on behalf of the Commission’ includes any employee or

contractor of the Commission, or employee of such contractor, to the extent that such employea

or contractor of the Commission, or employee of such contractor prepares, disseminates, or
provides access to, any information pursuant 1o his employment or contract with the Commission,

or his employment with such contrector.

 

 

Tull
L

CONTENTS

INTRODUCTION . .« v v v v v v v v s e v e e e e
DATA ACQUISITION AND REDUCTION , . . . . . . . .
2.1 Data Acquisition . + « . « « v o v o 0 00 0.

2.2 Data Reduction . &+ & v & v 4 v v ¢ v 4 o o o

THEORETICAL CALCULATIONS AND MODELING . . . .

3.1 Infroduction « .« « . . . . . o Lo oo oo L.
3.2 Development of the Model . . . . . . . . . . . .
3.3  Consideration of Possible Driving Functions. . . . .

MEASUREMENTS AND RESULTS . . .. . « . « . . . .
4.1 Establishment of Measurement Reproducibility and . .
Method of Spectrum Interpretation . . . . . . .

4,2 Results of Tests at 7 Mw Reactor Power . . . . . .
4,3  Results of Tests at 5 Mw Reactor Power . . . . . .
4.3.1 Neutron Noise Level vs Pump Bowl Level .
4.3.2 Neutron Noise Level vs Cover Gas Pressure
4,.3.3 Neutron Noise Level vs Average Reactor. .
Outlet Temperature . . « « .« « « « . .
4.3.4 Neutron Noise vs Net Reactivity . . . . .

4.4 Dominant Source of Observed Neutron Noise . . . .

CONCLUSIONS v & & v v vt v o i e e e e e e e e

FUTURE INVESTIGATIONS. . . . « . . v« o v o v . .

LEGAL NOTICE

This report was prepared as an account of Government sponsered work. Neither the United
States, nor the Commission, nor any person acting on behalf of the Commission:

A. Makes any warranty or representation, expressed or implied, with respect to the accu-
r?cy, c?nflpleteness, or usefulness of the information contained in this report, or that the fise
01 any information, apparatus, metho i
oeteately s o ::por . d, or process disclosed in this report may not infringe

B. Assumes any liabilities with reapect to the uge of, or for damages regulting from the
use of any information, apparatus, method, or process disclosed in this Teport.

As used in the above, *‘person acting on behalf of the Commission’ includes an em-
Ployee or contractor of the Commission, or empioyee of such contractor, to the exterfi: that
such employee or contractor of the Commission, or employee of such co’ntractor prepares
disseminates, or provides access to, any information pursuant to his erployment or contrac;;
with the Commission, or his emplovment with such contractor.

Page

N O

N SO 0O O

13

13
14
14
14
15

15
15
16
18

18

JISTRIOUTION OF YHS COCUMINT 15 UNLIMITED

i
i

i
1. INTRODUCTION ‘

Investigations were made at the MSRE to determine if the amount of helium
gas in the fuel salt can be measured using neutron noise analysis. In calculations
of reactivity balances that have been made by an on-line digital computer at the
Molten=-Salt Reactor Experiment (MSRE) since the start of power operation, there has
been an uncertainty concerning the concentration of ®%Xe in the circulating fuel
salt and, thus, an uncertainty in the reactivity balance computations. According
to Engel and Prince,? this uncertainty arises because an unknown and changing
amount of circulating voids (bubbles of undissolved helium gas believed to be
introduced by the xenon-stripping spray ring in the fuel pump tank) affects the
amount of ®°Xe in the system. Besides this uncertainty, these bubbles also introduce
a reactivity effect of -0.18% (A k/k) for 1 vol % of gas,since such voids affect the
neutron leakage and fuel inventory within the core.

Although measurements at zero power *## had indicated that no circulating
voids were present, pressure release tests’ performed later after operation at power
showed a small circulating void fraction at normal salt levels in the pump tank.

By comparing calculated and observed transient buildup of 135X e poisoning after

step changes in reactor power, Engel and Prince estimated the void fraction to be

between 0.1 and 0. 15 vol % and the bubble stripping efficiency to be between 50

and 100%. Later experience indicated that the circulating void fraction -
is dependent upon other reactor operating conditions, such as fuel pump tank level

for one.? Transient tests were performed to infer the void fraction and bubble

stripping efficiency, but because such tests interrupted the normal operation of the

MSRE, it seemed desirable to seek a nondisturbing method for determining the

helium void fraction and siripping efficiency.

Neutron noise analysis has been applied extensively as a nondisturbing method
for measuring reactor parameters at zero power and also for monitoring the dynamic
behavior of reactors operating at power. In an early application of this technique,

 

2
J. R. Engel and B. E. Prince, The Reactivity Balance in the MSRE,
ORNL-TM=1796 (March 10, 1967).

 

3
MSRP Semiann. Progr. Rept. Aug. 31, 1965, ORNL-3872, pp. 22-24.

4 ,
B. E. Prince et al., Zero-Power Physics Experiments on the Molten-Salt

Reactor Experiment, ORNL-4233 (Feb. 1968).

 

 

 
Hirota® concluded that gas effects were influencing the hydraulic behavior of the
Homogeneous Reactor Test (HRT) core; he compared calculated transfer functions
with the Fourier amplitudes of measured small (+3%) power deviations that occurred
during steady-state HRT operation. We therefore conjectured that neutron noise
analysis might be used to measure the amount of circulating void in the MSRE fuel
salt without disturbing the reactor operation in any way.

To determine the feasibility of using noise analysis to measure the void fraction,
three questions must be answered. First, do the helium bubbles, in passing through
the core, produce reactivity fluctuations that, in turn, cause neutron noise? Second,
if noise is generated by bubbles, is the frequency spectrum of the noise in a range
that is applicable to noise analysis and is the amplitude great enough to be detected
above background noise? Finally, is it possible to develop a quantitative relation-
ship between neutron power spectral density (NPSD) and the circulating void
fraction? The results of our investigation of these questions are presented in this
report.

To assure the maximum sensitivity of the measurement of neutron noise, special
data acquisition and reduction techniques were developed. After installation and
checkout of the equipment and analysis techniques, the neutron noise was measured
and analyzed at different reactor operating conditions, such as fuel-tank salt level,
average fuel outlet temperature, helium cover-gas pressure, etc., to determine the
effects of these variables on the NPSD. Concurrently with these measurements a
detailed theoretical model of the system was developed, and attempts were made to
understand what physical mechanisms could conceivably produce neutron density
fluctuations in the MSRE. These theoretical predictions were then compared with
the measured NPSD to infer that the most likely source of observed neutron noise in
the MSRE is the circulating helium bubbles. Finally, an attempt was made to
quantitatively relate the NSPD to the amount of void in the circulating fuel salt,
with some success.

The authors are grateful to C. B. Stokes for his assistance in performing the
measurements and to G. C. Guerrant for designing the ionization chamber assembly.
We also express our appreciation to the MSRE personnel for their help in performing
the tests, and especially to J. R. Engel for his many helpful discussions and
suggestions throughout the investigation.

2. DATA ACQUISITION AND REDUCTION

Figure 1 shows the essential elements that were used to obtain the power
spectral density of the neutron density fluctuations in the MSRE. The following two

 

5 . :
J. A, Thie, Reactor Noise, Rowman and Littlefield, New York, 1963,

pp. 214-20,
sections explain how the neutron ionization chamber signal was recorded and
analyzed to obtain the NPSD, or mean-squared noise per unit frequency.

2.1 Data Acquisition

To obtain an electrical signal proportional to the neutron density fluctuations,
a neutron-sensitive ionization chamber was placed in spare guide tube 7 of the
MSRE instrument penetration. This is a boron-coated, RSN-76A chamber filled
with a special gcs mixture (1500 mm *He and 50 mm CHy) to optimize it for neutron
noise analysis.® The chamber has a relatively high neutron sensitivity which is
deemed essential for noise analysis.

A two-conductor, low-capacitance, low-noise balanced cable carries the cur-
rent signal from the chamber and also the common-mode noise signal to the auxiliary
control room. The total capacitance to ground of the ~ 160 ft of signal cable plus
the chamber is 1880 pf. This capacitance, in parallel with the input resistance of
the noise amplifier, limits the highest frequency noise that can be analyzed with
this system. (This limitation is discussed further below.) The common-mode line is
terminated at the chamber with a capacitance to ground equal to that of the chamber.

The low-noise ac amplifier,” having a variable voltage gain from 200 to
10, 000 and a bandwidth of 0.013 to 3500 cps, was used to amplify the fluctuating -
portion of the chamber signal. 1t is a differential~input amplifier with a common~ -
mode rejection capability of 50 db for the elimination of unwanted electrical
noise picked up along the 160 ft of cable. Since the amplifier operates on voltage
signals, the current signal from the chamber must be converted to a voltage signal
at the amplifier input by using either a 2- or a 20-kilohm resistor to ground. Due
to the 20-kilohm resistor in parallel with the 1880-pf capacitance of the combined
cable and chamber, the practical upper frequency limit of the noise analysis
technique is 4240 cps, which is well beyond the range of interest for the MSRE; this
will be discussed inSection 2.2, The amplifier limits the lower frequency to
0.01 cps (-3 db), which is also consistent with the practical limitations on the lower
frequency imposed by statistical sampling laws.

The amplified ionization-chamber signal (~ 2 v p-p) at the amplifier output
is transmitted through shielded cable to the Bunker-Ramo 340 on-line digital

 

D. P. Roux, "Optimization of Reactor Shutdown Margin Measurements in
High Gamma Fluxes, " Nuclear Applications 3, p. 575 (Sept. 1967).

 

7
F. M. Glass, "Low-Noise Solid-State Differential Amplifier, " Instrumenta-
tion and Controls Div. Ann. Progr. Rept. Sept. 1, 1963, ORNL-3578, p. 113.

 
ORNL-DWG 68-8445

CAPACITANCE EQUAL . —— - e - q

f j TO CHAMBER CAPACITANGCE 0NSfl'\mlVIAF’GLR‘IEEDTIDCATT/}x -

 

 

  

| common-mope LINE

| |
l | 1
| | |
| ’(\‘r [ LOW-NOISE l '\ ' MODEL 340
SIGNAL LINE | I |
|
| | |

 

 

 

 

 

 

 

 

 

|
SUBAUDIO BUNKER~RAM N |
0 N |
AMPLIFIER DIGITAL COMPUTER S\ :
NEUTRON-SENSITIVE IONIZATION CHAMBER L AUXILIARY CONTROL ROOM | L MSRE COMPUTER ROOM ]
INSTALLED IN INSTRUMENT PENETRATION ~  ~——~—— ——————= b —— —— — — o — — — —
NO. 7

(o) DATA ACQUISITION AT MSRE

SAMPLED DATA
ON MAGNETIC TAPE

 

 

  

 

 

SPECTRAL DENSITY ANALYSIS CALCOMP PLOT
USING CDC1604 A OF FREQUENCY
COMPUTER SPECTRUM
L LUty gy

 

 

 

 

 

 

 

FREQUENCY
{6) OFF-LINE DATA REDUCTION

Fig. 1.  MSRE Neutron Noise Analysis.
computer, where the signal is sampled at a rate of 60 samples/sec and the digitized
data are recorded on magnetic tape for off-line analysis. The analog-to-digital
converter accepts signals in the range from -2 v to +2 v. For each test, data were
taken for 30 min, yielding 108,000 digitized values of the fluctuating neutron density.

2.2 Data Reduction

The frequency spectrum of the noise was determined by an off-line CDC
1604-A computer at the ORNL Computing Center, using a fast Fourier fransform
algorithm.® This method of analysis was chosen for its accuracy, speed, and good
frequency-resolution properties. However, because the Bunker-Ramo computer
had a fixed sampling rate of &0 samples/sec, the upper limit of the frequency that
could be analyzed was about 15 cps. As will be seen, this limit did not restrict
neutron noise analysis at the MSRE, because there is little useful information in
the neutron signal above 10 cps, but the 60 sample/sec rate did make it necessary
to filter the noise signal so that there would be negligible signal power at fre-
quencies greater than half the sampling rate (30 cps). This filtering was necessary
to prevent aliasing, or "folding, " of noise above 30 cps back into the frequency
range of interest (0 to 15 cps) which would have caused distortion of the frequency
spectrum of the neutron-induced current fluctuations in the ionization chamber.
Therefore, the bandwidth of the ionization-chamber-signal amplifier was changed
from 0.013 - 3500 cps to 0.013 - 35 cps (-6 db at 0.013 and 35 cps). Calibration
tests of the filtered signal showed that the amount of spectrum contamination caused
by aliasing was negligible. The bandwidth of the individual digital filters (fre-
quency spectrum resolution) chosen for the analysis was 0.117 cps. In the analysis
the frequency range from 0 to 15 cps was divided into 128 intervals of 0. 117 cps
each. This fine-frequency resolution was necessary to give an accurate picture of
the resonance structure of the NPSD curves. The standard deviation of the spectral
density determined for each of the 128 intervals for a 30 min sampling time was

+ 4%,

The data reduction program was coded to correct the NPSD for the amplifier
gain and to divide it by the square of the dc ionization chamber current (IDCZ).
(Earlier theoretical work® had pointed out that when the noise from reactors operating
at power is analyzed, variations are most easily interpreted when the NPSD is
normalized to the square of the dc component of the chamber current.) The
normalized NPSD functions were plotted using a CALCOMP plotter. A typical MSRE

 

®R. C. Kryter, "Application of the Fast Fourier Transform Algorithm to
On-=-Line Reactor Malfunction Detection," IEEE 15th Nuclear Science Symposium,
Monireal, Canada, Oct. 23-25, 1968.

?). C. Robinson, Analysis of Neutron Fluctuation Spectra in the Oak Ridge
Research Reactor and the High Flux Isotope Reactor, ORNL-4149 (Oct. 1967).
spectrum (computed in about 4 min on the 1604-A computer) is shown in Fig. 2.

The NPSD is plotted in absolute units of fractional mean-square reactor power fluctua-
tion per unit frequency interval in cycles/sec. Since in the frequency analysis it is
assumed that the spectral density is constant over the bandwidth of the effective
digital filter, we plotted the results in histogram form. The interpretation of the
spectra will be discussed in the next section.

3. THEORETICAL CALCULATIONS AND MODELING

3.1 Introduction

For a linear system driven by a single input, the output PSD, 3 o(w)’
related to the input PSD, @ (w) by

2
8 (w) = Gliw] 2.(w), (1)

where | G(jw)| is the modulus of the system frequency response function. Such an
output PSD, obtained from a neutron-sensitive ionization chamber at the MSRE
(operating at power), was presented in Fig. 2. To efficiently utilize the PSD
measurements in gaining insight into the dynamic properties of the reactor system,

it is necessary to understand the nature of the predominant input (driving function)
to the system., There are two possible approaches to the identification of this driving
function:

1.  Postulate a driving function and then attempt an experimental verification,
e.g., cross-correlation measurements or PSD measurements for a range of known
reactor conditions where various quantities are purposely modified, efc.

2.  Postulate a driving function, calculate analytically the appropriate system
transfer function, and then employ Eq. (1) to produce an expected output PSD
which can be compared to experimental results to validate or reject the postulate.

In principle, the cross-correlation techniques could lead to the unique identi-
fication of the driving function; however, in practice most cross-correlation measure-
ments are difficult. Therefore we adopted the combined approach of an analytic
model for the calculation of the frequency response function, coupled with PSD
measurements under controlled (as much as possible) reactor conditions.

Before a specific model was selected, the potential driving functions were
considered. In addition to potential driving functions commonly encountered, such
as coolant temperature fluctuations and rod vibrations,? it was known that helium gas
could be entrained in the fuel salt. Therefore, in addition to commonly encountered
10

potential driving functions, the possibility of fluctuations in the void fraction, which
could be induced by pressure or velocity fluctuations, was also considered.

A model had been developed by Ball and Kerlin!® for the calculation of the
power-to-reactivity frequency response function in the MSRE. Their model, basically
a multinode (fixed number of nodes) model becomes less valid as the frequency of
the disturbance increases.” Therefore, we decided that a distributed parameter mode!
would be necessary for the cases of interest to us. Furthermore, the BallKerlin
model treated a reactivity driving function only, whereas we wished to consider
several driving functions.

in the next section, the basic concepts of our model will be presented along
with the method of solution, and this will be followed by a section concerned with
the elimination of some potential driving functions and the identification of the
more probable driving functions.

3.2 Development of the Model

At the beginning of this study, we recognized that a model would be required
that would couple the neutronic and hydraulic states of the system; therefore, the
basic equations would be those of continuity of mass, momentum, and energy for the
fluid, along with the conservation of neutrons. If gas were present in the fluid fuel
(which was assumed to be the case), the fluid-gas system must be treated as a com-
pressible system, and, therefore, the conservation equations would contain the
dependent variables of pressure, velocity of the fuel salt, velocity of the gas,
temperature, void fraction of gas, density of the gas (we assumed that the density
of the salt would be constant), neutron flux, and precursor concentrations. Since
several dependent variables were involved, the following assumptions were made in
regard to fluctuations about the mean:

1. A "slug flow" model would be adequate to describe the hydraulics of the gas-
fluid system.

2.  The fluctuations about the mean would be small; therefore, the nonlinear
equations could be linearized.

3.  The state of the gas would be adequately described by the ideal gas law.

4.  Appreciable fluctuations in the density of the gas could be induced by
pressure fluctuations only.

5. The velocity of the gas would be proportional to the velocity of the fluid,
through a known "slip ratio" relationship.

 

05, J. Ball and T. W. Kerlin, Stability Analysis of the Molten-Salt Reactor
Experiment, ORNL-TM-1070 (Dec. 1965).

 
NPSD/IDC? (abs units)

ORNL-DWG 68-8416

REACTOR POWER=7.3 Mw
"KNET=—0.070 % Ak/k
FUEL TANK LEVEL=56%
“FUEL TEMPERATURE=1210°F
— PRESSURE=5psi

I I

| | |
. STANDARD DEVIATION OF HISTOGRAM (£4%)

 

 

5 © 7 8 9 10 14 12 13 14
FREQUENCY (cps)

Typical Neutron Noise Spectrum of the MSRE.

Ll
12

6. A one-dimensional, one-velocity neutron diffusion equation, which explicitly
accounts for delayed neutrons, would be adequate.

7.  The fluctuations in the fluid velocity could not significantly affect the
precursor balance equations.

With these assumptions, we devided the problem into two parts: o hydraulic
model which involves the momentum, mass of the gas, and mass of the fluid con-
servation equations, and a neutronic model which involves the energy, neutrons,
and precutsor conservation equations. The hydraulic model will be discussed first,
followed by the neutronic model.

The governing equations of the hydraulic model were reduced to a set of three
linear, coupled partial differential equations in space and time, with space-varying
coefficients. The time variable was removed by use of the Laplace transform, thus
obtaining a complex, coupled set of ordinary differential equations in space whose
coefficients are complex as well as space dependent. In mairix form the hydraulic
model can be written as

£ x29=AIxE9) ol

where % (z,s) is a column matrix (vector) whose elements are velocity fluctuations,
void fraction fluctuations, and pressure fluctuations; A(z,s) is a square matrix
whose elements depend on the system properties, steady=-state disfributions, efc;

z is the spatial variable; and s is the Laplace fransform parameter.

The frequency range of primary interest was about 0.1 to 20 cps. Since the
tatal loop time for the fuel salt is about 25 sec, we concluded that the details of
the salt loop external to the core would not be important; therefore, a simplified
physical system was chosen to present the more complex actual system (Fig. 3). In
particular, six regions (identified as L; through Lg in Fig. 3) were chosen:

1.  the region from the primary pump to the inlet of the downcomer including
the heat exchanger, L;;

2, the downcomer, Ly;
3. the lower plenum, Ls;

4.  a large number of identical parallel fuel channels'!, L

 

"The reactor actually consists of hydraulically different, parallel channels,
but no attempt has been made to model these.
13

ORNL-DWG 68 -8417

 

 

 

 

 

 

 

 

 

 

OUTLET
INLET T
16
£y UPPER PLENUM L5¢
YT

CHANNELED

REGION

{ bt
, |
|

I~
N

 

 

 

 

¥
L3 k_//' LOWER PLENUM l\\\_)

Fig. 3. Model Used to Approximate the MSRE Fuel Salt Loop.

 

 

 
14

5. the upper plenum, Ls;
6. the region from the reactor vessel to the primary pump, Lg.

The solution of Eq. 1 can be written for each region as

X i(s) = !\i(s)xi i(s) , (3)

I

where the subscripts o and i refer to the output and input respectively, subscript j
is for the | region, and Ai(s), given by

L.
Ai(S) = exp [ IJ'Ai(z,, s)dz] , (4)
0

is referred to as the transport kernel for the ifh region, The transport kernel is
evaluated for each region using matrix exponential techniques. Then coupling
equations are applied between each region; this permits the evaluation of the
overall system transport kernel, 1/4(s), then

X, 1) = 1ls) X ) (5)

The boundary conditions are then applied which (a) closes the loop and (b) inserts

a driving function, e.g., pressure fluctuations, between regions 6 and 1. The void-
fraction fluctuation, velocity fluctuation, and pressure fluctuation spatial distribu-
tions through the core are now obtained.

In the neutronic model the equations of interest are the conservation of energy,
the diffusion equation for neutrons, and the precursor balance equations.

The basic assumptions in addition to those previously stated were:
1. The flux introduced would be separate in space and time.

2.  Reactivity introduced into the system from temperature and void fluctuations
could be accounted for by using the appropriate coefficients of reactivity.

3.  The importance of the spatial insertion of reactivity would be properly
accounted for by using the variational principle.

Since the method of solution was similar to that presented in Ref. 9, it will
not be described here.
15

3.3 Consideration of Possible Driving Functions

The potential driving functions that have been considered are fluctuations in

1. the salt temperature at the entrance to the core region,
2,  the void fraction induced by salt velocity fluctuations at the pump,
3. the void fraction induced by fluctuations in the mass flow rate of gas entering

the salt at the pump bowl,

4.  the void fraction induced by pressure fluctuations introduced at the primary
pump or pump bowl,

5.  the reactivity caused by control rod vibrations.

The first three of these driving functions were eliminated because a comparison of
observed (experimental) PSD with the calculated (theoretical) PSD showed that
experimental PSD decreased about one decade in magnitude in the frequency range
of 0.2 to 1 cps, but the analytical PSD (for an assumed white driving function)
decreased about three decades in the same frequency range for each of these functions.
If any one of them had been primarily responsible for the observed PSD, the
magnitude of that particular function would have had to increase with frequency in
this frequency range. Physically, this behavior would be unreasonable.

The possibility of rod vibrations could not be eliminated analytically. Like-
wise, the fourth driving function could not be eliminated, because calculation shows
that a pressure fluctuation of 0.01 to 0.05 psi (which is physically realizable)
would be sufficient to produce the observed noise. The calculated PSD for an
assumed pressure driving function of unity magnitude in the frequency domain is
presented in Fig. 4 for two different mean steady-state void fractions. Although
the shapes of these calculated PSD curves are not precisely the same as the observed
PSD curves, we still regard pressure fluctuations as a highly probably driving
function for the following reasons:

1.  The required magnitude of the pressure fluctuations is not unreasonable.

2.  The analytical model developed is not expected to have sufficient detail
to predict exactly the observed frequency dependence.

As a further point of interest (see Fig. 4), the magnitude of the analytic
PSD (due to the fourth driving function possibility listed) is proportional to the
square of the steady=state void fraction existing in the core for frequencies below
4 cps (this observation will be applied in Section 4.4).
16

4. MEASUREMENTS AND RESULTS

4.1 Establishment of Measurement Reproducibility
and Method of Spectrum Interpretation

The theoretical studies described in Sect. 3 suggested several possible driving
functions for neutron noise. However, before studying the effects of particular
parameter changes, such as control rod position, etc., we observed the MSRE
neutron noise daily for several weeks to determine the reproducibility of the measure-
ments and how the shape and magnitude of the noise spectrum changed under normal
operating conditions. The results of these tests show that the measurements were
reproducible to within the anticipated £5%4. However, we did observe small con-
sistent changes in the spectrum, which indicated that the driving function (reactivity
fluctuation) was varying slightly from day to day. This variation was most pronounced
as an amplitude change in the vicinity of the 1-cps peak (see Fig. 2).

Because of this localized sensitivity of the spectrum, in addition to observing
the detailed shape and magnitude of the entire 0. 1- to 15-cps spectrum, we also
computed the noise level NPSD/(IDC)? averaged over the interval 0.5 to 2.0 cps.
(This averaged noise is defined as NPSD.) As a result of this averaging, changes in
the noise level were enhanced and the precision of the measurements increased to

+14.

Following these tests to establish confidence in this method of data acquisition
and reduction by using known test signals, a series of special tests was performed at
nominal reactor powers of 5 and 7 Mw to determine the effect of selected system par-
ameters on the neutron noise spectrum, the NPSD, and reactivity balance. (These
special tests were performed to better understand the origin of small reactivity changes
which had previously been indicated by the reactivity balance. ) The parameters
studied were: control rod position, B%(e poisoning, average fuel-outlet temperature,
helium cover-gas pressure and the fuel-salt level in the pumg tank. The P3Xe
poisoning was inferred from reactivity balance calculations.” The helium cover-gas
pressure in the fuel pump bowl was measured by a pressure transmitter in a helium
supply line outside the main secondary containment shell, approximately 15 ft from
the fuel pump bowl. The average fuel salt temperature at the reactor outlet was
calculated by the BR-340 computer from temperature readings from three thermo-
couples in the salt loop. Since significant changes in some of the parameters pro-
duced transient effects, the system was allowed to reach equilibrium (~48 hr) follow-
ing a change before noise measurements were made. The results of these tests are
presented in the following sections.

 

ZAMSRP Semiann. Progr. Rept. Feb. 29, 1968, ORNL-4254, pp. 3-7.

 
17

ORNL-DWG 68-8418

100

1071

10~3

FLUX TO PRESSURE ANALYTIC PSD (arbitrary units)

o VF=0,30%
e VF=0.095%

 

10~4
1074 100 Tok 102
FREQUENCY (cps)

Fig. 4. Calculated PSD for Void Fractions of 0.095 and 0.30 Vol %,

Assuming a White Pressure Driving Function.
18

4.2 Results of Tests at 7 Mw Reactor Power

Because of previous experience at the HFIR and ORR’, where we concluded
that a major portion of the neutron noise was caused by control rod vibration, we
speculated that the reactivity fluctuations in the MSRE might also be dependent on
the position of the rods. However, we found that NPSD varied only about 2% for
three regulating rod positions of 32.3, 36.4, and 38.6 in. This change borders on
significance because the statistical error of the NPSD is only £1%, but in practice,
other uncertainties either in equipment calibration or small changes in the reactor
system probably limit the reproducibility to more like +5%. Therefore, we conclude
that control rod vibration is not a significant source of neutron noise at the MSRE
in the frequency range of 0.1 to 15 cps.

Another series of tests was conducted to study the effects of equilibrium *°Xe
concentration on the NPSD, Following a period of operation at zero power during
which ¥%e was stripped out, the power was increased to 7 Mw and noise measure-
ments were taken at “°Xe poisoning levels of =0.012 and -0.261% ak/k. The
NPSD increased by 154, indicating some xenon dependence. However, because
the system was still in a transient condition due to the increase in power and because
of unavoidable changes in other parameterssuch as fuel tank salt level and helium
pressure, one cannot validly conclude that a cause and effect relationship exists.
We therefore suggest there might be some xenon dependence, but conclude that it
is not a significant contributor to neutron noise in the MSRE,

4,3 Results of Tests at 5 Mw Reactor Power

The results of tests performed at 5 Mw (to allow a wider range of outlet
temperature variation than would have been possible at 7 Mw) show the changes in
the NPSD as a function of changes in the operating conditions of the MSRE. To
illustrate the behavior of the noise as a function of each parameter varied, the same
data are presented in several figures. This is necessary because, in general, we
did not have exclusive control of only one variable at a time since the variables
such as fuel temperature, pump bowl level, pressure, etc., were related
interdependently.

4.3.1 Neutron Noise Level vs Pump Bowl| Level

 

Most of the time when the MSRE has been operated at power there has been a
small but continuous transfer of salt from the primary loop to the overflow tank, which
produced a change in salt level in the fuel pump bowl. Also, as the salt level in the
pump bow! was decreased below a certain level, considerably more helium bubbles
were apparently introduced into the circulating salt. Therefore noise measurements
19

were performed at different pump bowl levels in the normal range of operation

(5.2 - 6.0 in.) to determine the effect of salt level on noise amplitude. Figure 5
shows that as the salt level in the bow| decreased the noise level at about 1.0 cps
(NPSD) increased. Although this effect was small at the normal operating temperature
and pressure (1210°F and 5 psig), it was reproduced in many measurements over a

long time span. However, this effect almost vanished when the fuel outlet temper-
ature was increased to 1225°F, but Fig. 5 shows that when the temperature was
decreased to 1180°F the effect was much greater.

4,3.2 Nevutron Noise Level vs Cover Gas Pressure

 

Since the helium pressure in the fuel pump bowl is also a variable parameter,
the effect of the pressure on the NPSD is presented (Fig. 6). There was no change
in the NPSD for pressures between 3 and 9 psig at an operating temperature of 1225°F,
At the normal operating temperature of 1210°F, the NPSD increased with increasing
pressure. At the subnormal temperature of 1180°F, the NPSD increased markedly at
all pressures from 3 to 9 psig, with the highest noise occurring at the highest pressure

(9 psig).

4,3.3 Neutron Noise Level vs Average Reactor Outlet Temperature

 

The averaged neutron noise showed the most sensitivity to changes in the
reactor outlet temperature (Fig. 7). The largest effect was at the highest pressure
of 9 psig, where the NPSD increased by a factor of almost 50 when the temperature
was decreased from 1225 to 1180°F. At the normal operating pressure of 5 psig, the
noise increased by a factor of 15 for this same change in temperature. Although
data were taken only at 1225 and 1180°F at 3 psig, the trend of increasing noise
with decreasing temperature is seen to be consistent with the resulis at 5 and ¢ psig.
As a further illustration, Fig. 8 shows the change in the entire 0. 1- to 15-cps
noise spectrum in passing from a minimum to a maximum noise condition.

4,3.4 Neutron Noise vs Net Reactivity

 

The residual system reactivity? is determined by requiring a reactivity balance
between the calculated reactivity and the reactivity inserted by a calibrated control
rod. The net reactivity is computed only as a steady-state quantity, although it
does, of course, change slowly as system parameters are varied. As mentioned
earlier, the largest uncertainty in the reactivity calculations is believed to be the
lack of a measure of the circulating void fraction and its effect on *%Xe poisoning.
Therefore the residual reactivity (KNET) is dependent, among other things, on the
void fraction.
20

ORNL- DWG 68- 8419

 

1g©
5
AVERAGE FUEL QUTLET HELIUM PRESSURE
TEMPERATURE (°F) (psig)
2 O
07
@ 5
£ ®1180-5
w
L
S
3
g ¢ -
T—-a1180-3
108
5
210-5  y225-5
1225-9
2
1072
50 52 5.4 56 5.8 6.0 6.2 6.4

FUEL PUMP TANK LEVEL (in.)

Fig. 5. Neutron Noise Level vs Fuel Pump Tank Salt Level for the MSRE

at 5 Mw.
NPSD {(abs units)

0

-9

21

ORNL-DWG 68-8420

AVERAGE FUEL OUTLET TEMPERATURE
(°F)
8o
2.5 3.5 4.5 55 6.5 7.5 8.5

HELIUM PRESSURE (psig)

Fig. 6.  Neutron Noise Level vs Helium Cover Gas Pressure at 5 Mw.

 

95
22

ORNL-DWG 68- 8421

 

10 ©
5

o HELIUM PRESSURE

{psiq)
9

o’
TS
=
3
w
L0
S
0O
£ 2
=

108

5

2

10~3

1220 1230

n70 1180 190 1200 1210
AVERAGE FUEL OUTLET TEMPERATURE (°F)

Fig. 7.  Neutron Noise Level vs Average Fuel Temperature at Core Outlet

for the MSRE at 5 Mw.
NPSD/1DC? (abs units)

   
   

10

10

with Maximum Circulating Void Fractions.

23

ORNL-DWG 68-8422

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

— : _ B —
" OUTLET TEMPERATURE=
PRESSURE=9.2 psig

MAXIMUM CIRCULATING \OID

 
 
  
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o |
T QUTLET TEMPERATURE=12
~- -t — PRESSURE=3 psig
——~ MINIMUM CIRCULATING VOID

  

 

 

 

 

 

 

 

 

25°F T

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i

|
L

|

|

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fig. 8.

1

2 3 4

&

6 7

8 9 10 14 12 13 14 15

FREQUENCY (cps)

Neutron Noise Spectra of the MSRE at 5 Mw with Minimum and
24

In contrast to the static nature of net reactivity, neutron noise is a measure
of small fluctuations in reactivity; i.e., it is a dynamic quantity. However, if
we postulate that the neutron noise, like KNET, is also dependent on the volume
fraction of circulating helium bubbles, then there may conceivably be some con-
sistentrelationship between KNET and NPSD. The measurements show a variation
of NPSD with changes in KNET (Fig. 9). The data presented in Fig. 9 cover the
complete range of temperatures, pressures, and fuel pump bowl salt levels described
in the previous three sections of the tests conducted at 5 Mw. Although we con-
nected the data points with a smooth line, we ascribe very little significance to
the shape of the curve at this time., However, we can conclude that there is a
consistent, monotonic increase in the neutron noise in the range 0.5 to 2 cps for
a larger amount of poison in the reactor (as indicated by KNET),

These changes in KNET cannot be attributed to the reactivity effect of voids
alone, because this would require an unreasonably large void fraction: about
1.5 vol %, based on the -0.18% ak/k per vol % reactivity associated with gas in the
fuel.? This conclusion is substantiated by fuel volume inventory balances and zero-
power reactivity balance calculations maintained at the MSRE, which show the
maximum amount of fuel salt displaced by gas to be about 0.3 vol %. 3 Therefore,
it is now believed that the changes in KNET are predominantly due to increased
B% e poisoning as the amount of gas in the fuel is increased. This conclusion appears
to contradict the previously accepted belief that an increased amount of circulating
bubbles enhances the stripping of *%Xe from the fuel salt and,thus, decreases the
amount of xenon in the system. However, the original view failed to recognize the

interdependence of circulating void fraction with other system variables that can
also lead to higher zenon poisoning at higher void fractions,

4,4 Dominant Source of Observed Neutron Noise

The comparison of theoretical model predictions with experimental spectra in
Sect. 3 showed that the most likely sources of reactivity fluctuation in the MSRE are
control rod vibration and pressure fluctuation of circulating voids. Special tests in
which the rod position was varied provided a basis for concluding that rod vibration
is not the cause of significant neutron noise. However, further tests showed that
the magnifude of the noise is sensitive to pump bowl level, cover gas pressure, and
fuel outlet temperature, and it is known that changes in parameters alter the amount
of circulating void. Therefore, it seems plausible that the fluctuating helium
bubbles cause small reactivity fluctuation, which in turn causes reactor power
fluctuations and, hence, the observed neutron noise.

As noted in Sect. 3.3, our theoretical studies showed that, if the observed
neufron noise is due to pressure fluctuation of voids, the NPSD should be proportional

 

BM., p. 4.
10-7

P
o
!

NPSD {abs units)

109

8

 

-0.04

Fig. 9.

ORNL-DWG 68-8423

-0.08 -042 -0.46 -0.20 -0.24 -0.28
KNET (% Ak/K)

Neutron Noise Level vs Net Reactivity for the MSRE at 5 Mw.

¢¢
26

to the void fraction squared. Therefore, we attempted to compare the measured
NPSD (actually NPSD) with the estimated circulating void fraction at the time of
the measurement. 4

To make this comparison, we assume an expression of the form
NPSD - v =aV?, (6)

where NPSD is NPSD/(IDC)?, averaged over the frequency range 0.5 to 2 cycles/sec;
y is the unavoidable background noise produced by the random neutron detection
process; V is the vol % of circulating void; and o is a proportionality constant. Since
the available experimental information is the NPSD data and the change in void
fraction (6V =V - V) relative to some minimal condition V, we recast Eq. 6 in the
form

Y
ov=[ D2 Y ] -y %

In Eq. 7, we must estimate v, @, and V,, but we have only six estimates of §V.
Therefore, some of these parameters must be estimated by other means.

Since v is very small relative to NPSD, we neglect v, leaving only the two
parameters o and V. In principle, these can be obtained with a fit of Eq. 7 to the
data; however, ’rhe 5V data is estimated' to be accurate to approximately +50%
only (this is to be compared with a statistical precision of the NPSD data of 1-5%).
Because of this large uncertainty in 8V, we did not perform the two-parameter fit but
instead we assumed values for Vo which probably bracket the real value, and used the
experimental data only for the evaluation of o. Figure 10 shows the results for an
assumed Vg of 0.1 andzerovél 4 About all we can conclude from Fig. 10 is
(@) the minimal void fraction V_ appears to be more nearly zero; and (b) more
:mporfanfly, the NPSD does crppear to have the predicted squared dependence on
V, thus giving additional support to our conclusion that the neutron noise in the
MSRE is caused mainly by pressure-induced variations of the helium volume in the
reactor core. Although we have now established with some confidence that a
causal relationship exists, we do not attempt to explain how the helium bubbles are
introduced into the fuel salt or why the amount of void in the salt is so strongly
dependent on the fuel pump bowl level, helium cover gas pressure, or fuel salt
temperature.

 

MEngel has estimated the amount of circulating void present during the series
of tests at 5-Mw power. He is unable to supply an absolute void fraction, but rather
expresses his estimates in terms of a change in the void fraction from a minimum
(or base condition) void fraction, ~0.1 vol. % (MSRP Semiann. Progr. Rept. Feb. 29,

1968, ORNL-4254, p. 4),

 

155, R. Engel, ORNL, private communication.
0.28

0.24

0.20

o
o

Vo +8 V,VOID FRACTION (vol %)
O O
o -
© o))

0.04

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NPSD/3.7x10° (abs. units)

Fig. 10. Circulating Void Fraction vs NPSD for Assumed Minimum Void

Fractions of 0.1 and 0 Vol 4.

ORNL-DWG 68—-8424
| / ' ‘L“
/L / | [
/ |
2 Vo=0.14vol % |
Z 70 "
 NPSD =1.17x10°6 12
/ | |
- /y o DATA ASSUMING V= 0.1 vol % _
/ * DATA = o
7 ‘
o ’- // VO‘—" 0 ;IOI o !
° / _—TNPSD =4.34 x 108 12
/ o/
* /1 // 1
/ -
e i
0.01 0.02 0.03 0.04 0.05

LT
28

5. CONCLUSIONS

The results of experimental tests and analytical studies have shown that the
principal source of small neutron-density fluctuation observed in the MSRE is
helium bubbles circulating in the fuel salt.

We conclude from the limited amount of data that the neutron noise (NPSD)
in the frequency range from 0.5 to 2 cps varies as the square of the helium void
fraction; furthermore, the absolute void fraction corresponding to the minimum noise
condition observed is more nearly zero than the previously accepted value of 0. 1%.
These observations further substantiate that the neutron noise is extremely sensitive to
the helium void in the core. We therefore conclude that changes in the circulating
void fraction can be inferred with good sensitivity directly from neutron noise measure-
ments in the MSRE. Consequently, NPSD measurements can complement and enhance
the value of reactivity balance calculations in diagnosing residual reactivity in
the reactor by eliminating some of the reactivity uncertainty associated with an
unknown amount of circulating void.

Finally, we believe that these results have indicated the usefulness of neutron
noise analysis for on-line reactor diagnosis at the MSRE, but further work needs to
be done before the method can be fully implemented for measurement of voids.

6. FUTURE INVESTIGATIONS

Although all spectra reported here were computed off-line, the NPSD can be
computed on-line by the BR-340 computer,and a program similar to that written for
the CDC 1604-A computer is currently being tested. ' The estimated execution time
is approximately 8 min following acquisition of the ionization chamber data. This
on=line data reduction system will be used to repeat the tests described in this report
when the 22U-fueled MSRE reaches full-power operation.

Theoretical studies show that the absolute void fraction could be measured by
crosscorrelating the neutron noise signal with a pressure noise signal obtained from
transducer placed in the primary loop of the MSRE. The existing pressure sensor, as
pointed out earlier, is located ~ 15 ft from the pump bowl in a helium supply line.
This location could make it insensitive to the small pressure fluctuations that we
believe are causing the bubble volume to fluctuate in the core. It is probably not
feasible at this late date to place a pressure sensor in the primary loop of the MSRE,
but such a sensor should be seriously considered when instruments are designed for a
scaled-up version of this type reactor.

 

6C. L. Partain, ORNL, private communication.
29

However, lacking the proper pressure signal, there may still be a way to
establish a calibration between the neutron noise level and the absolute void
fraction. A deterministic pressure fluctuation experiment is capable of measuring
quantitatively the absolute void fraction. A preliminary experiment of this type has
been performed already in which a sawtooth pressure fluctuation was purposely
introduced and its effect on neutron level was investigated. 7 If the relationship
between neutron noise and absolute void fraction can be calibrated using such a
pressure test, then the NPSD should provide a relatively easy, on-line, non-
disturbing measurement of the absolute void fraction.

 

7). C. Robinson and D. N, Fry, Determination of the Void Fraction in the
MSRE Using Small Induced Pressure Perturbations, ORNL-TM-2318 (to be published).

 
45.
46.
47.
48,
49,
50.
51.
52.

31

ORNL-TM-2315

INTERNAL DISTRIBUTION

R. K. Adams

J. L. Anderson
S. J. Ball

A.E. G. Bates
H. F. Bauman
S. E. Beall

M. Bender

E. S. Bettis

D. S. Billington
F. T. Binford
E. G. Bohimann
C. J. Borkowski
G. E. Boyd

R. B. Briggs

J. B. Bullock
T. E. Cole

J. A. Cox

J. L. Crowley
F. L. Culler, Jr.
R. A. Dandl

H. P. Danforth
S. J. Ditto

W. P. Eatherly
J. R. Engel

D. E. Ferguson
A. P. Fraas

D. N. Fry

J. H. Frye, Jr.
C. H. Gabbard
R. B. Gallaher
W. R. Grimes
A. G. Grindell
G. C. Guerrant
R. H. Guymon
P. H. Harley
C. S. Harrill

P. N. Haubenreich
A. Houtzeel

T. L. Hudson
W. H. Jordan
P. R. Kasten

R. J. Kedl

M. T. Kelley

53.
54-63.
64.
65.

66.

67.

68.

69.

70.
71.
72.

73.

74.

/5.

76.
77.

78.
79-80.
81.
82.

83.

84.

85.

86.

87.

88.

89.

90.

?1.
92,

93.
94.
?3.

96.
97-98.
?9.
100-102.
103.
104,
105-119.

120,

121,

. |. Krakoviak

C. Kryter

. G. MacPherson

E. MacPherson

. D. Martin

. E. McCoy

L. Moore

L. Nicholson

C. Odkes

. G. O'Brien

. R. Owens

. W. Peelle

. M. Perry

B. Perez

P. Piper

E. Prince

L. Redford

. W. Rosenthal

. P. Roux

. S. Sadowski

unlap Scoftt

. Skinner

. Steffy

. Stokes

Tallackson

Thoma

. Trauger

. Walker

. R. Weir

K. W. West

A. M. Weinberg

M. E. Whatley

J. C. White

Gale Young

Central Research Library

Document Reference Section

Laboratory Records Department

Laboratory Records, ORNL R. C.

ORNL Patent Office

Division of Technical Information
Extension

Laboratory and University Division,

ORO

Nuclear Safety Information Center

Irm>xTNPI?®>»

Q@

>

o T ©

OQUZ

NOUFP=0OFZ
v @ T D o M)

-
~
122.
123.
124,
125.
126.
127.
128.
129.
130.

131.
132,
133.
134-135.
136.

137.
138.
139.
140.
141,
142.

143,
144.
145-154.
155.
156.
157.

158.
159.

160.
161.
162.
163.
164,
165.

32

EXTERNAL DISTRIBUTION

N. J. Ackermann, University of Tennessee, Knoxville, Tennessee

A. Z. Akcasu, University of Michigan, Ann Arbor, Michigan,

C. B. Deering, AEC-OSR

C. E. Cohn, Argonne National Laboratory, Argonne, lllinois

E. P. Epler, Oak Ridge, Tennessee

A. Giambusso, AEC-Washington

S. H. Hanauer, University of Tennessee, Knoxville, Tennessee

T. W. Kerlin, University of Tennessee, Knoxville, Tennessee

A. E. Klickman, Atomic Power Development Association, Detroit,
Michigan

B. R. Lawrence, Australian AEC (UKAEA, Risley, England)

F. C. Legler, Atomic Energy Commission, Washington, D. C.

G. H. McCright, Black and Veatch, Kansas City, Missouri

T. W. Mclntosh, AEC-Washington

M. N. Moore, San Fernando Valley State College, Northridge,
California

R. K. Osborne, University of Michigan, Ann Arbor, Michigan

C. L. Partain, University of Missouri, Columbia, Missouri

J. R. Penland, University of Tennessee, Knoxville, Tennessee

Joseph Pidkowicz, USAEC, Ock Ridge, Tennessee

C. A. Preskitt, Gulf General Atomic, San Diego, California

V. Rajagopal, Westinghouse Atomic Power Division, Pittsburgh,
Pennsylvania

. L. Randall, Atomics International, Canoga Park, California

. W. Ricker, Albion College, Albion, Michigan

. C. Robinson, University of Tennessee, Knoxville, Tennessee

. M. Roth, AEC-ORO

. F. Saxe, North Carolina State University, Raleigh, North Carolina

M. A. Schultz, Pennsylvania State University, University Park,
Pennsylvania

K. J. Serdula, Atomic Energy of Canada, Ltd, Chalk River, Ontario

J. R. Sheff, Battelle-Pacific Northwest Laboratory, Richland,
Washington

W. L. Smalley, AEC-ORO

J. A. Thie, Consultant, Chicago, lllinois

J. R. Trinko, Jr., EBR-Il Argonne National Laboratory, Idaho Falls, Idaho

C. O. Thomas, University of Tennessee, Knoxville, Tennessee

R. E. Uhrig, University of Florida, Gainesville, Florida

P. J. Wood, MIT Practice School

o T - N R