Annals of Nuclear Energy 114 (2018) 561–568

Contents lists available at ScienceDirect

Annals of Nuclear Energy

journal homepage: www.elsevier.com/locate/anucene

Technical note

Penn State University TRIGA Reactor Digital Reactivity Computer:

Development and Testing
James A. Turso
Penn State University Radiation Science and Engineering Center, University Park, PA 16802, USA

a r t i c l e i n f o a b s t r a c t

Article history: A digital control rod reactivity computer (DRC) has been developed using commercially available soft-
Received 6 November 2017 ware (National Instruments LabView) for use to determine control rod worth curves for the Penn State
Received in revised form 8 December 2017 University TRIGA research reactor. For low-worth reactivity steps, use of the prompt-jump approxima-
Accepted 15 December 2017
tion to the point kinetics equations enabled rapid calculation of rod reactivity insertions and minimized
Available online 10 January 2018
measurement noise effects. In contrast, use of the In-Hour method (an historically-effective way of deter-
mining control rod reactivity) would take several hours, compared to 30 min with the DRC for a complete
Keywords:
control rod worth curve. All of the data acquisition, reactivity calculations, and integral control rod worth
Digital Reactivity Computer
LabView
curve fitting are performed in LabView, which additionally provides for development of the graphical
Research Reactor Control Rod Reactivity user interface. Results demonstrate the LabView DRC developed provides a platform for these measure-
Worth Measurements ments that calculates accurate control rod reactivity measurements in significantly less time than the his-
torically accepted In-Hour method.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Eqs. (1) and (2) describe changes in reactor power below the so-
called point of adding heat, where temperature feedback effects
Core reactivity computers have been in-use for over 3 decades – are virtually nonexistent, which in the case of the Penn State TRIGA
initially as purely analog machines, then evolving into digital ver- research reactor is 1.0 kW. The system input being the core reactiv-
sions running on personal computers (referred to as Digital Reac- ity, qðtÞ, which describes the core’s deviation from criticality
tivity Computers, or DRC). Even with their availability (not K eff  1
necessarily commercially available), many facilities in the research qðtÞ ¼ ð3Þ
K eff
reactor community still rely on the In-Hour Equation/Stable Period
method of determining control rod effects on core reactivity, more Below the point of adding heat, qðtÞ, is composed of control rod
commonly known as control rod worth. To provide context for the reactivity, long-lived fission product poisons, and burnable poisons
development of the equations used in Penn State’s current version (depending on reactor design). For a positive step-change in reac-
of its DRC, a review of the In-Hour Equation/Stable Period method tivity, the behavior of the point kinetics equations is shown in
of reactivity determination is provided (Duderstadt and Hamilton, Fig. 1.
1976, Glasstone and Edlund, 1952, LaMarsh and Baratta, 2017). The In-Hour equation is derived from the solution to equations
The In-Hour Equation is essentially the solution to the point (1) and (2) for a step change in qðtÞ. For simplicity, the solution for
kinetics equations (with one (n = 1) or more (n = 6) delayed neu- 1 delayed neutron group shall be presented. If solutions of the form
tron groups), with the neutron kinetics equation provided by (1) for n(t) and c(t) are assumed

dn qðtÞ  b X
n¼6 nðtÞ ¼ Aext
¼ nðtÞ þ ki ci ðtÞ ð1Þ
dt K i¼1 CðtÞ ¼ Bext
And the delayed neutron precursor kinetics equations provided and inserted into equations (1) and (2)
by (2)
b xt
dci ðtÞ bi Bxext ¼ Ae  kBext
¼ nðtÞ  kk ci ðtÞ ð2Þ K
dt K
bA
B¼
E-mail address: [email protected] Kðx þ kÞ

https://doi.org/10.1016/j.anucene.2017.12.046
0306-4549/Ó 2017 Elsevier Ltd. All rights reserved.
562 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568

4. Determine the time reactor power takes to double. Record

time (this will be used in Eq. (5) after all data is acquired).
5. With power < 1 kW, insert remaining control rods to place
reactor in a critical condition at the original 100 W. This need
to be accomplished prior to the point of adding heat, so that the
reactor’s inherent temperature feedback mechanism does not
the reactor period measured.
6. Repeat 2) - 5) until calibrated control rod fully is withdrawn
from core.

The power evolution during the test is shown in Fig. 2. As Fig. 2

suggests, there is significant repetition in operator action and it
may take several hours to obtain all of the data points necessary
to create a single, accurate control rod worth curve. For 4 control
rods, the complete process would take 2–3 days, depending on
whether each data point was taken correctly.
The result of the calibration is the control rod reactivity vs. core
height curve, or the control rod worth curve, and is used by reactor
Fig. 1. Behavior of the point kinetics equations to a positive step change in qðtÞ at operators to accurately place the control rods for reactor startup,
10 s. low-power and high-power operations. Additionally, temperature
feedback effects may be quantified by using changes in control
rod critical position (and reactivity) due to changes in core temper-
ature. Fig. 3 is an example of a control rod worth curve.
dnðtÞ q  b xt
) Axext ¼ Ae þ kBext
dt K
2. Reactivity Computer Algorithm
ðq  bÞ kbA
Ax ¼ Aþ Inspection of Eqs. (1) and (2) would suggest that if reactor
K Kðx þ kÞ
power is measured, then Equation (1) could be rearranged to pro-
And recognizing that the reactor time constant, otherwise vide qðtÞ as a function of nðtÞ, ci ðtÞ, and dnðtÞ, resulting in Eqs. (6) and
dt
known as the reactor period (the time needed for the reactor
(7)
power to change by a factor of ‘‘e”) is ( )
K dnðtÞ Xi¼6

x¼
1 qðtÞ ¼ b þ   ki ci ðtÞ ð6Þ
s nðtÞ dt i¼1

The above may be solved for qðtÞ to provide the one delayed
dci ðtÞ bi
neutron group In-Hour equation ¼ nðtÞ  ki ci ð7Þ
dt K
K b
qðtÞ ¼ þ ð4Þ Eqs. (6) and (7) are the so-called Inverse point kinetics equa-
s 1 þ keff s
tions. Use of Eq. (6) may be problematic depending on power signal
where K is the prompt neutron lifetime, keff is the one delayed neu- quality i.e., electromagnetic noise as part of the reactor power sig-
tron group effective decay constant, and b is the fraction of delayed nal. Specifically, the power derivative, dnðtÞ
dt
, based on a noisy power
neutron emitted per fission. With s determined from data obtained signal, will amplify the signal noise. The gain of this amplification
from measuring the time period taken for the reactor power to dou- being the frequency corresponding to the noise, with higher fre-
ble i.e., the doubling time quency noise components resulting in large signal fluctuations that
0:693
s¼ ð5Þ
Doubling Time
the corresponding reactivity may be determined. In practice, a 6-
delayed neutron group version of Eq. (4) is typically utilized to
enhance accuracy of reactivity calculations (Toppel, 1959,
Glasstone and Sesonske, 1967).
An example procedure for determining the entire control rod
worth curve would be

1. Start CR worth test at 100 W. With reactor critical, put remain-

ing 3 control rods into manual mode of operation. The Penn
State reactor has the capability to have 3 of its 4 control rods
in automatic power control mode.
2. Remove control rod being calibrated. The reactor operator
removes the control rod as close to a step change in reactivity
as practicable. Typically many, small changes in control rod
position are required in order to accurately represent the non-
linear nature of the control rod worth curve.
3. Wait until a power rise with a stable reactor period is
observed. This may take several minutes depending on the
amount of control rod reactivity inserted. Fig. 2. Power evolution during a control rod calibration test.
 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568 563

Fig. 3. Example of a control rod worth curve.

are entirely unrelated to power production in the reactor. The

result being very noisy reactivity calculations. Filtering of the reac-
tor power signal would change the underlying dynamic character- Fig. 4. Comparison of a power transient for a step change in qðtÞ at t = 0.0 with and
istics of the signal, and result in erroneous reactivity calculations. without the Prompt-Jump Approximation (qðtÞ = 0.002).
To minimize the influence of noise on the reactivity calculated by
equations (6) and (7), the Prompt-Jump Approximation may be
measurements taken 2–3 s after the reactivity insertion (and sub-
applied to equation (6) to justify removal of dnðtÞ
dt
.
sequent no control rod movement).
Consider the normalized version of the point kinetics equations
Pi¼6
dnr ðtÞ ðqðtÞ  bÞ  nr ðtÞ i¼1 bi c ri ðtÞ 3. Digital Reactivity Computer (DRC) implementation
¼ þ ð8Þ
dt K K
It was determined that for ease of maintenance, the DRC would
dcri ðtÞ
¼ ki ðnr ðtÞ  cri ðtÞÞ ð9Þ be developed in a commercially-popular, user-friendly software
dt package – LabView software (National Instruments) – and
With deployed in National Instruments hardware. The graphics-base
programming capability of the LabView software package made
nðtÞ
nr ðtÞ ¼ development and testing straight-forward. The software was
nð0Þ designed so that other university research reactors may easily
adapt and use for their control rod calibrations. The motivation
cðtÞ for using National Instruments hardware and software are 1) Large
cri ðtÞ ¼
cð0Þ domestic customer base – solutions shared among customers 2) In-
If one assumes that immediately after the time of a step reactiv- house knowledge – Penn State Reactor staff can upgrade/alter as
ity insertion (positive or negative) that necessary 3) Software and hardware relatively inexpensive for aca-
demic end-users 4) LabView provides for Data Acquisition/reactiv-
dnr ðtÞ ity calculation/curve fitting/data archiving all in one program 5)
¼0
dt LabView enables stand-alone deployment, where an executable
then Eq. (8) reduces to may be produced and run on PC’s without LabView. The
Pi¼6 LabView-based reactivity computer calculates step changes in con-
i¼1 bi c ri ðtÞ trol rod reactivity based on the prompt-jump approximation,
nr ðtÞ ¼ ð10Þ
b  qðtÞ described previously, and integration of 6 delayed neutron group
precursor differential equations. For the control rods that are not
and Eq. (9) is left unchanged. Eq. (10) can then be rearranged to pro-
being calibrated during the test (which necessarily need to offset
vide nr ðtÞ as the input to the calculation and qðtÞ as the output i.e.,
the positive reactivity introduced by changes in calibrated rod’s
Pi¼6 position), the LabView software provides suggested positions
i¼1 bi cri ðtÞ
qðtÞ ¼ b  ð11Þ based on previous control rod worth curves. The following pro-
nr ðtÞ
vides a description of the software and hardware along with exam-
The Prompt-Jump Approximation is limited in validity to step ple results from control rod calibration tests. Fig. 5 is the DRC flow
 where b
changes in qðtÞ < 0.25b,  is the fraction of delayed neutrons chart providing the major functional aspects (highlights) of the
emitted per fission. Fig. 4 presents a comparison of a power tran- LabView code.
sient for a step change in qðtÞ at t = 0.0 with and without the Referring to Fig. 5, the rod position estimation portion of the
Prompt-Jump Approximation. DRC (4) uses curve fit equations for Rod Worth = f(Rod position)
The Prompt-Jump Approximation essentially results in an instan- and Rod Position = f(Rod Worth). The control rod worth data
taneous step to a new initial condition, with an eventual exponen- import (1) imports existing control rod worth calibration data from
tial rise due to the contribution of the delayed neutron precursors, previous year’s test, and uses LabView curve fitting function blocks
whose dynamics are dictated by ki , and are considerably slower (2) to provide Rod Worth = f(Rod position) and Rod Position = f
than the changes in the prompt neutron dynamics. After 200 ms, (Rod Worth). It is assumed that the rod worth data has not changed
the changes in power dictated by Eqs. (8) and (10) are virtually significantly (this test is performed annually) and is acceptable for
 Thus, use of
identical, again, for reactivity insertions qðtÞ < 0.25b. suggesting compensating positions of the other controls rods used
Eq. (11) to calculate qðtÞ is justified for relatively small qðtÞ and to offset the calibrated control rod’s positive reactivity insertion.
564 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568

Fig. 5. DRC flow chart providing the major functional aspects of the LabView code.

Fig. 6. LabView code corresponding to rod worth data import (Fig. 5 – (1)) and Rod Worth = f(Rod position) and Rod Position = f(Rod Worth) curve fitting (Fig. 5 – (2)) portion
of the DRC.

These operations are performed only during DRC initialization,  Used to provide operator with position placement guidance for
with the resulting curve fit equations used during reactivity step 3 CR used to offset calibrated rod reactivity insertion
data acquisition (via While Loop). Fig. 6 provides the LabView code  Only calculated once during initialization of DRC.
corresponding to this portion of the DRC.
In summary, the major aspects of this portion of the DRC are: The differential control rod reactivity calculation (correspond-
ing to each rod step out during the test) is provided in Fig. 7.
 ‘‘Forward” RW Curve Fit (rho = f(z)) This calculation is implemented continuously via a LabView for-
 ‘‘Backward” RW Curve Fit (z = f(rho)) mula node in a While Loop (3), and uses Eqs. (11) and (9). Volt-
 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568 565

Fig. 7. LabView implementation of power signal data acquisition and Eqs. (11) and (9) in a formula node (Fig. 5 – (3)).

Fig. 8. LabView code used to provide the reactor operator with control rod position placement guidance (Fig. 5 – (4)).

age corresponding to the power level is sampled at 10 ms inter- 4) Use ‘‘Backward” RW Curve to give new, suggested 3 CR esti-
vals, and provided to the function node where the normalized mated position
precursor differential equations are numerically integrated using
Euler integration. Additionally, Integral CRW is also calculated For calibrated CR:
continuously and is the summation of the current calculated dif-
ferential CR worth and the previous iteration’s integral CR 1) Operator identifies desired Delta-Rho insertion from cali-
worth. The running calculation of the integral control rod worth brated CR after DRC has converged on a differential reactiv-
is used by the control rod position estimator, which is used to ity value.
provide the reactor operator with position placement guidance
for the 3 CR used to offset for the calibrated rod reactivity
inserted.
Also in the While loop is the rod position estimator, used to pro-
vide the reactor operator with position placement guidance for the
3 CR used to offset for the calibrated rod reactivity inserted. This is
shown in Fig. 8. A LabView case structure is used to determine for
which control rods placement guidance is provided.
The process for use of this information by the reactor operator is
as follows:
For 3 CR used to offset calibrated CR reactivity inserted:

1) Operator inputs 3 CR leveled initial position (only once) –

‘‘Forward” RW Curve gives total worth of 3 CR.
2) Integral RW of calibrated CR subtracted from 1).
3) Subtract an additional amount to ensure sub-criticality upon
Fig. 9. LabView implementation of the integral control rod worth calculation in a
insertion. formula node (Fig. 5 – (5)).
566 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568

Fig. 10. User interface of the DRC, which uses LabView built-in interface features and the tab feature, which allows multiple interface capabilities on a single screen.

Fig. 11. User interface of the DRC. Post-data acquisition curve fitting of rod reactivity data.

2) Gets added to previous rod pull’s integral CR worth. Figs. 10 and 11 provide the user interface of the DRC, which uses
3) Uses ‘‘Backward” CR curve to estimate where to position cal- LabView built-in interface features and the tab feature, which
ibrated CR for next reactivity measurement. allows multiple interface capabilities on a single screen.

Upon acquiring all differential rod worth values for the control
rod being calibrated, the DRC exits the while loop and calculates 4. Results
values corresponding to the complete integral control rod worth
for the rod positions assigned during the test (rho = f(z)). This cal- The Process for determining the complete rod worth (rod reac-
culation is also implemented in a LabView formula node, shown in tivity) curve for use by reactor operators and the reactor control
Fig. 9, and is calculated once prior to executing the equation curve console is as follows:
fit in LabView. Prior to reactivity measurements
 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568 567

Fig. 14. Comparison between the unfiltered and filtered reactivity calculation.

Measurement sequence

1) Shim-out calibrated control rod

2) Wait for DRC to converge on constant delta-rho
Fig. 12. Compensated Ion Chamber used for DRC rod reactivity measurements. 3) With power < 1 kW, shim in-other 3 rods (in manual) and
wait for stable negative reactivity
4) Repeat 1) – 3) until calibrated control rod fully withdrawn

 Position Spare Compensated Ion Chamber (CIC) next to core An example of a single reactivity step measurement using the
(Fig. 12). DRC is provided in Fig. 13. To expedite the process of measuring
 Power-up CIC power supplies and connect pico-ammeter – this an entire integral control rod worth curve, each measurement is
instrument also converts CIC output current to a voltage, which started at a negative core reactivity, the calibrated control rod is
is acquired by the DRC. removed, and a corresponding positive reactivity inserted. In stark
 Reactor operator goes critical with 3 rods not being calibrated – contrast to use of the In-Hour method, a reactivity determination is
calibrated control rod fully inserted into the core. made within 5 s. The output of the reactivity computer has been
 Instrumentation linearity checks are performed between 100 W extensively benchmarked with the In-Hour method.
and 1 kW. As mentioned previously, the DRC uses Eq. (11), which uses a
 Start test at 100 W, put 3 rods (not being calibrated) into man- normalized version of the reactor power measurement. In order
ual mode of operation. to preserve the dynamics associated with the power change, the

Fig. 13. Example of a single reactivity step measurement using the LabView-based DRC.
568 J.A. Turso / Annals of Nuclear Energy 114 (2018) 561–568

Fig. 15. Power changes and corresponding reactivity calculations during a typical rod worth measurement sequence.

measurement (voltage from a pico-ammeter) is unfiltered, which, hours compared to 30 min with the DRC for a complete control
upon review of Eq. (11), would necessarily result in noisy reactivity rod worth curve. All of the data acquisition, reactivity calculations,
measurements. The reactivity is filtered after the calculation of and integral control rod worth curve fits are performed in LabView,
Eq. (10) in order to preserve the dynamic characteristics of the sig- which additionally provides for development of the graphical user
nal, yet provide a low-variability estimate of the reactivity. Fig. 14 interface. Results demonstrate the LabView DRC developed pro-
provided a comparison between the unfiltered and filtered reactiv- vides a platform for these measurements that calculates accurate
ity calculation. The filtered version of the reactivity is subsequently control rod reactivity measurements in significantly less time than
used to calculate the integral rod worth using the algorithm shown the historically accepted In-Hour method. The DRC is coded in a
in Fig. 9. commercially-popular, user-friendly software package, National
Fig. 15 illustrates power changes and corresponding reactivity Instruments LabView software, and is being deployed in National
calculations during a typical measurement sequence described Instruments hardware. The motivation for using National
above. This sequence is continued until the calibrated control rod Instruments hardware and software are 1) Large customer base –
is fully withdrawn. The same set of measurements would typically solutions shared among customers 2) In-house knowledge – Penn
take 2–4 h with the In-Hour (doubling time) method. Again, each State Reactor staff can upgrade/alter as necessary 3) Software and
filtered reactivity measurement with the DRC takes approximately hardware relatively inexpensive for academic end-users 4)
5 s. LabView provides for Data Acquisition/reactivity calculation/curve
fitting/data archiving all in one program 5) LabView enables stand-
5. Conclusion alone deployment, where an executable may be produced and run
on PC’s without LabView. Additionally, algorithms have been
A digital control rod reactivity computer (DRC) has been devel- developed so that the software provides suggested positions con-
oped using commercially available software (National Instruments trol rod positions based on previous control rod worth curves, fur-
LabView) for use to determine control rod worth curves for the ther expediting control rod worth measurements.
Penn State University TRIGA research reactor. For low-worth reac-
tivity steps, use of the prompt-jump approximation to the point References
kinetics equations enabled rapid calculation of rod reactivity inser-
Glasstone, Edlund, 1952. The Elements of Reactor Theory Chapter 10. D. Van
tions and minimized measurement noise effects. Comparisons Nostrand Co..
between the full implementation of the point kinetics equations Glasstone, Sesonske, 1967. Nuclear Reactor Engineering Chapter 5. D. Van Nostrand
and the prompt-jump approximation for reactor transient Co..
 typical of Duderstadt, Hamilton, 1976. Nuclear Reactor Physics. John Wiley & Sons Inc..
response due to reactivity insertion steps qðtÞ < 0.25b, Chapter 6.
those experienced during control rod worth tests, show good LaMarsh, J.R., Baratta, A., 2017. Introduction to Nuclear Engineering 4th ed. Chapter
agreement, and justify use of the approximation in the DRC. 7. Addison-Wesely.
Toppel, B.J., 1959. Sources of error in reactivity determination by means of
In contrast, use of the In-Hour method (an historically-effective asymptotic period measurements. Nucl. Sci. Eng., 88–93
way of determining control rod reactivity) would take several