An Approach to Autonomous Control for Space Nuclear Power Systems

109
intelligent control capability of the functional layer. The decision layer provides
functionality to break down goals into objectives, establish a sequential task ordering based
on the plant/system state and known constraints, and assess the capability of the functional
layer to implement those commands. At lower granularity within the decision layer,
executive functions such as procedure enforcement are dominant while, at higher
granularity, planning functions such as goal determination and strategy development are
dominant.
There is an architectural approach for nearly autonomous control systems that have been
applied through simulated nuclear power applications (see Fig. 1). As part of research into
advanced multi-modular nuclear reactor concepts, such as the International Reactor
Innovative and Secure (IRIS) and the ALMR, a supervisory control system architecture was
devised (Wood et al., 2004). This approach provides a framework for autonomous control
while supporting a high-level interface with operations staff, who can act as plant
supervisors. The final authority for decisions and goal setting remains with the human, but
the control system assumes expanded responsibilities for normal control action, abnormal
event response, and system fault tolerance. The autonomous control framework allows
integration of controllers and diagnostics at the subsystem level with command and
decision modules at higher levels.
Fig. 1. Supervisory control architecture for multi-modular nuclear power plants
The autonomous control system architecture is hierarchical and recursive. Each node in the
hierarchy (except for the terminal nodes at the base) is a supervisory module. The

Nuclear Power – Control, Reliability and Human Factors


functionality is maintained. It is possible to blend the decision and functional layers for this
application domain because the planning regime for nuclear power system operation is
much more restricted than for robotic or spacecraft applications. For example, while there
are a multitude of paths that a robot may traverse as it navigates to its next site, the states
are allowed for an SNPS are much more constrained. Even in the event of transients or
faults, the control system will try to drive the plant back to a known safe state. This
compression of the dual layers into a truncated three-sided pyramid allows for a deeper
integration of control, diagnostics, and decision to provide the necessary capability to
respond to rapid events and to adapt to changing or degraded conditions.
The granularity dimension is retained with more complexity shown at the lower hierarchical
levels. Additionally, the information and command flow reflects granularity as well. At
lower granularity, volumes of data are present. As the granularity increases moving up the
hierarchy, the data are processed into system state and diagnostic/prognostic information
that are subsequently refined into status and indicator information. On the command side,
the transition from the top is demands to commands to control signals with the resolution of
the plant/system control growing increasingly more detailed.

An Approach to Autonomous Control for Space Nuclear Power Systems

111
As with the supervisory control architecture, the bottom two levels of the hierarchy are the
equivalent of an automated control system. The embedded functionality that enables a
reliable, fault-tolerant implementation is indicated as a base intelligence. It is expected that
there will be some decision capability associated with the control/surveillance/diagnostics
level of that baseline system. The higher levels of the hierarchy assume greater degrees of
decision capabilities.
Fig. 2. Hierarchical framework to support SNPS control system autonomy

response to prospective actions, remaining useful life of components, and incipient
operational events or failures. The communication functionality involves control and
measurement signals to and from the field devices, information and commands within the
control system, and status and demands between the SNPS control system and spacecraft
or ground control. The fault management and configuration management functionalities
are interrelated and depend on two principal design characteristics. These are the ability
of the designer to anticipate a full range of faults and the degree of autonomy enabled by
the control system design.
Finally, the distribution of functions throughout the hierarchy must be established based on
the degree of autonomy selected, technology readiness, reliability and fault management
considerations, software development practices and platform capabilities, and the physical
architecture of the SNPS control system hardware. Because an autonomous control system
has never been implemented for a nuclear reactor and because several functional
capabilities remain underdeveloped (as seen in the overview of the state of the art), there is
clearly a critical need for further development and demonstration of a suitable architectural
framework.
4. Application of model-based control to Space Nuclear Power Systems
Key functionality that is necessary to establish the basis for autonomous control has been
demonstrated through a simulated space reactor application under university research
sponsored by DOE. These capabilities related to control elements within the lower layers of
the functional hierarchy. Specifically, the research conducted at UT involved development
of a highly fault tolerant power controller for the SP-100 space power reactor design
(Upadhyaya et al., 2007; Na & Upadhyaya, 2007).
The SP-100 design provides for a fast spectrum, lithium-cooled fuel pin reactor coupled with
thermo-electric converters (TE) with the waste heat removed through a heat pipe
distribution system and space radiators. The TE generator output is rated at 112 kW, with a
nominal reactor thermal power 2000 kW.
A lumped parameter simulation of a representative SNPS was developed based
on physics models specific to the SP-100 reactor, which were derived in prior academic
work at the University of New Mexico (El-Genk & Seo, 1987). The reactor system modules

Kinetics
Model
Reactivity Feedback
Model
Control
Drum
Model
+
TE
Model
Radiator
Model
Hx
Model
Core
Thermal
Model
Neutron
Kinetics
Model
Control
Drum
Model
+
Hx
Model
TE
Model
Core
Thermal

N, and M present and future control moves (M ≤ N) are computed to minimize the
quadratic objective function. Out of the M control moves that are calculated, only the first
control action is implemented. The prediction feature of the controller has an anticipatory
effect, and is reflected in the current control action. These calculations are repeated in the
next time step by appending the next measurement to the database. The new
measurements compensate for the unmeasured disturbances and model inaccuracies, both
of which result in the measured system output being different from that predicted by the
model. The MPC requires the on-line solution of an optimization problem to compute
optimal control inputs over the time horizon. The MPC calculates a sequence of future
control signals by minimizing a multi-stage cost function defined over a prediction
horizon.
The performance index for deriving an optimal control input is represented by the quadratic
objective function given in Eq. (1).


22
11
11
ˆ
(|)() ( 1)
22
NM
jj
JQytjtwtj Rutj



, (1)
subject to constraints
min max

known. N and M are the prediction horizon and the control horizon, respectively. The
prediction horizon represents the limiting time for the output to follow the reference
sequence. In order to obtain control inputs, the predicted outputs are first calculated as a
function of past values of inputs and outputs. The constraint,
(1)0forut j j M

   ,
indicates that there is no variation in the control signal after a certain time interval M < N,
where M is the control horizon.
min
u
and
max
u
are the minimum and maximum values of
input, respectively, and
max
u

is a maximum allowable control perturbation per time
step.
The applicability and the effectiveness of the MPC approach were demonstrated through its
simulated performance for several operational scenarios, including under degraded or ill-
characterized conditions (Upadhyaya et al., 2007). The effectiveness of the MPC controller
for tracking the TE power output is illustrated in Figure 6. Figure 6a shows the TE converter
set point profile and the actual TE generator power. The corresponding reactivity changes
(drum angle variations) are shown in Figure 6b.

An Approach to Autonomous Control for Space Nuclear Power Systems


response (i.e., reactivity control) in terms of the drum angle

Nuclear Power – Control, Reliability and Human Factors

116
The MPC approach was shown to provide a fast response and robustness under changing
system conditions. Specifically, fault tolerance and reconfigurability features of the control
approach were demonstrated in response to sensor faults, drum actuator anomalies, and
changes in model parameters (Upadhyaya et al., 2007; Na & Upadhyaya, 2007).
Consequently, it is observed that several of the capabilities and characteristics that are
necessary to enable autonomous control are provided by the MPC approach.
5. Conclusion
The control system for an SNPS will be subject to unique challenges as compared to
terrestrial nuclear reactors, which employ varying degrees of human control and decision-
making for operations and benefit from periodic human interaction for maintenance. In
contrast, the SNPS control system must be able to provide continuous, remote, often
unattended operation for a mission lasting a decade or more with limited immediate human
interaction and no opportunity for hardware maintenance. In addition to the inaccessibility
and periods of unattended operation, the SNPS control system must accommodate severe
environments, system and equipment degradation or failure, design uncertainties, and rare
or unanticipated operational events during an extended mission life. As a result, the
capability to respond to rapid events and to adapt to changing or degraded conditions
without near-term human supervision is required to support mission goals. Autonomous
control can satisfy essential control objectives under significant uncertainties, disturbances,
and degradation without requiring any human intervention. Therefore, autonomous control
is necessary to ensure the successful application of an SNPS for deep space missions.
Key characteristics that are feasible through autonomous control include
 intelligence to confirm system performance and detect degraded or failed conditions,
 optimization to minimize stress on SNPS components and efficiently react to
operational events without compromising system integrity,

be established to enable control system autonomy. Finally, development and demonstration
of a suitable architectural framework is also needed.
6. Acknowledgments
Portions of the work reported in this chapter were performed under the sponsorship of
NASA’s Project Prometheus and directed by DOE/National Nuclear Security
Administration (NNSA) Office of Naval Reactors. Other reported work was sponsored by
DOE Office of Nuclear Energy. Opinions and conclusions drawn by the authors are not
necessarily endorsed by the sponsoring organizations.
7. References
Alami, R., et al. (1998). An Architecture for Autonomy, International Journal of Robotics
Research
, Vol. 17, No. 4, (April 1998), pp. 315–337
American Nuclear Society (1993).
Proceedings of the 1993 ANS Topical Meeting on Nuclear Plant
Instrumentation, Control and Human-Machine Interface Technologies
, ISBN 0-89448-185-
1, Oak Ridge, Tennessee, USA, April 1993
American Nuclear Society (1996).
Proceedings of the ANS Topical Meeting on Nuclear Plant
Instrumentation, Control and Human-Machine Interface Technologies (NPIC&HMIT 96)
,
Vols. 1 & 2, ISBN 0-89448-610-1, State College, Pennsylvania, USA, May 1996
American Nuclear Society (2000). Proceedings of the ANS Topical Meeting on Nuclear Plant
Instrumentation, Control and Human-Machine Interface Technologies (NPIC&HMIT
2000)
, ISBN 0-89448-644-6, Washington, District of Columbia, USA, November 2000
American Nuclear Society (2004).
Proceedings of the ANS Topical Meeting on Nuclear Plant
Instrumentation, Control and Human-Machine Interface Technologies (NPIC&HMIT
2004)

Control System Magazine
, (October 1996), pp. 78–84
El-Genk, M. S. & Seo, J. T. (1987). SP-100 System Modeling: SNPSAM Update,
Transactions of
the 4th Symposium on Space Nuclear Power Systems
, Albuquerque, New Mexico, USA,
January 1987, pp. 513-516
Gat, E. (1998). Three-Layer Architectures, In:
Artificial Intelligence and Mobile Robots: Case
Studies of Successful Robot Systems
, D. Kortenkamp et al. (Eds.), pp. 195–210, MIT
Press, Cambridge, Massachusetts, USA
Mishkin, A. G., et al. (1998). Experiences with Operation and Autonomy of the Mars
Pathfinder Microrover,
Proceedings of the 1998 IEEE Aerospace Conference, ISBN 0-
7803-4311-5, Aspen, Colorado, USA, March 1998, pp. 337–351
Na, M. G. & Upadhyaya, B. R. (2007). Development of a Reconfigurable Control for an SP-
100 Space Reactor,
Nuclear Engineering and Technology, Vol. 39, No. 1, (February
2007), pp. 63-74
Rayman, M. D., et al. (1999). Results from the Deep Space 1 Technology Validation Mission,
Proceedings of the 50
th
International Astronautical Congress, American Institute of
Aeronautics and Astronautics,
Acta Astronautica, Vol. 47, pp. 475–488
Passino, K. (1995). Intelligent Control for Autonomous Systems, IEEE Spectrum, (June 1995),
pp. 55–62
Upadhyaya, B., et al. (2007).
Autonomous Control of Space Reactor Systems, DE-FG07-

1. Introduction
Optical fiber sensors (OFS) have a number of intrinsic advantages that make them attractive
for nuclear power plant (NPP) applications, including absolute explosion safety, extremely
low mass, small size, immunity to electromagnetic interference, high-accuracy, self-
calibration, and operation in extremely harsh environments, and it is a well-known fact.
Civil nuclear industry essentially encompasses the complete nuclear fuel cycle and therefore
the range of possible fiber applications both for communications and sensing is very broad
(Berghmans & Decreton, 1994), (Korsah et al., 2006).
In order to expand OFS applications in nuclear engineering it was necessary to overcome a
bias that some scientists and engineers used to have at the initial stage of using an optical
fiber for communication, about "darkening" of a fiber and sharp growth of optical
attenuation under the conditions of ionizing radiation, i.e. availability of convincing proofs
of radiation hardness of optical fibers and OFS.
Safety and long-term metrological stability of OFS for NPP assumes:
- Radiation hardness of fiber optic sensors and cables;
- Absence of mechanical resonances of the gauge at frequencies up to 200 Hz;
- Immunity to electromagnetic effects in the range of frequencies 200 kHz and 18 – 20
MHz,
- High reliability of a sensitive element of the OFS ;
- Temperature-insensitive measurements of pressure in the working range of
temperatures;
- Self-calibration of the gauge without stopping the process of measurement.
These requirements are satisfied by modern OFS, especially intellectual optical fiber sensors
which can self-calibrate, i.e. control themselves at the level of changing their internal (own)
parameters depending on the calibrated value (Buymistriuc & Rogov, 2009).
No optical measurement electronics will survive in, or near, an operating nuclear reactor
core. Therefore, OFS light emission must be guided to the measurement electronics located
in a well-controlled, benign environment. Several different implementations can be
employed to accomplish this, each with their own advantages and weaknesses. Recently
single material hollow-core optical fibers (referred to as photonic crystal fibers) have become

major challenge, a key element to the sustainability of future nuclear power use in most
countries. A first priority is obviously ensuring safety during operation under normal and
faulty conditions. With this, besides contributing to guarantee operational safety, systems
reliably monitoring the repository environment over several decades of years, whenever
possible maintenance free and in unattended mode, can become a key element in achieving
confidence on repository performance as well as public and regulatory acceptance.
Application of fiber optic technologies for monitoring SNF offers distinct advantages
compared with conventional systems. Optical fibers not only withstand chemical corrosion
and high temperatures much better than conventional systems, but their immunity to
electromagnetic interference and their large bandwidths and data rates ensure high
reliability and superior performance.
Due to this optical fibers are the preferred alternative for both: sensing and signal
transmission in long-term monitoring of NPP and SNF applications.
2. Background
A NPP generally uses about 200 to 800 pressure and differential pressure sensors to measure
the process pressure, level, and flow in its primary and secondary systems. For example, fig.
1 shows a typical pressure sensing (pulse) line inside a nuclear reactor containment (Lin K.
& Holbert K., 2010).

Radiation-Hard and Intelligent Optical Fiber Sensors for Nuclear Power Plants

121 Fig. 1. Typical pressure sensing (instrument) line inside a nuclear reactor containment.
Instrument lines can encounter a number of problems that can influence the accuracy,
response time of a pressure sensing system and decrease safety of NPP in consequence of
mechanical resonances which appear on frequencies up 200 Hz , for example, fig. 2 shows
transfer functions of a pressure sensing system (Lin K. & Holbert K., 2010).


When sensitive element 1 of an OFS placed in harsh environment can be moved away from
optoelectronic transceiver 2, which is under comfortable conditions of an equipment room,
at the distance up to 3000 meters by means of an optical cable option which replaces
undesirable pulse tubes very effectively.
The design of the sensitive element of pressure OFS working with the measured
environment of a nuclear reactor without application of pulse tubes is such that its resonant
frequency lies in the range of frequencies above 80 kHz, i.e. inadmissible resonances in
nuclear reactors at frequencies below 200 Hz are structurally excluded. In fact, the resonant
frequency of longitudinal vibration of optical fiber Fabry-Perot interferometer (FFPI) in the
form of a quartz glass core is defined as

1
4.91 E
f
L

 (1)
where Е - Young’ modulus of elasticity of a glass core, Pa
ρ – glass core density, kg/m
3

E/ρ - own rigidity, in particular for silica glass, 45х10
5
m.
Thus the sensor mechanical resonant frequency is defined by its length L = 0.001 … 0.1 m
and lies in the range f
1
= 10,4092 / L [kHz] = 104,092 … 10409,2 kHz.
Frequency resonant characteristic of a typical pressure OFS based on FFPI indicates Fig. 4.


Use of silica as a material for fiber optic sensors and measuring communication lines is an
effective solution both in terms of mechanical properties and radiation hardness of silica
fibers which are reached by modern manufacturers, for example, a method of entering and
retention of hydrogen in an optical fiber.
Another important factor of applying silica optical fibers under radiation conditions, in
particular for OFS based on silica optical fibers, is their low radiation induced losses in the
range of wavelengths between1150 nm and 1350 nm as is shown in fig. 7 (Fiedler et. al.,
2005).
Radiation hardness of OFS equaled earlier to a general dose of irradiation of about 1.2 MGy
with γ-radiation and 2.6•10
16
neutrons/cm
2
with a neutron fluence (Berghmans F. &
Decréton M., Ed.,1994) but now reaches doses of gamma radiation up to 23 MGy and
neutron flux 52 •10
16
neutrons/cm
2
(Fiedler et. al., 2005).
Photonic crystal fibers (PCFs) were also recently submitted to a number of nuclear
environments applications. In hollow core PCFs the light is essentially guided in air, which
may significantly decrease the radiation response of such waveguides compared to
conventional optical fibers. The structure used by us hole core PCF at 1000X and 10000X
imagnifying in a microscope is shown on fig. 8.
The permanent radiation induced attenuation (RIA) levels after radiation of PCF were found
to be very low. This was confirmed with hollow core PCF showing at least about 30 to 100
times lower RIA than the best present conventional optical fibers at 1550 nm with theoretical

Radiation-Hard and Intelligent Optical Fiber Sensors for Nuclear Power Plants

The principal innovation of this approach is to combine optical and fiber optical
measurement components in a form suitable for deployment in a nuclear reactor core.
The main needs for in-pile concern the assessment of creep and growth of cladding
materials, or nuclear fuel rod behaviour, which require elongation measurements or
diameter measurements of cylindrical samples. Among others adaptations to the nuclear
environment, these OFS will need a radiation resistant fixing.
Fig. 9. Hollow-Core Light Guide Concept.
As the light reaches lower radiation environments, several different optical transmission
technologies become possible:
hollow core PCF or conventional radiation-hard optical fibers.
For total doses up to about 10
4
Gy, pure silica core, fluorosilica clad, multimode optical
fibers are suitable light guides.
4. Enhanced reliability of OFS in harsh environments
А standard version of FFPI has no face reflecting coverings and works based on natural
Fresnel reflections in the amount about 4%. For changing the sensitivity and dynamic range
of the pressure OFS , we used TiO
2
reflecting coverings.
The fibers are placed in magnetron sputtering (MS) machine and coated with TiO
2
by
vacuum deposition. The reason for using TiO
2
is that it has high refractive index (~2.4, vs.
1.4 for the silica fiber) over visible and infrared spectral ranges and strong bonding on glass
Fig. 10. Microscopic view of optical fiber tip with TiO
2
magnetron sputtering
With processing by TiCl
4
steams the reaction on the surface proceeds as follows:
(≡Si-OH) + TiCl
4
→ (≡Si-O-)
2
TiCl
2
+ HCl (2)
With processing by water steams the reaction on the surface proceeds as follows:
(≡Si-O-)
2
TiCl
2
+ 2H
2
O → (≡Si-O-)
2
Ti(OH)
2
+ 2HCl (3)
At this stage of the ML process we obtain a hydroxylated surface again but now OH-
groups are linked not with the atoms of silicon of the initial matrix but with the atoms
which are part of the imparted functional groups. The hydroxylated surface is processed Fig. 11. Microscopic view of fiber tip with TiO
2
molecular layered nano-coating
An important advantage of such pressure OFS is its enhanced reliability determined by a
unitary structure of the sensor and extremely high adhesion of molecular coatings to silica
optical fibers. OFS fabricated with new ML technology possess the greatest reliability (on
distribution of Weibull) than usual MS method as shown on fig. 12.

Radiation-Hard and Intelligent Optical Fiber Sensors for Nuclear Power Plants

129 Fig. 12. Weibull distribution plots for basic fiber and fibers coated by using ML and MS
technologies
5. Principles and constructions of intelligent OFS
Information redundancy of optical fiber sensors, as well as possibility to their
programmable tuning in combination with a minimum structural redundancy allow to
develop the so-called intelligent sensors with a function of metrological self-checking. The
function of metrological self-checking of optical fiber sensors is provided with their
multimodality, i.e. with their similar dependence of an output signal on several variable
parameters, f.e., with their dependence on a variable pressure at a constant optical spectrum
of an input signal and, accordingly, on a readjusted optical spectrum of the signal at a
constant pressure. Construction of intelligent sensors of new generation assumes presence at
such sensors of structural (internal) and/or information (external) redundancy
(

By means of fast tuning of the spectrum of an optical source it is possible to make self-
calibration in the course of continuous work of the pressure gauge. Optical cables including
connectors, splices, and other components are tested by evaluating the optical losses
relationship along the cable.
OFS of physical quantity creates the multimodal output signal depending at least from two
parameters of the sensor, for example for OFS based on FFPI output signal I
s
depends on
change of length of optical resonator G and change of a wavelength of light λ:

0
0
4
1(;)
()
s
II Cos Gxt
t





 







On such principle has been realised the intelligent pressure OFS with function metrological
self-checking (FMSC) at long operation in extreme conditions

(Buymistriuc G. & Rogov
A.,2009).
Speed of tuning of a modern TSOF, for example the models “FFP-TF” from “Micron Optics,
Inc” or the acousto-optical tuneable filter models “AOTF” from “Fianium, Inc” is rather high
also the period of tuning time T on all set spectrum. For example in sequence T = t
3
- t
1
on
fig. 15, makes value of an order 0,1 … 0,4 microsecond that it is enough for the majority of
modes of measurement of pressure, deformation, vibration, temperature, level of etc.
controllable industrial equipment. Fig. 14. Intelligent pressure OFS
1 – FFPI; 2 – optical cable; 3 – optical coupler; 4 –light emitting diode; 5 –TSOF;
6 –photodiode; 7 – microcontroller
5.2 Algorithm
Procedure of self-checking of pressure OFS on the basis of FFPI consists of the following
consecutive steps.
Step 1. When OFS is manufactured, its calibration characteristic is measured:
I
с
= f(ΔG{P}) at λ

of calibration and corresponding correction factors are calculated.
Step 4. The measurement of pressure by means of OFS with using correction factors
obtained at step 3 is continued.
Thus, interference OFS is internally inherent necessary information and structural in
redundancy effectively to carry out FMSC.
5.3 Estimation of error
Error of measurement of the intelligent OFS, caused by instability of the central wavelength
of TSOF, i.e. shifted spectra TSOF and light emitting diode and casual displacement of
spectral characteristics of the filter us has been investigated by modeling and experimental
check.
Distortion
s’ of an output signal s, is described by following expression:

4
'1 1
2
c
o
I
ss Cos G
I








 







 



(8)

Radiation-Hard and Intelligent Optical Fiber Sensors for Nuclear Power Plants

133
A case non-centering (casual displacement) of a wavelengths of the optical filter and a
source of optical radiation in the course of long operation.
By means modeling and experiences is established that spectrum mismatching of the TSOF
with a light source poorly influences an error of measurements of OFS. Big shifted spectral
mismatch, above 3 nm, gives metrological error less then 0.02%.
Besides by tuning TSOF with feedback included on an output signal it is possible to provide
stability of a quiescent point (Q-point) interferometric OFS that is to realise the intelligent
gauge with self-correction.
Possible application of adaptive OFS is their use as additional measuring transducers of the
self-checking channel at construction of intelligent gauges of pressure with elastic elements
and gauges of level and other physical quantities for the purpose of substantial growth of
their actual interval between testing.
5.4 Intelligent acoustic emission OFS
Was investigated also methods of realization of intelligent OFS on other principles of
operation, in particular possibilities of intellectualizing of the acoustic emission OFS based

8- microcontroller