Transactions, SMiRT-26

Berlin/Potsdam, Germany, July 10-15, 2022

Division VIII

CONDITION MONITORING OF NUCLEAR EQUIPMENT-PIPING USING

DEEP LEARNING
Harleen Kaur Sandhu1 , Saran Srikanth Bodda2 , Serena Sauers3 , Abhinav Gupta4
1 Postdoctoral Research Scholar, CCEE, North Carolina State University, USA (hksandhu@ncsu.edu)
2 Research Faculty, CCEE, NCSU, USA
3 Graduate Student, CCEE, NCSU, USA
4 Director, Center for Nuclear Energy Facilities and Structures, NCSU, USA

ABSTRACT

With the resurgence of nuclear energy due to the ever-increasing demand for electricity and carbon free
power generation, ensuring safe operations at nuclear power plants (NNPs) is important. Nuclear safety
systems can undergo vibrations due to various normal operating loads such as pump operations,
flow-induced, water hammer, acoustical resonance, etc. Over the course of time, safety systems such as
piping-equipment systems experience degradation due to flow-accelerated erosion and corrosion.
Undetected degradation at hot-spot locations can be subjected to a build up of cyclic fatigue due to
operational vibrations and thermal cycles. To avoid fatigue and subsequent cracking, a condition
monitoring framework is required for the detecting degradation. This paper demonstrates the use of sensor
data, feature extraction and Artificial Neural Networks (ANNs) in designing a condition monitoring
framework for a nuclear piping-equipment system subjected to pump-induced vibrations during normal
operations. The proposed framework also incorporates uncertainty in the degradation severity
classifications, such as minor, moderate or severe. For the application case study, a piping system from
Experimental Breeder Reactor II nuclear reactor is selected. It is observed that a high prediction accuracy
is achieved when detecting degraded locations and their severity classification. Additionally, a strategy
based-assessment methodology is illustrated via an application example to provide recommendations for
“safe” pump speeds and allowable number of cycles, as per ASME design criteria, to avoid fatigue in
equipment and connected piping systems due to pump-induced vibrations.

INTRODUCTION

Operations and maintenance costs comprise about 60 to 70% of the overall generating cost in legacy light
water nuclear power plants (Wacker, 2007). The current practice involves performing maintenance of safety
systems through performance trending and maintenance practices, and of passive components through aging
management programs (U.S.NRC, 2010). In many plants, non-destructive testing (NDT) techniques like
ultrasound or thermal imaging are used to collect the data on structures, systems and components (SSCs)
degradation only during outage. However, this data collection is not comprehensive as sometimes it is not
possible to conduct conventional NDT techniques for scanning the entire system due to time constraints,
shutdown of the plant and subsequent loss of revenue. Due to these constrains, degraded locations may
pass undetected, and may cause multiple nuclear SSC failures (Wu, 1989). Therefore, it is important to
identify and retrofit any degraded locations to address the operational functionality of the nuclear safety
systems. An AI guided condition monitoring based on the sensor data collected from the system in real-
time can be powerful to determine the degraded locations and their severity. The diagnosis of safety systems
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
such as piping-equipment systems using a condition monitoring framework can be beneficial if current NDT
techniques can be first implemented at detected degraded locations. It may also help to reduce nuclear power
plant (NPP) outage time periods.
Currently, the nuclear industry is encouraging significant research (Lee and Kim, 2021; Lin et al.,
2022, 2021b; Wood et al., 2017) towards developing autonomous control systems for NPPs as well as
advanced nuclear reactors. The primary goal of an autonomous control system is to provide
recommendations to the operator during normal operations as well as beyond design basis events by
developing a digital twin of the nuclear reactor using artificial intelligence techniques (Lin et al., 2021a;
NAMAC, 2018). A condition monitoring framework for nuclear piping-safety systems is an important
component of any autonomous management and control system for nuclear reactors. One aspect of
developing a nearly-autonomous control system is to provide guidance on the constraints on operating the
pump at certain speeds that may cause relatively greater fatigue in the piping-equipment system when
compared to operating the pump at other speeds. Thus, a condition monitoring framework with the
enhanced ability to provide recommendations on pump operational speeds can prove to be an important
aspect of autonomous nuclear reactor control systems. Furthermore, comparing the stresses developed in
the piping-equipment systems due to operational vibration loads against the design fatigue curves criteria
as mentioned in the ASME BPVC II (American Society of Mechanical Engineers Boiler and Pressure
Vessel Committee, 2015) can provide strategic assessments, such as safe pump speeds and allowable
number of cycles, to the maintenance operators.
Sandhu (2021) highlights the limitations of extending existing health monitoring studies (Alamdari
et al., 2017; Erazo et al., 2019; Martini et al., 2015; Rezaei and Taheri, 2010; Rosafalco et al., 2020; Zhang
and Sun, 2021) conducted for civil structures such as buildings and bridges to piping-equipment systems.
These previously proposed methodologies, though powerful for their respective applications, cannot be
applied for developing a condition monitoring framework for nuclear piping-equipment systems because
of the difference in the acquired sensor response and dynamic characteristics. It seems like a vector of
damage-sensitive quantities might be a better degradation-sensitive quantity for the nuclear piping system’s
diagnostic application, to detect locations with their corresponding level of degradation including a minor,
moderate or severely degraded condition. Additionally, the distributed systems of nuclear equipment and
piping can generate large amounts of sensor data, making data interpretation the biggest challenge for a
condition monitoring framework (Saran et al., 2019). The use of artificial neural network (ANNs) and
convolutional neural networks (CNNs) as powerful deep learning algorithms is proposed for data handling
and processing. Thus, eliminating the need to formulate degradation indices based on structural behavior
assumptions. Raw sensor data can be processed and fed into the deep learning algorithms to predict degraded
locations and assess the structural life of nuclear piping-equipment systems.
This research focuses on three important aspects: (i) Extracting a novel vector of degradation
sensitive quantities from the sensor data to detect even minor degradation in nuclear piping-equipment
systems. (ii) Detecting degraded locations as well as their degradation severity by designing a deep
learning neural network and a sensor placement strategy (iii) Proposing a strategy based-assessment
methodology for recommending “safe” pump operational speeds, to avoid high-cycle fatigue build up in
the nuclear piping systems, which can be beneficial to the maintenance operators. Continuous condition
monitoring of such systems would result in lowering the maintenance costs along with extending the
operating lifetime for a nuclear power plant.

PROBLEM DESCRIPTION

The equipment and connected piping safety systems of nuclear facilities experience normal operational
vibrating loads caused by pump operations, flow-induced, or water hammer phenomenon (Jacimovic and
D’Agaro, 2019). The decision to operate the pumps at a certain speed depends on the thermal hydraulic
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
and power generation requirements of the NPP. Therefore, in the case of flow anomalies, different pumps
connected to the piping systems can operate at different speeds. The speed of pump operations can determine
the amount of vibrations and subsequent fatigue developed in the piping-equipment systems. Cyclic fatigue
for prolonged duration of time can cause cracks and leakage in the discontinuities of piping safety systems
(Gupta et al., 2021; Ju and Gupta, 2015; Ryu et al., 2016). Therefore, a condition monitoring framework
is designed by using a part of the piping-equipment system from EBRII (Davis et al., 1970; Lin et al.,
2021a; Sumner and Wei, 2012) coolant system called ‘Z-pipe’ system shown in Figure 1a. The Z-pipe
system is subjected to vibrational loads due to normal pump operations. Stress concentrations at detected
degraded sections of the Z-pipe system are determined and checked against the allowable ASME design
criteria (American Society of Mechanical Engineers Boiler and Pressure Vessel Committee, 2015) to provide
a scenario-based recommendation for safe pump operating speeds.

EBRII Z Piping-Equipment System

To develop the proposed condition monitoring methodology and demonstrate its capability for strategy-
based recommendations on avoiding fatigue build-up, the ‘Z-pipe’ system is selected from EBRII nuclear
reactor. The EBRII nuclear reactor is located at Idaho, USA and was operational from 1964 until 1994. The
Z-pipe system is used to carry hot sodium from the reactor core subassemblies into the intermediate heat
exchanger (IHX). It contains the auxiliary electromagnetic pump and is subjected to pump operational loads
due to pump-induced vibration as shown in Figure 1b. These steady-state vibrations can be harmonic with a
force amplitude and excitation frequency, as shown in Equation 1.

f (t) = F sin(ωf t) or F cos(ωf t) (1)

where, f (t) is the harmonic excitation load, F is the force amplitude and ωf is the excitation
frequency in radians/second. The excitation frequency ωf varies with the speed of pump operations, as
shown in Equation 2 (Gao et al., 2020; Qing et al., 2006).

nN P
ωf = 2π (2)
60
where, n is the frequency number, N is the number of blades/pump plungers/pistons in the pump
being considered and P is the rotation speed of the pump.
As a part of this proof-of-concept condition monitoring framework, the auxiliary pump is assumed
to operate between 620 rpm (rotations per minute) and 1000 rpm. Therefore, from Equation 2, the excitation
frequency range for pump-induced harmonic vibrations can be calculated as 10 − 17 Hz. A steady-state
harmonic analysis is carried out by subjecting the Z-pipe system to 70 harmonic excitations between the
range of excitation frequencies (10 − 17 Hz) with a 0.1 Hz frequency step and a unit amplitude of force.
Twelve sensors are assumed to be placed at discontinuities in the systems such as elbows and nozzles, as
well as long straight sections of pipe, as illustrated in Figure 1b.

Degradation and Uncertainty

Nuclear piping-equipment systems experience degradation in the form of pipe-wall thinning due to flow-
assisted erosion and corrosion (Wu, 1989). The loss of pipe-wall thickness can cause a subsequent reduction
in structural stiffness of the system. Any change in the structural stiffness causes a variation in the dynamic
characteristics and this is reflected in the acquired acceleration-time series sensor response. For developing
this framework, degradation is assumed to occur at one location at any given instant of time. However,
the amount of degradation severity at each location can vary depending on the amount of pipe-wall thinning
observed. This research classifies degradation severity in three distinct levels of minor, moderate and severe,
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
as a percentage loss of pipe-wall thickness due to erosion and corrosion phenomena. Due to uncertainty in
the amount of degradation severity at any given location, a uniform distribution with a lower bound and an
upper bound is assumed as [20%, 30%] for minor, [45%, 55%] for moderate, and [70%, 80%] for severe
degradation. Within these ranges, random severity values are generated using a Latin hypercube simulation
(LHS) and then utilized in the finite element model of the Z-pipe system for collecting simulated sensor
response from the degraded state of the system (Bodda et al., 2020a,b, 2019, 2020c, 2016; Sandhu, 2015).

(a) EBRII Primary Tank Layout (Sumner and Wei, 2012) (b) Z-pipe system

Figure 1. EBRII and Z pipe system

SIGNAL PROCESSING AND FEATURE EXTRACTION

The condition monitoring framework is first implemented by converting the acquired acceleration-time
series sensor response to it power spectral density (PSD) in the frequency domain. In this study, PSD is
used as a potential degradation-sensitive quantity of interest. Two different degradation-sensitive feature
extraction approaches are compared for their effectiveness in training the multilayer perceptron (MLP)
ANN. With the first simplistic approach, only one degradation-sensitive quantity is extracted from each
sensor’s response. The maximum value from the PSD curves, P SDmax , is extracted and saved in a
database repository to be used as the training feature for detecting degradation in the Z-pipe system. In the
second approach, feature extraction is implemented to obtain the maximum difference observed between
the non-degraded and degraded states of the Z-pipe system. This approach to extract a vector of four
degradation-sensitive quantities has shown potential benefits in Part I of this dissertation over using a single
degradation-sensitive quantity as the training feature for the AI algorithms. A representative vector for a
typical sensor response has been tabulated in Table 1.

Table 1: Feature Extraction: Four Quantities of Interest

QoIs Definition Sample Values

P SDmax Maximum PSD amplitude from degraded response 199.42
∆max P SD Maximum difference observed in significant PSD peaks 6.55%
ω∆ Frequency corresponding to ∆P SDmax 13.57 Hz
P SD0,∆ non-degraded PSD peak corresponding to ∆P SDmax 161.08
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
DEEP LEARNING AND PROPOSED CONDITION MONITORING FRAMEWORK

The equipment and connected piping systems at nuclear facilities can generate large amounts of sensor data,
making data interpretation the biggest challenge for a condition monitoring framework. The use of deep
learning algorithms such as the MLP ANN is proposed for data handling and processing. Thus, eliminating
the need to correlate different degradation-sensitive quantities and formulate degradation indices based on
structural behavior assumptions. Raw sensor data can be processed and fed into the deep learning algorithms
to predict degraded locations and assess the structural life of nuclear piping-equipment systems. The design
and architecture of an MLP ANN can affect its performance for condition monitoring applications. Hence,
Sandhu (2021) focuses on developing a generic design of the deep learning algorithm that can be applicable
to most nuclear piping-equipment systems. This part of the research focuses on applying the previously
proposed deep learning algorithm to piping-equipment systems subjected to vibrational operating loads.
Next, the proposed condition monitoring methodology is implemented on the Z-pipe system to
detect degraded locations as well as the corresponding degradation severity levels. Table 2 illustrates the
results obtained from testing the proposed framework for the various aforementioned feature extraction
techniques, sensor placement formulations and MLP ANN architectures.

Table 2: Accuracy of Proposed Condition Monitoring Framework

Model Predict Locations Predict Locations and Severity

12 Sensors 86% 74%
Training Feature: P SDmax
12 Sensors 99% 97%
Proposed Vector of Training Features
8 Sensors: Proposed Sensor Placement Strategy 98% 97%
Proposed Vector of Training Features

It is observed that using a vector of four degradation-sensitive quantities yields much higher
prediction accuracy when compared to using only one degradation-sensitive quantity of P SDmax .
Therefore, the deep learning algorithm is able to extract and learn much more beneficial information from
the proposed feature extraction technique than from previous methodologies built for structures such as
buildings, bridges, etc. This is true especially for predicting the degradation severity in addition to the
degraded locations where using only one quantity of interest resulted in a 86% prediction accuracy whereas
using the vector increased the prediction accuracy to 97%. For the sensor placement formulations, using 12
sensors yielded almost similar results than using only 8 at elbow locations of the system. The proposed
framework utilizing sensor data from only 8 sensors is able to predict degraded locations as well as the
degradation severity levels with 97% accuracy. This is considerable jump from the 74% accuracy obtained
by using P SDmax as the only degradation-sensitive quantity. These results provide confidence in the
proposed sensor placement formulation as well as the richness of degradation- sensitive data contained in
the extracted vector.

STRATEGY BASED ASSESSMENT FOR AVOIDING FATIGUE

Piping-equipment systems under repetitive or cyclic forms of loading can experience failure due to fatigue,
such as formation and propagation of cracks and leakages. Fatigue failures can occur due to cyclic loads
at significantly lower stresses than the yield stress of the material. At lower than yield stress values, cyclic
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
loading can cause microscopic cracks which over time and use can grow into macroscopic cracks. These
macroscopic cracks can result in structural failure of the material and component. The microscopic cracks
typically initiate at locations with structural discontinuities in the system such as nozzles, elbows, t-joints,
etc. for nuclear piping-equipment systems. Any defective welding in such systems can also be a potential
fatigue build-up hot spot location. Degradation in nuclear piping systems due to flow-assisted corrosion and
erosion can result in quicker transformation of a microscopic crack to a macroscopic structural defect. Thus,
a condition monitoring framework utilizing sensor data from the systems can act as an additional safety tool,
by detecting potential degraded locations well in advance of fatigue build-up and cracks.
The damage caused due to fatigue can be classified in two broad categories of high-cycle fatigue
and low-cycle fatigue. High-cycle fatigue is usually observed in the elastic regions of material behavior
whereas low-cycle fatigue is characterized by plastic behavior of materials. Vibrations in nuclear piping
systems, caused due to pump operations or thermal cycles, are characterized as high-cycle fatigue where the
amplitude of stress or excitation force is small and the number of cycles to reach fatigue failure are large.
As per the ASME operation and maintenance (O&M) (American Society of Mechanical Engineers, 2020)
design criteria for safety against fatigue failures, either the stresses developed in the piping system should
be monitored to be well below the material’s yield stress limits, or the cause of high-cycle vibrations must
be eliminated or reduced. This study proposes a novel approach that enhances the condition monitoring
framework by providing recommendations on eliminating or reducing the cause of high-cycle fatigue by
pump-induced vibrations.
The primary goal of this part of the research is to open a new pathway for integrating the fatigue-life
of piping-equipment systems subjected to vibrations during normal operations and the proposed condition
monitoring framework. The EBRII Z-pipe system is used in concordance with the proposed condition
monitoring framework to build an application example case study for the strategy based assessment with
recommendations for “safe” pump speeds. An example flowchart of the strategy based assessment is shown
in Figure 2.

Figure 2. Strategy Based Assessment Flowchart

The proposed condition monitoring framework is utilized to detect any degraded locations on the
Z-pipe system along with its degradation severity. Next, a transmissibility ratio plot is generated for the
degraded location subjected to various auxiliary pump operating speeds and corresponding frequencies of
excitation. From this transmissibility ratio plot, the range of frequencies with the highest transmissibility
ratio values are selected. Using high fidelity simulations, the bending stresses generated at the diagnosed
degraded location are extracted for the excitation frequency range with the highest transmissibility ratio
values.The next step involves capturing the maximum bending stress experienced at the degraded location
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
as well as the corresponding frequency of excitation due to pump operations. In this hypothetical example,
the maximum bending stress at Nozzle 1 of the Z-pipe system is found to be 22 ksi at 14.7 Hz of excitation
frequency. This excitation frequency is equivalent to an auxiliary pump operating speed of 882 RPM from
Equation 2.
The ASME BPVC II (American Society of Mechanical Engineers Boiler and Pressure Vessel
Committee, 2015) provides stress vs. number of cycles curve (S-N curve) which detail the allowable stress
values for various materials of boiler and pressure vessels (including the connecting piping systems) versus
the number of fatigue cycles before fatigue can cause cracks and leakages in the system. In this example,
the Z-pipe system is made up of Type 304 stainless steel schedule 40 pipe. The corresponding ASME S-N
Curve for austenitic steels is shown in Figure 3 where the y-axis represents the allowable stress value in
ksi, and the x-axis represents the number of cycles, N , before fatigue can cause failure in the system. In
this hypothetical example, it can be shown that a maximum number of 3 × 105 cycles can be obtained for
22 ksi allowable stress value from the ASME S-N Curve illustrated in Figure 3.

103
Value of allowable stress (ksi)

102

101

100 1
10 102 103 104 105 106 107 108
Number of Cycles, N

Figure 3. ASME S-N Curve for Type 304 Stainless Steel Pipe Material (American Society of Mechanical
Engineers Boiler and Pressure Vessel Committee, 2015)

The final step is to provide a strategy-based recommendation will be composed of two aspects.
One aspect relates to avoid a certain pump speed in RPM during normal operations. However, due to the
thermal hydraulic and power generation needs of a nuclear reactor, it is not always possible to avoid a certain
pump operating speed. Therefore, the second aspect of the proposed recommendation would detail the total
number of hours the pump can be allowed to operate at that speed, as permitted by ASME design criteria
for fatigue. For the example being considered, since the maximum bending stress at Nozzle 1 is developed
at a pump speed of 882 RPM, the first potential recommendation to the operator will be to avoid operating
the pump at 882 RPM. For the second potential recommendation, the allowable number of hours to operate
the pump at 882 RPM can be calculated as 5.6 hours from Equation 3.

Number of cycles from SN curve 3 × 105

Allowable hours, h = = = 5.6 hours (3)
Pump speed in RPM × 60 882 × 60
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
SUMMARY AND CONCLUSIONS

Critical vibrations caused by pressure pulsations in the fluid due to pump operations can cause cracks and
leakages in nuclear piping systems. Generally, these vibrations result in high-cyclic fatigue failure which is
difficult to detect with current NDT techniques and scanning processes prevalent in the nuclear industry. This
study is aimed at developing an AI based condition monitoring framework that can detect degraded locations
along with their degradation severity level for a nuclear piping-equipment system. The key conclusions of
the study are summarized as follows:

• It is shown that using a vector of degradation-sensitive quantities is better than using a single
degradation-sensitive quantity, since the vector can capture essential features from all modes of
vibration in piping systems. A loss of response from lower order modes of vibration can occur if a
single quantity is used for creating the ANN training database.

• A deep learning algorithm using an MLP ANN is designed along with a sensor placement strategy to
detect degradation due to flow-assisted erosion and corrosion with 98% accuracy.

• The importance of detecting degradation severity along with the degraded location is demonstrated
since different locations can experience varying intensities of degradation severity. The proposed
methodology achieved a 97% accuracy in predicting the degraded locations and classifying the
corresponding level of severity as minor, moderate and major. Uncertainty in these classifications
levels is also considered by assuming a uniform distribution with a lower and upper bound for each
of the levels.

• An enhancement to the proposed condition monitoring framework is demonstrated by including a

strategy based assessment to avoid high-cyclic fatigue at hot spot degraded locations of the piping
system. An application example is presented for the Z-pipe system which compares stresses at
degraded locations to the ASME design criteria for fatigue. Two potential recommendations are
specified on safe pump operational speeds as well as the allowable number of cycles or hours the
pump can be operated at certain speeds, before fatigue can cause cracking and leakage failure in the
piping system.

References

Alamdari, M. M., Rakotoarivelo, T. and Khoa, N. L. D. (2017). “A spectral-based clustering for structural
health monitoring of the Sydney Harbour Bridge.” Mechanical Systems and Signal Processing, 87,
pp. 384–400, ISSN 08883270, doi:10.1016/j.ymssp.2016.10.033.

American Society of Mechanical Engineers (2020). “ASME Operation and Maintenance of Nuclear Power
Plants.” Technical report.

American Society of Mechanical Engineers Boiler and Pressure Vessel Committee (2015). “ASME Boiler
and Pressure Vessel Code.” Technical Report Section II Materials, Part D.

Bodda, S. S., Gupta, A. and Dinh, N. (2020a). “Enhancement of risk informed validation framework for
external hazard scenario.” Reliability Engineering & System Safety, 204, p. 107140.

Bodda, S. S., Gupta, A. and Dinh, N. (2020b). “Risk informed validation framework for external flooding
scenario.” Nuclear Engineering and Design, 356, p. 110377, ISSN 0029-5493, doi:https://doi.org/10.
1016/j.nucengdes.2019.110377.
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
Bodda, S. S., Gupta, A., Ju, B. and Kwon, M. (2019). “Multi-hazard fragility assessment of a concrete
floodwall.” Reliability Engineering and Resilience, 1(2), pp. 46–66.

Bodda, S. S., Gupta, A., Ju, B. S. and Jung, W. (2020c). “Fragility of a Weir Structure due to Scouring.”
Computational Engineering and Physical Modeling, 3(1), pp. 1–15.

Bodda, S. S., Sandhu, H. K. and Gupta, A. (2016). “Fragility of a Flood Defense Structure Subjected to
Multi-Hazard Scenario.” In International Conference on Nuclear Engineering, volume 50046, p.
V004T14A015, American Society of Mechanical Engineers.

Davis, J., Deegan, G., Leman, J. and Perry, W. (1970). “Operating Experience with Sodium Pumps at EBR-
II.” Technical Report ANL/EBR-027, Argonne National Laboratory, Argonne, Illinois.

Erazo, K., Sen, D., Nagarajaiah, S. and Sun, L. (2019). “Vibration-based structural health monitoring
under changing environmental conditions using Kalman filtering.” Mechanical Systems and Signal
Processing, 117, pp. 1–15, ISSN 08883270, doi:10.1016/j.ymssp.2018.07.041.

Gao, P., Qu, H., Zhang, Y., Yu, T. and Zhai, J. (2020). “Experimental and Numerical Vibration Analysis
of Hydraulic Pipeline System under Multiexcitations.” Shock and Vibration, 2020, pp. 1–13, ISSN
1070-9622, 1875-9203, doi:10.1155/2020/3598374.

Gupta, A., Dubey, A. and Cho, S.-G. (2021). “Seismic Fragility of Piping Nozzles in Nuclear Power Plants:
A Case for Updating the Current State-of-Practice.” Journal of Pressure Vessel Technology, 143(6),
doi:10.1115/1.4051136.

Jacimovic, N. and D’Agaro, F. (2019). “On Piping Vibration Screening Criteria.” Journal of Pressure Vessel
Technology, 142, p. 6.

Ju, B. S. and Gupta, A. (2015). “Seismic fragility of threaded Tee-joint connections in piping systems.”
International Journal of Pressure Vessels and Piping, 132-133, pp. 106–118, ISSN 03080161, doi:
10.1016/j.ijpvp.2015.06.001.

Lee, D. and Kim, J. (2021). “Autonomous Emergency Operation of Nuclear Power Plant Using Deep
Reinforcement Learning.” pp. 522–531, Springer International Publishing, Cham, ISBN 978-3-030-
80624-8.

Lin, L., Athe, P., Rouxelin, P., Avramova, M., Gupta, A., Youngblood, R., Lane, J. and Dinh, N. (2021a).
“Development and assessment of a nearly autonomous management and control system for advanced
reactors.” Annals of Nuclear Energy, 150, p. 107861, ISSN 03064549, doi:10.1016/j.anucene.2020.
107861.

Lin, L., Athe, P., Rouxelin, P., Avramova, M., Gupta, A., Youngblood, R., Lane, J. and Dinh, N. (2022).
“Digital-twin-based improvements to diagnosis, prognosis, strategy assessment, and discrepancy
checking in a nearly autonomous management and control system.” Annals of Nuclear Energy, 166,
p. 108715, ISSN 03064549, doi:10.1016/j.anucene.2021.108715.

Lin, L., Bao, H. and Dinh, N. (2021b). “Uncertainty quantification and software risk analysis for digital
twins in the nearly autonomous management and control systems: A review.” Annals of Nuclear
Energy, 160, p. 108362, ISSN 03064549, doi:10.1016/j.anucene.2021.108362.

Martini, A., Troncossi, M. and Rivola, A. (2015). “Automatic Leak Detection in Buried Plastic Pipes of
Water Supply Networks by Means of Vibration Measurements.” Shock and Vibration, 2015, pp. 1–13,
ISSN 1070-9622, 1875-9203, doi:10.1155/2015/165304.
 26th International Conference on Structural Mechanics in Reactor Technology
Berlin/Potsdam, Germany, July 10-15, 2022
Division VIII
NAMAC (2018). “Development of a Nearly Autonomous Management and Control (NAMAC) System for
Advanced Reactors.”

Qing, M., Jinghui, Z., Yushan, L., Haijun, W. and Quan, D. (2006). “Experimental studies of orifice-induced
wall pressure fluctuations and pipe vibration.” International Journal of Pressure Vessels and Piping,
83(7), pp. 505–511, ISSN 03080161, doi:10.1016/j.ijpvp.2006.03.010.

Rezaei, D. and Taheri, F. (2010). “Health monitoring of pipeline girth weld using empirical mode
decomposition.” Smart Materials and Structures, 19(5), p. 055016, ISSN 0964-1726, 1361-665X,
doi:10.1088/0964-1726/19/5/055016.

Rosafalco, L., Manzoni, A., Mariani, S. and Corigliano, A. (2020). “Fully convolutional networks for
structural health monitoring through multivariate time series classification.” Advanced Modeling and
Simulation in Engineering Sciences, 7(1), p. 38, ISSN 2213-7467, doi:10.1186/s40323-020-00174-1.

Ryu, Y., Gupta, A., Jung, W. and Ju, B. (2016). “A Reconciliation of Experimental and Analytical Results
for Piping Systems.” International Journal of Steel Structures, p. 14.

Sandhu, H. K. (2015). Flooding Fragility of Concrete Gravity Dam-Foundation System. Master’s thesis,
North Carolina State University.

Sandhu, H. K. (2021). Artificial Intelligence Based Condition Monitoring of Nuclear Piping-Equipment

Systems. Ph.D. thesis, North Carolina State University.

Saran, B., Ankit, D., Daniel, V., Abhinav, G., Mohammed, F. and Brian, D. (2019). “Seismic Behavior of
Coupled Reactor Pressure Vessel and Reactor Coolant Loop.” IASMiRT.

Sumner, T. and Wei, T. (2012). “Benchmark Specifications and Data Requirements for EBR-II Shutdown
Heat Removal Tests SHRT-17 and SHRT-45R.” Technical Report ANL-ARC-226 Rev. 1, Nuclear
Engineering Division, Argonne National Laboratory.

U.S.NRC (2010). “Generic Aging Lessons Learned (GALL) Report - Final Report.” Technical Report
NUREG–1801, Rev 2.

Wacker, J. (2007). “FY06 Annual Report on the Progress and Path Forward for the NA-22 Funded Project.”
Technical Report PL06-AUT308-PD01: PNNL-16527.

Wood, R. T., Upadhyaya, B. R. and Floyd, D. C. (2017). “An autonomous control framework for advanced
reactors.” Nuclear Engineering and Technology, 49(5), pp. 896–904, ISSN 17385733, doi:10.1016/j.
net.2017.07.001.

Wu, P. C. (1989). “Erosion/Corrosion-Induced Pipe Wall Thinning in U.S. Nuclear Power Plants.” Technical
Report NUREG–1344, 6152848, doi:10.2172/6152848.

Zhang, Z. and Sun, C. (2021). “Structural damage identification via physics-guided machine learning: a
methodology integrating pattern recognition with finite element model updating.” Structural Health
Monitoring, 20(4), pp. 1675–1688, ISSN 1475-9217, 1741-3168, doi:10.1177/1475921720927488.