IEC TR 63468:2023

IEC TR 63468 ®
Edition 1.0 2023-05
TECHNICAL
REPORT
colour
inside
Nuclear facilities – Instrumentation and control, and electrical power systems –
Artificial Intelligence applications
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form
or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from
either IEC or IEC's member National Committee in the country of the requester. If you have any questions about IEC
copyright or have an enquiry about obtaining additional rights to this publication, please contact the address below or
your local IEC member National Committee for further information.

IEC Secretariat Tel.: +41 22 919 02 11
3, rue de Varembé info@iec.ch
CH-1211 Geneva 20 www.iec.ch
Switzerland
About the IEC
The International Electrotechnical Commission (IEC) is the leading global organization that prepares and publishes
International Standards for all electrical, electronic and related technologies.

About IEC publications
The technical content of IEC publications is kept under constant review by the IEC. Please make sure that you have the
latest edition, a corrigendum or an amendment might have been published.

IEC publications search - webstore.iec.ch/advsearchform IEC Products & Services Portal - products.iec.ch
The advanced search enables to find IEC publications by a Discover our powerful search engine and read freely all the
variety of criteria (reference number, text, technical publications previews. With a subscription you will always have
committee, …). It also gives information on projects, replaced access to up to date content tailored to your needs.
and withdrawn publications.
Electropedia - www.electropedia.org
IEC Just Published - webstore.iec.ch/justpublished
The world's leading online dictionary on electrotechnology,
Stay up to date on all new IEC publications. Just Published
containing more than 22 300 terminological entries in English
details all new publications released. Available online and once
and French, with equivalent terms in 19 additional languages.
a month by email.
Also known as the International Electrotechnical Vocabulary

(IEV) online.
IEC Customer Service Centre - webstore.iec.ch/csc

If you wish to give us your feedback on this publication or need
further assistance, please contact the Customer Service
Centre: sales@iec.ch.
IEC TR 63468 ®
Edition 1.0 2023-05
TECHNICAL
REPORT
colour
inside
Nuclear facilities – Instrumentation and control, and electrical power systems –

Artificial Intelligence applications

INTERNATIONAL
ELECTROTECHNICAL
COMMISSION
ICS 27.120.20; ICS 35.020 ISBN 978-2-8322-6901-5

– 2 – IEC TR 63468:2023 © IEC 2023
CONTENTS
FOREWORD . 4
INTRODUCTION . 6
1 Scope . 8
2 Normative references . 8
3 Terms and definitions . 8
4 Abbreviated terms . 13
5 AI overview from a nuclear perspective . 14
5.1 Brief history of AI . 14
5.2 Major concepts of AI . 15
5.2.1 AI definition . 15
5.2.2 Levels of intelligence and autonomy . 16
5.2.3 Data processing and specifications . 18
5.2.4 Symbolic and sub-symbolic approaches . 19
5.3 Specific fields of AI applications . 27
5.3.1 Data mining . 27
5.3.2 Natural language processing . 27
5.3.3 Computer vision . 28
5.4 Challenges of AI applications in nuclear facilities . 28
5.4.1 General . 28
5.4.2 Trustworthiness . 28
5.4.3 AI verification and validation . 29
6 Some potential nuclear AI applications . 29
6.1 General . 29
6.2 Virtual sensors . 29
6.3 Intelligent control . 30
6.4 Ageing management . 30
6.5 Preventive maintenance . 30
6.6 Anomaly detection . 31
6.7 Operational decision support. 31
6.8 Cyber security . 31
6.9 Human factor engineering . 32
7 Proposed structure for SC 45A AI standards . 32
7.1 General . 32
7.2 Key criteria for structure design . 32
7.2.1 Technical coverage. 32
7.2.2 Hierarchical levels . 33
7.2.3 Entry point documents . 33
7.2.4 Reference to other standards . 33
7.3 Structure of AI standard series . 33
8 Near-term development priorities . 34
9 Organizational challenges and recommendation . 35
9.1 Cross-cutting characteristics of nuclear AI standards . 35
9.2 Organizational challenges . 36
9.3 Recommendations . 36
9.3.1 General . 36
9.3.2 Title and scope of the proposed new working group . 37

9.3.3 Liaison with external organizations . 37
Annex A (informative) AI applications beyond the SC 45A scope . 38
A.1 General . 38
A.2 Nuclear research and development . 38
A.2.1 Overview . 38
A.2.2 Material property prediction . 38
A.2.3 Thermal-fluid phenomena . 38
A.3 Nuclear reactor system design . 39
A.3.1 Overview . 39
A.3.2 Steam explosion analysis . 39
A.3.3 Thermal fatigue analysis . 39
A.3.4 DNBR prediction . 39
A.3.5 Fuel assembly design . 40
A.4 Nuclear project construction . 40
A.4.1 Overview . 40
A.4.2 Monitoring concrete quality . 40
A.4.3 Non-destructive testing . 40
A.5 Plant operation and maintenance . 41
A.5.1 Overview . 41
A.5.2 Accident identification . 41
A.5.3 Transient identification . 41
A.5.4 Fuel management . 42
A.5.5 Component inspection . 42
A.5.6 Water chemistry management. 42
A.5.7 Reactor uprate support . 42
A.5.8 Physical protection . 43
A.5.9 Probabilistic risk assessments . 43
Bibliography . 44

Figure 1 – Brief history of AI . 15
Figure 2 – Functional view of a nuclear AI system. 16
Figure 3 – Major approaches to AI . 20
Figure 4 – Machine learning pipeline (workflow) . 23
Figure 5 – Major approaches to machine learning . 24
Figure 6 – Proposed structure for SC 45A AI standards . 34

Table 1 – Example of autonomy levels for nuclear facilities (Referring to the
categorization scheme of autonomy levels from SAE) . 18
Table 2 – Working groups of IEC SC 45A. 35
Table 3 – Cross-cutting between nuclear AI applications and SC 45A working groups . 35
Table 4 – Cross-cutting between general AI topics and SC 45A working groups. 36

– 4 – IEC TR 63468:2023 © IEC 2023
INTERNATIONAL ELECTROTECHNICAL COMMISSION
____________
NUCLEAR FACILITIES – INSTRUMENTATION AND CONTROL, AND
ELECTRICAL POWER SYSTEMS – ARTIFICIAL INTELLIGENCE
APPLICATIONS
FOREWORD
1) The International Electrotechnical Commission (IEC) is a worldwide organization for standardization comprising
all national electrotechnical committees (IEC National Committees). The object of IEC is to promote international
co-operation on all questions concerning standardization in the electrical and electronic fields. To this end and
in addition to other activities, IEC publishes International Standards, Technical Specifications, Technical Reports,
Publicly Available Specifications (PAS) and Guides (hereafter referred to as “IEC Publication(s)”). Their
preparation is entrusted to technical committees; any IEC National Committee interested in the subject dealt with
may participate in this preparatory work. International, governmental and non-governmental organizations liaising
with the IEC also participate in this preparation. IEC collaborates closely with the International Organization for
Standardization (ISO) in accordance with conditions determined by agreement between the two organizations.
2) The formal decisions or agreements of IEC on technical matters express, as nearly as possible, an international
consensus of opinion on the relevant subjects since each technical committee has representation from all
interested IEC National Committees.
3) IEC Publications have the form of recommendations for international use and are accepted by IEC National
Committees in that sense. While all reasonable efforts are made to ensure that the technical content of IEC
Publications is accurate, IEC cannot be held responsible for the way in which they are used or for any
misinterpretation by any end user.
4) In order to promote international uniformity, IEC National Committees undertake to apply IEC Publications
transparently to the maximum extent possible in their national and regional publications. Any divergence between
any IEC Publication and the corresponding national or regional publication shall be clearly indicated in the latter.
5) IEC itself does not provide any attestation of conformity. Independent certification bodies provide conformity
assessment services and, in some areas, access to IEC marks of conformity. IEC is not responsible for any
services carried out by independent certification bodies.
6) All users should ensure that they have the latest edition of this publication.
7) No liability shall attach to IEC or its directors, employees, servants or agents including individual experts and
members of its technical committees and IEC National Committees for any personal injury, property damage or
other damage of any nature whatsoever, whether direct or indirect, or for costs (including legal fees) and
expenses arising out of the publication, use of, or reliance upon, this IEC Publication or any other IEC
Publications.
8) Attention is drawn to the Normative references cited in this publication. Use of the referenced publications is
indispensable for the correct application of this publication.
9) Attention is drawn to the possibility that some of the elements of this IEC Publication may be the subject of patent
rights. IEC shall not be held responsible for identifying any or all such patent rights.
IEC TR 63468 has been prepared by subcommittee 45A: Instrumentation, control and electrical
power systems of nuclear facilities, of IEC technical committee 45: Nuclear instrumentation. It
is a Technical Report.
The text of this Technical Report is based on the following documents:
Draft Report on voting
45A/1458/DTR 45A/1472/RVDTR
Full information on the voting for its approval can be found in the report on voting indicated in
the above table.
The language used for the development of this Technical Report is English.

This document was drafted in accordance with ISO/IEC Directives, Part 2, and developed in
accordance with ISO/IEC Directives, Part 1 and ISO/IEC Directives, IEC Supplement, available
at www.iec.ch/members_experts/refdocs. The main document types developed by IEC are
described in greater detail at www.iec.ch/publications.
The committee has decided that the contents of this document will remain unchanged until the
stability date indicated on the IEC website under webstore.iec.ch in the data related to the
specific document. At this date, the document will be
• reconfirmed,
• withdrawn,
• replaced by a revised edition, or
• amended.
IMPORTANT – The "colour inside" logo on the cover page of this document indicates that it
contains colours which are considered to be useful for the correct understanding of its
contents. Users should therefore print this document using a colour printer.

– 6 – IEC TR 63468:2023 © IEC 2023
INTRODUCTION
a) Technical background, main issues and organization of the technical report
Artificial intelligence(AI) is transforming many fields including nuclear industry drastically. It
has been explored and deployed for many years in the nuclear industry and recent advances
in AI have enabled many more potentials. Wide adoption of AI calls for standardization
efforts to minimize the risks and optimize the efficiency in developing and deploying AI
applications. Due to its nature as an enabling technology, the topic of AI applications will
cross-cut with almost all working groups within SC 45A, which entails discussions on the
setting up of a new working group to dedicate to this new technical field.
This technical report overviews AI technologies from a nuclear perspective, and summaries
potential AI application scenarios in nuclear facilities. Based on these inputs, a three-tiered
structure for nuclear AI standards within the framework of SC 45A is proposed, and
development priorities are discussed. This document then moves on from technical
discussions to the organizational challenges in SC 45A. It analyses the cross-cutting nature
of AI applications in nuclear facilities and makes the recommendation the setting-up a new
working group, whose title and scope are also proposed. Possibility of SC 45A liaison with
other technical subcommittees is explored and recommendation is given accordingly.
b) Situation of the current technical report in the structure of the IEC SC 45A standard
series
The technical report IEC TR 63468 is a fourth level IEC SC 45A document. This document
overviews the fundamentals of artificial intelligence (AI) and its potential applications in
nuclear facilities to foster better understanding and adoption of AI technologies within such
facilities. It also proposes a structure for future SC 45A standard series on nuclear AI
applications.
For more details on the structure of the SC 45A standard series, see item d) of this
introduction.
c) Recommendations and limitations regarding the application of this technical report
This document is the first of its kind within SC 45A, intended to pave the road for extensive
and systematic efforts in the standard development activities with regard to AI applications.
It helps stakeholders to understand the main benefits and challenges of AI from a nuclear
perspective. More documents are expected to follow in this direction in the coming years.
It is important to note that a technical report is entirely informative in nature, and it
establishes no requirements.
d) Description of the structure of the IEC SC 45A standard series and relationships with
other IEC documents and other bodies documents (IAEA, ISO)
The IEC SC 45A standard series comprises a hierarchy of four levels. The top-level
documents of the IEC SC 45A standard series are IEC 61513 and IEC 63046.
IEC 61513 provides general requirements for instrumentation and control (I&C) systems and
equipment that are used to perform functions important to safety in nuclear power plants
(NPPs). IEC 63046 provides general requirements for electrical power systems of NPPs; it
covers power supply systems including the supply systems of the I&C systems.
IEC 61513 and IEC 63046 are to be considered in conjunction and at the same level.
IEC 61513 and IEC 63046 structure the IEC SC 45A standard series and shape a complete
framework establishing general requirements for instrumentation, control and electrical
power systems for nuclear power plants.
IEC 61513 and IEC 63046 refer directly to other IEC SC 45A standards for general
requirements for specific topics, such as categorization of functions and classification of
systems, qualification, separation, defence against common cause failure, control room
design, electromagnetic compatibility, human factors engineering, cybersecurity, software
and hardware aspects for programmable digital systems, coordination of safety and security
requirements and management of ageing. The standards referenced directly at this second
level should be considered together with IEC 61513 and IEC 63046 as a consistent
document set.
At a third level, IEC SC 45A standards not directly referenced by IEC 61513 or by IEC 63046
are standards related to specific requirements for specific equipment, technical methods, or
activities. Usually these documents, which make reference to second-level documents for
general requirements, can be used on their own.
A fourth level extending the IEC SC 45 standard series, corresponds to the Technical
Reports which are not normative.
The IEC SC 45A standards series consistently implements and details the safety and
security principles and basic aspects provided in the relevant IAEA safety standards and in
the relevant documents of the IAEA nuclear security series (NSS). In particular this includes
the IAEA requirements SSR-2/1 , establishing safety requirements related to the design of
nuclear power plants (NPPs), the IAEA safety guide SSG-30 dealing with the safety
classification of structures, systems and components in NPPs, the IAEA safety guide SSG-
39 dealing with the design of instrumentation and control systems for NPPs, the IAEA safety
guide SSG-34 dealing with the design of electrical power systems for NPPs, the IAEA safety
guide SSG-51 dealing with human factors engineering in the design of NPPs and the
implementing guide NSS42-G for computer security at nuclear facilities. The safety and
security terminology and definitions used by the SC 45A standards are consistent with those
used by the IAEA.
IEC 61513 and IEC 63046 have adopted a presentation format similar to the basic safety
publication IEC 61508 with an overall life-cycle framework and a system life-cycle
framework. Regarding nuclear safety, IEC 61513 and IEC 63046 provide the interpretation
of the general requirements of IEC 61508-1, IEC 61508-2 and IEC 61508-4, for the nuclear
application sector. In this framework, IEC 60880, IEC 62138 and IEC 62566 correspond to
IEC 61508-3 for the nuclear application sector.
IEC 61513 and IEC 63046 refer to ISO 9001 as well as to IAEA GSR part 2 and IAEA GS-
G-3.1 and IAEA GS-G-3.5 for topics related to quality assurance (QA).
At level 2, regarding nuclear security, IEC 62645 is the entry document for the IEC/SC 45A
security standards. It builds upon the valid high level principles and main concepts of the
generic security standards, in particular ISO/IEC 27001 and ISO/IEC 27002; it adapts them
and completes them to fit the nuclear context and coordinates with the IEC 62443 series. At
level 2, IEC 60964 is the entry document for the IEC/SC 45A control rooms standards,
IEC 63351 is the entry document for the human factors engineering standards and
IEC 62342 is the entry document for the ageing management standards.
NOTE 1 It is assumed that for the design of I&C systems in NPPs that implement conventional safety functions
(e.g. to address worker safety, asset protection, chemical hazards, process energy hazards) international or
national standards would be applied.
NOTE 2 IEC TR 64000 provides a more comprehensive description of the overall structure of the IEC SC 45A
standards series and of its relationship with other standards bodies and standards.

– 8 – IEC TR 63468:2023 © IEC 2023
NUCLEAR FACILITIES – INSTRUMENTATION AND CONTROL, AND
ELECTRICAL POWER SYSTEMS – ARTIFICIAL INTELLIGENCE
APPLICATIONS
1 Scope
This document overviews the fundamentals of artificial intelligence (AI) as it could potentially
be applied within nuclear facilities and identifies proven or potential applications, with the
objective to foster better understanding and adoption of AI technologies within such facilities.
With the objective of supporting future standard development work of IEC SC 45A in this
technical area, this document takes the initiative to propose a structure for SC 45A standard
series on nuclear AI applications, and recommends setting up a new dedicated working group
to be responsible for and coordinate standard development efforts in this particular area, taking
into account its cross-cutting nature.
As some technical aspects of AI are still evolving, and the regulatory framework from nuclear
regulators is not yet established, this document focuses on AI applications in nuclear facilities
that are non-safety related. However, this approach does not necessarily exclude the
applicability of AI technologies in safety applications in nuclear facilities where the technology
itself and the related regulatory framework support such potentials.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content
constitutes requirements of this document. For dated references, only the edition cited applies.
For undated references, the latest edition of the referenced document (including any
amendments) applies.
IEC 61513, Nuclear power plants – Instrumentation and control important to safety - General
requirements for systems
IEC 63046, Nuclear power plants – Electrical power system – General requirements
IEC TR 63400, Nuclear facilities – Instrumentation, control and electrical power systems
important to safety – Structure of the IEC SC45A standards series
ISO/IEC 22989:2022, Information technology – Artificial intelligence – Artificial intelligence
concepts and terminology
ISO/IEC 23053, Framework for Artificial Intelligence (AI) Systems Using Machine Learning (ML)
ISO/IEC TR 29119-11, Software and systems engineering – Software testing – Part 11:
Guidelines on the testing of AI-based systems
3 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
ISO and IEC maintain terminology databases for use in standardization at the following
addresses:
• IEC Electropedia: available at https://www.electropedia.org/
• ISO Online browsing platform: available at https://www.iso.org/obp

3.1
artificial intelligence
AI
research and development of mechanisms and applications of AI systems
Note 1 to entry: This definition is further expanded in Clause 5 of ISO/IEC 22989.
Note 2 to entry: For the purpose of this document, this definition can be supplemented by the definition given in
Wikipedia, where Artificial Intelligence is defined as “intelligence—perceiving, synthesizing, and inferring
information—demonstrated by machines, as opposed to intelligence displayed by animals and humans”.
[SOURCE: ISO/IEC 22989:2022, 3.1.3]
3.2
artificial intelligence system
AI system
engineered system that generates outputs such as content, forecasts, recommendations or
decisions for a given set of human-defined objectives
[SOURCE: ISO/IEC 22989:2022, 3.1.4]
3.3
autonomy
autonomous
characteristic of a system that is capable of modifying its operating domain or goal without
external intervention, control or oversight
[SOURCE: ISO/IEC 22989:2022, 3.1.5]
3.4
automatic
automation
automated
pertaining to a process or system that, under specified conditions, functions without human
intervention
[SOURCE: ISO/IEC 22989:2022, 3.1.7]
3.5
Bayesian network
probabilistic model that uses Bayesian inference for probability computations using a directed
acyclic graph
[SOURCE: ISO/IEC 22989:2022, 3.3.1]
3.6
continuous learning
continual learning
lifelong learning
incremental training of an AI system that takes place on an ongoing basis during the operation
phase of the AI system life cycle
[SOURCE: ISO/IEC 22989:2022, 3.1.9]
3.7
data mining
computational process that extracts patterns by analyzing quantitative data from different
perspectives and dimensions, categorizing them, and summarizing potential relationships and
impacts
– 10 – IEC TR 63468:2023 © IEC 2023
[SOURCE: ISO 16439:2014, 3.13, modified – “identifies” has been replaced by “extracts”]
3.8
data sampling
process to select a subset of data samples intended to present patterns and trends similar to
that of the larger dataset being analyzed
Note 1 to entry: Ideally, the subset of data samples will be representative of the larger dataset.
[SOURCE: ISO/IEC 22989:2022, 3.2.4]
3.9
dataset
collection of data with a shared format
Note 1 to entry: Datasets can be used for validating or testing an AI model. In a machine learning context, datasets
can also be used to train a machine learning algorithm.
[SOURCE: ISO/IEC 22989:2022, 3.2.5]
3.10
decision tree
model for which inference is encoded as paths from the root to a leaf node in a tree structure
[SOURCE: ISO/IEC 22989:2022, 3.3.2]
3.11
declarative knowledge
knowledge represented by facts, rules and theorems
Note 1 to entry: Usually, declarative knowledge cannot be processed without first being translated into procedural
knowledge.
[SOURCE: ISO/IEC 22989:2022, 3.1.12]
3.12
deep learning
deep neural network learning
approach to creating rich hierarchical representations through the
training of neural networks with many hidden layers
Note 1 to entry: Deep learning is a subset of ML.
[SOURCE: ISO/IEC 22989:2022, 3.4.4]
3.13
expert system
AI system that accumulates, combines and encapsulates knowledge provided by a human
expert or experts in a specific domain to infer solutions to problems
[SOURCE: ISO/IEC 22989:2022, 3.1.13]
3.14
general AI
artificial general intelligence
AGI
type of AI system that addresses a broad range of tasks with a satisfactory level of performance
Note 1 to entry: Compared to narrow AI.

Note 2 to entry: AGI is often used in a stronger sense, meaning systems that not only can perform a wide variety
of tasks, but all tasks that a human can perform.
[SOURCE: ISO/IEC 22989:2022, 3.1.14]
3.15
label
target variable assigned to a sample
[SOURCE: ISO/IEC 22989:2022, 3.2.10]
3.16
life cycle
evolution of a system, product, service, project or other human-made entity, from conception
through retirement
[SOURCE: ISO/IEC/IEEE 15288:2015, 4.1.23]
3.17
long-short-term-memory network
LSTM
type of neural network that processes sequential data with a satisfactory performance for both
long and short span dependencies
[SOURCE: ISO/IEC 22989:2022, 3.4.7, modified – “long-short-term” has been replaced by
“long-short-term-memory”]
3.18
machine learning
ML
process of optimizing model parameters through computational techniques, such that the
model's behavior reflects the data or experience
[SOURCE: ISO/IEC 22989:2022, 3.3.5]
3.19
machine learning model
mathematical construct that generates an inference, or prediction, based on input data or
information
[SOURCE: ISO/IEC 22989:2022, 3.3.7]
3.20
model
physical, mathematical, or otherwise logical representation of a system, entity, phenomenon,
process or data
[SOURCE: ISO/IEC 18023-1:2006, 3.1.11, modified –"or data" added]
3.21
narrow AI
type of AI system that is focused on defined tasks to address a specific problem
Note 1 to entry: Compared to general AI.
[SOURCE: ISO/IEC 22989:2022, 3.1.24]

– 12 – IEC TR 63468:2023 © IEC 2023
3.22
neural network
neural net
artificial neural network
NN
network of one or more layers of neurons connected by weighted links with adjustable weights,
which takes input data and produces an output
Note 1 to entry: Neural networks are a prominent example of the connectionist approach.
Note 2 to entry: Although the design of neural networks was initially inspired by the functioning of biological
neurons, most works on neural networks do not follow that inspiration anymore.
[SOURCE: ISO/IEC 22989:2022, 3.4.8]
3.23
prediction
primary output of an AI system when provided with input data or information
Note 1 to entry: Predictions can be followed by additional outputs, such as recommendations, decisions and actions.
Note 2 to entry: Prediction does not necessarily refer to predicting something in the future.
Note 3 to entry: Predictions can refer to various kinds of data analysis or production applied to new data or historical
data (including translating text, creating synthetic images or diagnosing a previous power failure).
3.24
reinforcement learning
RL
earning of an optimal sequence of actions to maximize a reward through interaction with an
environment
[SOURCE: ISO/IEC 22989:2022, 3.1.27]
3.25
sample
atomic data element processed in quantities by a machine learning algorithm or an AI system
[SOURCE: ISO/IEC 22989:2022, 3.2.13]
3.26
semi-supervised machine learning
machine learning that makes use of both labelled and unlabelled data during training
[SOURCE: ISO/IEC 22989:2022, 3.3.11]
3.27
sub-symbolic AI
AI based on techniques and models that use an implicit encoding of information, that can be
derived from experience or raw data.
Note 1 to entry: Compared to symbolic AI. Whereas symbolic AI produces declarative outputs, sub-symbolic AI is
based on statistical approaches and produces outputs with a given probability of error.
[SOURCE: ISO/IEC 22989:2022, 3.1.34]
3.28
supervised machine learning
machine learning that makes use of labelled data during training
[SOURCE: ISO/IEC 22989:2022, 3.3.12]

3.29
symbolic AI
AI based on techniques and models that manipulate symbols and structures according to
explicitly defined rules to obtain inferences
Note 1 to entry: Compared to sub-symbolic AI, symbolic AI produces declarative outputs, whereas sub-symbolic AI
is based on statistical approaches and produces outputs with a given probability of error.
[SOURCE: ISO/IEC 22989:2022, 3.1.33]
3.30
test data
evaluation data
data used to assess the performance of a final model
Note 1 to entry: Test data is disjoint from training data and validation data.
[SOURCE: ISO/IEC 22989:2022, 3.2.14]
3.31
training
model training
process to establish or to improve the parameters of a machine learning model, based on a
machine learning algorithm, by using training data
[SOURCE: ISO/IEC 22989:2022, 3.3.15]
3.32
unsupervised machine learning
machine learning that makes only use of unlabelled data during training
[SOURCE: ISO/IEC 22989:2022, 3.3.17]
3.33
validation data
development data
data used to compare the performance of different candidate models
Note 1 to entry: Validation data is disjoint from test data and generally also from training data. However, in cases
where there is insufficient data for a three-way training, validation and test set split, the data is divided into only two
sets – a test set and a training or validation set. Cross-validation or bootstrapping are common methods for then
generating separate training and validation sets from the training or validation set.
Note 2 to entry: Validation data can be used to tune hyper-parameters or to validate some algorithmic choices, up
to the effect of including a given rule in an expert system.
[SOURCE: ISO/IEC 22989:2022, 3.2.15]
4 Abbreviated terms
AI Artificial Intelligence
BWR Boiling Water Reactor
CAP Corrective Action Program
CFD Computational Fluid Dynamics
CNN Convolutional Neural Network
CRDM Control Rod Drive Mechanism
CV Computer Vision
– 14 – IEC TR 63468:2023 © IEC 2023
DL  Deep Learning
ENIQ European Network for Inspection and Qualification
FFD Fitness For Duty
GOFAI Good Old-Fashioned Artificial Intelligence
HST Hot Spot Temperature
I&C Instrumentation and Control
KG Knowledge Graph
ML Machine Learning
NDT Non-Destructive Testing
NLP Natural Language Processing
NN Neural Network
NPP Nuclear Power Plant
RUL Remaining Useful Life
SAE Society of Automotive Engineers
SSC Systems, Structures, Components
V&V Verification and Validation
5 AI overview from a nuclear perspective
5.1 Brief history of AI
The history of AI dates back to the 1940s when Warren McCulloch and Walter Pitts suggested
connected neuron networks could learn. Alan Turing proposed the Turing test, machine learning,
and reinforcement learning in his 1950’s article “Computing Machinery and Intelligence.”
The term “artificial intelligence” was coined at a Dartmouth workshop in 1956. The next two
decades were golden years for AI and the field received extensive government funds for its
promising potential for logic-based problem solving. However, by 1974, overly high expectations
and limited capabilities led to the first “AI winter”. The rise of knowledge-based expert systems
brought new successes in the 1980s and following years. But the second “AI winter” started
with the identification of expert systems limitations in 1987. AI returned to favor in 1993 with
the help of increased computational power. From 2012, unprecedented availability of data and
computational power enabled breakthroughs in machine learning, in particular deep machine
learning, and ushered in greater success for AI. In the last decade, AI applications in such fields
as image analysis, speech recognition and autonomous driving have greatly changed people’s
daily lives. With deep learning and other advanced techniques, AI now can outperform humans
over a range of specific tasks such as image recognition and also beat human champions at
games such as go. An intuitive development of different periods of AI is shown in Figure 1.

Figure 1 – Brief history of AI
The nuclear industry’s effort in pursuing the application of AI techniques dates back to, at least,
the early 1980s, when the AI itself was experiencing the second boom with the success of
expert systems. As the industry shifted into improving the safety aspects of operating NPPs,
following the Chernobyl accident, the interest in automation was overshadowed by the need to
improve safety. Over decades of operations, the industry resorted to increasing staffing levels
to meet increasing and emerging safety requirements, accompanied by increasing use of
systematic processes and procedures in all aspects of industry operations. This increase of
staffing requirements offset the economic advantage of the relatively low energy cost of nuclear
fuel in comparison to other forms of baseload energy sources, i.e., mainly fossil energy.
However, the drop in oil and gas prices following the 2008 global financial crisis and the
associated crash of fossil energy prices resulted in a substantial economical challenge to
nuclear power, especially in non-subsidized nuclear energy markets. An NPP operating at 1 000
MWe could have more than 10 folds the number of a staff as a fossil plant operating at the
same power level. This disadvantage of nuclear energy, coupled with the urge to fight climate
change using nuclear energy as the main baseload clean source of energy, rejuvenated the
interest in automation to offset manual activities performed by an NPP staff and enable the
industry to become more economically sustainable.
In parallel to this realization by the industry, AI as a science have benefited from advancements
in computational power and presented unique capabilities to enable the needed automation for
the nuclear power industry. Given, all those factors, AI has found an exponentially growing
number of applications in the nuclear industry. While NPPs operation was one of the main
drivers of the nuclear industry leveraging AI, the broader nuclear scientific and professional
community rapidly adopted AI too. AI is now used in reactors design, fuel optimization,
intelligent control, preventive maintenance, ageing management, non-destructive testing,
physical protection, cybersecurity, and many other related fields.
5.2 Major concepts of AI
5.2.1 AI definition
Historically, AI has been approached from a number of different perspectives and accordingly
diverse definitions have been established. Some definitions of AI have defined intelligence in
terms of fidelity to human performance, while others prefer an abstract, formal definition of
intelligence called rationality.

– 16 – IEC TR 63468:2023 © IEC 2023
This document will leverage the AI definition from ISO/IEC 22989, which views AI systems as
engineered systems that generate outputs such as content, forecasts, recommendations or
decisions for a given set of human-defined objectives. It is worthwhile to note that this definition
has been further expanded in Clause 5 of ISO/IEC 22989:2022 to make it more concrete. For
the purpose of this document, the definition can be supplemented by the one given in Wikipedia,
where AI is viewed as “intelligence—perceiving, synthesizing, and inferring information—
demonstrated by machines, as opposed to intelligence displayed by animals and humans”. This
supplemental definition emphasized the machine intelligence aspects of ISO/IEC 22989
definition.
To put this definition into context, a simplified functional view of an AI system is presented in
Figure 2, which is illustrated from a nuclear facility application perspective. The workflow for a
deployed nuclear AI system is very straightforward, which is comprised of three major
components, namely inputs, AI processing and outputs. Taking the AI system for identifying
plant transients as an example, it takes inputs from plant instrumentation and control systems,
performs intelligent analysis and inference, and then generates some outputs that predict the
type of transients the NPP is experiencing.
Behind the abovementioned workflow, is an AI model that has been designed and developed
by hum
...