**2. Approach and features of the JAEA KMS**

Recognition of the importance of knowledge management by JAEA led to establishment of a new Knowledge Office with the remit to develop and implement an advanced KMS tailored to the requirements of the Japanese geological disposal programme. This programme is coordinated with initiatives led by the responsible Government department 'ANRE' (Agency of Natural Resources and Energy) of METI (Ministry of Economy, Trade and Industry) and will, eventually, be formally linked to other relevant organisations in Japan that may be either producers or users of knowledge. Within JAEA, the Knowledge Office not only develops the new tools and organisational structures required to implement the KMS, but also provides executive support for synthetic analysis and evaluation of knowledge (including meta-analysis and top-level quality management).

It is important to emphasise here the particular boundary conditions for this KMS. At the most fundamental level, JAEA is charged with providing scientific and technical support to both implementing and regulatory organisations and also interested stakeholders – including the general public. This requires that all parties recognise JAEA to be a competent and unbiased organisation and that the KMS incorporates a rigorous Quality Management System (QMS) that is accepted by all.

For geological disposal in Japan, the implementer 'NUMO' (Nuclear Waste Management Organisation of Japan) has selected a volunteering approach for initiation of the siting process, which introduces technical challenges in terms of assuring flexibility of site characterisation planning, repository concept development and associated PA (e.g. Kitayama et al., 2005). Additionally, attracting volunteers and developing dialogue with local communities involve challenges in public communication.

On the regulatory side, development of guidelines and licensing procedures for deep geological repositories is ongoing, but one clear aim is to maximise consistency with already established regulations for near surface disposal facilities for lower level radioactive waste (NSC, 2004). This is complicated in Japan due to its location on an Island Arc, with its associated high tectonic activity and geological complexity (e.g. Apted et al., 2004).

Although current emphasis is on existing inventories of radioactive waste, the Japanese long-term commitment to nuclear power and a reference development programme over the next 100 years has to be considered in order to develop a consistent and compatible longterm waste management programme (Makino et al., 2009b). In addition, various types of

the problems of information overload were recognised during a comprehensive assessment of High-Level Waste (HLW) disposal feasibility at the turn of the century (JNC, 2000). This problem is exacerbated by a Japanese volunteering approach to siting of a deep geological repository, which requires particular flexibility in the tailoring of site characterization plans, repository concepts and associated Performance Assessments (PAs). Recognition of this situation led, in 2005, to initiation by Japan Atomic Energy Agency (JAEA) of an ambitious project to develop an advanced KMS aimed to facilitate its role as the supplier of background R&D support to both regulators and implementers of geological disposal. The chapter introduces the background to this initiative and the basic approach selected, and then review progress to date in this work, with emphasis on tailoring of existing Knowledge Engineering tools and methods to radioactive waste management requirements, and outline future developments and challenges (Umeki et al., 2009; Osawa et al., 2009b;

Recognition of the importance of knowledge management by JAEA led to establishment of a new Knowledge Office with the remit to develop and implement an advanced KMS tailored to the requirements of the Japanese geological disposal programme. This programme is coordinated with initiatives led by the responsible Government department 'ANRE' (Agency of Natural Resources and Energy) of METI (Ministry of Economy, Trade and Industry) and will, eventually, be formally linked to other relevant organisations in Japan that may be either producers or users of knowledge. Within JAEA, the Knowledge Office not only develops the new tools and organisational structures required to implement the KMS, but also provides executive support for synthetic analysis and evaluation of

It is important to emphasise here the particular boundary conditions for this KMS. At the most fundamental level, JAEA is charged with providing scientific and technical support to both implementing and regulatory organisations and also interested stakeholders – including the general public. This requires that all parties recognise JAEA to be a competent and unbiased organisation and that the KMS incorporates a rigorous Quality Management

For geological disposal in Japan, the implementer 'NUMO' (Nuclear Waste Management Organisation of Japan) has selected a volunteering approach for initiation of the siting process, which introduces technical challenges in terms of assuring flexibility of site characterisation planning, repository concept development and associated PA (e.g. Kitayama et al., 2005). Additionally, attracting volunteers and developing dialogue with

On the regulatory side, development of guidelines and licensing procedures for deep geological repositories is ongoing, but one clear aim is to maximise consistency with already established regulations for near surface disposal facilities for lower level radioactive waste (NSC, 2004). This is complicated in Japan due to its location on an Island Arc, with its

Although current emphasis is on existing inventories of radioactive waste, the Japanese long-term commitment to nuclear power and a reference development programme over the next 100 years has to be considered in order to develop a consistent and compatible longterm waste management programme (Makino et al., 2009b). In addition, various types of

associated high tectonic activity and geological complexity (e.g. Apted et al., 2004).

Semba et al., 2009; Makino et al., 2009a; Makino et al., 2011).

knowledge (including meta-analysis and top-level quality management).

local communities involve challenges in public communication.

**2. Approach and features of the JAEA KMS** 

System (QMS) that is accepted by all.

waste caused by Fukushima Dai-ichi accident occurred in 2011 will become significant subject to the Japanese waste management programme.

Taken together, these constraints on the KMS lead to a requirement for a holistic approach that will form the core of a programme to be implemented over a period extending beyond the 21st century. Clearly, major developments in technology are to be expected over this period, although these are inherently unpredictable in detail. Emphasis is thus placed on development of a fundamental KMS concept that will evolve in line with advancing technology, supported by tools and approaches that are, to the maximum extent possible, scale and platform independent.

The overall structure and key components of the KMS are illustrated in Figure 1. It should be noted that the remit of the Knowledge Office is very wide and the KMS thus includes all aspects of tacit knowledge management (e.g. focused training and experience transfer schemes – often denoted as human resource management), focused quality management (discussed in more detail below) and anticipation of technology developments and future requirements (e.g. using a think tank approach – elsewhere often part of strategic planning or requirements management). Therefore, the combination of the challenging boundary conditions and wide remit led to the decision to establish a support team (Knowledge Office) composed of staff with wide experience of radioactive waste management, and knowledge engineering, who can tailor established methodology and tools effectively and efficiently to the various requirements needed for development of the KMS. In retrospect, this strategic decision probably contributed significantly to the successes achieved to date.

Fig. 1. Outline KMS concept: Structure and key elements

The various different types of knowledge involved and the management functions required are summarised in Table 1. This table also notes which functions have the potential to be, at least partially, automated or facilitated using advanced knowledge engineering tools. Automation and/or computer support of knowledge management functions is a key to

A Challenge on Development of an Advanced Knowledge Management

tailored to the needs of different knowledge users.

a unified approach to quality management.

or Argumentation Model JAEA Knowledge Base

Project Knowledge

• Secure, traceable record of all knowledge applications (database freezing) • Specified reference data to compare against future developments, potential

QA, integration – synthesis, R&D planning, archiving

• Identification of gaps to allow focused searches / knowledge creation

Fig. 2. The knowledge base and its interfaces to users and knowledge resources

Associated Knowledge

Partner Knowledge -intranets

World Knowledge -internet

Focused searches

RMS (Requirements Management System)

Role of JAEA KB:

Communication interfaces

triggers for change assessment

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 169

A key component of the KMS is the JAEA Radwaste Knowledge Base (KB). This is a dynamic entity, which will be constantly supplemented by new input – through focused production by Japanese (or international partner) R&D programmes or autonomically generated by directed searches of the internet. Unlike traditional databases, there is no imposed structure on the KB: it is simply an electronic library of all relevant information and documentation that is applicable to specific radioactive waste applications. It will be used to generate application-specific sub-databases, which are frozen and archived as required to ensure transparent documentation of the background to all major project milestones or decisions that utilise this KB. Utilisation of the KB will be facilitated by a user-friendly interface that allows the mode of access and the form and technical level of output to be

The information explosion noted in the introduction is the key problem when considering how the KB will be utilised by the main user organisations such as implementer and regulator. Although huge volumes of material are already available and very much more will be generated in coming decades, much of this is (or will become) irrelevant for actual implementation or regulation of a repository project. A strict filtering process to develop application-specific subsets of the KB is thus essential for practical use – and also to develop

The process of developing a project-specific KB from the requirements specified by end users is illustrated in Figure 2 (Umeki et al., 2009; Umeki et al., 2010). Ideally, this would be facilitated with a direct interface to a formal Requirements Management System (RMS),

implementation of this novel approach – providing the potential to respond to the exponentially increasing rate of information production, but also giving the greatest challenge to the Knowledge Office team. The essence of what is aimed for has been summarised as an 'intelligent assistant' – an electronic toolkit that allows project leaders and coordinators to manage the huge fluxes of data that they have to control and to efficiently use their time by carrying out only the weighting and top-level synthesis and decision making that cannot be automated.


Note: \*Of the required developments, those highlighted in bold text may be supported by advanced IT/KE technology

Table 1. Typical contents and structure of a KMS based on initial studies carried out by JAEA

implementation of this novel approach – providing the potential to respond to the exponentially increasing rate of information production, but also giving the greatest challenge to the Knowledge Office team. The essence of what is aimed for has been summarised as an 'intelligent assistant' – an electronic toolkit that allows project leaders and coordinators to manage the huge fluxes of data that they have to control and to efficiently use their time by carrying out only the weighting and top-level synthesis and decision making that cannot be

> - solicited data (external) - processed data





coordination team

(interactive – allowing

Note: \*Of the required developments, those highlighted in bold text may be supported by advanced

Table 1. Typical contents and structure of a KMS based on initial studies carried out by JAEA

team - expert systems



dialogue)

*Content Planned/ongoing* 

**- robust archive**

**- autonomic QA** 

**- robust archive**

**- robust archive - autonomic change management** 

**knowledge** 

**(Think Tank)** 

**knowledge**

**information** 

**cross-referencing** 

**mining** 

*developments\**

**- internal and external data** 

**- autonomic data processing** 

**- formal approaches for QA** 

**- autonomic QA/cataloguing/** 

**- formal approaches for QA** 

**- use of expert systems to capture and transfer tacit** 

**- advanced training and experience transfer** 

**- formal description of key integration processes - formal approach for QA** 

**- prediction of requirements** 

**- process for filling key gaps in** 

**- high-end graphical methods for presenting complex** 

automated.

*Form of knowledge* 

Experience and methodology

Documents Document

Software Software

Synthesis Knowledge

Guidance Knowledge

Presentation User/producer

IT/KE technology

*Management functions*

management

management

Resource management

integration

coordination

dialogue

Data Data management - raw data (internal)

A key component of the KMS is the JAEA Radwaste Knowledge Base (KB). This is a dynamic entity, which will be constantly supplemented by new input – through focused production by Japanese (or international partner) R&D programmes or autonomically generated by directed searches of the internet. Unlike traditional databases, there is no imposed structure on the KB: it is simply an electronic library of all relevant information and documentation that is applicable to specific radioactive waste applications. It will be used to generate application-specific sub-databases, which are frozen and archived as required to ensure transparent documentation of the background to all major project milestones or decisions that utilise this KB. Utilisation of the KB will be facilitated by a user-friendly interface that allows the mode of access and the form and technical level of output to be tailored to the needs of different knowledge users.

The information explosion noted in the introduction is the key problem when considering how the KB will be utilised by the main user organisations such as implementer and regulator. Although huge volumes of material are already available and very much more will be generated in coming decades, much of this is (or will become) irrelevant for actual implementation or regulation of a repository project. A strict filtering process to develop application-specific subsets of the KB is thus essential for practical use – and also to develop a unified approach to quality management.

The process of developing a project-specific KB from the requirements specified by end users is illustrated in Figure 2 (Umeki et al., 2009; Umeki et al., 2010). Ideally, this would be facilitated with a direct interface to a formal Requirements Management System (RMS),

Fig. 2. The knowledge base and its interfaces to users and knowledge resources

A Challenge on Development of an Advanced Knowledge Management

acceptance.

counterexamples.

argumentation model editor.

safety case structure;

Additional functions of Scarab include:

search cases similar to the one at hand;

common FAQ might be necessary for understanding of citizens.

explaining the reason for changes;

and set priorities for associated R&D.

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 171

contributes to the safety case". This puts any lower level input clearly into context. - The initial claim must be supported by arguments, which can be usefully classified into different types. At present, different classes of arguments are used: these range from consideration of "hard" laws of science or defined exclusion criteria to "softer" assessments of common understanding and requirements for public



The formalised method involves classification of arguments, which allows the relative strength of associated evidence to be assessed (e.g. arguments based on scientific laws vs. those based on empirical data) and also allows associated typical 'critical questions' to be reviewed to check if the argumentation model is complete (Figure 4). Inevitably, argumentation models become increasingly complex as they go into finer technical detail and rapidly lead to cross-linking between different sub-systems. To manage the network, software tools are essential and after a number of existing packages were examined, a tailored argumentation model editor called Scarab (**S**upporting tool for **Co**nstructing **AR**gumentation models with **A**ssociated knowledge-**B**ase) has been developed (Osawa et al., 2009b). Figure 5 shows an example of screenshots of the

a. Storing existing argumentation models in a case-base, allowing users to key-word

d. Supporting discovery of new counter-arguments by using "deep" knowledge of the

b. Storing base information of arguments which is called knowledge note (Figure 6); c. Recording all the revisions made to each argumentation model, with comments

e. Link with group-ware that provides a collaborative internet working environment; Finally, it should be mentioned that an even simpler form of argumentation modelling has proven useful in developing new communication methods, particularly aimed at nontechnical (or, at least, non-expert) audiences. Here the argumentation model is developed not as a network as shown in Figure 4, but in the form of a dialogue between cartoon characters. Such presentation is very familiar in Japan and this option will be implemented in both static (*manga*) and animated (*anime*) forms. In addition, presentation form like

at a particular programme milestone" or, at a lower level, "a particular system component

which provides a hierarchy of project needs, identifies potential conflicts and establishes priorities or weightings (possibly as a function of programme progress). Although requirements management is recognised to be essential by NUMO and a formal system is under development (NUMO, 2007), it is not yet operational. To initiate work in the absence of clearly defined requirements, therefore, a prototype project-specific KB is being developed using a basic structure of a geological repository safety case, which will certainly be a key requirement for the implementer and a focus of programme review by the regulators (JNC, 2005; Umeki et al., 2008). In this context, the safety case can be clearly a common frame for R&D activities carried out in all relevant organisations. The safety case is also of great interest for all stakeholders.

A safety case is an extremely complex synthesis of a vast array of different information. Although this is supported by much of the technical work carried out by JAEA and other R&D organisations, development of the safety case is actually carried out by NUMO and reviewed by the regulators. To make the benefits of this advanced KMS by JAEA apparent to all relevant parties, it is important that it eases the work of specialists by highlighting the relevance of their research to the safety case and simplifying the processes of accessing relevant literature, synthesising data, producing and reviewing documentation and communicating their results to interested parties (see Section 3). Here argumentation modelling mentioned in Section 3.1 have played an additional key role of overviewing the essence of the safety case in an easily understandable manner.

### **3. Tools developed in the JAEA KMS project**

This section will discuss about development and application of tools comprising the JAEA KMS with practical examples.

#### **3.1 Argumentation modelling**

The safety case can be seen as the top-level goal of all works carried out within a geological disposal project. Knowledge can be classified in terms of its role in the safety case and prioritised in terms of its impact on overall safety case argumentation (Figure 3). The resultant argumentation model can be developed in a top-down manner, highlighting the constraints on decisions set by upper level requirements and the consequences of decisions on all interlinked topics.

Argumentation modelling is a well-established tool in Knowledge Engineering (e.g. Kirschner et al., 2003) and can be implemented in a number of different ways. In all cases, an initial claim is analysed to determine possible counter-arguments which are, in turn, analysed to identify supporting arguments that counter these. The process iterates until unambiguous hard evidence is provided or an open question is identified. Although such processes are well established in areas such as philosophy and law, they are less often used in technical fields. Nevertheless, experiences to date within the JAEA KMS project have shown that this approach is well suited to breaking down complex multidisciplinary problems in radioactive waste management.

Use of an argumentation model to represent key components of a safety case has certain advantages:


which provides a hierarchy of project needs, identifies potential conflicts and establishes priorities or weightings (possibly as a function of programme progress). Although requirements management is recognised to be essential by NUMO and a formal system is under development (NUMO, 2007), it is not yet operational. To initiate work in the absence of clearly defined requirements, therefore, a prototype project-specific KB is being developed using a basic structure of a geological repository safety case, which will certainly be a key requirement for the implementer and a focus of programme review by the regulators (JNC, 2005; Umeki et al., 2008). In this context, the safety case can be clearly a common frame for R&D activities carried out in all relevant organisations. The safety case is

A safety case is an extremely complex synthesis of a vast array of different information. Although this is supported by much of the technical work carried out by JAEA and other R&D organisations, development of the safety case is actually carried out by NUMO and reviewed by the regulators. To make the benefits of this advanced KMS by JAEA apparent to all relevant parties, it is important that it eases the work of specialists by highlighting the relevance of their research to the safety case and simplifying the processes of accessing relevant literature, synthesising data, producing and reviewing documentation and communicating their results to interested parties (see Section 3). Here argumentation modelling mentioned in Section 3.1 have played an additional key role of overviewing the

This section will discuss about development and application of tools comprising the JAEA

The safety case can be seen as the top-level goal of all works carried out within a geological disposal project. Knowledge can be classified in terms of its role in the safety case and prioritised in terms of its impact on overall safety case argumentation (Figure 3). The resultant argumentation model can be developed in a top-down manner, highlighting the constraints on decisions set by upper level requirements and the consequences of decisions

Argumentation modelling is a well-established tool in Knowledge Engineering (e.g. Kirschner et al., 2003) and can be implemented in a number of different ways. In all cases, an initial claim is analysed to determine possible counter-arguments which are, in turn, analysed to identify supporting arguments that counter these. The process iterates until unambiguous hard evidence is provided or an open question is identified. Although such processes are well established in areas such as philosophy and law, they are less often used in technical fields. Nevertheless, experiences to date within the JAEA KMS project have shown that this approach is well suited to breaking down complex multidisciplinary

Use of an argumentation model to represent key components of a safety case has certain


also of great interest for all stakeholders.

essence of the safety case in an easily understandable manner.

**3. Tools developed in the JAEA KMS project** 

KMS with practical examples.

**3.1 Argumentation modelling**

on all interlinked topics.

advantages:

problems in radioactive waste management.

at a particular programme milestone" or, at a lower level, "a particular system component contributes to the safety case". This puts any lower level input clearly into context.


The formalised method involves classification of arguments, which allows the relative strength of associated evidence to be assessed (e.g. arguments based on scientific laws vs. those based on empirical data) and also allows associated typical 'critical questions' to be reviewed to check if the argumentation model is complete (Figure 4). Inevitably, argumentation models become increasingly complex as they go into finer technical detail and rapidly lead to cross-linking between different sub-systems. To manage the network, software tools are essential and after a number of existing packages were examined, a tailored argumentation model editor called Scarab (**S**upporting tool for **Co**nstructing **AR**gumentation models with **A**ssociated knowledge-**B**ase) has been developed (Osawa et al., 2009b). Figure 5 shows an example of screenshots of the argumentation model editor.

Additional functions of Scarab include:


Finally, it should be mentioned that an even simpler form of argumentation modelling has proven useful in developing new communication methods, particularly aimed at nontechnical (or, at least, non-expert) audiences. Here the argumentation model is developed not as a network as shown in Figure 4, but in the form of a dialogue between cartoon characters. Such presentation is very familiar in Japan and this option will be implemented in both static (*manga*) and animated (*anime*) forms. In addition, presentation form like common FAQ might be necessary for understanding of citizens.

A Challenge on Development of an Advanced Knowledge Management

**The longevity can be specified by determination of an overpack thickness needed for mechanical integrity and adding an allowance for corrosion expected during the period for which integrity is required to be maintained.**

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 173

**Could corrosion rate increase with time leading to early failure?** 

**The long-term corrosion rates measured in experiments under relevant conditions are well below the reference values of** 

**0.01mm/y**

Fig. 5. Screenshots of the argumentation model editor (Scarab)

**Realistic corrosion rates from long-term corrosion experiments (less than 2μm/y).**

Fig. 3. Role of argumentation models in developing the safety case knowledge base

Fig. 4. Representing the safety case as an argumentation model in order to structure the KB

Arguments based on knowledge of geological environment

"Established" knowledge-base

Evidence "Deep" knowledge

> Plan Progress

materials Experience &

**Counter Defence Counter Defence**

・・・

・・・

1

Working memory

Plan

Knowledge production/use cycle

See Do

・・・

・・・

**Creation of Knowledge Sets: Structuring knowledge based on a general concept of the** 

**safety case**

Arguments based on disposal concept and repository design

Arguments based on understanding

of system behaviour

Fig. 3. Role of argumentation models in developing the safety case knowledge base

Data Document Software Synthesis Guidance Presentation

Argumentation network model

methodology

**Safety case at a given stage in disposal system planning and development Purpose and context of the safety case**

**Requirements**

**Claim of Safety**

**Safety strategy**

Siting and design strategy Management strategy Assessment strategy

**Assessment basis**

Scientific and technical information and understanding

System concept Methods, models,

computer codes and database

Fig. 4. Representing the safety case as an argumentation model in order to structure the KB

**R&D needs Knowledge manipulation**

**Sub-claim**

**Sub-claim**

**Sub-claim**

**Evidence, analyses and arguments** 



**Synthesis**

Key findings and statement of confidence *vis-à-vis* purpose and context

Safety strategy

Claim based on established knowledge Claim based on working hypothesis Counter argument

**Creation of Knowledge Components in Userfriendly form** 

**Components of KB**

Fig. 5. Screenshots of the argumentation model editor (Scarab)

A Challenge on Development of an Advanced Knowledge Management

System (ISIS) (Osawa et al., 2009a; Semba et al., 2009).

reasoning as major constituent elements (Figure 7).

Evidence for decision making

Total management

Site Investigation

knowledge using rule-based or case-based approaches.

Planning of Site Investigation

⇒ Making planning report of site investigation

Construction of Geological Environment Model

⇒ Making explanatory notes of GEM

will occur.

Manager

Person in charge of Safety Assessment and Repository Design

Fig. 7. Basic concept of ISIS

of the following tools:

Repository Design

Evaluation of feasibility of geological disposal

Safety Assessment

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 175

data handling that may be easily automated, much of it requires input of "tacit knowledge" (Nonaka & Takeuchi, 1995; IAEA, 2005), which involves the experience of expert staff. In particular, planning and managing the characterisation programme results in challenges, due to the inherent uncertainty in site understanding and the inevitable surprises that

To provide support for NUMO, which may need to run several field programmes in parallel - and also the regulator, which is expected to follow these and provide input for key decisions - JAEA is attempting to capture both Japanese and international geosynthesis experience within a KMS component, termed the Information Synthesis and Interpretation

ISIS is being developed by applying advanced electronic information technology and knowledge engineering approaches; it will include an extensive knowledge base, expert systems utilising an inference engine and an archive for rule-based and/or case-based

Information Synthesis and Interpretation System (ISIS)

ISIS Management Cockpit ・ Visualisation of GEM

Planning staff

⇒ Making working paper *ES for Site Investigation*

Although the Japanese knowledge base will need to be complemented by input from international partner programmes with wider practical experience, all indications to date suggest that development of such an intelligent system is feasible with existing technology. Based on the requirements and goals specified above, ISIS has investigated implementation

1. Expert system (ES) development tools: particularly focused on capturing tacit

ES: Expert System *Plan* Site investigation staff

ES Rule-based reasoning

Case-based reasoning

*ES for construction of GEM*

Modelling staff

> *ES for development of Site Investigation*

・ Supporting tasks in regard to geological investigations with ES ・ ES development interface

・ Documentation support tool ・ Problem-solving tool (e.g. Evidential Support Logic, Analytic

Hierarchy Process)

・ Progressing

Fig. 6. Illustration of a knowledge note

#### **3.2 Support tool/method for geosynthesis process**

The characterisation of potential repository sites is one of the most resource-intensive and politically-sensitive tasks facing the Japanese geological disposal programme. This work will process huge volumes of information that must be corrected, quality assured, integrated, analysed, documented and archived in a rigorous and efficient manner, a process often referred to as "geosynthesis".

A geosynthesis methodology has been developed to facilitate integration of site characterisation information flow, incorporating feedback from design and PA users. Trial application of this approach is now ongoing within JAEA studies at two generic underground research laboratory (URL) sites in order to synthesise the huge amount of practical experience and data into a consistent geological environment model (GEM) . The methodology of site characterisation has evolved from initial studies of simple geosynthesis data flow diagrams, which traced how measurements during site investigation generated data sets for end-users in a transparent and quality-assured manner.

A particular feature of JAEA's activities involves R&D in two URL projects: one at Mizunami (Saegusa & Matsuoka, 2010), central Japan, in crystalline rock and the other at Horonobe (Ota et al., 2010), north of Japan, in sediments. These URLs are generic research facilities and thus distinct from site-specific underground facilities that will be constructed by NUMO at volunteer sites, during the detailed investigation stage of site characterisation (NUMO, 2004).

If the experience and know-how obtained in the URL projects is to be used at other sites (for example by NUMO), it is necessary to have flexibility to respond to considerable differences in boundary conditions. While some of the site characterisation tasks involve rather routine

・The amount of weight loss, and the amount of the hydrogen emergence has got bigger as the

・The corrosion rate which corresponds to the hydrogen emergence amount is about 30% of the

・The sum up of the corrosion rate obtained by the reduce of the 'Fe(III)/Fe(II)' ratio & the corrosion ratio obtained by the hydrogen emergence is almost the same as the corrosion rate

・As Table1 shows, when magnetite corrosion occurs, the increase of the average corrosion depth is 0.26mm (2.05mm-1.79mm), and the increase of the maximum corrosion depth is 1mm (12.8mm-11.8mm). Even if we set the ration of Fe(III) reduction reaction & the hydrogen emergence reaction as 5:5, the increase of the maximum corrosion depth is about 2mm. In addition, against the existing max corrosion depth '32mm', the corrosion allowance '40mm' has the room of '8mm', so we can conclude that the magnetite behavior doesn't effect to overpack longevity. Fig.1 Schematic of experimental procedure using the glass ampoules

[1]N. Taniguchi, Effect of Magnetite as a Corrosion Product on the Corrosion of Carbon Steel Overpack, European Federation of Corrosion Publications NUMBER 36, Prediction of Long Term Corrosion Behaviour in Nuclear Waste Systems Proceedings of an International

The characterisation of potential repository sites is one of the most resource-intensive and politically-sensitive tasks facing the Japanese geological disposal programme. This work will process huge volumes of information that must be corrected, quality assured, integrated, analysed, documented and archived in a rigorous and efficient manner, a

A geosynthesis methodology has been developed to facilitate integration of site characterisation information flow, incorporating feedback from design and PA users. Trial application of this approach is now ongoing within JAEA studies at two generic underground research laboratory (URL) sites in order to synthesise the huge amount of practical experience and data into a consistent geological environment model (GEM) . The methodology of site characterisation has evolved from initial studies of simple geosynthesis data flow diagrams, which traced how measurements during site investigation generated

A particular feature of JAEA's activities involves R&D in two URL projects: one at Mizunami (Saegusa & Matsuoka, 2010), central Japan, in crystalline rock and the other at Horonobe (Ota et al., 2010), north of Japan, in sediments. These URLs are generic research facilities and thus distinct from site-specific underground facilities that will be constructed by NUMO at volunteer sites, during the detailed investigation stage of site characterisation

If the experience and know-how obtained in the URL projects is to be used at other sites (for example by NUMO), it is necessary to have flexibility to respond to considerable differences in boundary conditions. While some of the site characterisation tasks involve rather routine

The corrosion rate increases because the excess of trivalent iron works as an oxidant.

If this trivalent iron will be consumed, the corrosion rate reduces.

which is calculated from the amount of weight loss (Fig.3).

Workshop, C adaarche, France, pp.424-438 (2003).

・Error cause & the uncertainty should be considered

**3.2 Support tool/method for geosynthesis process** 

data sets for end-users in a transparent and quality-assured manner.

corrosion rate which is calculated from the amount of the weight loss. ・The ratio of 'Fe(III)/Fe(II)' in the magnetite has reduced after the test.

Argument(based on experimental data)

area magnetite density is bigger (Fig.1).

The abstract of the evidence [1]

References

[2] ・・・

(NUMO, 2004).

Complementary information ・Experimental conditions ・The background scenarios

Fig. 6. Illustration of a knowledge note

process often referred to as "geosynthesis".

data handling that may be easily automated, much of it requires input of "tacit knowledge" (Nonaka & Takeuchi, 1995; IAEA, 2005), which involves the experience of expert staff. In particular, planning and managing the characterisation programme results in challenges, due to the inherent uncertainty in site understanding and the inevitable surprises that will occur.

To provide support for NUMO, which may need to run several field programmes in parallel - and also the regulator, which is expected to follow these and provide input for key decisions - JAEA is attempting to capture both Japanese and international geosynthesis experience within a KMS component, termed the Information Synthesis and Interpretation System (ISIS) (Osawa et al., 2009a; Semba et al., 2009).

ISIS is being developed by applying advanced electronic information technology and knowledge engineering approaches; it will include an extensive knowledge base, expert systems utilising an inference engine and an archive for rule-based and/or case-based reasoning as major constituent elements (Figure 7).

Fig. 7. Basic concept of ISIS

Although the Japanese knowledge base will need to be complemented by input from international partner programmes with wider practical experience, all indications to date suggest that development of such an intelligent system is feasible with existing technology.

Based on the requirements and goals specified above, ISIS has investigated implementation of the following tools:

1. Expert system (ES) development tools: particularly focused on capturing tacit knowledge using rule-based or case-based approaches.

A Challenge on Development of an Advanced Knowledge Management

within tasks of site

site investigation was solved in the past, to suggest ways to handle similar problems in the

investigation, represented using IF...THEN format

Rule-base Decision-making rules

Case-base Cases how a problem in

future

Fig. 9. Knowledge modelling and ES development

and resulting output.

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 177

a. Identification of relevant knowledge elements

b. ES development interface for the rule-base

knowledge management technologies and tools should increase efficiency, traceability, transparency and repeatability of such PA tasks (Makino et al, 2009a). They will also ease the process of auditing contents of PA and support QA of the input, analytical methodology

e.g. Selection of drilling methods

If full core recovery needed

be selected (tricon bit etc.)

be selected

YES

NO

e.g. Troubleshooting of drilling fluid loss

Then a wireline drilling method should

then a non-core drilling method should


The first stage of ES development involves a formal process of knowledge acquisition. This can be illustrated as a detailed task flow diagram resulting from expert elicitation. An example is shown in Figure 8, which shows the sequence of tasks for establishing on-land seismic reflection surveys resulting from interviews with experts.

The next stage involves knowledge modelling; carrying out syntax analysis of the component rules and cases (Figure 9a). This allows representation of knowledge in a form that is accessible to methods from the field of knowledge engineering.

The final stage is creation of the operational ES; this involves establishing a user interface based on established templates (Figure 9b) and formulating knowledge elements as a menu, based on the results of the syntax analysis. The interface is used to create the rule-based and case-based procedures, by selecting appropriate knowledge elements and syntax. A knowledge engineering tool then produces the ES automatically, based on this formalised input.

Fig. 8. An example of flow of elicitation of task knowledge

#### **3.3 Support tool/method for performance assessment**

Regarding total system performance assessment, it was recognised that routine tasks in PA, e.g. development of input datasets, groundwater flow and transport modelling, interpretation of model output, integration within a total system PA, etc., are repeated many times. This may occur whenever there are changes in assessment scenarios, geological environment models, repository design, relevant databases, etc. Introduction of advanced

3. Problem-solving methodology: formal approach for identifying and resolving

The first stage of ES development involves a formal process of knowledge acquisition. This can be illustrated as a detailed task flow diagram resulting from expert elicitation. An example is shown in Figure 8, which shows the sequence of tasks for establishing on-land

The next stage involves knowledge modelling; carrying out syntax analysis of the component rules and cases (Figure 9a). This allows representation of knowledge in a form

The final stage is creation of the operational ES; this involves establishing a user interface based on established templates (Figure 9b) and formulating knowledge elements as a menu, based on the results of the syntax analysis. The interface is used to create the rule-based and case-based procedures, by selecting appropriate knowledge elements and syntax. A knowledge engineering tool then produces the ES automatically, based on this formalised

> 1-1. Study of existing information 1-2. Setting for survey object

1-4. Check to the line situation

1-5. Planning of the investigation parameters

detailed task

1-6. Create the investigation protocol

2-1. Set the line and elevation survey

3-1. Preparation of acquisition data

3-3. Preparation of other survey data

4-1. Interpretation and synthesise with other

3-2. Data processing

geophysical data

Regarding total system performance assessment, it was recognised that routine tasks in PA, e.g. development of input datasets, groundwater flow and transport modelling, interpretation of model output, integration within a total system PA, etc., are repeated many times. This may occur whenever there are changes in assessment scenarios, geological environment models, repository design, relevant databases, etc. Introduction of advanced

2-2. Acquisition parameter test 2-3. Recording activities

1-3. Setup survey line

2. Management cockpit: integration of ISIS tools and communication tool.

seismic reflection surveys resulting from interviews with experts.

detailed task

detailed task

detailed task

detailed task

Fig. 8. An example of flow of elicitation of task knowledge

**3.3 Support tool/method for performance assessment** 

that is accessible to methods from the field of knowledge engineering.

conflicting requirements.

input.

1. Prior preparation

2.Investigation

3.Processing

4.Interpretation


#### a. Identification of relevant knowledge elements


b. ES development interface for the rule-base

Fig. 9. Knowledge modelling and ES development

knowledge management technologies and tools should increase efficiency, traceability, transparency and repeatability of such PA tasks (Makino et al, 2009a). They will also ease the process of auditing contents of PA and support QA of the input, analytical methodology and resulting output.

A Challenge on Development of an Advanced Knowledge Management

Easy access and operation with browser

**≈**

**Table(s)**

**Figure(s)**

**3.4 CoolRep** 

**≈**

**An example of e-PAR on web**

**Do you want to reanalysis with different value(s)? Yes!**

different disciplines

Main features and advantages of e-PAR in practical viewpoints are:

Recording all the changes in a systematic manner for future reference

static, one-way and, to some extent, user-unfriendly.

System (KMS) for Radioactive Waste Disposal: Moving from Theory to Practice 179

 Easy execution of PA tasks in an easy and understandable operation, which allow application of PA tools by non-PA experts and expedite communication among

 Storing context and procedure of PA tasks with all relevant data, information and knowledge (both domain knowledge and task knowledge) in an easily accessible format

Dynamic, interactive and user-friendly, while a conventional paper/electronic report is

**Editing text on screen**

**parameter value(s) Browsing text Text on screen**

・ **Comparison and analysis of outputs** ・ **(if needed) e-PAR** 

**Specification of relevant operator and/or parameter set** **Reanalysis**

**Modification of** 

**Representation of output (graph, table)**

**Specification of output format**

**updating,**

As indicated in Figure 12, at a top hierarchical level the key tools utilised are, in addition to

'CoolRep' is an advanced, internet-based approach to management of documentation and providing an interface with users – both technical and non-technical. It allows the vast volumes of relevant information to be presented in a user-friendly manner, with different access options for different stakeholder groups. Since March 2010, the CoolRep 2010 version in Japanese language is available (http://kms1.jaea.go.jp/CoolRep/) and demonstrates the capability of JAEA (in the future, together with other Japanese R&D organisations) to support production and review of safety cases for deep geological disposal (Figure 12). An

Fig. 11. Flow of functions and operations on e-PAR (browsing, editing and reanalysis)

Scarab for argumentation model development, CoolRep and a smart search engine.

The overall toolkit under development is termed PAIRS (Performance Assessment Integrated Report System) (Figure 10). A significant feature of the PAIRS is that this comprises a linked set of not only text, tables and figures corresponding to conventional PA report but also applications that have been used through PA tasks. The key components of this system are as follows:


Fig. 10. Schematic overview of PAIRS and component tools

Figure 11 shows an example of e-PAR and outline of operation on e-PAR (browse, edit and reanalysis). Functions of operator library and calculation management on e-PAR are currently being tested with an existing post-closure PA code (for example, a code based on GoldSim (Golder Associates, 2001)), but will be flexible enough to incorporate other existing codes and next generation models/databases when they arise.

Main features and advantages of e-PAR in practical viewpoints are:

Easy access and operation with browser

178 New Research on Knowledge Management Technology

The overall toolkit under development is termed PAIRS (Performance Assessment Integrated Report System) (Figure 10). A significant feature of the PAIRS is that this comprises a linked set of not only text, tables and figures corresponding to conventional PA report but also applications that have been used through PA tasks. The key components of

 Electronic Performance Assessment Report (e-PAR): the electronic report developed with support of PAIRS, which contains a core of quality-assured text, tables and figures linked to calculation by encapsulated tasks for a specific dataset (e-PAR case-base

Operator: an encapsulated application library that includes numerical tools and

 User interface: an intelligent user interface facilitates reconfiguring the input dataset and encapsulated tasks, management of calculations and then editing the resultant output to create/modify contents of e-PAR; this will be supported by change management tool to provide a top-level overview of all changes implemented, their

 Application knowledge base: a central library of all PA data (both raw and derived) together with all relevant information supporting their selection and application,

> **Text Tables (Input) Operators (Calculations) Figures (Output) Text . . .**

elements in e-PAR User interface :

text in e-PAR User interface:

Figure 11 shows an example of e-PAR and outline of operation on e-PAR (browse, edit and reanalysis). Functions of operator library and calculation management on e-PAR are currently being tested with an existing post-closure PA code (for example, a code based on GoldSim (Golder Associates, 2001)), but will be flexible enough to incorporate other existing

*e-PAR* e-PAR case-base

*background knowledge*

Knowledge-base *PA data*

Editing of

*background knowledge*

together with an associated historical record of all changes and modifications.

assessment softwares used for certain tasks in a user friendly form

justification and their consequences in terms of overall performance

Operator library *numerical tools, assessment software*

Fig. 10. Schematic overview of PAIRS and component tools

codes and next generation models/databases when they arise.

*Operators Input dataset etc.,* 

User interface: Configuration of

*Management of execution and input/output handling* 

*PAIRS*

Management of calculations in e-PAR

this system are as follows:

means a library of developed e-PARs)

