`SAIC-95/1028
`Vol. 3
`
`
`
`Guidelines for the
`Verification and Validation of
`Expert System Software and
`Conventional Software
` RecelvE>
`‘OST
`
`Survey and Documentation of Expert System
`Verification and Validation Methodologies
`
`
`
`Prepared by
`E. H. Groundwater, L. A. Miller, S. M. Mirsky
`
`Science Applications International Corporation
`
`Prepared for
`U.S. Nuclear Regulatory Commission
`
`and
`
`Electric Power Research Institute
`
`.
`'
` SSee
`
`CONFIGIT 1033
`
`
`DISTRIBUTION OF THIS DOCUMENT 1S UNLIMITED
`
`CONFIGIT 1033
`
`1
`
`
`
`AVAILABILITY NOTICE
`
`Availability of Reference Materials Cited in NRC Publications
`
`Most documents clted In NRC publications will be avaliable frorn one of the following sources:
`
`1.
`
`The NRC Public Document Room, 2120 L Street, NW., Lower Level, Washington, DC 20555-0001
`
`The Superintendent of Documents, U.S. Government Printing Office, P. O. Box 37082, Washington, DC
`20402-9328
`
`The Natlonal Technical Information Service, Springtleld, VA 22161-0002
`
`Although theIlsting that follows represents the majority of documents cited in NAC publications, it is not In-
`tended to be exhaustive.
`
`Referenced documents avallable for Inspection and copying for a fee from the NAC Public Document Room
`include NRC correspondence and Internal NRC memoranda; NRCbulletins, circulars, information notices, In-
`spection andInvestigation notices; llcensee event reports; vendor reports and correspondence; Commission
`Papers; and applicant and llcensee documents and correspondence.
`
`The following documents In the NUREGseries are available for purchase from the GovernmentPrinting Office:
`formal NRC staff and contractor reports, NRC-sponsored conference proceedings, international agreement
`reports, grantee reports, and NAC booklets and brochures. Also available are regulatory guides, NRC regula-
`tlons In the Code of Federal Regulations, and Nuclear Regulatory Commission Issuances.
`
`Documents avallable fromm the National Technical Information Service Include NUREG-serles reports and tech-
`nical reports prepared by other Federal agencies and reports prepared by the Atomic Energy Commission,
`forerunner agency to the Nuclear Regulatory Commission.
`
`Documents available from public and special technicallibraries include all open literature items, such as books,
`Journal articles, and transactions. Federal Register notices, Federal and State legislation, and congressional
`reports can usually be obtained from theselibraries.
`
`Documents such as theses, dissertations. foreign reports and translations, and non-NRC conference pro-
`ceedings are avallable for purchase from the organization sponsoring the publication cited.
`
`Single coples of NAC draft reports are available free, to the extent of supply, upon written requestto the Office
`of Administration, Distribution and Mall Services Section, U.S. Nuclear Regulatory Commission, Washington,
`De 20555-0001.
`
`by such third party would not infringe privately owned rights.
`
`Coplesof industry codes and standards used In a substantive mannerIn the NRC regulatory process are main-
`tained at the NAC Library, Two White Flint North, 11545 Rockville Pike, Rockville, MD 20852-2738, for usa by
`the public. Codes and standards are usually copyrighted and may be purchased from theoriginating organiza-
`tlon or, If they are American National Standards, from the American National StandardsInstitute. 1430 Broad-
`way, New York, NY 10018-3308.
`
`DISCLAIMER NOTICE
`
`This report was prepared as an account of work sponsored by an agencyof the United States Goverment.
`Neitherths United States Governmentnorany agency thereof, norany oftheir employees, makes any warranty,
`expressed or implied, or assumes any legalliability or responsibility for any third party’s use,or the results of
`such use, ofany information, apparatus, product, or processdisclosedin this report, or represents thatits use
`
`2
`
`
`
`DISCLAIMER
`
`
`
`Portions of this document maybeillegible
`in electronic image products.
`Images are
`produced from the best available original
`document.
`
`3
`
`
`
`NUREG/CR-6316
`SAIC-95/1028
`Vol. 3
`
`Guidelines for the
`Verification and Validation of
`Expert System Software and
`Conventional Software
`
`Survey and Documentation of Expert System
`Verification and Validation Methodologies
`
`Manuscript Completed: February 1995
`Date Published; March 1995
`
`Prepared by
`E. H. Groundwater, L. A. Miller, S. M. Mirsky
`
`Science Applications International Corporation
`1710 Goodridge Drive
`McLean, VA 22102
`
`Prepared for
`Division of Systems Technology
`Office of Nuclear Regulatory Research
`U.S. Nuclear Regulatory Commission
`Washington, DC 20555-0001
`NRC Job Code L1530
`
`Palo Alto, CA 94303
`
`and
`
`Nuclear Power Division
`Electric Power Research Institute
`3412 Hillview Avenue
`
`DISTRIBUTION OF THIS DOCUMENT IS UNLI
`
`MITED
`
`4
`
`
`
`5
`
`
`
`ABSTRACT
`
`This report is the third volumein the final report for the Expert System Verification and Validation (V&V)
`project which was jointly sponsored by the Nuclear Regulatory Commission andthe Electric Power Research Institute.
`The ultimate objective is the formulation of guidelines for the V&V ofexpert systems for use in nuclear power
`applications. The purpose ofthis activity was to survey and document techniques presently in use for expert system
`V&V.
`
`The survey effort included an extensive telephone interviewing program,site visits, and a thorough
`bibliographic search and compilation, The majorfinding was that V&V ofexpert systems is not nearly as established or
`prevalent as V&V ofconventional software systems. When V&V was used for expert systems, it was almost always at
`the system validation stage after full implementation andintegration usually employing the non-systematic dynamic
`method of“ad hoc testing." There were few examples ofemploying V&Vin the early phases of developmentand only
`weak sporadic mention ofthe possibilities in the literature. There is, however, a very active research area concerning
`the developmentofmethods andtools to detect problemswith, particularly, rule-based expert systems. Four suchstatic-
`testing methods were identified which were not discovered in a comprehensive review of conventional V&V methods in
`an earlier task.
`
`
`
`6
`
`
`
`7
`
`
`
`TABLE OF CONTENTS
`
`ABSIRAGT)
`
`ih boi cxanSeeeri eae aie ais AER PERO ST ICG EF PEERED EER RRTSE VEDEYZCLEPIIT EC ETE ESAS OE a
`
`iii
`
`EXECUTIVESUMMARY 0.555 oFeesp ns ascaeesSoeauloeabontercdedavera¥ers eopnaspeeurgeseaneee
`
`ix
`
`TOUINTRODUCTION: <5 <5; csv sewwhw ate asa pg ad eo aviscpe bey et es CACBLE SS EERE RPV NGGurapaeeeg estes OFe
`1,1
`Purpose and Scope ofActivity? .......... 00000-2222 eee eee eee eee eee eee eee
`1,2
`Reniort Orennization fv
`.lo sac 2slawise, estes, xalee dnp cdelein dale tte clans ae godt alsawioela's lass etafeatste’s
`
`1
`1
`1
`
`2,0 “TECHNICAL APPROACH iu cctes os os clgactae ere eh wousiare tcaeeet risk iee ss deesticaWgacegan 3
`2.1
`Overall Approach ......... 2.00 ee eee ee eee eee pening ota species ely cnietewew wg tend ewe
`3
`2.2
`TelephoneInterviews and Data Collection ..............2..2222 22222222 e cece e eee ees
`3
`2.3
`Site Selection arid: Vists:
`. ca. vc ve ve ee see epee opeis e ee eeg we ees cabs ea eaeeereeernres
`12
`2.4
`Technique Characterization and Analysis .......--.......---2+2:2e0ceeec eee cece eee eeee
`17
`
`3,0 REFERENCE LIFECYCLE FOR THIS ACTIVITY -...2....-..202sc0cscssee eee ee anseeeeeeeenaues
`
`23
`
`25
`4.0 EXPERT SYSTEM V&V TECHNIQUE DESCRIPTIONS...- 2... cc en ecceceseeeeeeeeeteneeenceeees
`4.1
`Types oF Tedhiiqnes 4 fcc cg os ds 6S ee acess da eed Sa Neda hada ne edaaaigenbeebeaeaces 25
`4.2
`Requirements and Design Testitig <. 2. sccade daca ctvteagedadadadavaruncrvesereseubaes
`25
`43
`Dlatic POSich $6955 See Gh SEES oa lta etayae 4s coo sagadap hy aed peeps ease exe ond hone 29
`44
`Beynarine Testing jt pssciata!le gp ssc gle siete stellate eatab este feleralelelateletatale a tatale ofaralalatarsla cla altattt 40
`4.5
`The Stite-omihe-Arb sig: caccs nei hehe gigi tiniekii e555 2596 cee tases 53
`
`SU AUTOMATED TOOUS 252565 25.5596 28 pag SDSS ake e re cece enh hrh preter CRS Guba Neen ys 55
`5.]
`Syntax Checking Toole’: 325 08353 ¢iscgitiaieia ese bigad ai miatawke b2 aah P aie GRA eee 55
`3,2
`Semantic Checking TOs 6.40: schoo ee aed Sa aa as Eade Rake Gee EE 63
`5.3
`Knowledge Acquisition/Refinement Tools ............-.-.+-2---+-+-+-- picnae theta leas 64
`5.4
`Intelligent Compilers ............2.2-22002 00 ude va ates we deta Mamdeagaclcwere sees clea 65
`aA
`Other Dynamic Testing Tools ......-.2. 20.2222 eee eee eee eee eee 65
`5.6
`Subnary On AMA TOS ss on Ss ses ce er epee een Sees pie Grab Maceuenswar rads 66
`
`6,0. CATEGORIZATION AND ANALYSIS OF TECHNIQUES AND TOOLS .......-.--.....----------- 67
`6.1
`Components ofExpert Systems .....-- 2-20.22 222202 eee eee eee ee eee phatclesssre tects Siecle ibe ss 67
`6.2
`Expert System Faults: Anomalies or Invalidities ..... 2.2.2. 22-22. cece ee eee cece 70
`
`TO CONCLUSIONS 2 se: c2 55543555 AS OEE eRe ge eee eee al aes ee aa eeaaragi aes
`
`75
`
`APPENDIX A.BIBLIOGRAPEDY siecle caer eens eelsca dor acivlnlnetne dba stwacsesedikeaaladadnas 77
`
`v
`
`
`
`8
`
`
`
`List of Figures
`
`Figure ES-1
`
`Classes of V&V Techniques Which Have Been Applied to
`Conventional Systems: 2025 c522 pees ese Ses 4 SSUES Lak Sola asco la saad a cwawres x
`
`Figure 2.1-1
`
`USNRC/EPRI Expert System V&V Project
`Detailed Activity 2 Work Plan ....... 02... c cece cece rete cnet tenet tence net etege a
`
`Figure 2.2-1
`
`USNRC/EPRI V&V Interview Questionnaire .... 2.2.2.0. eee nee 9
`
`Figure 2.2-2
`
`Survey Form: Nuclear Industry Expert Systems
`that Have Been Tested . 0.0.6... cece cece eer e ee eee eee eeeee Te eee ETT eT Eee Te ee 13
`
`Figure 2.2-3
`
`Activity 2 Telephone Interviews .........----- eee ee cee ce een eet c eect heen es eebees 14
`
`Figure 2.4-1
`
`Tools/Techniques Worksheet
`
`.. 2.2.2.5 eee eee ee eee eee eee eens 20
`
`Figure 3.0-1
`
`Relationship of V&V Activities to
`Generio-Projedt Activites c..5.4 5.5 hides hea 24 cae ed bal ta troedeebsaSebamediy nd Eaatnygde we aan 24
`
`Figure 4.1-1
`
`Classification of V&ViTesting: Techniques 35s iaec iad aecdessdceada dedceasaveransias 26
`
`Figure 6.2-1
`
`Comprehensive Expert System V&V Matrix ...........:20s cece ce ete e eee eee ee eeeeeseees 71
`
`vi
`
`9
`
`
`
`List of Tables
`
`Table 2,2-1]
`
`Persons Contacted for Telephone Interviews ..........6.066 5.6 sce cc cece eee eee ete 5
`
`Table 2.3-1
`
`Preliminary Site List for Visit Evaluation ...........-......2.. 02-202 eee eee ee eee eee eee evils
`
`Table 2,3-2
`
`Site Selection Crtferia:s oyds ger cges aes g ogee essa eee esac Sch nb bb nase pues 16
`
`Table 2.3-3
`
`Relative Site Rankings UsingSite Selection
`OHS onic h anki ca dew ae sea ca nd dane shad Heed Mae os een eyed tip sis gapigneSenmeanie 18
`
`Table 2.3-4
`
`Recommended Site Visits Based on Site
`Selection Criteria Ranking... 25.522 -nyvevwwyecey eyes he cece st er etre ep eaves erenegees 19
`
`Table 4.2-1
`
`Table 4,3-]
`
`Requirements and Design V&V Techniques
`Applied ty Expert Systems: Lo. 52266 6h GaSS Gash Ses F epee par eeaesradsedvereugeasecau' 27
`
`Static Testing V&V Techniques Applied to
`Bxmert Sostetnis: wails oauu teu 2co%2 pet pa betee is peAineaed eeatatea dena sans dels 30
`
`Table 4.3-2
`
`Types of Syntactic Errors Found by Automatic
`Rule Base Syntax Checkers, 0... cc inccccecee eco becececneubectbesagaceaaucsmangueeens 39
`
`Table 4.4-1
`
`Dynamic Testing V&V Techniques Applied to
`Expert Systeme i. (cisici gaa Sige sete ee iveeeiea oe ee caveewteiads rates cia aia eada. 4)
`
`Table 5.0-1]
`
`Description ofAutomated Tools for Expert
`iE Vase CES, Lides cds ple Baas Gees lees Gucacees recarseaeSeaneranescels Oaks pane 56
`
`Table 6.1-1
`
`Components and GeneralFeatures ofKnowledge
`Based Systems | OP oi coisas ca we ea aa bis Dees Graie eis pie Cape's bas Coe a vowel we ce bane tee 68
`
`Table 6,1-2
`
`Components and Typical Testing-Related
`Features ofKnowledge-Based Systems, with
`Testing Rectoririendations',.62525%.0etadcecdencdens ae cetapanecenteeaerbaee bane dave 69
`
`vii
`
`10
`
`10
`
`
`
`11
`
`
`
`EXECUTIVE SUMMARY
`
`This report is the third volumein the final report for the Expert System Verification and Validation (V&V)
`project, which wasjointly sponsored by the Nuclear Regulatory Commission (USNRC)andthe Electric Power Research
`Institute (EPRI), Theultimate objectiveis the formulation ofguidelines for the V&V ofexpert systemsfor use in
`nuclear power applications. The purpose ofActivity 2 was to survey and documenttechniques for expert system V&V.
`The survey used the results ofActivity |, a survey oftechniques for conventional software V&V,to determine which of
`these techniques are being applied to expert systems, and what new techniques have been developed solely for expert
`system V&V.
`
`The survey effort included: 1) an extensive telephone interviewing campaign to over 130 points of contact, 2)
`site visits to nine institutions conducting research in or applying expert system V&V,and 3) the collection of an
`extensivelibrary ofwell over 300 bibliographic references. The survey encompassed work doneboth within the nuclear
`powerindustry and in other industries as well. Contacts included corporations, universities, governmentagencies, and
`utilities, Within the last four to five years, there has been an explosive growth ofinterest and work in the field. It has
`nowreached a level ofmaturity where expert system V&V techniques are being implemented in automated tools and
`being applied to operational expert systems development and maintenanceefforts.
`
`As can be seen in Figure ES-], many of the classes of V&V techniquesidentified in Volume 2 as being applied
`to conventional software systems are also being researched for, or applied to, expert systems. Thisis particularly true in
`the areas ofStatic Testing (tests performed directly on the codeitself) and Dynamic Testing (tests performed by running
`the code and evaluating the results). Fewer formal techniques are applied during the Requirements and Design phases
`ofexpert systems development(only five outoften possible methods) and then only infrequently. This is primarily
`because the activities performed during these phases for expert systems are usually informal themselves. Requirements
`and Design documents for expert systems are often not written atall, or written after-the-fact, and thus cannotbe used as
`a basis for V&V activities. When they are written, usually no more is done with them than to review them and,possibly,
`trace requirements to design elements.
`
`Fifteen of 58 possible Static Testing techniques were researched or applied for expert systems (including four
`new ones). Most of the workin Static Testing of expert systems has focused on the development of automated toolsto
`perform sophisticated syntactic checking ofrule bases. Thetypesoferrors that may be found by such checkers include
`redundant or subsumed rules (one rule's conditions are a subset of another's), rule cycles (there is a path from a rule back
`to itself), unreachable or dead-end mules, inconsistent rules, and incompleteness (e.g., not all possible input values are
`covered), Someofthe mule base checkers will perform semantic checks ofthe rule base using meta-constraints defined
`by the programmer, and others will perform checking on thefly during knowledge acquisition and/or refinementofthe
`rule base. Other work in Static Testing has included conducting various kinds ofinspections (e.g., structured walk-
`throughs and expert panel reviews), performance ofdependency analyses ofthe output valueson the inputs, and
`attempts at applying program proving techniques. A point ofview becoming strongly accepted is that it may not be as
`vital to provethat a safety-critical expert system is totally error-free as it is to prove thatifit fails, it will not fail badly
`(i.e., compromise safety).
`
`In Dynamic Testing,there is a wide range of activities: 38 of67 techniques have been researched or applied to
`expert systems. The state-of-the-art in the operational expert system worldis still Ad Hoc Testing, or defining test cases
`at whim, with no systematic guidance. Newer work has focused on more systematic methods for specifying test case
`sets, such as Structural Testing (attempting to cover all mules or rule pathsin the expert system), Random Testing
`(attempting to covera representative sample ofthe possible inputs), and Performance Testing (to assure timing,
`memory, and other constraints are met). Some operational expert systems, such as those developed for safety-related
`
`ix
`
`12
`
`12
`
`
`
`JBEIjSHe}S/[P419Usy
`
`
`
`Hurjsayjeuoyoun.
`
`
`
`Buysayonsyeay
`
`
`
`sishjeuyJosjuog
`
`
`
`sishjeuye}eq
`
`
`
`sishjeuywyywobly
`
`sjuswenbay
`
`
`
`UOHeSHUS//eOjewWayjey
`
`
`
`BUISS8D04obenbue’7sjuawasinbay
`
`sisfjeuy
`
`aBenbHue’
`
`
`Buljsalssaijs
`
`
`(mau)siskjeuyyoajeq
`
`
`
`BuseyAouajadwo5g
`
`
`
`Bunseluojnnsexy
`
`
`
`Burjsayaoejiazu]
`
`
`
`Buyjsayjeanjonsjs
`
`
`
`Bursa,volonpodjufsouzy
`
`suonoadsu}
`
`
`
`
`
`aben6ue7uoijduosagwesbolg
`
`
`
`malAsyUBisagjeuuo4
`
`
`
`uoveinwigubiseg
`
`
`
`
`
`sisfjeuymoj4/Bulull,LOUD
`
`
`
`
`
`sishjeulyeoueljdiwoyubiseq
`
`
`
`Buldes,sjuawisiinbay
`
`
`BuijsaldoUeWIOLIdd
`
`sisfjeuydinjesjjne4
`
`
`MAalABYSJUaWaIINbayjeULIO4
`
`13
`
`
`
`
`
`swiaysAgpadxgJojpasnusaqaaeysanbiuysespjog{pusba7)
`
`
`
`
`
`
`
`
`
`
`
`
`
`sulajsASyadxzjojpasnuaagaaeysanbiuyoa}Jey)aouapiAeONoye)
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`SUl9}SAs[PEUOHUSAUOS0}paljddeuaeqsAeYYIMSenbiuYyd9}AEAJOSesseiOL-Syoni,
`
`
`
`13
`
`
`
`
`functions (e.g,, NASA space shuttle diagnostics), do undergo various formsofRealistic Testing using scenariofiles,
`simulators, or actualfield conditions. Lastly, there are a few automated tools to support generating, managing, or
`scoring test cases.
`
`Uponanalysis ofthe V&V techniques being applied to expert systems, it was found thatthere is sufficient
`coverage across all the components ofexpert systems andacross all error types(static vs. dynamic, anomalies vs.
`invalidities), The challenge is in selecting the appropriate combination of techniques to use for performing V&V ona
`particular expert system that is both effective and cost efficient.
`
`
`
`
`
`14
`
`14
`
`
`
`1, INTRODUCTION
`
`This report is the third volumein the final report for the Expert System Verification and Validation (V&V)
`project. The ultimate objective is the formulation of guidelines for V&V ofexpert systems for use in nuclear power
`applications. This workis jointly sponsored by the Nuclear Regulatory Commission (USNRC)and the Electric Power
`ResearchInstitute (EPRI).
`
`1.1
`
`Purpose and Scopeof Activity 2
`
`The purpose ofActivity 2 was to survey and document techniques for expert system V&V. This report is a
`companion to Volume2 that surveys techniques for conventional software V&V. As will be seen, there is (and should
`be) considerable overlap in the techniques being applied to both types ofsoftware. Thus, this report will reference and
`draw upon the contents of Volume 2 considerably.
`
`The survey included both techniques being applied in the field to operational expert systems andthose being
`researched in Artificial Intelligence (AI) laboratories. With the help ofDr. John Bernard, from the Massachusetts
`Institute ofTechnology (MIT), we surveyed V&V techniques being applied to expert systems for nuclear power
`applications (Bernard & Washio, 1989). However, the survey also encompassed work in otherfields such as space
`operations, manufacturing, military, and otherutilities. We contacted a diverse range oforganizations including
`government agencies and laboratories, universities, contractors and other commercial concerns, and power utilities. We
`attempted to comprehensively cover the work being performed in the United States and opportunistically included work
`done abroad.
`
`As in Activity 1, we covered both lifecycle management and testing techniques, focusing primarily on the
`testing techniques, Again, as in Activity 1, we examined V&V techniques applied to all phases ofthe development
`lifecycle, versus justto the testing phase. Finally, we examined both manual and automated techniques, providing a
`separate description ofdetailed automated tools.
`
`As part ofthe survey effort, nine sites, where work was being performed in V&V ofexpert systems, were
`selected andvisited,
`
`1,2
`
`Report Organization
`
`The next section, 2.0, describes our technical approach to the Activity 2 survey, beginning with a description of
`our overall approach, then telephone surveys andsite visits, followed by a description ofour characterization and
`analysis ofthe techniques. Section 3.0 describes the reference lifecycle to be used for discussing and characterizing the
`techniques found in the survey. Section 4.0 presents a briefdescription ofeach ofthe techniques found. Section 5.0
`describes separately the automated tools for expert system V&V that were found. A categorization and analysis ofthe
`techniques and tools follows in Section 6.0. This is followed by a summary in Section 7.0, which primarily contains
`recommendations for how the Activity 2 results can be applied in subsequentactivities. Appendix A contains the
`Bibliography of materials collected over the course ofthe survey.
`
`15
`
`15
`
`
`
`16
`
`
`
`2. TECHNICAL APPROACH
`
`2.1
`
`Overall Approach
`
`Thedetailed work plan for Activity 2 is shown in Figure 2.1-1. Three main "threads" can be seen in the
`Activity 2 work plan diagram. Thefirst onein the left and lower middle involvestelephone interviewing and reference
`documentcollection, and will be described below in Section 2.2. The second, in the upper middle involves site selection
`and survey, and will be described in Section 2.3. Thelast, on the right involves characterizing and analyzing the
`techniques, and will be discussed in Section 2.4.
`
`2.2
`
`Telephone Interviews and Data Collection
`
`Thefirst step in conducting telephone interviews was to developalist of people to call. Names, addresses, and
`phone numbers ofknowledgeable practitioners and researchers came from a numberofsources throughoutthe activity
`period;
`
`®
`
`®
`
`®
`
`®
`
`Team members'existing professional contacts,
`
`Referrals from Dr. Bemardofpeople involved in nuclear power expert systems development and
`testing,
`
`Authors ofpapers on operational nuclear power expert systems (Artificial Intelligence and other
`Innovative Computer Applicationsin the Nuclear Industry, 1988, EPRI 1989b, 1988a,d, f, 1987d,
`Motoda, 1990, Moradian et al, Nelson,.1989, Osborne, 1986, Proceedings ofthe Intemational
`WorkshoponArtificial Intelligence for Industrial Applications, 1988),
`
`Attendees and speakers at the 1988, 1989, and 1990 AAAI, and IJCAI Workshops on V&V ofExpert
`Systems,
`
`@ Membersofstandards organizations,
`
`@
`
`®
`
`@
`
`Authors ofpapers collected from automated bibliographic search,
`
`Other references and acknowledgements in the papers we collected, and
`
`Referrals from other telephone contacts.
`
`Thelist ofnames was organized into a Point ofContact (PoC) List, which was continuously updated and
`distributed to team members during the activity period. A list ofthe 97 names and organizations ofthe contacts is
`shown in Table 2.2-1,
`
`Interview forms were prepared for collecting information from the telephoneinterviewees. After a few trial
`calls with the first draft ofthe form, it was shortened and modified to the one shown as Figure 2.2-1. The first page was
`followed by a totally blank page, on which answers to the discussion points on the bottom ofthefirst page could be
`transcribed. The Activity 2 team members weretrained in structured interviewing and the use of the form,and the
`
`
`
`17
`
`17
`
`
`
`
`
`
`
`avid42047Qlaqoypapejagpeya0DAVAWeyskgprsdxyHAT/OUNSMI-17sans
`
`
`
`
`
`
`
`eur
`
`yodayyzAwanoy
`
`yeid
`
`yoday
`
`auMLojoy
`
`Ayqiqeoyddy
`
`xXUJeAT
`
`
`
`B@quosaqIsr]
`
`‘sonbrayoay,
`
`
`
`puno.jsjooy,
`
`dojaaoq
`
`anbiuyoa],
`
`wonezuodaye~)
`
`doyaaaq
`
`Aydexsouqua
`
`angdopaasg
`
`Sas199]9g
`
`a[Npayospue
`
`SUSTA
`
`uolDI]9S
`
`eno)
`
`
`
`BE SMOIAIOU]
`
`amoyd
`
`aredaig
`
`MAIAIOIU]
`
`Salevonsang)
`
`SsurpNG
`
`yoday
`
`jeulydn-Mo[joy
`
`
`wpal[oD
`
`aouaIayey
`
`squamEND0q]
`
`18
`
`
`
`18
`
`
`
`
`
`
`
`
`
`
`Table 2.2-1 Persons Contacted for TelephoneInterviews
`
`
`
`| Point of Contact[iiation
`
`Adelman, Leonard
`
`George Mason University
`
`Bahill, Terry
`
`University of Arizona
`
`Bartschat, Steffen
`
`Ultrasystems
`
`Basti, D.W.
`
`Bayse,Al
`
`Bernard, John
`
`Bloom, Howard
`
`Bond, David
`
`Boose, John H.
`
`Bray, Mike
`
`Forschungsgelande (Germany)
`
`Federal BureauofInvestigation (FBI)
`
`MassachusettsInstitute of Technology
`
`NationalInstitute of Standards & Technology (NIST)
`
`SAIC, COMSYSTEMSDivision
`
`Boeing Computer Services
`
`EG&G IdahoInc.
`
`Buchanan, Bruce G.
`
`University of Pittsburgh
`
`Renesselaer Polytechnic Institute
`
`Carbonara, Joe
`
`Chee, Christine
`
`Cohen, Paul R.
`
`Combs, Jacqueline
`
`Cragun,Brian J.
`
`Cross, Steve
`
`Culbert, Chris
`
`Duckworth, Jim
`
`Edwards, Robert
`
`Fausett, Mark
`
`Franklin, Randolph
`
`Consolidated Edison - Indian Point 2
`
`BD Systems,Inc.
`
`University of Massachusetts
`
`Lockheed Missiles and Space Company, Inc.
`
`IBM
`
`Defense Advanced Research Projects Agency (DARPA)
`
`NASA/Johnson Space Center
`
`Worcester Polytechnic Institute
`
`Pennsylvania State University
`
`Rome Laboratory/COES
`
`19
`
`19
`
`
`
`Table 2.2-1 (Continued).
`
`Point of Contact
`
`Freeman, Michael
`
`Friedland, Peter
`
`Fujii, Roger U.
`
`Fussel, Louise
`
`NASA
`
`NASA-Ames Research Center
`
`Logicon, Inc.
`
`Rockwell Space Operations Company
`
`Gabrielian, Armen
`
`Thomson-SCF,Inc./Pacific Rim
`
`Geissman,Jim
`
`Gelperin, David
`
`Garrett, Randy
`
`Gilstrap, Lewey
`
`Ginsberg, Allen
`
`Gowens, Jay
`
`Griebenow, Ronald
`
`Griesmer, James
`
`Hajek, Brian K.
`
`Hamilton, David
`
`Harder, Bob
`
`Harrison, Patrick
`
`Abacus Programming Corporation
`
`Software Quality Engineering
`
`Institute for Defense Analysis
`
`Computer Science Corporation
`
`AT&T Bell Labs
`
`U.S. ArmyInstitute for Research in Management
`Information
`
`NUS Corporation
`
`Thomas Watson Research Center
`
`The Ohio State University
`
`IBM
`
`USAEPG; STEEP-ET-S
`
`U.S. Naval Academy
`
`Hayes-Roth, Frederick
`
`Cimflex TeknowledgeInc.
`
`Klein Associates
`
`U.S. Army Missile Command Research, Development &
`Engineering Center
`
`Heindel, Troy
`
`Hirschberg, Morton
`
`Holmes, Willard
`
`Johnson, Sally C.
`
`Kiguchi, Takashi
`
`Kiss, Peter
`
`Klein, Gary A.
`
`NASA/Johnson Space Center
`
`U.S. ArmyBallistic Research
`
`NASA
`
`Hitachi, Ltd.
`
`Sentar, Inc,
`
`20
`
`20
`
`
`
`Table 2.2-1 (Continued).
`
`Point of Contact
`
`Affiliation
`
`
`
`Intellicorp, Inc.
`
`University of Michigan
`
`George Mason University
`
`Microelectronics and Computer Corp.
`
`Rochester Gas & Electric Company
`
`George Washington University
`
`Advanced Decision Systems,Inc.
`
`University of Southern Louisiana
`
`Chalk River Nuclear Laboratories
`
`Digital Equipment Corporation (DEC)
`
`Dupont Corporation
`
`George Mason University
`
`Westinghouse Electronic Corporation
`
`University of Wisconsin-Milwaukee
`
`Georgia Power Company
`
`RensselaerPolytechnicInstitute
`
`University of Southern California
`
`Data Systems Technology
`
`Westinghouse Electric Corporation
`
`University of California, Irvine
`
`Laning, David
`
`Lee, John Cc,
`
`Lehner, Paul
`
`Lenat, Doug
`
`Leoni, Nicholas
`
`Liebowitz, Jay
`
`Linden, Theadore
`
`Loganantharaj, R.
`
`Lupton, Lawrence
`
`Lutsky, Patty
`
`Mahler, Ed
`
`Michalski, R.S.,
`
`Moradian,Ali
`
`Nazareth, Derek
`
`Nelson, Robert
`
`O'Keefe, Robert
`
`O'Leary, Daniel
`
`Odubiyi, Jide B.
`
`Osborne, Robert
`
`Owens,Jerry
`
`Owre,Fridtjov
`
`Parsaye, Kamran
`
`Pazzani, Michael
`
`Navy Center for Applied Researchin Artificial Intelligence
`
`institutt fur Engergeteknikk
`
`Intelligence Ware
`
`
`
`21
`
`21
`
`
`
`|
`
`Point of Contact
`
`Plant, Robert T.
`
`Preece, Alun
`
`Table 2.2-1 (Continued).
`
`University of Miami
`
`Concordia University
`
`Affiliation
`
`|
`
`Rossomando,Philip J.
`
`General Electric Corporation
`
`Rousset, Marie-Christine
`
`L.R.I. - University’ d'Orsay
`
`Rushby, Dr. John
`
`St. Clair, Daniel
`
`Sharma, Ravi S.
`
`Sizemore, Nick L.
`
`Stewart, Tammy
`
`Sudduth, Al
`
`Surko, Pam
`
`Sztipanovits, Dr.
`
`Takahaski, Makoto
`
`Terano, Takoa
`
`Touchton, Robert
`
`Ulvila, Jacob
`
`Vesonder, Gregg
`
`SRIInternational
`
`McDonnel Douglas Corporation
`
`University of Waterloo
`
`COMARCO,Inc.
`
`USAEPG
`
`Duke Power Company
`
`Vanderbilt University
`
`Tohoku University
`
`The University of Tsukuba, Tokoyo
`
`Pathfinder Advanced Computing,Inc.
`
`Decision Sciences Consortium,Inc.
`
`AT&T Beli Labs
`
`Japan Atomic Energy Research
`
`Science Applications International Corporation (SAIC)
`
`Vignollet, Laurence
`
`University of Savoie
`
`Watson, David
`
`Williams, Robert
`
`Williamson, Keith
`
`Yen, John
`
`Yokobayaski, Masao
`
`Martin Marietta
`
`U.S. Army Electronic Proving Ground
`
`Boeing Computer Services
`
`Texas A&M University
`
`22
`
`
`
`22
`
`
`
`Figure 2.2-1 USNRC/EPRI V&V Interview Questionnaire
`
`| am X from SAIC, working under contract to the NRC and EPRI, on a survey task. We are interested in
`finding out what, if anything, you might be doing in the area of verification, validation or testing of expert
`systems or knowledge-based systems (ES/KBS V&V).
`
`SAIC Interviewer:
`
`DATE/TIME:
`
`Person(s) Interviewed:
`
`Contactlist entry correct? Yes__=—s- No____
`
`FAX:
`
`E-MAIL:
`
`PHONE:
`
`Title/Role:
`
`Type of Work: Research ___ ES Development ____ Services ___
`
`Study ____ Standards
`
`Project/System Name:
`
`Length of Work:
`
`Numberof People:
`
`Customers? Yes___ (see referrals) No___
`
`Funding source:
`
`Can we visit? Yes___-~No__
`
`Project/System Description: (next page)
`
`- Development/productplans
`
`- Who should beinterested (industry/ES type)
`
`- Problem areas encountered
`
`- Tool/technique needsidentified
`
`- Success?
`
`23
`
`
`23
`
`
`
`Figure 2.2-1 (Continued).
`
`Expert Systems Tested:
`
`SYSTEM NAME
`
`OPpsYS
`PLATFRM
`1)DOS
`1)PG
`02) Apple
`2)Apple
`3)SUN 3)Unix 2)Embed
`4)VAX 4)IBM 3)Stand
`5)Other
`5)Other
`
`SOFTWARE
`TYPE
`1)Real
`Time
`
`SIZE
`ENV.
`1)LISP,
`Prolog
`2)Shell
`3)Other
`Alone
`
`Rul
`1)Small(<50)
`2)Med(<500)
`3)Lge (<3000)
`4)Very Las
`(>3000)
`
`Testing Techniques:
`
`TECH
`NAME
`
`1)EVA
`2)random
`testing
`3)other
`
`ES COMPON
`
`1)KB
`2)infEng
`3)MMI
`4)Shell
`5)Other
`
`TYPE
`ERRORS
`
`Stai(1)
`Dynam(2)
`Anom(3)
`Valid(4)
`
`AUTOMATED
`TOOLS?
`
`(¥ or N)
`
`EASEof
`SET-UP
`
`4(Io)
`7(hi)
`
`POWE
`BR
`Abllity
`to find
`errors
`1(lo)
`7(hi)
`
`
`
`
`
`Automated Toois:
`
`TOOLNAME AVAIL SOURCE PLATFORM ENV,
`
`SOFTWARE
`
`
`
`
`(Y or N)
`1)PG
`1)DOS
`1)LISP, Prolog
`2)Apple
`2)Apple
`2)Shell
`3)SUN
`3)Unix
`3)Other
`4) VAX
`4)IBM
`5)Oiher
`5)Other
`
`
`24
`
`
`
`Referrals (Colleagues/Customers):
`Name
`Affiliations
`Topic
`
`Address
`
`Phone
`
`Figure 2.2-1 (Continued).
`
`Publication/Documentation References:
`
`Action Items:
`
`PERSON
`
`REQUIRED ACTION
`
`DATE REQUIRED
`
`
`
`ll
`
`25
`
`
`
`25
`
`
`
`contacts were distributed among them. Weekly meetings were held during the heavy period oftelephoningto collect
`referrals and other changes to the PoClist and to share information and interviewing hints.
`
`A separate form was prepared for Dr. Bernard tofill out on operational or nearly operational expert systems
`within the nuclear industry. This form is reproduced as Figure 2.2-2. The aim was to draw uponhis experience in
`writing his book, Expert Systems Within the Nuclear Industry, to gain an understandingofthe state-of-the-art of expert
`system V&V within the nuclear industry. He sent along references with the forms, and if needed, follow up contacts
`were made.
`
`In all, 138 PoCs were contacted which yielded the 97 doing current work mentioned above. However, many
`more people were called to generate these PoCs, This is because we would ofien have one nameas an entry into the
`organization, and would chase through a number ofreferrals to obtain the best and most knowledgeable PoCin that
`organization. Also, some people were not doing work in the field themselves, but gavereferrals to those wha were.
`Then there were referrals by the referrals. We got to the point in the survey where PoCs were referring to each other and
`we had both the funder/sponsor and contractor/university PoCsoffunding relationships on our list. This fact, and our
`limited resources, led us to limit the telephone survey, except for PoCs we knew were important, at some point so we
`could move on with the activity. A breakdown of the PoCsis shown in Figure 2.2-3.
`
`Publications were collected from a number of sources. These included:
`
`e
`
`a
`
`®
`
`*
`
`a
`
`Keyword-based search of the DIALOG and Defense Technical Information Center (DTIC) on-line
`computerized bibliographic services, followed by obtaining the mostsuitable publications,
`
`Conference and workshop proceedings and reports already on hand,
`
`Publications in Dr. Bernard's possession,
`
`Publications sent to us by PoCsafter telephone interviews, and
`
`Publications collected atsite visits.
`
`The result was a very extensivelibrary of materials on experi sysiem V&V (well over 300 references). The
`bibliography forthis library is included as Appendix B.
`
`2.3
`
`Site Selection and Visits
`
`As a result of the telephone interview and data collection process which was described in Section 2.2, the
`project team determinedthat a numberofsites offered the potential for obtaining significant additional information on
`expert system V&V techniques and tools. This preliminary list of sites was chosen after analyzing telephone interview
`data sheets and papers that were collected. Only those locations with robust ongoing expert system V&V activities that
`required an onsite, face-t