Web-based Weather Expert System (WES) for …

1 Web-basedWeatherExpert System (WES) forSpaceShuttleLaunch @NASAAmesResearch centerMS BardinaNASAAmesResearch centerMS TheWeb- based WeatherExpert System (WES) is a critical moduleof the VirtualTest Bed de-velopmentto support go/nogo decisionsfor SpaceShuttleoperationsin the IntelligentLaunchand RangeOperationsprogram of of the environmentrelated tothe launchingor landingsite,the timeof the day ornight,the pad or runwayconditions,the missiondu-rations,the runwayequipmentandlandingtype. Ex-pert systemrulesare derived from weathercontingencyrules,whichwere developed overseveral years by chaining,a goal-directed inference method isadopted, becausea particularconsequence or goal clauseis evaluated first,andthenchained backward throughthe a rule is satisfied or true,thenthat par-ticularruleis fired and the decisionis expressed. Theexpert systemis continuouslyverifyingthe rulesagainstthe past one-hourweatherconditionsand the decisionsare normalprocedure of operationsrequires aformalpre-launchweatherbriefingheldon Launchmi-nus 1 day,whichis a specificweatherbriefingfor allareas of Space this paper,the Web-based WeatherExpert Systemof the IntelligentLaunchand rangeOperationsprogram is :WeatherTracking, Expert System ,shuttlelaunch and IntroductionThelaunch weatherguidelinesinvolvingthe SpaceShuttleand expendablerockets are similarin many ar-eas, but a distinctionis madefor the individualcharac-teristicsof each [3].

2 Thecriteriaare broadlyconserva-tive and assureavoidanceof possiblyadverse the SpaceShuttle,weatherforecastsare providedbythe ForceRangeWeatherOperationsfacilityat Cape Canaveral beginningat Launch Minus 3 days incoordinationwiththe NOAA NationalWeatherServiceSpaceFlight MeteorologyGroup(SMG)at the JohnsonSpaceCenter in Houston[13, 14]. Theseincludeweathertrendsand theirpossibleeffectson launch day. A formal 0-7803-7952-7/03/$ weatherbriefingis heldon Launch minus oneday which is a specificweatherbriefingfor all areasofSpaceShuttlelaunch weatherfore-casts,groundoperationsforeca stsandlaunch weatherbriefingsfor the MissionManagement TeamandtheSpaceShuttleLaunch Directorare preparedby theRangeWeatherOperationsFacility. Theseincludeallemergencylandingforecasts and the end of missionfore-castsbriefedby SMGto the astronauts,the Flight Di-rectorand MissionManagement Team[13].

3 Thebasicweatherlaunch commitcriteriafocusesan ambient tem-perature,windspeed,precipitation[9, 11, 12], lightning[8], type of clouds,and cloudcharacteristics(cloudtem-peraturean d thickness).Thesefactorsare usedas rulesfor the weatherexpert many rulesand conditionsdependingupon different the needfor anexpert System ,components of the Expert System ,andfutureresearch Needfor a WeatherExpertSystem (WES) Intelligent Launch andRangeOperationsVirtualTest Bed(ILRO-VTB)is software for prototypingoflarge-scalemulti-discipline ddynamicsystems[1, 4].Withthe increasingpressureto integrateandto buildcontrol systemsand services,thereis a desperateneedto testandevaluatethe complexinteractionslikely tooccurundernormaland adverse( ) is usedto simulatethemission,control,ground-vehicl e,launch launch and rangeoperations[1, 4], weatherplays a the weatherexpert systemis to make expertiseavailableto decisionmakers whoneedanswers quickly and rapidlyin a globalmanner(macroscale).

4 Theexpert System [5] can assistwithsituationassessment and launch the Expert systemcan vary acrosstext, systemsaves moneyby reducingthe timeinvolved in weatheranalysisandthe management can make smarterdecisionsand workmoreproductively. Thecomplexity andresponsibilityof weatheroperationsraisesthe needfor an automatedexpert systemto analyzeandprovideexpertiseto Components of Expert SystemAn Expert System [5] is a computerprogramdesignedto simulatethe problem-solvingbehaviorof a humanwhois an Expert in a narrow domainor systemis normallycomposedof a knowledgebase(information,heuristics,etc .),inferenceengine(ana-lyzesthe knowledgebase),andthe end userinterface(acceptinginputs,generating outputs).Thepaththatleadsto the development of Expert systemsis differentfromthatof Expert systemdevelopment comefromthesubject domainof artificialintelligence(AI),and requirea conventionalprogramcon-sistsof an algorithmicprocess to reach a AI programis madeup of a knowledgebaseand aprocedureto inferan answer.

5 Expert systemsare ca-pableof deliveringquantitative information,much ofwhich has been developed throughbasicandappliedresearch as well as heuristicsto interpretqualitativelyderived values,or for use in lieu of quantitative thatexpert systemscan addressimpreciseand incompletedatathroughthe assignmentof confidencevaluesto inputsand mostpowerfulattributesof Expert systemsis theability to systemremem-bers its logicalchainof reasoning,a usermay ask for anexplanationof a recommendationandthe systemwilldisplay the factorsit consideredin providinga particu-lar the recommendationandacceptanceof theexpert Expert usesto solve a problemmust be represented in a fashionthatcan be coded intothe computerand thenbe availablefor decisionmakingby the Expert be rep-resented by aconditionor premisefollowed by an actionor conclusion(IF ).Productionrulespermitthe relationshipsthatmake up the knowledgebaseto bebroken down into consul-tation,the rule baseis searched for conditionsthatcanbe satisfiedby factssuppliedby the is performedby the of theconditions( partsof rules)of a rule are matched,the ruleis executedandthe appropriateconclusionisdrawn.

6 Basedupon the conclusionsdrawn and the factsobtainedduringconsultation,the inferencemechanismdetermineswhich questionswill be asked andin variousinferencingmethods availableto performthe tasksof searching,matching,and distinctive characteristicof Expert systemsthatdistinguishesthemfromconventi onalprogramsis theirability to utilizeincompleteor partialdataset, an Expert is likely to have lessthanabsolutecertainty in his certainty can be quantifiedin relative termsand in-cludedin the valuesareassignedby the Expert duringthe knowledgeacquisitionphaseof developingthe incorporatingrulesin the knowledgebasewithdifferent certainty values,the systemwill be ableto offersolutionsto problemswithouta completeset of the rulesfor the weatherexpert systemarederived fromweathercontingencyrulesdeveloped overseveral yearsby NASA[3]. An exampleof weatherruleis presented 36<Temperature<98 and0<WindSpeed<24 andPrecipitaion= No andLightning= No andCloudTemperature>32 and Cloudheight>20000ftand Cloudthickness<4500ft and Cumulus= No and Cumulonimbus= No thenLaunch= GO Theabove rule has nineantecedent clausesjoinedbya conjunction and and has a singleconsequent clause(Launch).

7 The rule is triggered,if all antecedent clausesare set to be derived foreach vehicletype. Dependingupon the launch vehicle,the rulesare slightly rule variablesremainconstant for mostof the launch rulesfor GO and upon the prevailingweatherconditions, at the goalvariableof theexpert fire thatruleby provingthe conditionsof the ruletrueusingasequenceof rulesin the the systemdoesn thave enoughstoredknowledgeto prove a ruletrue,it will ask the end userto continues to search the knowledgebaseandask theuserfor informationuntil a valuecan be determinedforthe , whenit has assembledtheknowledgeneededto fire a rulethatestablishesa valuefor the goalvariable,the Expert decisionis paper, the backward chainingmechanismis adoptedfor the backward chainingis morefocussedbecauseit onlyprocessesrulesthatare relevantto the simplytraversesthe rule basetryingto prove thatclausesare truein a mostwidelyusedfor interactionwithcomputersbecauseof theirease,superiority, effi-ciency, userfriendlinessand robustnessas theyallow theuserto interactby manipulationof weatherweb Expert System ,the userinterfaceis au-tomatedin such a way thatthe inputsto the expertsystemare downloadedand fed to the systemin a peri-odic no needof humaninterventioninthe Expert systemand decisionsfor launch are automat-icallydisplayed as a web expertsystemis basedon Java technologyandweb enabled,which can be viewed fromany partof the WeatherExpert SystemTheweatherexpert systemis launchedon the webby a dedicatedtomcatserver[2,10].

8 In this section,thedetailsof weatherweb Expert systemand acquisitionofdetailedinputsare supportedby Java servlets[6, 7]wherethe usercan inputthe datafor servlettechnologyprovidesa simple,con-sistent mechanismfor extendingthe functionality of aWeb servletcanalmostbe thought of asan appletthatrunson the server component- based ,platform-independent methodfor buildingWeb-basedapplications,withoutthe per-formancelimitationsof figure1,thereare 16 different dealon the US andNorthAmericancontinen-tal providesinformationon globalweathersystemincludingtropi-cal Launch Decision buttonactivatesthe Expert systemand providesthe Expert decisionforthe shuttlelaunch. Exceptthe launch decisionbutton, Figure1: Front-endof WeatherWeb serverall otherbuttonsinvoke the dataor imagesfromvarioussourcesacrossthe US. TheUS weatherbuttonprovidesa 7 dayweatherforecastfor a given zip code in the continentalUS.

9 It providesa Nationalweatherserviceradarimageand ,specificweatherdetailslike humidity, windspeed,barometricpressure,heatin-dex, and dew point are updatedin hourlyintervals. TheUS cloudclassificationis providedby Naval postgradu-ate school at Monterey, providedby NationalLightningDetectionNetwork(NLDN)a ndat 30 minute intervals lightningstrikesacrossthe US are surfacewindspeed (knots)are providedbythe stateanalysisisprovidedto the userto understandthe boosterrocketrecovery. Weathercriteriafor an emergencylandingatthe TransOceanicAbort LandingSites(TALS)is mon-itoredby NOAA in SpainandNorthAfrica[13]. Thedatais downloadedevery hourfromvariousagenciesin the formatof imagesand downloaded,the imagesare processedand spe-cificnumericalvaluesare derived for usedas inputsto the Expert Launch Decisionbutton ,an Expert systeminferenceenginechecks the valuesagainstthe weatherrulesand it suggeststhe shuttlelaunch decisionby GOor figure2, the Expert decisionsfor the Shuttlelaunchare shown below the the GO situationwhereasred GO shouldoccur,everyvaluein the lower frameshouldbe Figure2: Outputof Weatherexpert systemexpert systemprovidesthe decisionfor a genericshut-tle launch.

10 For a specificshuttlelaunch, morestringentruleshave to be addedto the EndSystemThecrondaemonis a longrunningprocessthatex-ecutescommandsa t a timeweatherdatais obtainedfromdifferent federalweathermonitoringagenciesand it is shown in illustratesthe flow of datato the Expert CHARACTERISTICSUS THREAT WATCHTEMPERATURE, WIND SPEEDRADAR , : Real-timeinputdatafor weatherexpert appli-cationswhich downloadimagesand othertypes of imagesand ex-tractsandconvertsto inputsfor Expert dataand the expertsystemprocessesthe informationandprovidesa deci-sionfor the mainJavaclassand it actsas an applicationfor the Expert set of controlsforselectingbackward chaininferencing,the GoalvariableandResultwhenbackward chainingis ,theRule-Variableclass,and theRuleBaseclassas well as sup-ports classessuch asClause. TheRuleclassis usedtodefinea singlerule and also containsmethods thatsup-port the Rulehas a namedatamember, a referenceto the owningBooleanRuleBaseobject,an array of antecedent clauses,and a singlecon-sequent struthvalueis storedin a Booleanobject,not an rule has been firedor several ruleconstructors,each requiringa referenceto theBooleanRulebaseinstance,the Rulename,oneor moreantecedent or left-hand-side(LHS)clauses,andthe singleconsequent or right-hand-side(RHS) constructorallocatesthe correctnumber of entriesJava (cron job)Java (cron job) (cron job)Back end Data AcquisitionBack end Data SystemLaunch Decision Figure4: Flow of datafor weatherexpert systemin the antecedents array, and also also addsthe clausesto its datamem-bers.