Open Research OnlineThe Open University’s repository of research publicationsand other research outputs

Advanced knowledge system for coatings and the gasturbine MRO industryConference or Workshop ItemHow to cite:

Chandler, Peter; Hall, Wendy; Shadbol, Nigel; Alani, Harith and Szomszor, Martin (2008). Advanced knowledgesystem for coatings and the gas turbine MRO industry. In: International Thermal Spray Conference & Exposition:Thermal Spray Crossing Borders (DVS-ASM), 2-4 Jun 2008, Maastricht, The Netherlands.

For guidance on citations see FAQs.

c© 2008 The Authors

Version: Accepted Manuscript

Link(s) to article on publisher’s website:http://asmcommunity.asminternational.org/portal/site/www/AsmStore/ProductDetails/?vgnextoid=8e8288fbd8e4c110VgnVCM100000621e010aRCRD

Copyright and Moral Rights for the articles on this site are retained by the individual authors and/or other copyrightowners. For more information on Open Research Online’s data policy on reuse of materials please consult the policiespage.

oro.open.ac.uk

Advanced knowledge system for coatings and the gas turbine MRO industry P E Chandler, W Hall, N R. Shadbolt, H Alani and M Szomszo. Southampton/UK The growth of data generated within thermal spraying is, for many, a daunting business. Yet, this growing resource represents a largely untapped and potentially valuable asset capable of providing “knowledge” rather than just “information”. Many companies already use a range of Web based tools. However, the Web itself is changing and the vision for the future, the “Semantic Web”, is set to revolutionise how business will be done. One important aspect of this Web “future” is that web pages will be greatly enriched and data will have additional information (tags) which help to describe it and more significantly, put the data into a context. This will enable machine readability and the use of query languages to ask direct questions. Following on from ideas introduced at ITSC 2007, a proof of concept demonstrator has been built for thermal spray coatings used in the Maintenance Repair and Overhaul (MRO) of gas turbines. A system has been built which stores and manipulates a range of data including; aircraft deliveries, RSS feeds of aircraft sales, engine types, MRO business details, thermal spray coatings and market dynamics. This paper presents the development of this system and discusses its future potential. 1 Introduction Like most business afflictions the problem of “data-overload” can be looked at as an important and even significant commercial opportunity. In order to make this turn-around happen a method is needed which enables the consolidation of data held in a wide range of formats and sources plus the ability to interrogate this vast data source in order to extract knowledge. Given the exponential growth of data with time any system must also be fully expandable and, of course, be capable of automatic operation! The initial reaction to this issue may be to ask “why not just use conventional spread sheets, databases and web tools?” This is understandable but to just “Google it” does not actually solve a basic problem namely how does one piece of data relate to another? There is no knowledge behind a simple word search. A common example often used to show this is to search on the word “apple”. The results will include references to computers and fruit in order of commercial sponsorship, and a mix of citations, number of hits, meta data and word searches. The only real knowledge involved will lie with the user who has to refine the search terms. What is needed is a way to add some knowledge to the data itself. In fact this aspect is currently being worked on by developers of what is seen to be the next generation of the Web, the Semantic Web [1,2]. The new vision will be to enrich data by incorporating additional information (tags) which both help to describe it but more significantly, put the data into a context. Furthermore, this Semantic Web content will be fully machine readable and allow the use of query languages to ask direct questions. In essence, this future generation of the Web will contain knowledge as well as documents. The term “semantic” simply means the meaning of things. In this context it is used

to define the meaning of data but more importantly how they interrelate. Establishing interelationships however is in itself a complex field of study. A useful starting point may be the development and use of taxonomies which aim define and organise objects. The ASM use a taxonomy as a browse feature “to provide relevant results from all of content sources in the ASM Community“ [3]. Adding further information about relationships goes one step further and creates what is called an ontology. This idea was introduced at ITSC 2007 [4] and is a central feature of the system discussed in this paper. A more detailed treatment of this field may be found elsewhere [5]. Some examples of some defined relationships within an aviation ontology are provided in table 1. Table 1: Some examples of ontology classes and relationships, instance triples in aviation and MRO shops.

Ontology classes and relationships

Aero-Space Companies

Manufacture Aircraft

Aero Engines Power Aircraft Airlines Operate Aircraft

Gas Turbine Is a Aero Engine Examples of instance triples - aviation

Boeing Manufacture 777-200 Airbus Manufacture Airbus 320

BA Operate 777-200 777-200 Is a Aircraft

IAE Manufacture Aero Engines V2500 Is a Aero Engine

Examples of instance triples – MRO Shops MRO shops Overhaul Aero Engines MTU, Zhuhai Operate An MRO shop MTU, Zhuhai Is located in Far East MTU, Zhuhai Overhaul V2500

InsrelknoBomaae WhtheonlanFraacc In codesecdamaIntof Shproenfleenethethewit Thtecof preFinde 2 ThusedeMRnuAirnuanthefor

stances arisations can owledge. Fo

oeing is an akes the 77rospace com

hat makes these instancetology can

nguages suamework (Rcessed unive

terms of auld potentialmonstrate itctor a small dta on aircr

arkets to proerestingly, Rthis market

hop Visit Plaovide a plangine repair aet managergotiate to s

ese emergine whole manthin this sect

is paper setschnology bei

the demonesentation onally, suggesveloped are

Semant

is section ped to build scribing airc

RO shops thmber of pubrbus and Boember of plad PDF. The

e MRO-Assisrmat, and the

• Airbus-ordere

• Boeingmodel,ordere

• Enginenumbe

• EngineName,

• CIA Fdemog

sing from tbe used

or exampleaerospace

7-200, the 7mpanies mak

he Semantices of obje

be expreuch as thRDF) [6] thaersally.

application, lly be appliets potential demonstratoraft, airlinesovide a busi

Rolls-Royce at and have

anner [7, 8]. nning and scand overhaulrs and otheschedule eng technologi

nufacturing, ctor.

s out a brief ng used andnstrator thaf some of thstions for hoalso discuss

ic Technolo

rovides an othe MRO-A

raft types, thhat can repalic data soureing provide nes orderedlist below ou

st applicatione information

-Orders: Ad / delivered

g-Deliveries: , engine fitd (Excel).

e Yearbook 2er of enginese Yearbook Address, Ap

Fact-book: Cgraphic / eco

hese sorts to infer va, it can becompany fr

777-200 is ke aircraft.

c Web so apcts and re

essed in Whe Resourcat allows th

this emerged to any fiein the surfa

or has been bs, engines, iness tool foare working a

recently deIn this case

cheduling syl. The agent

er stakeholdngine shop ies will evencustomer an

introduction d how it workat has beenhe results a

ow this technsed.

ogy

overview of Assist demo

he engines thir them, is a

rces. Manufadetailed info in formats,

utlines the dan, including tn they contain

Airlines, aird / operationa

Airline, cotted, order

2007: Aircra, compatible

k 2007: MRpprovals (pdf

Country coornomics (rdf).

of ontologarious sorts e inferred trom the facan aircraft a

ppealing is telations in Web standce Descripthe data to

ing technoloeld. In order

ace engineerbuilt which us

coatings aor MRO shoat the other eemonstrated e “Web agenystem for aets, representers, adaptivvisits. Clea

ntually embrand supply ch

to the semanks, a descriptn built and

and capabilitinology could

the technoloonstrator. Dhey use and available fromacturers suchormation on such as Ex

ata sets usedthe source dn:

rcraft, numal (Excel). ountry, regi

date, num

aft, model, ty engines RO, Compaf). rdinates, fla.

ical of

that ct it and

that the ard tion be

ogy r to ring ses and ops. end

its nts” ero-ting vely arly ace

hain

ntic tion

d a ies. be

ogy Data

the m a h as the

xcel d in

data

ber

ion, ber

ype,

any

ags,

Thcrenutherelbe On ThcasodifAmcothemasoAmcothemainfspairorden

Fig

he data was eate a suitabmber of bes

ese data soevant sourc

elow.

ntology Cons

he first stagepture the surce. While

fferent sourcmerican Airlincepts (e.g.e granularityakes the conurces, both

merican Airlincepts (e.g.e granularityakes the cfeasible. For ecifies the d

rline made thder was madgine was fitte

gure 1 Ontol

then procesble ontology, spoke scriptsources to Rces. These s

struction

e was to creemantics anmuch of thees, both in tenes is an a aircraft, en

y of informanstruction of h in terms nes is an a aircraft, en

y of informaconstruction example, the

date when ahe order, forde, how maned.

logy map

ssed in threeconvert the

s were develoRDF) and lsteps are de

eate a suitand structure

e data is replerms of insta

airline from tgine), subtle

ation and coa domain on

of instancairline from tgine), subtle

ation and coof a do

e Boeing delan order war which modny were orde

e-stages whdata to RDF

oped to convink it to otescribed bri

ble ontologye of each dlicated betweances (e.g. tthe U.S.A) ae differencesonceptualisatntology differces (e.g. tthe U.S.A) ae differencesonceptualisatmain ontololiveries data

as made, whdel of plane ered, and wh

ich: F (a vert ther efly

y to data een that and s in tion rent that and s in tion ogy set

hich the

hich

Wipromodeinfotypfrocuforwedalevthe RD Widadatherepto Smtheapin-Cosimto for (Nocorelof sys Da Oncosalanprodife ThdainsBoknoNoacma

th the Airbuovided for eodel has belivered, and ormation is p

pes that wereom each datarrently in oper each dataere created, ta set that c

vel. Figure (e Boeing deli

DF Generatio

th suitable ta source, tta to RDF. F

e content wpresentation facilitate com

mall Java proe Jena APpropriate comemory mo

onversion of milar pattern

massage thrmat.

ote that lanntent as tationship opthese are foustem is used

ata Linking

nce the sourncept instanme entity w

nguage, OWoperty is useerent datase

is linking prota sources s

stances of oeing and Airowledge bas

orth America hieve this linanually creat

• airbus_

• boeing

• cia_co

• overha

s orders dateach airline een orderedhow many a

provided by Ae fitted. Whia set to estimeration, the c

a source. Twith low comcan be linke1) provides iveries ontolo

on

ontologies ihe next stagFor data setwas first e(comma sep

mpatibility wiograms were PI, to parsoncept instanodels were f PDF docu

but requiredhe source te

nguages suctriples cons

perator and aund in Table

d to provide e

rce data hadces from eawere linkedL [10].In thised to link id

ets as illustra

ovides the posimultaneousairlines and

rbus data sese for all oror Europe),

nking, a numted, as summ

_engine_ma

g_manufactu

untry_mappi

aul_engine_m

ta set, only stating how d, how maare currently Airbus regarle it is possi

mate the numcalculations herefore, sm

mplexity, for ed together a

a visual reogy.

in place to ge was to cs provided i

exported to parated valuth programmthen written

se the sounces and pro

then expoments to Rd more manext to a ma

ch as RDF sisting of an object. S

e 1. A “ triple easy access

d been convech data set t

d using a s language, tdentical termted in Figure

ower to quesly. For examd countriests it is possibrders made by country, o

mber of mapmarised in the

pping.rdf

rers_mappin

ing.rdf

mapping.rdf

summaries many of ea

ny have beoperational. ding the engble to use d

mber of enginwill differ submall ontologeach individat the ontolopresentation

describe eaconvert the rn Excel form

a text-bases in this ca

ming languag, making userce data ioperties. Theorted to RD

RDF followedual intervent

achine-reada

represent tha subject,

Some exampstrore ” storato the data [

erted into RDthat refer to Web Ontolothe owl:same

ms found wite (2).

ry over multimple, by link between ble to query by region (eor by airline.

pping files we list below:

ng.rdf

are ach een No

gine data nes btly

gies dual ogy

n of

ach raw

mat, sed

ase) ges. e of into ese DF. d a tion

able

heir a

ples age 9].

DF, the ogy eAs thin

iple king the the

e.g. To

were

Fig Th OnweStolancawhstaPHthetripusreq 3 ThMR

Thtec

gure 2 Links

he Overall sy

nce the RDFere created, ore. This pronguages couse the quer

hich was ablatistics needHP+ Javascre applicationple store. Aed to improquired large

The MRO

here are thrRO-Assist de

• To cheterorepresinformaindivid

• To proexplore

• To foreby esengine

he first aim wchnologies d

s between on

stem

F data, ontoall RDF wa

ovided an enduld be used tory languagee to provideded for theript front-endn interface

A number of ove performamounts of p

O Assist De

ee principalemonstrator:

consolidate geneous daentation thatation that ual sources.

ovide a grape data. ecast future

stimating thees in operatio

was realiseddescribed ab

ntologies

ologies, andas imported d-point throuo extract info

e used was e summariese forecastingd was develousing HTTP

f caching prmance since

processing.

emonstrator

objectives

data fata sources t facilitates tis not poss

phical interfa

MRO re-coe number oon at a region

d using the bove. The se

d mapping finto the Tr

ugh which quormation. In t

SPARQL [s as well as g algorithm.oped to provP to query rocedures w

many quer

r

underlying

from multinto a sin

he extractionsible from

ce to view a

oating businof planes anal level.

Semantic Wecond two w

files iple

uery this [11] the

. A vide the

were ries

the

iple ngle n of the

and

ess and

Web were

acbasto Da In foraccFigthesuche Figen A tbroalsof a splasupasho Thexabuto qucu RetheCIA

hieved throused applica

ored data.

ata Presentat

order to aid r the interfaccessing and gure 3 showese can be ch as the cld.

gure 3 Displagine capabil

tabular data owsers such so supplied tthis, shown summary of anes they opplied themrticular airlinowing types

e maps anamples of ht are actually

provide sestions abourrently has th

egional Infoe dataset, usA factbook.

ugh the deveation that ex

tion

visualisationce. This prointerrogating

s the locatioselected to company de

ay Interface ities/approva

browser (muas mSpace o show furthin the screenairlines and own and wm. Going dne, the engin

and number

nd tabular bhow the story based on qspecific ansut the MRO he following

ormation. Masing coordina On select

lopment of axploits the

n, a Google vides a cong the system

ons of MRO view additio

etails and re

showing MRals

uch like cont[12] and Ta

her information shot in figuthe number

which aircraftdeeper by ne inventory rs currently in

browser intered data canquestions poswers to sindustry. Thcapabilities:

arkers for evate data extrting a part

a graphical, wintegrated a

map was usnvenient waym. For exampshops. Each

onal informatepair approv

RO shops

temporary Rabulator [13]on. An exam

ure 3, which hof engines a

ft manufactuhighlightingcan be view

n service.

erface are jn be presensed by the usome detaie demonstra

very countryracted from ticular coun

web and

sed y of ple, h of tion vals

RDF ) is

mple has and urer a

wed

just nted user iled ator

y in the

ntry,

adpothaGDpo Reeapothe CograMRussiz

Figtyp Covaeausregdis A usalg In tog(onlinusreqtheresanforto is enbu

ditional infoopulation, numat it would bDP per headopulation who

epair Shopsach repair shossible to viee engine app

oatings Infoaph for eachRO businesser may click

ze, more deta

gure 4 Histope (1981-200

oatings Pie lue of the M

ach region. Ter can easilygions compasplayed, to se

configurationer to tweak

gorithm.

this study a gether usingntologies, triked data caning SPARQquired knowe Airbus andstricted to a estimate for

r that regionexamine whpossible to gines in opesiness that w

ormation (sumber of airpobe easy to a which is an

o take flights

s This map hop. By clickw the compa

provals.

ormation Thh region on s generated

k on the grapailed version

rical growth 05)

This tab isMRO re-coatThe size of ey see how t

ares to each elect a year

n screen is athe paramet

range of datg the newly ple store, linn be easily inL query lan

wledge. For d Boeing daparticular rer the total nu. By using t

hich engines determine

eration and twould be gen

uch as the orts, etc.) is add financiaindicator of

.

view displayking on a reany name, it

his tab will the map sh

d for engineph snap-shotn.

in MRO bus

s used to dting busines

each chart isthe business

other. A timfrom 1965 –

also provideters used in

ta sources hadeveloped

nks etc). In tnterrogated (

nguage [11])example byatasets for gion, it is po

umber of planhe Aircraft-Eare fitted to the numberherefore estnerated.

country’s fldisplayed. N

al data such numbers of

ys a marker epair shop it's address, a

place a smhowing the toe re-coatingt to obtain a

iness by eng

isplay the toss generateds relative so s in each of me slider is a

2025.

d to enable the forecast

ave been linksemantic to

this system (for example) to extract y querying bairplane ord

ossible to obtnes in operatEngine ontolothese planesr and typestimate the M

lag, Note

as the

for t is and

mall otal . A full

gine

otal d in the the

also

the ting

ked ools the

e by the

both ders tain tion ogy s, it

s of RO

FigFousege 4. Asscaeaanspecouseuseestgealgin theenpre Thscrforthr

gure 5 Coatior values in ed to estinerated.

The MRO

s already mealable. Any sily integratd links wherecific compombustors aned on these ed to answtimate the Mnerated in

gorithm to esoperation in

e future. Frogines that edict the bus

e screen-shreen with alrecasting algree categorie

[a

[b

ng business the future,

mate the

Coating Bu

ntioned, thisamount of

ed by addinre necessaryonents to ennd turbine bla

componentswer questionMRO re-coat

the futurestimate the nn a particulaom this, wewill require

siness genera

hot of figure l the param

gorithm. Theses:

a.]

b.]

(a) 2007 ana forecastinMRO coat

usiness

system is bfurther data

ng extra ony. So, for exgines such a

ades and dets can readilyns about ming business, we use umber of plar region at a

e can infer e overhaul, ated.

6 shows theters that ase paramete

d (b) 2025 ng algorithmtings busine

both flexible aa sets can

nlological terxample, addas fan casintails of coatiny be added aarket size.

s that would a forecast

anes that willa given datethe numberand theref

he configuratre used by

ers are split i

m is ess

and be

rms ding ngs, ngs and To be

ting l be e in r of fore

tion the into

Fig Aiallotha NepeneDr Refor

Fig

gure 6 Confi

rcraft Manuows the userat would be h

ew Aircraft ercentage of ew models (sreamliner).

egional Grorecast the nu

gure 7 Confi

guration Scr

factures Thr to modify thheld by Airbu

Models Thiplanes purc

such as the

owth Rates umber of plan

guration Scr

reen

e slider bar he percentagus and Boein

s sliders bachased in thAirbus A380

Growth ratenes sold in e

reen

in the upper ge of the marng in the futu

ar dictates we future will 0 or the Boe

es are usedach region.

left rket re.

what be

eing

d to

The first step in the forecasting algorithm extracts historical plane order data for each region from the triple store using SPARQL queries. These results are used to calculate the number of planes and engines currently in operation, as well as the regional preferences towards a particular engine manufacturer. The second step estimates the number of planes that would be ordered in the future using the three parameters (outlined above) set by the user. The regional growth rate is used to predict the number of new aircraft that will be ordered. Using 5 years of historical data, and the user parameters that set the number of new aircraft that would be ordered and the Airbus – Boeing market share, an estimate for the aircraft models that would be ordered is generated. The final step uses the regional preferences towards a particular engine manufacturer to estimate the engines that would be fitted to the newly ordered aircraft. From this figure, it is possible to predict the total number of engines that will be in operation and therefore the MRO coatings business generated. 5. Summary and Discussion The MRO-Assist application has shown how a number of datasets can be integrated to provide an intelligent knowledge base for the civil engine MRO business. That this can be expanded to cover any aspect of the business has been demonstrated for one particular activity i.e. coatings. However, there is potential for using this technology as a platform for any number of applications. Companies such as Airbus already have a simple system for stock control whereby parts are bar-coded. When used they are scanned and this information is passed to the supply chain in order to optimise stock control. Using this new technology much more ambitious systems could be realised. Companies could integrate their own data about parts, materials, processes, quality, approvals, commercial detail and management systems and generate a common system for the whole company and/or supply chain. This could form web based tool which was expandable in real time. Another, more general example could be the generation of a system for Surface Engineering. There is a wealth of data available on surface engineering processes, materials, coatings, wear, corrosion etc which lies in a disparate set of sources. Increasingly, data from processing is being generated as well as in-service data. In the latter case, data mining is becoming an increasingly important diagnostic tool. The ability to integrate all of these aspects into a single, intelligent system would provide designers and users of surface engineering and coatings with a very important tool.

6. References 1. Berners-Lee, T., Hendler, and O. Lassila, “The Semantic Web”, Scientific Am., May 2001,pp. 34–43. 2. Shadbolt, N., T. Berners-Lee, T. and W. Hall, W. , 2006 The Semantic Web Revisited. IEEE Intelligent Systems, May/June 2006. 3. ASM.http://asmcommunity.asminternational.org/ portal/ site/asm/ResearchLibrary/ 4. Chandler P. E.,“Semantic technology as a core business activity for thermal sprayers“. In Proc. ITSC 2007, Beijing. 14-16 May 2007. Poster Session. 5. Guarino, N. Formal Ontology and Information Systems. Proceedings of FOIS'98, Trento, Italy, 1998 6. RDF. www.w3.org/TR/rdf-primer. 7. Stranjak,A., Duttag, P.S., Ebden, M.,Rogers,A., Vytelingum, P.“ A Multi-Agent Simulation System for Prediction and Scheduling of Aero Engine Overhaul”, to be presented at AAMAS-08. 8. Rolls-Royce. Demonstration of “Shop Visit Planner“ made at ARGUS II Conference, BERR Conference Centre, London. 28 February 2008. 9. Steven Harris S. And Gibbins N., Triple-Store: Efficient bulk RDF storage. Proc. 1st Int. Workshop on Practical and Scalable Semantic Systems (PSSS'03), Sanibel Island, FL, USA, 2003. 10. OWL. http://www.w3.org/TR/owl-features/ 11. SPARQL. http://www.w3.org/TR/rdf-sparql-query/ 12. schraefel, m. c., Karam, M. and Zhao, S. (2003) “mSpace: interaction design for user-determined, adaptable domain exploration in hypermedia. In: AH 2003: Workshop on Adaptive Hypermedia and Adaptive Web Based Systems, August 26, Nottingham, UK 13. Berners-Lee T., Chen Y., Chilton L., Connolly D., Dhanaraj R., Hollenbach J., Lerer A. and Sheets D.. “Tabulator: Exploring and Analyzing linked data on the Semantic Web”. 3rd International Semantic Web User Interaction Workshop, Collocated with the 5th Int. Semantic Web Conf., Athens, Georgia, USA, 2006.