Availability Allocation Using Systems Engineering Principles

Transcription

1 Proceedngs - Internatonal Conference on Industral Engneerng and Operatons Management, Kuala Lumpur, Malaysa, March 8-10, 2016 Avalablty Allocaton Usng Systems Engneerng Prncples Abraham Almaw Jgar, Cecla Haskns and Mary Ann Lundtegen Department of Producton and Qualty Engneerng Norwegan Unversty of Scence and Technology Trondhem, Norway abraham.a.jgar@ntnu.no, cecla.haskns@ntnu.no, mary.a.lundtegen@ntnu.no Abstract Avalablty s one of the key performance ndcators for reparable systems, usually expressed as the average proporton of tme the system s n a state capable of performng the requred functon. Avalablty can be enhanced by ncreasng the relablty and/or reducng the restoraton tme of components. In the desgn phase, a mathematcal method s appled to optmze the avalablty mprovement (or target) at the component level wth some relevant constrants such as cost and complexty. Ths paper s a contnuaton of the work carred out by the authors to address such mathematcal aspects. System desgn s a multdscplnary actvty subject to conflct of nterest among stakeholders such as system ntegrator, equpment suppler, and operator. The focus of ths paper s to extend the prevous work by lookng nto avalablty allocaton from the perspectve of systems engneerng. Systems engneerng s a systematc approach to enable the realzaton of a successful soluton. The paper uses a systems engneerng methodology to demonstrate ts applcablty as an ntegral part of the avalablty allocaton process. The concluson s that a systems engneerng methodology can be appled successfully such that the system avalablty target and the stakeholders needs can be met. Keywords Avalablty allocaton; Allocaton method; Systems engneerng; Reparable systems; Stakeholders I. INTRODUCTION Avalablty s a useful performance ndcator for reparable systems, usually expressed as the average proporton of tme the system s n a state capable of performng the requred functon [1]. In the early desgn phase of a new product development or modfyng an exstng one, avalablty s calculated to estmate the performance of the system and to check whether the target/requrement s met. Avalablty, n ts smplfed form, s a functon of relablty and restoraton tme. If a desgn soluton does not meet the requred avalablty target, an mprovement can be found through ncreasng the relablty of components and decreasng ther restoraton tme. To acheve the mprovement wth a mnmum effort (or cost), an avalablty allocaton method can be used [2-5]. Avalablty allocaton s an teratve process that nvolves both objectve judgments (e.g., component performance from relablty data) and subjectve judgments (e.g., operatonal and mantenance phlosophes and other decsons from stakeholders). Developng a complex system nvolves a number of stakeholders and some stakeholders may have the power/urgency to nfluence avalablty mprovement alternatve solutons [6]. Whether or not an mprovement alternatve s optmal may depend on both the level of effort the soluton would requre and whether t would meet all stakeholders needs throughout the lfecycle of the system. The former s an optmzaton problem and a number of avalablty allocaton methods have exsted and used for many years [7]. The prevous work by the authors of ths paper [5] addressed the mathematcal aspects of avalablty allocaton. For complex systems, such as subsea producton and processng systems, the avalablty allocaton process needs a systematc (qualtatve) approach along wth mathematcal methods to take nto account stakeholders needs. To ths end, the purpose of ths paper s to ntroduce a systems engneerng method that enables the analyst to look holstcally at the avalablty mprovement process [8]. Despte the popularty n ndustry and academa, there s no a generally accepted defnton for systems engneerng, and people use the term dfferently from the perspectve of ther expertse [9]. In ths paper, we suffce wth the defnton provded by Internatonal Councl on Systems Engneerng [10], an nterdscplnary approach and means to enable the realzaton of successful systems. Frutful efforts by specalzed areas/departments at lowest levels and n the ntegraton of them to form a whole wll brng about a successful soluton. In our case, the relablty engneer, who s responsble for the avalablty allocaton process, s to mplement a systems engneerng method/thnkng to brng about a successful system/soluton (wth respect to system avalablty). But, the relablty engneer as a system engneer does not own the allocaton problem or the soluton but s the enabler [8] who makes sure the system meets the stakeholders needs n a hgh qualty, trustworthy, cost effcent and schedule complant manner throughout the system desgn and development process [10]. Many systems engneerng methods have been proposed snce the early 70 s wth general agreement on the tasks they consdered relevant. The SIMILAR process s orgnally developed by extractng common tasks from several systems IEOM Socety Internatonal 1488

2 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 engneerng methods and s comprsed of seven tasks [11]. These are presented below wth ther descrpton (quoted from [12]): State the problem. Statng the problem s the most mportant systems engneerng task. It entals dentfyng customers, understandng customer needs, establshng the need for change, dscoverng requrements and defnng system functons. Investgate alternatves. Alternatves are nvestgated and evaluated based on performance, cost and rsk. Model the system. Runnng models clarfes requrements, reveals bottlenecks and fragmented actvtes, reduces cost and exposes duplcaton of efforts. Integrate. Integraton means desgnng nterfaces and brngng system elements together so they work as a whole. Ths requres extensve communcaton and coordnaton. Launch the system. Launchng the system means runnng the system and producng outputs -- makng the system do what t was ntended to do. Assess performance. Performance s assessed usng evaluaton crtera, techncal performance measures and measures -- measurement s the key.. Re-evaluaton should be a contnual and teratve process wth many parallel loops. Accordngly, a systems engneerng process starts wth dentfyng customers needs and goes through tasks from 1 to 6 n an teratve manner wth contnuous Re-evaluaton at each step untl the product/ system lfecycle s completed (Fg.1). Although SIMILAR provdes a good overall nsght on what a systems engneerng method entals, t s not qute sutable for avalablty allocaton process. Because, whle the avalablty allocaton process s performed prmarly n the desgn phase, the SIMILAR spans from cradle to grave/rebrth. Customer needs State the problem Investgate alternatves Model the system Integrate Launch the system Assess performance Product and Process Fg.1. The SIMILAR framework [11] Haskns [8] provded a breadth and depth overvew of systems engneerng methods, ncludng SIMILAR. The author dentfed 38 actvtes from the twelve most often-cted methods. Followng a detal analyss, a smple and jargon-free method, called SPADE emerged wth the ntenton that t can be appled wthout requrng years of systems engneerng tranng. Ths makes the method sutable for our purpose as avalablty allocaton s normally carred out by relablty engneers. SPADE s an acronym for Stakeholder (.e., dentfcaton of stakeholders and recognton of ther mportance and needs), Problem formulaton, assessng Alternatves (or Analyss), Decson-makng and Evaluaton. Lke SIMILAR, the method conveys, as shown n Fg.2, that the realzaton of a successful system/soluton s an teratve process comprsng a careful evaluaton at each step/mlestone. Parallel to SPADE, Haskns [13] proposed a 6Cs meta-framework,.e., Communcaton, Comprehenson, Contnuty, Collaboraton, Cooperaton and Coordnaton. The ntent of 6Cs metaframework s to complement the SPADE method. Because systems engneerng addresses problems n nterdscplnary and teratve manner, these 6C attrbutes are essental to enable the work of systems engneerng. The 6Cs are essental but not dscussed n ths paper and reference s made to [13]. Fg. 2. The SPADE framework [8] IEOM Socety Internatonal 1489

3 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 Ths paper demonstrates the applcaton of the SPADE method n the avalablty allocaton process alongsde the applcaton of the avalablty allocaton method proposed by Jgar et al. [5]. Therefore, attrbutes assocated wth the SPADE method n ths paper are delmted wthn the context of the appled avalablty allocaton method,.e., the method n [5]. The outlne of the paper follows the SPADE method after ntroducng the fundamentals of the avalablty calculaton n Secton II. Secton III dscusses stakeholder dentfcaton and problem formulaton actvtes. These actvtes are followed by Alternatves, presented n Secton IV, to choose an alternatve soluton that addresses the formulated problem havng the am to meet the dentfed stakeholders needs. Actvtes assocated wth decson-makng and evaluaton are presented n Secton V. Fnally, Secton VI presents some concludng remarks and ndcatons for further research work. II. AVAILABILITY CALCULATION Because avalablty has dfferent meanng and nterpretaton for safety systems and producton systems, t s mportant to dstngush them here. Most safety systems are dormant and are requred to perform ther functon upon demand. Part of the falures s hdden and s revealed only durng regular proof testng [14-16]. As a result, the avalablty formula for safety systems s dfferent from producton systems, as falures n a producton system reveal themselves mmedately when the system stops or shows some knd of state transton. The focus of ths paper s only on producton systems, such as subsea producton and processng systems. In order for the desgn of such systems to meet the avalablty target, the restoraton actvtes throughout the system lfecycle ncludng the avalablty, needs, and requrements of the stakeholders who are performng restoraton actvtes should be taken nto account. Fg. 3 demonstrates the man actvtes requred to put a subsea producton and processng system back to normal operaton upon a crtcal component falure. Four man locatons are shown: () where the system s operatng (subsea), () servce base located onshore where captal spares are placed and mnor refurbshment and test actvtes are carred out and, () component suppler (dependng on the component t may be located outsde the country) where major refurbshment actvtes are carred out, and (v) where the nterventon vessel(s) s located, whch s not often statc unless dedcated vessel(s) s avalable. When a falure occurs, the vessel s sent to the servce base to pck-up the spare and, at the same tme, the test actvty at the servce base ntates, snce t s "rsky" to put a spare nto operaton wthout proof testng. When the test actvty s completed and the vessel arrves, the spare s transported to the ste and the component replacement actvty ntates mmedately. The elapsed tme untl the system s restored s referred to as restoraton tme. Now, the servce base lacks one spare. As sparng has an mpact on the avalablty of the system (.e., f a lke component fals before the faled one s restocked) the tme requred to refurbsh the faled component needs to be consdered. Dependng on the severty of the falure, the refurbshment can be performed at the servce base or at the suppler's workshop. Thus, refurbshment tme s the perod from when the module s retreved from the servce based untl t s restocked. Is major repar requred? Interventon vessel Component suppler Replacement Actvty Servce base (Testng and loadng of the spare actvty) Sea surface Subsea producton and processng system Fg. 3. Man actvtes consdered n avalablty calculaton Avalablty ( A ) of a component s a functon of tme to falure and the tme needed to put the component back to operaton after falure [1]. The (average) tme requred to put a component back to normal operaton s referred to as mean tme to IEOM Socety Internatonal 1490

4 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 restoraton (MTTR), whch covers all techncal and logstc actvtes. The steady state avalablty for component wll then be MTTF A = (1) MTTF + MTTR where mean tme to falure (MTTF) and MTTR are assumed to be exponentally dstrbuted wth rate λ and μ. For tractablty purpose, t may also be reasonable to assume that the system comprses n ndependent components connected n seres and all components are equally crtcal for the (un)avalablty of the system. The assumpton of seres archtecture may stem from the fact that the allocaton s performed n the early desgn phases where a detal system archtecture s not yet avalable. Further, snce the assumpton s that each component s equally crtcal for the system, the probablty that the system s functonng gven one out of n components falure s zero. Hence, the system avalablty s n = n n 1 MTTF μ 1 σ = 1 MTTF + MTTR = 1 λ + μ = 1 (2) A = = + where σ = λm can be regarded as the current unavalablty contrbuton of component, and for brevty purpose M s used nstead of MTTR. Ths avalablty formula s referred to as nherent avalablty or desgn-n avalablty [4], and a soluton based on ths method s hgh-level n the sense that the allocaton s carred out by mprovng the MTTF and MTTR wthout detalng how. III. IDENTIFICATION OF STAKEHOLDERS AND PROBLEM FORMULATION Identfcaton of stakeholders s dependent on how the engneer perceves the problem and how s/he ntends to solve t [17]. It s thus mportant that the system provder frst understands the problem well from dfferent perspectves. Accordng to Fet [18], a systems engneerng process begns wth dentfyng the needs, whch s carred out by answerng three questons: 1) What s needed? 2) Why s t needed? 3) How may the need be satsfed? As ponted out by Haskns [8], there s an overlap between stakeholder dentfcaton and problem formulaton tasks and these two can be performed n parallel. Therefore, n the followng, we wll address the problem formulaton n Secton A, based on ths, stakeholder dentfcaton and mappng of avalablty mprovement constrans to the dentfed stakeholders wll be presented n Secton B, and, fnally, how may the [stakeholders ] needs be satsfed? wll be dscussed n Secton C. A. What s needed? Why s t needed? In the early desgn phase, desgn and operatonal concepts should be assessed carefully so that stakeholders needs can be understood and acheved. Where avalablty mprovement s necessary, e.g., when the avalablty target s not met, engneers are challenged by how the necessary mprovement can be translated down to the level of component relablty and restoraton tme wth a mnmum cost. The target s generally set out by the owner of the system and the system ntegrator/desgner strves to translate the target nto component level wth mnmum cost and effort [2]. The frst step perhaps would be to breakdown the system desgn (or concept) to module, subsystem, tem, element or components level, dependng on the progress of the system desgn process. Fg.4 llustrates the translaton of system avalablty target (.e., expressed as the proporton of tme the system s avalable for producton) down to component relablty goals and restoraton tme goals. IEOM Socety Internatonal 1491

5 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 SYSTEM AVAILABILITY TARGET Subsystem 1... Subsystem 2 Subsystem j... Component 1 Relablty goal (1) Component 2 Restoraton tme goal (1) Relablty goal (2) Restoraton tme goal (2)... Component n Relablty goal (n) Restoraton tme goal (n) Fg.4. System breakdown The problem can mathematcally be formulated as follows: If At be the avalablty target and σ a be the allocated unavalablty contrbuton of component, = 1,", n, then the allocaton problem can be stated as [5] n Mn f (σ, σ a ) =1 1 n Subject to 1 + σ = At =1 where f s the cost or complexty functon to reduce σ to σ a. (3) B. Stakeholders dentfcaton Accordng to the stakeholder framework suggested n [6], stakeholder dentfcaton attrbutes can be seen from three nonmutually exclusve perspectves: power, legtmacy, urgency. A stakeholder possesses one or more of these attrbutes, but, n a general context, we can see stakeholders as those who drectly or ndrectly affect or are beng affected by the system durng ts lfecycle [19]. As mentoned, hgh avalablty can be acheved by ncreasng the relablty (preferably) and/or decreasng the restoraton tme of components. Relablty can be ncreased through hgh-relablty components and/or hgh level of redundancy [20]. Restoraton tme s dependent on the tme requred to make mantenance supports readly avalable to perform the repar/replacement actvty (.e., logstcs such as spare, mantenance crew and ther moblzaton) and the tme requred to perform the actual repar/replacement actvty (Fg. 3). To reduce the restoraton tme the followng optons can be consdered: mnmze warehouse-testng tme: Spares often need a proof-test before they are moblzed to replace a faled module. The duraton can be mnmzed through, for example, ncreasng the test and refurbshment frequency of spares whle n the warehouse, ncrease the crew sze to perform the test upon demand as well as mprovng concurrency. mnmze logstc delays, such as spare and vessel/rg moblzaton tmes, and other admnstratve delays. mprove spare avalablty through acqurng suffcent amount and/or mnmzng lead-tmes. mnmze module replacement tme, for example through effectve modularzaton, ncreasng crew sze and mprovng concurrency. Preventve mantenance may also have a postve mpact on relablty but t mproves relablty at the expense of avalablty. Determnng preventve mantenance schedule and cost to maxmze the avalablty s an optmzaton problem per se [21]. Consderng a subsea producton and processng system, Fg. 5 llustrates some of the dentfed stakeholders and ther nteractons. The system ntegrator s oblged to provde the system wth a statstcal evdence about the avalablty IEOM Socety Internatonal 1492

6 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 performance of the system as specfed by the operatng company. These two partes, namely system ntegrator and operatng company, can be consdered prmary stakeholders. Because some desgn and operatonal specfcatons should be n-lne wth government regulatons, the socety/government also s regarded as prmary stakeholder. Further, the relablty data for the avalablty calculaton s obtaned from equpment supplers, and the restoraton tme data manly from the operatng company. As mentoned, restoraton tme s mpacted by servce provdng companes such as vessel supplers. Therefore, companes such as an equpment suppler and servce provder are consdered secondary stakeholders. The fgure ndcates also the teratve process of avalablty allocaton, whch can be related to the SPADE method depcted n Fg.2. Equpment Suppler Vessel suppler Relablty data Moblzaton tme Restoraton tme System Desgner/ Integrator Requrements Operator Requrements Standards and regulatons Avalablty allocaton Feedback wth updated data Fg. 5. Relatonshp among man stakeholders Once stakeholders are dentfed, ther needs should be recognzed and followed-up throughout the lfecycle of the system. Care should be taken when new stakeholders are dentfed wth ther own needs and perhaps wth an alternatve soluton as t may cause conflct of nterests wth exstng ones. The followng can serve as examples for stakeholders needs. No more than one restoraton actvty n 5 years assocated to component X The avalablty mprovement cost for the system should not be greater than $1.5M Total mantenance cost durng the system s msson tme should not be greater $100M No relablty mprovement should take place on component X Part of the problem formulaton and recognzng stakeholders needs s to map the avalablty mprovement optons, opportuntes and constrants to the relevant stakeholders (Fg. 6). To demonstrate the mappng, consder any sngle component from Fg. 4. As mentoned, the preferred avalablty mprovement route s through bult-n relablty, whch can be performed through one of the followng optons: purchasng a commercally avalable component that s more relable than the currently avalable one, ntroducng redundancy or carryng out a relablty development program [20]. Ths mprovement route s related to captal expendture (CAPEX), prmarly performed by the system ntegrator although t s of hgh nterest to the operatng company. The other route s through restoraton tme (and preventve mantenance), whch s an operatonal expendture (OPEX) and s of prmary concern to the operatng company. The am of ths mappng s to generate alternatves to mprove the avalablty of a component, wth the assocated estmates for mprovement cost as nput to the avalablty allocaton model. C. How may the need be satsfed? In order to protect the Evaluaton from beng flexble and stuaton-dependent, measures of effectveness (MOEs) need to be stated n unambguous and measureable manner durng problem formulaton. MOEs are standards aganst whch the "goodness" of the soluton may be assessed [22], and are drectly related to the objectve, purpose or target of the system under consderaton. MOEs can also be seen as measure of performance and success crtera for evaluatng the progress and achevement of the process that brngs the system nto beng [8]. Although they are somehow subjectve to the stakeholder who has the rght to set or nfluence the crtera, MOEs measure the qualty that stakeholders needs are to be met, or, smply, the qualty of the soluton. An example of a MOE can be that the confdence of attanng the avalablty target wth mnmum IEOM Socety Internatonal 1493

7 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, 2016 cost (proposed by the soluton) should be greater than 95%. As can be notced, MOEs affect the decson on How may the need be satsfed?.e., the type of avalablty allocaton method (Alternatve) that needs to be appled. Avalablty mprovement for component Secondary stakeholder Prmary stakeholder Yes Can mprovement be acheved through commercally avalable component Operator No Component suppler System ntegrator CAPEX Redundancy Servce provder Relablty development Operator System ntegrator Reducng restoraton tme OPEX Preventve mantenance Combnaton of two or more optons from the above The avalablty mprovement should be compensated by other components No Can the avalablty mprovement be acheved by one of the optons Yes Estmate the cost for nput to the avalablty allocaton model Fg. 6. An llustraton for mappng avalablty mprovement optons to the relevant stakeholders IV. ALTERNATIVE Durng problem formulaton, stakeholders may propose alternatve solutons to address the problem [8]. However, these alternatve solutons should be analysed for optmalty n the sense that stakeholders needs and MOEs are met [18]. Avalablty allocaton methods have exsted for several decades wth varyng degrees of complexty [7]. Jgar et al. [5] argue that n the early desgn phase smple and ntutve allocaton methods are more practcal than manstream allocaton methods. Consequently, the authors proposed a new method that s adopted from a well-known relablty allocaton method,.e., mnmum effort algorthm. In ths paper, as a contnuaton of Jgar et al. [5], we suffce wth descrbng the procedure by whch the allocaton s performed and for detal dscusson of the method reference s made to the artcle. A. Allocaton procedure (cost based) 1. Arrange the current unavalablty contrbutons n descendng order,.e., σ 1 σ 2 σ 3 " σ n. IEOM Socety Internatonal 1494

8 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, Determne weakest components that need to be mproved, and that can be done by fndng the maxmum value of such that n +1 σ > where σ n Ut σ = +1 t = 0 and Ut = 1/ A 1 =ψ (4) The allocated unavalablty contrbutons are ψ k σ a = σ > k Calculate the unavalablty contrbuton reducton factor for component,.e., β = σ a / σ. Calculate the relatve relablty and restoraton tme mprovement costs as rλ = 6. Cλ2 ( Cλ + CM ) k Cλ =1, rm = CM2 ( Cλ + CM ) k CM (6) =1 Determne θ such that β = ( rλ rm 7. (5) ) θ ln( β ) ln(rλ rm ) (7) M a = rmθ M (8) θ = The allocated λ and M wll then be λa = rλθ λ, and the rest n k reman unchanged. As ndcated n Fg. 6, there may be stuatons where there s no budget (or desre) for mprovement for the falure rate or the restoraton tme of a component. In ths case, the value of the assocated relatve cost r should be 1. The relatve cost calculaton for other components should not take ths falure rate (or restoraton tme) nto account. Ths can be done by settng the mprovement cost of the relablty or restoraton tme equals to zero whle calculatng the relatve costs for others, and then assgn 1 as ts value for r. Further, n stuatons where mprovement s known to be mpractcal (or undesred) for both the falure rate and restoraton tme of a component, the followng modfcaton s necessary. Suppose no mprovement s allowed for component l, the unavalablty contrbuton reducton factors should be adjusted as k σ ka σ l, l β* = (9) σ B. Allocaton procedure (complexty based) If component mprovement cost s dffcult to estmate or allocatng the avalablty target wth mnmum cost s of less relevance, t may be mportant nstead to consder relatve complexty. Jgar et al. [5] have proposed a calbraton method that may replace component s mprovement cost (C ) by a complexty score ( Z ). The complexty score for component measures how dffcult t s to perform an mprovement compared to other components. Wth ths complexty-based allocaton, the ntenton s to mnmze the total mprovement effort (or complexty ). The steps presented n [5] are repeated below: IEOM Socety Internatonal 1495

9 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, Apply steps 1 to 3 n Secton A to dentfy the weakest components. 2. Determne, among the dentfed components, the most complex (dffcult) falure rate or restoraton tme to mprove. Wth respect to falure rate, the complexty may be the dffculty to ncrease the MTTF per hour and for restoraton tme to ncrease the restoraton rate per hour. 3. Assgn a maxmum score ( Z) assocated wth ths falure rate or restoraton tme. The maxmum score can be 10 or any other order. Note that the maxmum score does not mply that mprovement s not gong to take place. If there s a desre however to not allow mprovement, must be omtted or can be set to zero. 4. Assgn scores for the remanng falure rates and restoraton tmes relatve to ths maxmum score. 5. Apply equaton 6 to calculate the relatve complexty factors for falure rates and restoraton tmes. The same procedure as n Secton A can then be appled to perform the allocaton. V. EVALUATION AND DECISION-MAKING Decson-makng s a process that could suffer from bas f two or more stakeholders present conflctng objectves. The trade-off s to choose among the many alternatve avalablty mprovement solutons [23]. The allocaton method presented above may help to fnd a balanced soluton gven stakeholders' needs. The method allocates system avalablty target down to the level of component relablty and restoraton tme such that the total mprovement effort (cost/complexty) s mnmzed. Ths s also to mean that a certan "amount" of performance mprovement wth the assocated cost wll be allocated to a stakeholder such that the requred system performance s acheved wth mnmum total mprovement cost. As clearly ndcated n [8], decson-makng s not a once and done event. The decson made at a certan step should be evaluated, communcated to relevant stakeholders and refned as desgn of the system progresses and more data become avalable. As can be seen from the SPADE method, Fg 2, evaluaton s located n the centre to convey contnuous evaluaton whle the allocaton process s evolvng. Each task, namely stakeholder dentfcaton and recognton of ther needs, problem formulaton, alternatve assessment and decson-makng, needs to be contnuously evaluated. Accordng to Haskns (2008a: 28), contnuous evaluaton means beng open to recognzng new stakeholders, allowng ther late-arrvng vewponts to nfluence the framng of the problem and the lst of alternatve optons. Decsons are evaluated based upon measures of performance and success crtera dentfed at the tme of problem formulaton. VI. CONCLUSION Followng a systems engneerng framework, the paper proposes a new perspectve to carry out avalablty allocaton for complex systems. The paper uses the systems engneerng SPADE method, and the beneft s that t helps to conduct avalablty mprovement analyss n an teratve and systematc manner. The central dea of the method s to structure the avalablty allocaton problem and assessment of the soluton n such a way that stakeholders needs are met wth the qualty as specfed n the measures of effectveness (MOEs). Measures of effectveness (MOEs) are performance crtera that are set n the problem formulaton phase so that the avalablty mprovement process strves towards fndng a soluton that satsfes that crteron. Contnuous evaluaton of the progress of the avalablty allocaton process s also another core dea of the method. The proposed perspectve s a concept that has not been tested by usng a realstc case study and t s our suggeston for further research work to mplement t aganst a realstc case study. In addton, the only focus n ths paper has been on producton systems. However, system avalablty s a key measure for safety systems as well and consderaton of the systems engneerng perspectve for such systems would also be of nterest. Further, mplementng effectve modularzaton s a common and effcent measure to mprove avalablty through reducng the restoraton tme. Modularzaton requres a holstc perspectve snce modularzng a certan subsystem affects the remanng subsystems. The proposed method can be further developed to take nto account modularzaton ssues n detal. REFERECES 1. Rausand, M. and A. Høyland, System relablty theory: models, statstcal methods, and applcatons. Vol : John Wley & Sons. 2. Elegbede, C. and K. Adjallah, Avalablty allocaton to reparable systems wth genetc algorthms: a multobjectve formulaton. Relablty Engneerng & System Safety, (3): p Amar, S.V. and V. Hegde. New allocaton methods for reparable systems. n Relablty and Mantanablty Symposum p IEOM Socety Internatonal 1496

10 Proceedngs of the 2016 Internatonal Conference on Industral Engneerng and Operatons Management Kuala Lumpur, Malaysa, March 8-10, McNchols, R.J. and G.H. Messer Jr, A cost-based avalablty allocaton algorthm. Relablty, IEEE Transactons on, (3): p Jgar, A., M.A. Lundtegen, and Y. Lu, A new avalablty allocaton method, n ESREL. 2015, Taylor & Francs Group, London: Zürch, Swtzerland. p Mtchell, R.K., B.R. Agle, and D.J. Wood, Toward a theory of stakeholder dentfcaton and salence: Defnng the prncple of who and what really counts. Academy of management revew, (4): p Kuo, W. and R. Wan, Recent advances n optmal relablty allocaton, n Computatonal Intellgence n Relablty Engneerng. 2007, Sprnger. p Haskns, C., Systems engneerng analyzed, syntheszed, and appled to sustanable ndustral park development. 2008, Norwegan Unversty of Scence and Technology. 9. Fraser, J. and A. Gosav. What s systems engneerng? n Amercan Socety for Engneerng Educaton INCOSE. cted November 30, 2015; Avalable from: Bahll, A.T. and B. Gssng, Re-evaluatng systems engneerng concepts usng systems thnkng. Systems, Man, and Cybernetcs, Part C: Applcatons and Revews, IEEE Transactons on, (4): p Bahll, A.T. and F.F. Dean. What s systems engneerng? A consensus of senor systems engneers. n INCOSE 6th Annual Int Symp. 1996: p Haskns, C., Usng systems engneerng to address soco-techncal global challenges. Proceedngs CSER08, Lu, Y. and M. Rausand, Relablty effects of test strateges on safety-nstrumented systems n dfferent demand modes. Relablty Engneerng & System Safety, : p Lundtegen, M.A. and M. Rausand, Relablty of safety nstrumented systems: Where to drect future research? Process safety progress, (4): p Rausand, M., Relablty of Safety-Crtcal Systems: Theory and Applcatons. 2014: John Wley & Sons. 17. Shanee, M., et al., Desgnng offshore fsh cages usng systems engneerng prncples. Systems Engneerng, (4): p Fet, A.M., Systems engneerng methods and envronmental lfe cycle performance wthn shp ndustry. Doktor Ingenøravhandlng, Clarkson, M.E., A stakeholder framework for analyzng and evaluatng corporate socal performance. Academy of management revew, (1): p Murthy, D., T. Østerås, and M. Rausand, Component relablty specfcaton. Relablty Engneerng & System Safety, (10): p Tsa, Y.-T., K.-S. Wang, and L.-C. Tsa, A study of avalablty-centered preventve mantenance for multcomponent systems. Relablty Engneerng & System Safety, (3): p Sproles, N., Formulatng measures of effectveness. Systems Engneerng, (4): p Olver, D.W., T.P. Kellher, and J.G. Keegan Jr, Engneerng Complex Systems. McGraw-Hll Companes, BIOGRAPHY Abraham Almaw Jgar s a PhD canddate n the department of producton and qualty engneerng at the NTNU. Hs man research topc s relablty assessment of safety-crtcal systems. He earned BSc. n statstcs from the unversty of Gondar, Ethopa, and masters n mathematcs from NTNU. Abraham has worked as a relablty engneer at Aker Solutons, Norway, for one year before he started hs PhD. Cecla Haskns s an Assocate Professor n systems engneerng at NTNU. Her educatonal background ncludes a BSc n Chemstry from Chestnut Hll College, and an MBA from Wharton, Unversty of Pennsylvana. She has been recognzed as a Certfed Systems Engneerng Professonal snce After earnng her Phd from NTNU, she teaches and conducts research on nnovatve applcatons of systems engneerng to soco-techncal problems such as those encountered n sustanable manufacturng and global producton systems. Mary Ann Lundtegen s a professor n relablty assessment at NTNU wth the man research focus on relablty assessment of safety-crtcal systems. She has a combned academc and ndustral background, prmarly related to the ol and gas ndustry. She has a MSc. n engneerng cybernetcs and a PhD n RAMS (relablty, avalablty, mantenance, and safety). Currently, she s headng the RAMS group at NTNU, and she s also the co-drector of the new Norwegan Center of Excellence for subsea producton and processng (SFI SUBPRO). Snce 2006, she has also been a member of the techncal commttee for updatng IEC 61511, a key standard that frames the desgn and operaton of safety-nstrumented systems for the process ndustry. IEOM Socety Internatonal 1497