A Fuzzy Model to Assess Resilience for Safety Management

Size: px
Start display at page:

Download "A Fuzzy Model to Assess Resilience for Safety Management"

Transcription

1 A Fuzzy Model to Assess Reslence for Safety Management Cláudo H. S. Grecco 1, Maro C. R. Vdal 2, Carlos A. N. Cosenza 2, Isaac J. A. L. dos Santos 1 and Paulo V. R. Carvalho 1 1 Comssão Naconal de Energa Nuclear/Insttuto de Engenhara Nuclear, Cdade Unverstára - Ilha do Fundão, Ro de Janero, RJ, Brazl, paulov@en.gov.br 2 Unversdade Federal do Ro de Janero/ Programa de Engenhara de Produção, Cdade Unverstára - Ilha do Fundão, Ro de Janero, RJ, Brazl, Abstract. The Most of methods to assess reslence cannot fully solve the subjectvty of reslence evaluaton. In order to remedy ths defcency, the am of ths research s to adopt a Fuzzy Set Theory (FST) approach to establsh a method for reslence assessment n organzatons based on leadng safety performance ndcators, defned accordng to the reslence engneerng prncples. The method uses FST concepts and propertes to model the ndcators and to assess the results of ther applcaton. To exemplfy the method we performed an exploratory study at the radopharmaceutcals dspatch package sector of a radopharmaceutcals producton faclty. 1 INTRODUCTION Contemporary vew on safety based on reslence engneerng (RE) prncples emphaszes that safety crtcal organzatons should be able to proactvely evaluate and manage safety of ther actvtes. Ths new safety paradgm must be endorsed by the organzatonal safety management to be successful. Therefore we need new methods to measure safety accordng to RE prncples, consderng that safety s a phenomenon that s hard to descrbe measure, confrm, and manage. Scentsts n the feld of safety crtcal organzatons state that safety emerges when an organzaton s wllng and capable of workng accordng to the demands of ther tasks, and when people understand the changng vulnerabltes of ther work actvtes (Reman & Odewald, 2007). Thus, safety management reles on a systematc antcpaton, montorng the evoluton of trade-offs, n whch varous safety ndcators play a key role n provdng nformaton on current organzatonal safety performance. An ncreasng emphass has been placed also on the role of ndcators n provdng

2 nformaton to be used n antcpaton and evoluton of organzatonal performance. These ndcators are called leadng ndcators. The safety performance ndcators that have commonly been used n tradtonal safety management have often been laggng ndcators, measurng outcomes of actvtes or thngs and events that have already happened (e.g., njury rates, radaton doses and ncdents). These ndcators are reasonably objectve, easy to quantfy, and that they can be used requrng small changes n the exstng system. However, t can be questoned whether they really ndcate the actual safety of organzaton processes, because there s no sharp causal lnk between past events and the current safety performance. Laggng ndcators may be more useful to confrm effects after a whle, n long term, than to manage mmedate changes n dynamc envronments. Montorng should not rely solely on laggng ndcators but also on ndcators of current actvtes and the potental of the organzaton to succeed n the future. To quckly montor trade-offs, the effects of good work practces, as well as, to antcpate vulnerabltes, the organzatons should defne leadng ndcators. Those should be able to grasp organzatonal practces and processes that lead changes n safety performance of the people n the organzaton. Several reasons for usng leadng ndcators are: (a) they provde nformaton on where to focus mprovement efforts; (b) they drect attenton to proactve measures of safety management rather than reactve follow up of negatve occurrences or trendng of events; (c) they provde early warnng sgns on potental weak areas or vulnerabltes n the organzatonal rsk control system or technology; (d) they focus on precursors to undesred events rather than the undesred events themselves; (e) they provde nformaton on the effectveness of the safety efforts underway; and (f) they tell about the organzatonal health, not only sckness or absence of t. The am of ths research s to adopt a Fuzzy Set Theory (FST) approach to establsh a method for safety assessment n organzatons based on reslence engneerng usng leadng performance ndcators, as the bass for a safety management system. To exemplfy the method we performed an exploratory case study at the process of radopharmaceutcals dspatch package n a radopharmaceutcals producton faclty. 2 METHOD FOR RESILIENCE ASSESSMENT The method has the followng phases: (1) Selecton of the leadng ndcators; (2) Determnaton of a reslence deal pattern; (3) Assessment of the actual reslence level compared wth the pattern. 2.1 Selecton of leadng ndcators Selecton of leadng ndcators should always start from the consderaton of what

3 are the key ssues to montor, manage and change (EPRI, 2000). The leadng ndcators are utlzed as part of the reslence management process, not as an ndependent goal or functon as such. The operatonalzaton of an ndcator s called metrc. A metrc denotes how the ndcator s measured, whereas an ndcator denotes somethng that one wshes to measure wth the use of one or more metrcs. The selecton of the reslence themes addressed and leadng ndcators used n radopharmaceutcals dspatch package sector was based on prevous ergonomc study (Grecco et al., 2010) and are descrbed n table 1. Table 1. Themes and Leadng ndcators. Themes Indcators Themes Indcators Top-level commtment 1.1 Human resources 1.2 Materal resources Awareness 4.1 Reports of problems 4.2 Informaton securty 1.3 Safety commtment 4.3 Communcaton 1.4 Safety polcy 4.4 Team work 1.5 Procedure management 1.6 Tranng programs 1.7 Competence selecton 4.5 Workload 4.6 People relatons 4.7 Tasks and sklls 4.8 Awareness of lmtatons 4.9 Preventve mantenance 4.10 Proactve actons Learnng culture 2.1 Informaton dssemnaton 2.2 Informaton flow 2.3 Work management 2.4 Actual workng practces Just culture 5.1 Reportng of devatons/worres 5.2 Understandng of errors 5.3 Percepton of errors 5.4 Actons are not puntve 5.5 Peer assessments 2.5 Local adaptatons 5.6 Professonal recognton 2.6 Content of the documentaton 2.7 Avalablty of the documentaton 2.8 Analyss of ncdents 2.9 Investgatons of ncdents and accdents

4 Flexblty 3.1 Ablty to cope wth unexpected 3.2 Capacty for flexblty 3.3 Safe workng lmts 3.4 Reports on adaptatons 3.5 Incorporaton of adaptatons Preparedness 6.1 Emergency plan 6.2 Identfcaton of rsks 6.3 Safety equpments 6.4 Alarm system 6.5 Proactve procedures 6.6 Emergency tranng 2.2 Reslence deal pattern The second phase of the method s to obtan from experts n radopharmaceutcals producton and reslence engneerng ssues the degree of mportance of each ndcator metrc, so that the organzaton sector can be consdered reslent. Ths means that the degree of mportance assgned to each ndcator by the specalst, should show how the sector should be to acheve an deal reslence level. Thus, t s not evaluatng the sector, but the deal of reslence that t should have. The phase has the followng steps: 1) Experts selecton, 2) Calculaton of each expert relatve mportance, based on knowledge and experence, 3) Choce of lngustc terms and membershp functons, 4) Determnaton of the mportance degree of each ndcator, 4) Aggregaton of fuzzy opnons, 5) Reslence pattern. Calculaton of experts relatve mportance. The relatve mportance of the expert was calculated on the bass of experts attrbutes (experence, knowledge of radopharmaceutcals producton safety and knowledge of the dspatch package radopharmaceutcals). We used a questonnare (Q) to dentfy the profle. Each questonnare contans nformaton of a sngle expert. The relatve mportance (RI) of expert E ( = 1, 2, 3,, n) s a subset μ (k) Є [0,1] defned by equaton 1. Referrng to equaton 1, tq, s the total score of the expert. RI n tq 1 tq (1) Choce of lngustc terms and membershp functons. Each leadng ndcator can be seen as a lngustc varable, related to a lngustc terms set assocated wth membershp functons. These lngustc terms are represented by trangular fuzzy numbers to represent the mportance degree of each ndcator. It s suggested that the experts employ the lngustc terms, U (Unmportant), LI (Lttle Important), I (Important) and VI (Very Important) to evaluate the mportance of each ndcator.

5 Aggregaton of the fuzzy opnons. The smlarty aggregaton method proposed by Hsu and Chen [21] s used to combne the experts opnons whch are represented by trangular fuzzy numbers. The agreement degree () between expert E and expert Ej s determned by the proporton of ntersecton area to total area of the membershp functons. The agreement degree () s defned by equaton 2. mn ( x), ( x) Ñ Ñj x max ( x), ( x) Ñ Ñj x If two experts have the same estmates, that s, =1. In ths case, the two experts estmates are consstent, and then the agreement degree between them s one. If two experts have completely dfferent estmates, the agreement degree s zero. If the ntal estmates of some experts have no ntersecton, then we use Delph method to adjust the opnon of the experts and to get the common ntersecton at a fxed α level cut (Hsu & Chen, 1996). The hgher the percentage of overlap, the hgher the agreement degree. After all the agreement degrees between the experts are calculated, we can construct an agreement matrx (AM), whch gve us nsght nto the agreement between the experts. 1 AM 1 n n2 1 j j nj dx dx (2) 1 1n n The relatve agreement (RA) of expert E ( = 1, 2, 3,, n) s gven by equaton 3. RA 1 n 2 ( ) (3) j n 1 j1 Then we calculate the relatve agreement degree (R) of expert E ( = 1, 2, 3,, n) by equaton 4 and the consensus coeffcent (CC) of expert E ( = 1, 2, 3,, n) by equaton 5. R n 1 RA RA (4) n R RI 1 R RI Let Ñ be a fuzzy number of combnng expert s opnons. So, Ñ s the fuzzy value of each leadng ndcator whch s also trangular fuzzy number. By defnton of the consensus coeffcent (CC) of expert E ( = 1, 2, 3,, n), Ñ can be defned by equaton 6. Referrng to equaton 6, ñ, s the trangular fuzzy number relatng to the lngustc terms, U (Unmportant), LI (Lttle Important), I (Important) and VI (Very Important). CC (5)

6 Ñ n CC ñ 1 (6) Reslence pattern. The reslence pattern as a reference for assessng organzatonal reslence of the s establshed by calculatng the normalzed mportance degree (NID) of each leadng ndcator that make up each property relevant to reslent organzatons. The normalzed mportance degree (NID) of each leadng ndcator s gven by deffuzfcaton of ts trangular fuzzy number Ñ (a, b, c), where b represents the mportance degree. Then, NID can be defned by equaton Reslence assessment NID NID (7) hgh value of b Ths thrd phase of the method s to assess reslence level compared to the reslence pattern. In ths phase, the lngustc values are used to assess the attendance degrees of the leadng ndcators to the radopharmaceutcals dspatch package sector gven by workers. It s suggested that the workers employ the lngustc terms, SD (Strongly Dsagree), PD (Partally Dsagree), NAND (Nether Agree Nor Dsagree), PA (Partally Agree), SA (Strongly Agree). Table 4 shows the attendance degrees and trangular fuzzy numbers for lngustc terms. Usng center of area defuzzfcaton method we calculate the attendance degree () to the reslence pattern by equaton 8. Referrng to equaton 8, adj, s the attendance degree of the leadng ndcator j of the theme n the dspatch package radopharmaceutcals sector. 3. RESULTS k j1 NID k j1 j. ad j (8) NID j The reslence assessment of the radopharmaceutcals dspatch package sector was performed by seven workers and results are presented n fgure 1. The average evaluaton of the reslence based on each ndcator was computed and showed n fgure 2. We consder satsfactory an attendance degree greater than or equal to 0.6. The result of the average evaluaton showed that the radopharmaceutcals dspatch package sector presented satsfactory learnng culture, flexblty awareness, just culture and preparedness. However, ths sector presented problems related to the toplevel commtment.

7 Fg. 1. Result of the evaluaton of the reslence by the seven workers. Fg. 2. Average evaluaton of the reslence performed by the seven workers.

8 4. CONCLUSIONS In ths paper we descrbed a method for organzatonal reslence assessment was used. We proposed a method that uses leadng ndcators and concepts and propertes of Fuzzy Sets Theory. We develop a reslence pattern usng a smlarty aggregaton method to aggregate fuzzy ndvdual opnons, consderng the dfference of mportance of each expert. A plot study n the radopharmaceutcals producton faclty shows that ths method based on leadng ndcators and fuzzy logc offers nterestng perspectves for the mplementaton of reslence engneerng prncples. Ths assessment method can be a proactve tool to provde a bass for acton wthout watng for events. Through the use of the method we dentfy problems related to leadng ndcator metrcs of the top-level commtment theme. These problems can be nvestgated n order to mplement actons to make the process of radopharmaceutcals dspatch package more effcent and secure besdes to mprove the reslence n ths sector. REFERENCES EPRI. (2000). Gudelnes for tral use of leadng ndcators of human performance: the human performance assstance package. Palo Alto, CA: Electrc Power Research Insttute. Grecco C. H. S.; Vdal M. C. R. Bonfatt, R. (2010). Análse ergonômca do trabalho no Setor de Expedção de Radofármacos de um Insttuto de Pesqusas. Proceedngs of the Congress of Brazlan Ergonomcs Assocaton. Ro de Janero. (n Portuguese) Hsu, H. M.; Chen, C. T. (1996) Aggregaton of fuzzy opnons under group decson makng. Fuzzy Sets and Systems, v. 79, pp Reman, T., Oedewald, P., (2007). Assessment of Complex Socotechncal Systems Theoretcal ssues concernng the use of organzatonal culture and organzatonal core task concepts. Safety Scence 45,