Accident Analysis and Prevention 45S (2012) 4144

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier.com/locate/aap

Research and guidelines for implementing Fatigue Risk Management Systems for the French regional airlines
Philippe Cabon a, , Stephane Deharvengt b , Jean Yves Grau c , Nicolas Maille d , Ion Berechet e , Rgis Mollard a
a Universit Paris Descartes, Laboratoire Adaptations Travail Individu (LATI), Centre Henri Piron, UFR Institut de Psychologie, 71, Avenue Edouard Vaillant, 92774 Boulogne-Billancourt Cedex, France b Direction des Services de la Navigation Arienne (DSNA), Direction Gnrale de lAviation Civile, 50 rue Henry Farman, 75020 Paris Cedex 15, France c SynRjy, Conseil Formation en Facteurs Humains, 10 rue Nicephore NIEPCE, 91410 Dourdan, France d Ofce National dEtude et de Recherche Arospatiale DCSD PSEV, Centre de Salon de Provence Base Arienne 701, 13661 Salon Air, France e Air France Consulting, Air France Consulting SISPIA, 18, Alle Henri Dunant, 94300 Vincennes, France

a r t i c l e

i n f o

a b s t r a c t
This paper describes research that aims to provide the overall scientic basis for implementation of a Fatigue Risk Management System (FRMS) for French regional airlines. The current research has evaluated the use of different tools and indicators that would be relevant candidates for integration into the FRMS. For the Fatigue Risk Management component, results show that biomathematical models of fatigue are useful tools to help an airline to prevent fatigue related to roster design and for the management of aircrew planning. The Fatigue Safety assurance includes two monitoring processes that have been evaluated during this research: systematic monitoring and focused monitoring. Systematic monitoring consists of the analysis of existing safety indicators such as Air Safety Reports (ASR) and Flight Data Monitoring (FDM). Results show a signicant relationship between the hours of work and the frequency of ASR. Results for the FDM analysis show that some events are signicantly related to the fatigue risk associated with the hours of works. Focused monitoring includes a website survey and specic in-ight observations and data collection. Sleep and fatigue measurements have been collected from 115 aircrews over 12-day periods (including rest periods). Before morning duties, results show a signicant sleep reduction of up to 40% of the aircrews usual sleep needs leading to a clear increase of fatigue during ights. From these results, specic guidelines are developed to help the airlines to implement the FRMS and for the airworthiness to oversight the implementation of the FRMS process. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 21 April 2011 Received in revised form 8 August 2011 Accepted 11 August 2011 Keywords: Fatigue Risk Management System Aircrew Sleep

1. Introduction Fatigue in aviation is widely acknowledged as a major risk for safety and has been on the most wanted list of the National Transportation Safety Board (NTSB) since 1990. In order to manage this specic risk, several approaches have been proposed, from the education and training of the crews related to the management of fatigue risk to the integration of scientic knowledge in Flight and Duty Time Limitations (FTL). New rules have recently been proposed by the Federal Aviation Administration (FAA) and the European Aviation Safety Agency (EASA). Furthermore, the International Civil Aviation Organization (ICAO) will shortly release

Corresponding author. Tel.: +33 1 42 86 20 34; fax: +33 1 55 20 57 12. E-mail addresses: philippe.cabon@parisdescartes.fr (P. Cabon), stephane.deharvengt@aviation-civile.gouv.fr (S. Deharvengt), synrjy@wanadoo.fr (J.Y. Grau), nicolas.maille@onera.fr (N. Maille), ion.berechet@sispia.fr (I. Berechet), regis.mollard@parisdescartes.fr (R. Mollard). 0001-4575/$ see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2011.09.024

Standard and Recommended Practices (SARP) and guidelines to support the aviation industry in developing Fatigue Risk Management System (FRMS). In the context of the new harmonised regulation on FTL (EUOPS-sub-part Q), France has issued a specic regulation relating to crew rest requirements. More specically, the directive applies to two specic cases, split duties and reduced rests. In the split duty provision, if a ight duty includes a break period between 03:00 h and 09:59 h, the ight duty period can be increased by half the break time, less 15 min. The reduced rest provision allows rest periods to be reduced to a minimum of 7 h and 30 min if the rest is taken in a hotel located less than 15 min from the airport. The ight duty period prior to the reduced rest period must not include more than ve sectors and the next ight duty must be limited to three sectors. A maximum of two reduced rests can be scheduled between two normal rest periods. The directive stipulates that companies wishing to benet from these specic provisions must set up a Fatigue Risk Management System (FRMS). This system, part of the operators Safety Management System (SMS) (Graeber, 2008) as dened

42

P. Cabon et al. / Accident Analysis and Prevention 45S (2012) 4144

by ICAO, is intended to ensure that these specic provisions allow for maintaining a level of safety that meets the requirements of European regulations (EU-OPS) in regard to fatigue risk. A FRMS has been dened as a scientically based, data-driven, exible alternative to prescriptive ight and duty time limitations that forms part of an operators Safety Management System and involves a continuous process of monitoring and managing fatigue risk (Graeber, 2008). In fact, a prescriptive approach has been under increased scrutiny regarding its efciency to prevent fatigue (Cabon et al., 2002; Dawson and McCullough, 2004). The main assumption is that regulations are not able to take into account the complexity of fatigue (Dawson and McCullough, 2004). For instance, because of the circadian rhythms, a break will not have the same recovery value depending on the time of the day, the timing of the break being more important than the duration of the break itself. This is the reason why alternatives to a prescriptive approach, such as FRMS, are becoming more popular. Most FRMS are based on a risk management approach applied at various levels of the organisation (Gander et al., 2011). Most also use biomathematical models that predict the risk of fatigue occurrence associated with a specic working hours pattern. Several software programs have been developed that may be useful to evaluate the risk of fatigue associated with work schedules (Dawson et al., 2011; Mallis et al., 2004). To date, several industries and airlines in the world have moved towards a nonprescriptive approach focusing on the fatigue risk management rather than on the compliance to a ight time limitation. In aviation, New Zealand has the longest experience in the development of FRMS. In 1995, the regulations were altered so that air operators could either comply with a standard prescriptive scheme or by having an alternative, company-specic scheme approved. In this last case, the operator has to take into account additional factors that may result in fatigue (Signal et al., 2008) (e.g. rest prior to duty, effects of time zone change, etc.). The introduction of Ultra Long Range ights by Singapore Airlines in 2003 is another example of a successful FRMS application (Spencer et al., 2004). In Europe, easyJet became the rst major airline to be granted exception from the current FTL in 2005 (Stewart, 2006) by means of the implementation of a FRMS. The development of these new approaches raises several issues, both theoretical and practical, such as the complex links between fatigue and safety, the multi-factorial nature and sources of fatigue, and the management of this specic risk at the various levels of the organization. In order to develop specic guidelines to allow French regional airlines to implement FRMS, the French Airworthiness Directive has commissioned the present project run by a multidisciplinary consortium and the partnership of three French regional airlines. This paper presents the methodology and the main results.

SMS
Policy and safety objecves SafetyManagement

FRMS
Dene the policy regardingaircrewfague Assign responsibilies

Design of hours of work by the means of predicve models

Safety data : Air SafetyReport, Flight Operaon QualityAssurance Fague data: sleepand fague assessment, inight observaons, website survey

SafetyAssurance

SafetyPromoon

Educaon and training on sleep and fague

Fig. 1. The FRMS four essential elements. The aspects in bold are developed in the present paper.

Karolinska Sleepiness Scale (KSS). Values of 7 or higher are associated with intrusion of sleep and an increased risk of impaired performance (kerstedt and Gillberg, 1990). For the safety assurance part, the research has evaluated the sensitivity to fatigue of two safety indicators routinely collected by airlines, the Air Safety Reports (ASR) and the Flight Data Monitoring (FDM). 3. Results 3.1. Safety management: evaluation of hours of work by the means of a predictive model of fatigue The application of the FRI model to all the duty that followed a reduced rest showed a large variability of fatigue scores ranging from 4.05 to 42.76. A more detailed analysis shows that this variability is mainly due to the position of the reduced rest in the sequence of the work schedule and therefore to a cumulative effect. For example, the fatigue risk increases dramatically from a reduced rest falling at the beginning of a week compared to a reduced rest falling at the end of the week. This suggests that reduced rests should be planned by taking into account the cumulative effects induced by the succession of disruptive hours of work. Interviews with the scheduling ofcers in the airlines suggest that this cumulative effect is not currently taken into account and that the rosters are processed as isolated blocks. 3.2. Safety assurance: ASR and FDM ASR are mandatory reports written by the captain whenever a safety event has occurred during a ight. FDM is the systematic analysis of ight parameters to detect signicant safety events, identify areas of operational risk and quantify current safety margins. Fig. 2 shows the comparison of the frequency of ASR for morning ights (starting between 06:00 h and 07:00 h) occurring during duties that followed a standard rest versus duties that followed a reduced rest. The results show a clear interaction between the length of the previous rest and the length of the duties. For short duties, the length of the rest has no signicant impact on the frequency of the ASR. Between 3 and 5 h of duty, the frequency of ASR is signicantly higher during the ights that followed a reduced rest compared to ights that followed a standard rest. Surprisingly, for the longest duties, frequency of ASR is no longer signicantly different between reduced rest and standard rests. These results suggest a link between the tours of duty, the duty length and the frequency of ASR.

2. Methods The general methodology of the project is structured around the four essential elements of FRMS (Fig. 1) that are supposed to mirror the SMS structure. This paper is focused on aspects relating to safety management and safety assurance. For the safety management aspect, the evaluation of the fatigue risk associated with the reduced rests has been conducted by the means of a predictive model of fatigue, the Fatigue Risk Index (FRI) (Spencer and Robertson, 2007). This model has been selected as it comprises a cumulative component that enables evaluation of the effect of sequences of duties rather than of isolated duty. All sequences of duties that comprised a reduced rest have been extracted from a 12-month period of work schedules of all crews and input into the model. The model output is the estimated average probability, multiplied by 100, of a value of 7 or more on the

P. Cabon et al. / Accident Analysis and Prevention 45S (2012) 4144

43

Frequency of ASR (occurrence per 1000 h of flight)

3.0 2.5 2.0 1.5 1.0 0.5 0.0

- - -morning duties after reduced rests

morning duties after standard rests

0 to 3 h

03h-05h

>5h

Time Since Duty Start (h)

Fig. 2. Frequency of ASR during duties after reduced and standard rests as a function of duty length (n = 563 ASR).

For the purpose of FDM analysis, all individual aircrew schedules have been categorized according to three levels of fatigue risk (high, moderate, low) on the basis of a statistical analysis of six fatigue criteria: % of reduced rests % of split duties Number of backward duties % of duties starting <06:00 h % of more than ve consecutive duty days Number of ights/day

Once each individual aircrew schedule was categorized, they were combined for each ight to take into account the schedule of the captain and the rst ofcer. Therefore, each ight was categorized on a scale of ve levels of crew fatigue risk: Low/Low: fatigue 2 Low/Moderate: fatigue 1 Moderate/Moderate: fatigue 0 High/Low: fatigue +1 High/High: fatigue +2

Fig. 3. Frequency of FDM events per 1000 ights (Panel A all events; Panel B Class 3 events) as a function of fatigue risk levels. Fatigue 2: very low risk of fatigue for both pilots, fatigue +2: very high risk of fatigue for both pilots.

The next step was the extraction of the events associated with each ight from the FDM database. These events are automatically categorized in three classes depending on the level of exceedance (speed, altitude): Class 1: low (65,541 events). Class 2: moderate (2846 events). Class 3: high (338 events). Then the frequency of events per 1000 ights was computed. Fig. 3 shows the results for the ve classes of fatigue risk levels. Results show that when all the events are taken regardless of the exceedance level, the frequency of events is signicantly lower for the highest fatigue risk level. On the contrary, when taking only the Class 3 events, an increase in frequency is observed for the highest fatigue risk level, with a signicant increase in Class 3 events. 4. Discussion and conclusion This paper presents results of a project aimed at dening the principles of a FRMS that will be required to monitor the impact of reduced rests on fatigue and safety in the context of the European regulation on Flight and Duty time. The FRMS components are based both on existing data that are already integrated into the

current or future Safety Management Systems of airlines to manage safety and specic tools that are more fatigue-oriented. One of the key features of this FRMS is the introduction of a predictive model of fatigue into the aircrew scheduling process. The estimation of fatigue associated with rosters that include a reduced rest using a biomathematical model shows a large variability in the predicted risk of fatigue depending on the position of the roster in the duty schedules sequence. This clearly shows that the scheduling process should integrate the cumulative aspect of fatigue. Several airlines have started to use these models to improve aircrew scheduling and rst attempts have been proposed to connect them with existing aircrew scheduling software (Klemets and Romig, 2009). However, the introduction of predictive models still raises issues and should be used carefully. In fact, the use of predictive models by non-experts leads to a risk of oversimplication and overreliance (Dawson, 2009). Therefore the use of these models requires appropriate training of users that will provide the necessary knowledge of the rationale and the limitations of biomathematical models. With regard to the management of risk, one of the challenges of this research is the identication of the underlying links between fatigue and safety. In fact, in low-reliability systems such as car driving, fatigue is directly responsible for around 30% of accidents (Cabon et al., 2001). In more complex and reliable system such as aviation, the link between fatigue and safety is much less linear as individual fatigue can be partially compensated by crew cooperation and automation. The non-linearity between fatigue and risk is well illustrated by our data on the impact of duty schedules on ASR and FDM events frequency. For the duties that followed a reduced rest, there is a clear increase of the frequency between 3 and 5 h of the duties but

44

P. Cabon et al. / Accident Analysis and Prevention 45S (2012) 4144

a subsequent reduction after 5 h of duty (i.e. 11:00 h to 12:00 h). This trend is similar to previous results published by Folkard (1997) that revealed a time on task effect with a pronounced but transient 24 h peak in risk. These results raise several questions regarding the meaning of the ASR with regard to aircrew fatigue. In fact, the decrease in the frequency of ASR with duty length could be explained by the fact that when fatigue is reaching a high level, aircrews tend to develop strategies to protect their performance and hence decrease the risk (e.g. rely more on automation, adhere more strictly to standard procedures, apply greater control over their own performance, or monitor more closely the performance of colleagues). The FDM data conrm a signicant effect of duty schedules. Surprisingly, the frequency of events signicantly decreased with the more disruptive schedule. One possible explanation is that tired aircrews tend to rely more on the autopilot and therefore reduce the frequency of Class 1 and 2 events. However, when the risk of fatigue is high, more serious exceedance levels are likely to occur. Therefore, other models have to be considered to account for the complex relationship between fatigue and safety. One of the key factors that could explain this complex link would be fatigue awareness. In fact, being aware of ones own fatigue may lead to the development of strategies to either reduce fatigue or to protect performance. These performance protection strategies developed by aircrew when tired were observed by Petrilli et al. (2006). On the contrary, being unaware would not lead to the development of these strategies and thus impacting performance and consequently safety. This level of awareness is likely to be low at intermediate levels of fatigue, leading to an increase of safety events. From a FRMS perspective, these results show that both safety and fatigue data should be collected to better evaluate the actual risk of operations. Acknowledgments This research is supported by a DGAC grant. It is carried out in cooperation with Airlinair, Britair and Regional in partnership with Airbus. The authors wish to express their gratitude to all the crews who participated in the data collection.

References
kerstedt, T., Gillberg, M., 1990. Subjective and objective sleepiness in the active individual. The International Journal of Neurosciences 52, 2937. Cabon, P., Cointot, B., Coblentz, A., 2001. Baisse de vigilance et accidents sur autoroutes. Urgence Pratique 47, 7578. Cabon, P., Bourgeois-Bougrine, S., Mollard, R., Coblentz, A., Speyer, J.J., 2002. Flight and duty time limitations in civil aviation and their impact on crew fatigue: a comparative analysis of 26 national regulations. Human Factors and Aerospace Safety: An International Journal 2, 379393. Dawson, D., 2009. A critical review of current fatigue and performance models and their use. In: Proceedings of the 2009 International Conference on Fatigue Management in Transportation Operations, Boston. Dawson, D., McCullough, K., 2004. Managing fatigue as an integral part of a fatigue risk management system. In: Proceedings of the 2004 International Society of Air Safety Investigators, Gold Coast, pp. 114. Dawson, D., Noy, Y.I., Hrm, M., kerstedt, T., Belenky, G., 2011. Modelling fatigue and the use of fatigue models in work settings. Accident Analysis and Prevention 43, 549564. Folkard, S., 1997. Black times: temporal determinants of transport safety. Accident Analysis and Prevention 29, 417430. Gander, P., Hartley, L., Powell, D., Cabon, P., Hitchcock, E., Mills, A., Popkin, S., 2011. Fatigue risk management: organizational factors at the regulatory and industry/company level. Accident Analysis and Prevention 43, 573590. Graeber, C., 2008. Fatigue Risk Management System within SMS. In: Proceedings of the Aviation Fatigue Management Symposium Partnership for Solutions, Tysons Corner. Klemets, T., Romig, E., 2009. Flight crew scheduling based on fatigue risk guidance. In: Proceedings of the 2009 International Conference on Fatigue Management in Transportation Operations, Boston. Mallis, M.M., Mejdal, S., Nguyen, T.T., Dinges, D.F., 2004. Summary of the key features of seven biomathematical models of human fatigue and performance. Aviation, Space and Environmental Medicine 75, A4A14. Petrilli, R.M., Thomas, M.J.W., Dawson, D., Roach, G.D., 2006. The decision-making of commercial airline crews following an international pattern. In: Proceedings of the Seventh International AAvPA Symposium, Manly. Signal, T.L., Ratieta, D., Gander, P.H., 2008. Flight crew fatigue management in a more exible regulatory environment: an overview of the New Zealand aviation industry. Chronobiology International 25, 373388. Spencer, M., Robertson, K., 2007. The application of an alertness model to ultra-longrange civil air operations. Somnologie 11, 159166. Spencer, M.B., Robertson, K.A., Cabon, P., Mollard, R., Akerstedt, T., Guillberg, M., Simon, R., Valk, P., Samel, A., Gundel, A., 2004. Aircrew alertness on the SingaporeLos Angeles route: nal report. QinetiQ Report No. QINETIQ/KI/CHS/CR050022/1.0, December 2004. Stewart, S., 2006. An integrated system for managing fatigue risk within a low cost carrier. In: Proceedings of the International Aviation Safety Seminar, Flight Safety Foundation, Paris, France, October 2326, 2006.