53 pixels 4/4/16, 9:11 am a scientific approach for the fatigue … · 2017. 8. 21. · alfredo...
TRANSCRIPT
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16
A Scientific Approach for the Fatigue Risk Management in the Brazilian Civil Aviation
1
Alfredo Menquini2, Arthur Lobo2, Luciano Baia4, Paulo Licati2, Philipe Pacheco4, Raul Bocces3, Tiago Bertalot3, Tiago Rosa1, Tulio Rodrigues3,5, Victor Casseta4
1Crew Member National Union - SNA 2Brazilian Civil Aviation Pilots Association - ABRAPAC
3Gol Airline Crew Member Association - ASAGOL 4TAM Airline Crew Member Association - ATT
5Physics Institute, University of São Paulo
In partnership with:
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 2
1) Overview of the SAFTE-FAST model
2) Results:
• Pilot fatigue in Brazil
• Risk analyses: scenarios, fatigue risk & risk exposure
• FDT Limits: a science based proposal for minimum crew
3) Conclusions
4) Future plans
Outline
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 3
(*) Adapted from S. Hursh et al., Aviat Space Environ Med 2004; 75(3 Suppl): A44-53
Homeostatic Process
Circadian Rithmin
Sleep Inertia
Effectiveness ~ (Reaction Time)-1 Hours awake
Reservoir ê Hours of sleep
Reservoir é
Sleep reservoir
Sleep Intensity
Overview of the SAFTE-FAST model*
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 4
Results/pilot fatigue in Brazil
(*) Licati, PR; Rodrigues, TE; Wey, D; Fischer, FM; Menna-Barreto, L. (2015) Conexão Sipaer, Vol. 6, No. 1, pp. 7-17.
14
40 50 60 70 80 90 1000
10
20
30
40
50
60
E
vent
s
Effectiveness (%)
This work (Licati et al., 2015) Gaussian FIT
Figure 4: Distribution of pilot effectiveness at the time of fatigue sensation
according with the SAFTE-FAST predictions (Licati et al., 2015).
Combining the information presented in figures 4, 5 and 6 one can
easily verify that the sample of pilots that was studied presented
low levels of effectiveness with almost 50% of the reports
concentrated at 10:00 in the morning and with a very small
average wakefulness of 7 hours. In order to provide a suitable
explanation for this apparent contradiction, the researchers
decided to investigate the correlation between the time of the
fatigue sensation (clock time) and the wake-up time (start of
wakefulness).
Average effectiveness: 73.8 ± 0.8% (E = 77.5% à BAC ≈ 0.05%)
Pilot Effectiveness*
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 5
Pilot fatigue in Brazil
WOCL Shoulder…
Fatigue clock time - FT
0 2 4 6 8 10 12 14 16 18 20 22 240
10
20
30
40
50
60
70
80
8%49%43%
Even
ts This work Sum of three gaussian shapes <FT>1 = 3.1 ± 0.4 (h) <FT>2 = 9.8 ± 0.5 (h) <FT>3 = 21.3 ± 0.4 (h)
Fatigue "clock time" (h)
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 6
Pilot fatigue in Brazil
15
0 2 4 6 8 10 12 14 16 18 20 22 240
10
20
30
40
50
60
70
80
8%49%
43%
Eve
nts
This work Sum of three gaussian shapes <FT>
1 = 3.1 r 0.4 (h)
<FT>2 = 9.8 r 0.5 (h)
<FT>3 = 21.3 r 0.4 (h)
Fatigue "clock time" (h)
Figure 5: Distribution of fatigue reports as a funtion of the time of the day.
0 5 10 15 20 25 300
20
40
60
80
100
47%53%
Wakefulness prior to fatigue sensation (h)
Eve
nts
This work Sum of two gaussian shapes <AT>
1 = 7.0�r�0.6 (h)
<AT>2 = 18.0 r 0.5 (h)
Figure 6: Distribution of wakefulness prior to fatigue sensation.
7h? 53% of the pilots
manifested fatigue after
only 7 hours of wakefulness
Wakefulness - AT
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 7
Pilot fatigue in Brazil 18
2 3 4 5 6 7 8 90
20
40
60
80
100
E
ven
ts
Pilot reports Gaussian Fit (average sleep of 5.05 h)
Amount of sleep in the last 24 hours (h)
Figure 9: Distribution of the amount of sleep reported by the pilots within the
last 24 hours prior to the fatigue sensation.
4 6 8 10 120
20
40
60
80
100
E
ven
ts
FAST (301 fatigue reports) Sum of two gaussian shapes Average value = 7.41 r 0.22 (h)
Chronic sleep debt in the last 72 hr (h)
Figure 10: Distribution of chronic sleep debt within the last 72 hours prior to
the fatigue sensation.
18
2 3 4 5 6 7 8 90
20
40
60
80
100
E
ven
ts
Pilot reports Gaussian Fit (average sleep of 5.05 h)
Amount of sleep in the last 24 hours (h)
Figure 9: Distribution of the amount of sleep reported by the pilots within the
last 24 hours prior to the fatigue sensation.
4 6 8 10 120
20
40
60
80
100
E
ven
ts
FAST (301 fatigue reports) Sum of two gaussian shapes Average value = 7.41 r 0.22 (h)
Chronic sleep debt in the last 72 hr (h)
Figure 10: Distribution of chronic sleep debt within the last 72 hours prior to
the fatigue sensation.
• Insufficient sleep in the last 24 hours (5.05 h) and • chronic sleep debt in the last 72 hours (7.4 h)
Sleep history
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 8
Mean values (h)
Wakefulness Fatigue Time
18.0 ± 0.5 3.1 ± 0.4 7.0 ± 0.6 9.8 ± 0.5
Pilot fatigue in Brazil 15
0 2 4 6 8 10 12 14 16 18 20 22 240
10
20
30
40
50
60
70
80
8%49%
43%
Eve
nts
This work Sum of three gaussian shapes <FT>
1 = 3.1 r 0.4 (h)
<FT>2 = 9.8 r 0.5 (h)
<FT>3 = 21.3 r 0.4 (h)
Fatigue "clock time" (h)
Figure 5: Distribution of fatigue reports as a funtion of the time of the day.
0 5 10 15 20 25 300
20
40
60
80
100
47%53%
Wakefulness prior to fatigue sensation (h)
Eve
nts
This work Sum of two gaussian shapes <AT>
1 = 7.0�r�0.6 (h)
<AT>2 = 18.0 r 0.5 (h)
Figure 6: Distribution of wakefulness prior to fatigue sensation.
15
02
46
810
1214
1618
2022
2401020304050607080
8%49
%
43%
Events
Thi
s wor
k Su
m of
thre
e gau
ssian
shap
es <
FT> 1 =
3.1 r
0.4
(h)
<FT
> 2 = 9.
8 r 0
.5 (h
)
<FT
> 3 = 21
.3 r
0.4 (
h)
Fatig
ue "c
lock
tim
e" (h
)
Figu
re 5:
Dist
ribut
ion
of fa
tigue
repo
rts as
a fu
ntio
n of
the t
ime o
f the
day.
05
1015
2025
30020406080100
47%
53%
W
akefu
lnes
s prio
r to f
atigu
e sen
satio
n (h
)
Events
Thi
s wor
k Su
m of
two g
auss
ian sh
apes
<AT
> 1 = 7.
0�r�0
.6 (h
)
<AT
> 2 = 18
.0 r
0.5 (
h)
Fi
gure
6: D
istrib
utio
n of
wak
efuln
ess p
rior t
o fat
igue s
ensa
tion.
Fatigue time x wakefulness
0 5 10 15 200
5
10
15
20
25
30
Fatigue Time, FT (h)
Wak
eful
ness
, AT
(h) Fatigue Reports
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 9
Pilot fatigue in Brazil
16
Figure 7 shows the fatigue clock time as a function of the wake-up
time for all the data, where it is clear two quite distinct behaviors.
The left side plot shows the fatigue events that took place one day
after the start of wakefulness, while the right hand plot presents
the events that occurred during the same day of the awakening. In
the first case it is clear that the fatigue was felt essentially around
04:00 and almost independent of the start of wakefulness. In the
second case, there is a strong variation in the fatigue time with the
wake-up time. Figure 8 presents an exponential fitting (solid red
line) with its respective limits (dashed blue lines) obtained via the
propagation of uncertainties of the fitted parameters (Licati et al., 2015).
-24 -20 -16 -12 -8 -4 0 4 8 12 16 20 240
4
8
12
16
20
24
51%49%
day beforefatigue sensation
same day offatigue sensation
Data points
Wake-up time
Fat
igu
e cl
ock
-tim
e
Figure 7: Fatigue reported clock-time versus wake-up time.
17
0 4 8 12 16 20 240
4
8
12
16
20
24
Wake-up time
day beforefatigue sensation
Data points Exponential Fit Statistical limits
4 8 12 16 20 24
Fat
igu
e cl
ock-
tim
e
same day offatigue sensation
Figure 8: Relationship between the wake-up time and the fatigue clock-time
(solid squares). The solid red line represents the exponential fitting with its
respective statistical limits shown by the dashed blue lines.
This result demonstrates the contribution of two distinct effects.
The first one is related with the higher probability of fatigue
reports during the WOCL and is almost independent of the wake-
up time. The second effect is probably related with the rostering
structure that generates a progressive sleep debt in consecutive
early starts duties without adequate sleep opportunity. The latter
can be further verified by the inspection of the amount of sleep
reported by the pilots within the last 24 hours prior to the fatigue
sensation (Figure 9), as well as with the analysis of the chronic
sleep debt accumulated in the last 72 hours (Figure 10).
Exponential relationship between FT and AT
Fatigue time x awake time
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 10
Scenario 1: first late night duty*
26
x Intrinsic software commute 4: zero
x Minimum sleep: 60 minutes
The following tables present all the scenarios and initial conditions
studied in this work.
Scenario 1: Crew member checks in fully recovered
Check-in (h) ' = 2h ' = 3h Duty time
(h) # results
02:00 M1 M3 10 2 04:30 M5 M7 10 2 05:30 M9 M11 12 2 12:30 M13 M15 13 2 14:30 M17 M19 12 2 15:00 M21 M23 12 2 15:30 M25 M27 11 2 19:30 M29 M31 12 2 22:30 M33 M35 11 2 23:30 M37 M39 10 2
Table 3: Initial conditions adopted in the present analysis for a crew member
commencing the duty period fully recovered. A portion of the flight can elapse
between 0000 and 0600 in this scenario.
According with table 3, two different wakefulness prior to check-in
were considered, one with 2 hours (' = 2h) and another one with
three hours (' = 3 h). The labels M1, M3, M5, etc... refer to the
respective initial condition and are used for notation purposes
only.
4 The commute in the FAST software was set to zero since the simulations already assume pre‐defined and realistic wakefulness hypotheses by the check‐in time.
Δ: elapsed time between wake-up and check-in
(*) “late night duty” indicates that any fraction of the duty can elapse between 0000 and 0600
Risk Analyses
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 11
27
Scenario 2: Crew member starts the second work day after completing the scenario 1 in the first day
Check-in (h) ' = 2h ' = 3h Duty period (h)
# results
02:00 M2 M4 10 2 04:30 M6 M8 10 2 05:30 M10 M12 12 2 12:30 M14 M16 13 2 14:30 M18 M20 12 2 15:00 M22 M24 12 2 15:30 M26 M28 11 2 19:30 M30 M32 12 2 22:30 M34 M36 11 2 23:30 M38 M40 10 2
Table 4: Initial conditions for the crew member that starts the second day
after completing the first day in the scenario 1. A portion of the flight can
elapse between 0000 and 0600.
Scenario 3: Crew member checks-in fully recovered
Check-in (h) ' = 2h ' = 3h ' = 4hDuty
period (h)# results
06:30 M41 M43 -- 13 2 07:30 M45 M47 -- 14 2 09:30 M49 M51 M53 14 3
Table 5: Initial conditions for a crew member commencing the duty period
fully recovered. No portion of the flight elapses between 0000 and 0600.
These are typical early-starts conditions.
Scenario 2: second late night duty Risk Analyses
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 12
27
Scenario 2: Crew member starts the second work day after completing the scenario 1 in the first day
Check-in (h) ' = 2h ' = 3h Duty period (h)
# results
02:00 M2 M4 10 2 04:30 M6 M8 10 2 05:30 M10 M12 12 2 12:30 M14 M16 13 2 14:30 M18 M20 12 2 15:00 M22 M24 12 2 15:30 M26 M28 11 2 19:30 M30 M32 12 2 22:30 M34 M36 11 2 23:30 M38 M40 10 2
Table 4: Initial conditions for the crew member that starts the second day
after completing the first day in the scenario 1. A portion of the flight can
elapse between 0000 and 0600.
Scenario 3: Crew member checks-in fully recovered
Check-in (h) ' = 2h ' = 3h ' = 4hDuty
period (h)# results
06:30 M41 M43 -- 13 2 07:30 M45 M47 -- 14 2 09:30 M49 M51 M53 14 3
Table 5: Initial conditions for a crew member commencing the duty period
fully recovered. No portion of the flight elapses between 0000 and 0600.
These are typical early-starts conditions.
1st day 28
Scenario 4: Crew member in the third consecutive day after the completion of two successive early-starts
Check-in (h) ' = 2h ' = 3h ' = 4hDuty time
(h) # results
06:30 M42A M44A -- 13 2 07:30 M46A M48A -- 14 2 09:30 M50A M52A M54A 14 3
Table 6: Initial conditions for a crew member in the third consecutive day
after the completion of two successive early-starts.
Scenario 5: Crew member in the sixth consecutive day after the completion of five successive early-starts
Check-in (h) ' = 2h ' = 3h ' = 4h Duty time (h) # results
06:30 M42 M44 -- 13 2 07:30 M46 M48 -- 10 2 09:30 M50 M52 M54 12 3
Table 7: Initial conditions for a crew member in the sixth consecutive day after
five successive early-starts.
The simulations were carried out entirely by the IBR team for all
the initial conditions depicted in Tables 3 to 7. The plots and the
results were gently provided by Lauren Waggoner, PhD (IBR) and
are presented below in Figures 12 to 17.
28
Scenario 4: Crew member in the third consecutive day after the completion of two successive early-starts
Check-in (h) ' = 2h ' = 3h ' = 4hDuty time
(h) # results
06:30 M42A M44A -- 13 2 07:30 M46A M48A -- 14 2 09:30 M50A M52A M54A 14 3
Table 6: Initial conditions for a crew member in the third consecutive day
after the completion of two successive early-starts.
Scenario 5: Crew member in the sixth consecutive day after the completion of five successive early-starts
Check-in (h) ' = 2h ' = 3h ' = 4h Duty time (h) # results
06:30 M42 M44 -- 13 2 07:30 M46 M48 -- 10 2 09:30 M50 M52 M54 12 3
Table 7: Initial conditions for a crew member in the sixth consecutive day after
five successive early-starts.
The simulations were carried out entirely by the IBR team for all
the initial conditions depicted in Tables 3 to 7. The plots and the
results were gently provided by Lauren Waggoner, PhD (IBR) and
are presented below in Figures 12 to 17.
3rd day
6th day
Scenarios 3, 4 & 5: early-starts Risk Analyses
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 13
Data points: S. Hursh, T. Raslear, A. Kaye and J. Fanzone, Jr. (2006). Validation and Calibration of a Fatigue Assessment Tool for Railroad Work Schedules, Summary Report. (Report No. DOT/FRA/ORD-06/21). Washington, DC: U.S. Department of Transportation.
9
The results are presented in Figure 3, together with the fitted
function2 R(E) = a + b/E, with a= 0.27 ± 0.20, b = 0.58 ± 0.19 (F2
= 3.03 and n.d.f. = 3).
40 50 60 70 80 90 100
0.8
1.2
1.6
2.0H
F a
ccid
en
t rela
tiv
e l
ikeli
ho
od
SAFTE-FAST predicted effectiveness (%)
Hursh (2006)
Fitted function (~1/E)
Figure 3: Human-Factor (HF) accident relative likelihood as a function of
SAFTE-FAST predicted effectiveness. Details in the text.
The data of Figure 3 (Hursh et al., 2006) represent the relative
probability of accidents caused by human errors normalized by the
amount of work hours (exposure). The error bars were estimated
to be N1/2, where N is the number of total accidents for a given
effectiveness interval.
The risk exposure was calculated as a function of the area of the
effectiveness curve along the duty period below a threshold value,
herein fixed at 80%. This limit is close to the level of 77%, which
corresponds to a blood alcohol concentration (BAC) of 0.05% (the
2 For the statistical analyzes we have adopted the Least Squares Method described elsewhere (Helene, 2013).
Fatigue risk versus Effectiveness Risk Analyses
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 14
Risk Analyses 32
with the original histogram from the IBR team. The average
effectiveness for this initial condition was 78.33%.
2 4 6 8 10 1250
60
70
80
90
100
Eff
ecti
ven
ess
(%)
Time of the day (h)
M1 (Eave
= 78.33%)
Interpolation
Figure 18: Interpolation of SAFTE-FAST predicted effectiveness during the
first night shift with check-in at 0200 (M1). The average effectiveness Eave =
78.33% was calculated by the integral of E(t) from the check-in until the
check-out.
Using the relationship presented in Figure 3 one can estimate the
risk due to fatigue along the duty period. The result for M1 is
shown in Figure 19.
In the next step we evaluate the hazard area (HA) below our
threshold value of effectiveness ETh=80%. As seen by the
inspection of Figure 20, the total costs for human-factor accidents
vary quite significantly when the effectiveness drops below 77%,
supporting our strategy to map the area below 80%.
33
2 4 6 8 10 120.92
0.96
1.00
1.04
1.08
1.12
M1: <RM1
> = 1.015
R(E) ~1/E
Rela
tiv
e F
ati
gu
e R
isk
Time of the day (h)
Figure 19: Relative fatigue risk during the first night shift with check-in at
0200 (M1). The average relative risk of 1.015 was calculated by the integral of
R(t).
Figure 21 shows a comparison between the first and the second
night shifts (M1 and M2) with its respective hazard areas, where
one clearly verifies an increase of almost a factor of five in the risk
exposure comparing the first and the second night shift (M2/M1 ~
5). Table 8 summarizes all the results obtained for the six scenarios
and more than 60 different initial conditions.
(*) SAFTE-FAST predicted Effectiveness were generously provided by Lauren Waggoner from IBR. The plots refer to the first night duty with check in at 02:00.
Effectiveness* à Risk
Eave = 78.33 % <R> = 1.015
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 15
Risk Analyses
34
Figure 20: Total cost of human-factor accidents as a function of SAFTE-FAST
predicted effectiveness. The plot was extracted from Hursh et al. (2011).
2 4 6 8 10 120
20
40
60
80
100
HA ratio between M2 and M1: 4.7
Time of the day (h)
Eff
ecti
ven
ess
(%)
M1 (HA = 0.265 h) M2 (HA = 1.246 h) Threshold value of 80%
Figure 21: Comparison between the hazard areas (HA) obtained in the first
(M1) and in the second (M2) night shifts with check-in at 0200. Details in the
text.
The hazard areas (HA) were calculated by the fraction of the duty period with SAFTE-FAST Effectiveness below a threshold of 80%.
Effectiveness & Hazard Area (HA)
5HAHA
M1
M2 ≈
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 16
Risk Analyses 37
0.8 0.9 1.0 1.1 1.2 1.3 1.40
1
2
3
4
Eave
= 77%Eave
= 90%
Low Risk Zone (HA = 0 and Eave
t 90%)
Caution Zone (77% d�Eave
< 90%)
Danger Zone (Eave
< 77%)
Haz
ard
Are
a, H
A (
h)
Average Relative Risk
Figure 23: Relationship between the average relative risk and the hazard area
(HA) for all the initial conditions considered in the simulations. Details in the
text.
Obviously that the linear relationship between <R> and HA for the
average effectiveness below 77% reflects our choice for the
threshold value of 80% to calculate the areas. It is worth-
mentioning, however, that this value is not supposed to provide a
go no go decision, but to establish a reference limit where the risk
exposure starts to build up.
So, in order to verify that the proposed method is in fact useful for
risk mitigation, we have included in the present analysis two
famous fatigue-related accidents that were largely studied with the
help of the SAFTE-FAST model. The first was the AIA 808 that
crashed in Guantanamo Bay, Cuba in 1993 and the second one was
the Comair 5191 (2006) in Lexington, USA.
Risk exposure: average risk versus HA
GREEN: LOW RISK
AMBER: CAUTION
RED: DANGER
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 17
Risk Analyses
41
undoubtedly that this parameter has to be taken into account very
carefully during the risk analysis and mitigation.
0.8 0.9 1.0 1.1 1.2 1.3 1.40
1
2
3
4
<E> = 77%<E> = 90%
Low Risk Zone Caution Zone Danger Zone AIA 808 - Captain AIA 808 - First Officer AIA 808 - Flight EngineerComair 5191 - Air Traffic Controller
Haz
ard
Are
a, H
A (
h)
Average Relative Risk
Figure 26: Relationship between the average relative risk and hazard area
obtained for the 61 initial conditions of this work, together with the results
obtained for actual accidents in Guantanamo Bay (blue symbols) and
Lexington (magenta square).
SAFTE-FAST Effectiveness calculations: (1) AIA 808 (Guantanamo Bay): Wesensten & Belenky (private comm.)
(2) Comair 5191: Pruchnicki, Wu & Belenky (2011)
Risk exposure: including real accidents (1,2)
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 18
FDT Limits A science based proposal for minimum crew
Start with CAO’s-48 (scenarios 2 and 4), but limited to 12 hours*
(*) Details at http://www.asagol.com.br/files/_dirtecnica/FRMS/FRMS%20Report%20Part%20II.pdf
Duty &me limits (1 or 2 sectors) in scenarios 2 and 4* with Δ = 2 hours
Check-‐in (h) Ini?al condi?on label
Maximum duty ?me (h)
Average risk Hazard area (h)
0500-‐0559 M10 11 1.059 0.667
0600-‐0659 M42A(*) 12 1.089 0.974
0700-‐0759 M46A(*) 13 à 12 0.973 0
0800-‐1059 M50A(*) 14 à 12 0.894 0
1100-‐1359 M14 13 à 12 0.871 0
1400-‐1459 M18 12 0.869 0
1500-‐1559 M26 11 0.869 0
1600-‐2259 M30 10 0.978 0.174
2300-‐0459 M2 10 1.135 1.246
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 19
FDT Limits A science based proposal for minimum crew
Current (Federal Law 7.183/84) versus “future” limits (one or two sectors)
0 2 4 6 8 10 12 14 16 18 20 22 246
7
8
9
10
11
12
13
14
D
uty
time
limit
(h)
Check-in time (h)
HA = 0.174
HA = 0HA = 0.974
HA = 0.667
HA =1.246
Federal Law 7.183/84
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 20
FDT Limits A science based proposal for minimum crew
Safety x Productivity Fatigue Risk As Low As Reasonably Practicable (ALARP)
0 2 4 6 8 10 12 14 16 18 20 22 246
7
8
9
10
11
12
13
14
D
uty
time
limit
(h)
Check-in time (h)
HA = 0.174
HA = 0HA = 0.974
HA = 0.667
HA =1.246
Federal Law 7.183/84
Red zone: FDT reduced by 1 hour
Amber zone: FDT maintained
Green zone: FDT increased to 12 hours
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues -‐ Montreal 2016 4/7/16 21
FDT Limits A science based proposal for minimum crew
Duty and flight time limitations
Check-in (h) Number of flight sectors
1-2 3-4 5 6 7+ 0500-0559 10(9) 9(8) 9(8) 9(8) 9(8) 0600-0659 11(9) 10(9) 9(8) 9(8) 9(8) 0700-0759 12(9.5) 12(9) 11(9) 11(9) 10(8) 0800-1059 12(10) 12(9.5) 12(9) 11(9) 11(9) 1100-1359 12(9.5) 12(9) 11(9) 11(9) 10(8) 1400-1459 12(9) 11(9) 10(8) 10(8) 9(8) 1500-1559 12(9) 11(8) 11(8) 10(8) 10(8) 1600-2259 10(8) 9(8) 9(8) 9(8) 9(8) 2300-0459 9(8) 9(8) 9(8) 9(8) 9(8)
ü Limits in black: the same as CAO’s-48, but limited to 12 hours ü Limits in red: one hour lower than CAO’s-48 ü Limits in blue: one hour higher than CAO’s-48
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues – Montreal 2016 4/7/16 22
Conclusions
1. Licati et al. (2015) found a scenario of chronic fatigue in Brazil with an average effectiveness of 73.8 ± 0.8 %. Roma et al. (2012) found a much higher value of 87.9% in USA. Comparing these experiments, the average fatigue risk in Brazil is 14% higher than in the USA.
2. More than 50% of the pilots manifested fatigue after only 7 hours of wakefulness. This result was explained by an insufficient sleep in the last 24 hours (average of 5 h) and chronic sleep debt in the last 72 hours (average of 7.4 h).
3. Single days-off should be avoided as they do not allow for a complete recovery. If they are needed it is not recommended that the crew members show to work before 10:00 in the next day.
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues – Montreal 2016 4/7/16 23
Conclusions
4. The Hazard Areas (HA) in the second late night duty (CAO’s-48 or FAR’s-117 FDT limits) is roughly 5 times higher than in the first one;
5. The HA during the third successive early-start before 07:00 is similar to the one found in the second late night duty;
6. The daily averaged HA of FAR’s-117 is 26% (9%) higher than CAO’s-48 in scenarios 1 and 3 (2 and 5);
7. The relative human-factor accident risk increases by 45% (90%) as the time on duty increases from 10 (8) to 12 hours [(Goode, 2003), reanalysis]. Is it really necessary to go beyond 12 hours for minimum crew?
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues – Montreal 2016 4/7/16 24
Conclusions
I. The importance of an effective participation of Crew Member Representatives in de decision making of the FSAG;
II. Review of Prescriptive Limits as they play a major role for the fatigue risk management (ICAO’s Annex 6). A step back may be needed;
III. The State FRMS Regulations can not rely 100% in a “copy
and paste” strategy.
Lessons learned
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues – Montreal 2016 4/7/16 25
Future Plans
1. Member Institutions: SNA, ABRAPAC, ASAGOL and ATT
2. Objective: To monitor the alertness level of the Brazilian crew members during their duty periods, providing data analyses for the identification of potential hazards, as well as safety recommendations for risk mitigation
3. Partnerships with IFALPA (HUPER Committee), Universities, Research Centers, FRMS Specialists, etc…
4. When to start: hopefully by the end of 2016!
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg
Tulio E. Rodrigues – Montreal 2016 4/7/16 26
Acknowledgments
v Crew Member National Union (SNA) v Brazilian Civil Aviation Pilots Association (ABRAPAC) v GOL Airline Crew Member Association (ASAGOL) v TAM Airline Crew Member Association (ATT) In Partnership with: q University of São Paulo (USP) q Institutes for Behaviour Resources, Inc Collaborating Stakeholders: ü National Centre of Investigation and Prevention of Aeronautical
Accidents (CENIPA) ü National Commission of Human Fatigue (CNFH) Special Thanks to: Steven Hursh (IBR), Lauren Waggoner (IBR), Reid
Blank (IBR), Gregory Belenky (WSU) and Nancy Wesensten (WRAIR)
Thank you!
4/4/16, 9:11 AMfrmsforum_logo.jpg 248×53 pixels
Página 1 de 1http://137.117.64.219/www.icao.int/safety/fatiguemanagement/frms2011/publishingimages/frmsforum_logo.jpg