Title: CGAR STATISTICAL ANALYSIS FOR FATAL GENERAL AVIATION ACCIDENTS Ken Knopp, FAA Massoud Bazargan, ERAU
1CGARSTATISTICAL ANALYSIS FOR FATAL GENERAL
AVIATION ACCIDENTSKen Knopp, FAAMassoud
Bazargan, ERAU
2Scope of the Project
- Analyze existing data on General Aviation
accidents resulting in fatalities. - Identify associations and patterns between flight
elements and fatalities
3NTSB Database
- Every accidents/incidents receives a record
(event ID). Each aircraft involved in the
accident receives an NTSB report number. - 600 MB
- Around 60,000 events accidents/incidents (total).
- 50,000 events relate to GA. (83 of all
accidents). - 9200 events 9380 aircraft fatal GA accidents
(18.4 of total GA events).
4GA Fatal Accidents
5GA Accidents Accidents per 100,000 flight hours
(est.)
6Phases I II
- Phases I II analyzed and reported 5, 10 and 20
year fatal accidents. - Identified patterns for national and nine regions
as defined by the FAA (a summary presented at UND
and UAF). - Provided statistical analyses on single factors
for national and 9 regions (75 factors).
7Some General Statistics Phase III
8Time of Accidents (1983-1992)National
9Time of Accidents (1993-2002)National
10Time of Accidents (1983-1992)Arizona
11Time of Accidents (1993-2002)Arizona
12Arizona vs. National (1983 1992)
Differences across the categories are NOT
SIGNIFICANT
13Arizona vs. National (1993 2002)
Differences across the categories are NOT
SIGNIFICANT
14Pilot Hours of Experience(1983-1992) National
15Pilot Hours of Experience(1993-2002) National
16Pilot Hours of Experience(1983-1992) Arizona
17Pilot Hours of Experience(1993-2002) Arizona
18Pilot Gender Fatal Accidents (1993-2002)Nationa
l
19Pilot Gender Fatal Accidents (1993-2002)Arizona
20Pilot Age Fatal Accidents (1993-2002)National
21Pilot Age Fatal Accidents (1993-2002)Arizona
22Analyses
- Primary causes
- Statistical analyses
23Primary Causes
- The NTSB reports up to 5 occurrences for each
accident. - Each occurrence has up to 10 sequence of events
showing chronologically what happened at each
occurrence. - We searched for the primary cause that triggered
all the events.
24More than 400 causes
25Fatal GA Aircraft Accidents
- Nationwide Analysis of 5-Year Periods
- Top 10 primary causes
- 1983 1987
- 1988 1992
- 1993 1997
- 1998 2002
2656 of all fatal accidents
2750 of all fatal accidents
2853 of all fatal accidents
2957 of all fatal accidents
30Comparison of 5-Year Periods
Categorized as Aircraft Handling
31Primary Cause Aircraft Control
NTSB Code 24566
32Primary Cause Airspeed
NTSB code 24506
33First Primary- VFR flight into IMC
NTSB code 24015
34Non-Fatal GA Aircraft Accidents
35Pattern among Fatal and Non-Fatal GA Aircraft
Accidents
36Risk Factor Comparison
Rank First Cause 83-88 88-92
93-97 98-02
- Aircraft Control 2.792 3.586 4.200 2.497
- Airspeed 1.667 2.120 2.865 2.7
00 - VFR Flight Into IMC
16.818 18.415 17.416 15.666 - Altitude
4.306 4.080 5.242 5.199 - In-flight Planning 1.778 1.380
0.865 1.334 - Visual Lookout 1.393
1.400 1.561 1.357
37General Aviation Accident Analysis
- Part 91 GA accidents analyzed in 5-year periods
- 1983 1987
- 1988 1992
- 1993 1997
- 1998 -- 2002
- Single engine aircraft analysis separated by
engine horse-power - Quartile 1
- Less than 150 HP
- Quartile 2
- 150 200 HP
- Quartile 3
- 200 300 HP
- Quartile 4
- Greater than 300 HP
38Comparison of 5-Year PeriodsSingle engine
reciprocating lt150HP
Categorized as Aircraft Handling
39Comparison of 5-Year PeriodsSingle engine
reciprocating 150-200 HP
Categorized as Aircraft Handling
40Comparison of 5-Year PeriodsSingle engine
reciprocating 200-300 HP
Categorized as Aircraft Handling
41Comparison of 5-Year PeriodsSingle engine
reciprocating gt300 HP
Categorized as Aircraft Handling
42Statistical Analyses on combined multi-Factors
Contributing to Fatalities in General Aviation
Accidents
43National Analyses
- Sample 35,960 GA airplane accidents
- From 1983 to 2002.
- More than 420 data fields
- 28 variables related to pilot and airplane
characteristics selected. - Sample was meticulously screened for
- accuracy,
- missing variables,
- outliers,
- Absence of multicollinearity problems
44National AnalysesVariables Included
- FATAL is the fatal accident (dependant variable)-
Binary Logistic Regression - VFR flight plan
- Light condition
- Instrument meteorological condition
- Cross-country flight
- Wind velocity in the area
- Retractable landing gear
45National Analyses Variables Included
- engines
- Pilots type rating variable
- Second pilot on board
- Seat belt
- Shoulder harness
- Medical certificate
46National AnalysesVariables Included
- Gender
- Age
- Pilots total flight experience
- Phase of the flight (taxi, takeoff, climb,
descent, ..) - YEAR of the accident
47National Analyses
- Please Pay Attention You will have a test on
the following equation after this!!
SPSS Logistic Regression
48National AnalysesOnly those significant
- Only 10 variables found to be significant
(combined). - Lighting Condition positive and significant,
indicating that darker environment - increased
probability of an accident to be fatal - Instrument Meteorological Condition (IMC)
significant, highest odds ratio.
49National Analyses
- Cross Country Flight positive and significant
- Wind Condition negative and significant -higher
wind velocity - decreases the probability of the
accident to be fatal !! - Literature also reports that stronger winds lead
to more GA accidents, but these accidents most
commonly result in aircraft damage and minor
injuries
50National Analyses
- Retractable Landing Gear positive, significant.
- Second Pilot Onboard Positive, significant
(?!!!). - More than 13 percent of the accident airplanes in
the sample had a second pilot on board.
51National Analyses
- Conventional wisdom suggests that an extra pilot
on board would help the flying pilot in case of
unexpected emergency. - However, for this help to be constructive, GA
pilots should understand how to properly use
human resources in flight.
52National Analyses
- U.S. airlines have implemented Crew Resource
Management (CRM) training with an emphasis on
principles and concepts that improve crew
performance and flight safety, - A typical GA pilot has never taken a CRM course.
Our results imply that such a course would
decrease the number of fatal GA accidents.
53National Analyses
- Seatbelts significant
- Phase of flight CLIMB and DESCENT significant
- Pilots flight time over the last 30 days
- less than 5 hours (rusty!) significant
- More than 100 hours (tired!) significant
- Year Negative and significant
54Research in Progress
- Machine learning approaches to find patterns
among fatal/non-fatal accidents - Continue with logistic regression for all 9
regions. - Text processing analysis.
- Finalize phase III.
- Start phase 4 on statistical analyses on
rotorcraft and helicopters.
55