Railroad Accident Rates for Use in Transportation Risk Analysis Paper No' 042341

1
Railroad Accident Rates for Use in Transportation
Risk Analysis(Paper No. 04-2341)

• Presented at the 83rd Annual Meeting of the
• Transportation Research Board
• Washington, D.C.
• January 14, 2004
• Robert T. Anderson
• Christopher P.L. Barkan
• University of Illinois at Urbana-Champaign

2
Background

• Accident rates essential for transportation risk
  analysis of hazmat shipments
• Detailed information needed to accurately access
  risk for particular hazmat shipment
• FRA data allows for more resolution than
  published statistics

3
Outline

• Train Accident Rates
• Track Class-Specific Derailment Rates
• Sensitivity of Results
• Example Risk Calculation
• Conclusions

4
Train Accident Rates

• Ratio of number of independent accidents and the
  total number of train-miles
• e.g. In 2001,
• 3,023 train accidents
• 711.55 million train-miles
• 4.25 accidents per million train-miles
• Overall rate a combination of several independent
  rates

5
Distinctions

• Type of track
• Mainline, Yard, etc.
• Type of accident
• Derailment, Collision, etc.
• Railroad
• Class I vs. Non-Class I
• Track Class

6
Mainline vs. Yard Accident Rate

• FRA calculated accident rates
• Yard accident rate
• Other accident rate
• Overall increase in accident rate b/w 1997-2001
  largely due to increase in yard accident rate
• Yard accident rate increased 27
• Mainline accident rate increased 12

7
Distinctions

• Type of track
• Mainline, Yard, etc.
• Type of accident
• Derailment, Collision, etc.
• Railroad
• Class I vs. Non-Class I
• Track Class

8
Mainline vs. Yard Derailment Rate

• Important to hazmat risk assessment due to
  frequency of occurrence and severity of
  consequences
• e.g. In 2001, derailments accounted for
• 28 of the 32 train consists that had hazmat
  released
• 51 of the 57 hazmat cars that released
• Overall increase in derailment rate largely due
  to increase in yard accident rate
• Yard derailment rate increased 38.9
• Mainline derailment rate increased 4.4

9
Distinctions

• Type of track
• Mainline, Yard, etc.
• Type of accident
• Derailment, Collision, etc.
• Railroad
• Class I vs. Non-Class I
• Track Class

10
Class I vs. Non-Class IRailroad Derailment Rates

• Majority of hazmat traffic mileage carried by
  Class I freight railroads
• e.g. In 2001, Class I RRs carried
• 83 of all trains that were carrying hazmat and
  involved in an accident
• 93 of all hazmat cars that released their
  contents
• Rate calculated using freight train-miles as more
  appropriate denominator
• Excludes passenger ( other) train-miles

11
Uses total less yard switching train-miles as
denominator
Uses freight train-miles as denominator
Differs by factor of 2
Differs by factor of 5
Nearly constant over ten years
12
Distinctions

• Type of track
• Mainline, Yard, etc.
• Type of accident
• Derailment, Collision, etc.
• Railroad
• Class I vs. Non-Class I
• Track Class

13
Track Class-Specific Derailment Rates

• FRA track class as proxy variable for statistical
  derailment probability estimation
• Higher track classes have more stringent
  requirements (resulting in better track quality)
  and higher maximum speeds
• Train Derailment Rate
• Ri of derailments per million freight
  train-miles or per billion freight car-miles, on
  class i track
• Car Derailment Rate
• Ri of cars derailed per billion freight
  car-miles, on class i track

14
Derailments per Million Freight Train-Miles
2 orders of magnitude difference b/w lowest and
highest track class
Derailments per Billion Freight Car-Miles
class 3 track rate twice total rate while class 4
track rate about half total rate
Cars Derailed per Billion Freight Car-Miles
15
Uncertainty of Results

• Annual Variation
• Average over ten years
• Traffic Percentages
• Assumption of constant traffic distribution among
  track classes over ten year period
• Traffic Changes
• Abandonment and sale of lower class track
• Shift in yard operations/track
• Capacity speed upgrades

16
Sensitivity Analysis

• Assume Estimated traffic distribution
  percentages representative for years 1992-1994
• Calculate 3-yr. derailment rate by track class
• Assume Derailment rate has changed little over
  ten-year period that 3-yr. rate for 1999-2001
  is nearly the same as that for 1992-1994

17
Sensitivity Analysis

• Calculate Traffic needed for constant derailment
  rates between 1992-1994 and 1999-2001 using
  actual derailment counts
• Determine New traffic distribution percentages
  using calculated traffic from above
• Calculate New derailment rates using new
  estimated traffic distribution percentages and
  actual total mileage

18
Little changeslight decrease
Traffic shift towards higher class track
19
(Derailments per Million Train-Miles)

• New traffic percentages
• Increase in derailment rates for classes 3 and 4
  (and lower) track
• Decrease in class 5 6 derailment rate

• Old traffic percentages
• Decrease in derailment rates for classes 3 and 4
  (and lower) track
• Increase in class 5 6 derailment rate

20
(No Transcript)
21
Hazardous Materials Release Probability

• Higher hazmat release rates on higher track
  classes?
• Higher derailment speeds
• More cars derailed at higher speeds

22
(No Transcript)
23
Example Risk Calculation

• Route 1,000 miles
• 65 class 4 track, 35 class 5 track
• Train 100 cars
• Rate Calculation
• Ri Derailment rate on class i track
• mi Mileage (train-miles or car-miles) on class
  i track
• Across all n track classes traversed

24
Train Derailment Probability

• per Train-Mile
• 650 train-miles on class 4 track
• Ri 0.53 derailments per million train-miles
• 350 train-miles on class 5 track
• Ri 0.32 derailments per million train-miles
• Pr(der) (0.53x10-6 x 650)4 (0.32x10-6 x 350)5
• 4.6x10-4 (12,000)

25
Train Derailment Probability (cont.)

• per Car-Mile
• 65,000 car-miles on class 4 track
• Ri 7.8 derailments per billion car-miles
• 35,000 car-miles on class 5 track
• Ri 4.9 derailments per billion car-miles
• Pr(der) (7.8x10-9 x 65,000)4
• (4.9x10-9 x 35,000)5
• 6.8x10-4 (11,500)
• Nearly 50 higher derailment probability due to
  longer than average train length

26
Car Derailment Probability

• per Car-Mile
• 65,000 car-miles on class 4 track
• Ri 77 cars derailed per billion car-miles
• 35,000 car-miles on class 5 track
• Ri 42 cars derailed per billion car-miles
• Pr(der) (77x10-9 x 65,000)4
• (42x10-9 x 35,000)5
• 6.5x10-5 (115,000)
• 10 probability of a single car being derailed
  given that a derailment has occurred

27
HazMat Release Probability

• Suppose now that 10 of the 100 cars were
  hazardous materials cars
• Probability of having a hazmat release
• Car derailment probability 6.5x10-5
• Average release rate 13
• Pr(rel) (6.5x10-5) x 10 x 0.13
• 8.45x10-5 (112,000)

28
Conclusions

• Rail transportation risk analysis relies on the
  accurate estimation of accident rates
• Class I railroad mainline derailment rate most
  pertinent to hazardous materials shipments
• Derailment rates vary by track class with
  substantial differences between lower and higher
  track classes
• Potential for over- or under-stating risk
  associated if track class is not factored into
  analysis
• Conservative risk assessment would use the
  greater of the two derailment probabilities (i.e.
  per Train-Mile or per Car-Mile)

29
Future Work

• Need to obtain updated track class traffic
  distribution estimates
• Provide more detailed rates breaking down
  accidents by cause groups (train-mile or car-mile
  correlated)
• Analyze other derailment attributes (e.g. speed,
  tonnage, position in train) and their effect on
  hazmat risk

30

• Thank you for your attention
• Paper No. 04-2341
• CD-ROM TRB2004-000341.pdf

31
Assigning Unique Accident Identification Numbers

• Distinguishing independent accidents when
  multiple consists/railroads involved
• Method to link all reports filed for a particular
  independent accident
• Compare FRA field codes
• RR3, INCDTNO3, STCNTY, Date
• Check JOINTCD information
• ACCAUSE, ACCTRK, ACCTRKCL, ACCDMG
• 1992-2001 FRA Data
• 34,061 reports
• 27,850 independent accidents
• Matching the number of accident reports with a
  joint code value of one (1)

Accessed on June 2, 2003.
32
Assigning Unique Accident Identification Numbers

• 1992-2001 FRA Data
• 34,061 reports
• 27,850 independent accidents
• 21,896 accidents with 1 report filed
• 5,726 accidents with 2 reports filed
• 212 accidents with 3 reports filed
• 9 accidents with 4 reports filed
• 4 accidents with 5 reports filed
• 1 accident with 6 reports filed
• 1 accident with 7 reports filed
• 1 accident with 8 reports filed

33
Track Class-Specific Derailment Rates

• Consider FRA track class as a proxy variable for
  statistical derailment probability estimation
• Higher track classes have more stringent
  requirements (resulting in better track quality)
  and higher maximum speeds
• Train Derailment Rate
• Ri DERi / TM x TMi, or Ri DERi / CM x
  CMi
• Ri derailment rate for class i track
• DERi number of derailments on class i track
• TM total freight train-miles (millions), or
• CM total freight car-miles (billions)
• TMi percent of total freight train-miles on
  class i track, or
• CMi percent of total freight car-miles on
  class i track
• Car Derailment Rate
• Ri DERi / CM x CMi
• Ri car derailment rate for class i track
• DERi number of cars derailed on class i track
• CM total freight car-miles (billions)
• CMi percent of total car-miles on class i track

34

35
Sensitivity Analysis

• Assume Estimated traffic distribution
  percentages representative for years 1992-1994
• Calculate 3-yr. derailment rate by track class
• Ri (92-94) DERi (92-94) / TMi (old) x TM
  (92-94), for class i track
• Assume Derailment rate has changed little over
  ten-year period that 3-yr. rate for 1999-2001
  is nearly the same as that for 1992-1994
• Calculate Traffic needed for constant derailment
  rates between 1992-1994 and 1999-2001 using
  actual derailment counts
• TMi (est) DERi (99-01) / Ri (92-94)
• Determine new traffic distribution percentages
  using calculated traffic from above
• TMi (est) TMi (est) / S TMi (est)
• Calculate new derailment rates using new
  estimated traffic distribution percentages and
  actual total mileage
• Ri (99-01) DERi (99-01) / TMi (est) x TM
  (99-01)

36

37
Hazardous Materials Release Probability