Tank Car Safety Design Optimization to Reduce Hazardous ...
Transcript of Tank Car Safety Design Optimization to Reduce Hazardous ...
9/11/2009
1
Slide 1ILLINOIS - RAILROAD ENGINEERING
Tank Car Safety Design Optimization to Reduce Hazardous Materials Transportation RiskHazardous Materials Transportation Risk
M. Rapik Saat, Ph.D.p ,The William W. Hay Railroad Engineering Seminar Series
11 September 2009
9/11/2009
2
Slide 2ILLINOIS - RAILROAD ENGINEERING
OutlineOutline
• Overview of hazardous materials transport risk by rail
• Background on vehicle structural design optimization & railroad hazardous material transportation
• A Two-Phase Tank Car Safety Design Optimization Processy g
• Generalized Bicriteria Model for Optimizing Railroad Tank Car Safety Design
Risk Based Tank Car Safety Design Optimization Model• Risk-Based Tank Car Safety Design Optimization Model
• Application of tank car safety design optimization to reduce the environmental risk
• Future research
• Summary
9/11/2009
3
Slide 3ILLINOIS - RAILROAD ENGINEERING
Influence diagram showing relationships offactors affecting hazardous materials transportation safety
Financialand
SafetyFinancial Impact
g p y
yImpacts
TrafficVolume
People, Safety
Accident Rate
Routing
Property, orEnvironment
TrackCondition
Safety Impact
Release I id t R t
Hazard Area
g Condition
Mileage
Incident Rate
Infrastructure
Product
Release Probability
Car DesignRelease Quantity
Car Design
Decision Node Uncertainty Node Deterministic Node Value/Objective Node
Quantity
9/11/2009
4
Slide 4ILLINOIS - RAILROAD ENGINEERING
Structural optimization in vehicle design
• Vehicle structural designs are subject to both performance requirements and cost constraints
• For aircraft and most other aerospace systems, conceptual design optimization has typically been based on achieving efficient aerodynamics while minimizing weight configuration subject to structural requirements
• With regard to automotive design, crashworthiness criteria to maximize vehicle structural integrity for occupant safety in the event of a crash has been
id d t th ith th bj ti t i i i i ib ti h hconsidered together with the objectives to minimize noise, vibration, harshness (NVH), and weight or other cost constraints
9/11/2009
5
Slide 5ILLINOIS - RAILROAD ENGINEERING
Tank car weight vs. capacity tradeoffMaximum Gross Rail Load (GRL) = Lading Capacity + Light (Empty Weight)
Sulfuric Acid
Density = 14 26 lbs /gallon
Maximum Gross Rail Load (GRL) = Lading Capacity + Light (Empty Weight)
Density = 14.26 lbs./gallonca. 13,000 gallon tank
AlcoholDensity = 6.58 lbs./gallonca 29 000 gallon tankca. 29,000 gallon tank
• Tank cars can be made safer by increasing tank thickness and adding various• Tank cars can be made safer by increasing tank thickness and adding various protective features, but these increase the weight and cost of the car and reduce its capacity and consequent transportation efficiency
• Formal consideration of this tradeoff between tank car safety and transportationFormal consideration of this tradeoff between tank car safety and transportation efficiency, and use of optimization techniques to address this tradeoff represent the first phase involved in tank car safety design optimization
9/11/2009
6
Slide 6ILLINOIS - RAILROAD ENGINEERING
Railroad hazardous materials transportation
• More than 70% of about 2 million annual rail shipments of hazardous materials inMore than 70% of about 2 million annual rail shipments of hazardous materials in the U.S. & Canada are transported in tank cars
• Actual hazard posed by these materials varies widely in terms of both the nature and magnitude of the hazardand magnitude of the hazard
• In order to allocate safety enhancement resources in the most efficient manner possible requires quantitative understanding of the consequent risks and benefits
Th d h f th t k f t d i ti i ti d l dd• The second phase of the tank car safety design optimization model addresses chemical-specific hazard and its consequent risks and benefits
9/11/2009
7
Slide 7ILLINOIS - RAILROAD ENGINEERING
Tank car safety design optimization model
• Phase 1: A Generalized Bicriteria Model for Optimizing Railroad Tank Car Safety Design
• Addresses the tradeoff between safety and transportation efficiencyy p y
• Phase 2: Risk-Based Tank Car Safety Design Optimization
• Accounts for chemical-specific hazard levels and the consequent benefits and costsbenefits and costs
9/11/2009
8
Slide 8ILLINOIS - RAILROAD ENGINEERING
A generalized bicriteria model for optimizing railroadA generalized bicriteria model for optimizing railroad tank car safety design
T id tif P t ti l t f t k d i b d th t d ff b t• To identify a Pareto-optimal set of tank car designs, based on the tradeoff between safety and transportation efficiency
• Consideration of Risk Reduction Options (RROs)
- tank car safety design features
• Use of a statistical model to estimate tank car safety performance
• Development of a tank car weight & capacity model• Development of a tank car weight & capacity model
• Enumeration of tank car weight and safety performance metric
• Identification of a set of Pareto-optimal solutions
9/11/2009
9
Slide 9ILLINOIS - RAILROAD ENGINEERING
Tank car risk reduction options (RROs)
• Principal approaches considered to enhance tank car safety design:– Thicker/stronger head and/or head shield– Thicker/stronger shell– Adding top fittings protection– Removing bottom fittings
• Stronger tank and better-protected fittings improve accident performance• Also increase weight and cost, thereby reduce transport efficiencyg , y p y• Thus there is a tradeoff between enhanced safety and transport efficiency
9/11/2009
10
Slide 10ILLINOIS - RAILROAD ENGINEERING
Estimating tank car safety performance• More than 40 thousand records of tank cars involved in accidents have been
recorded since 1970 in the RSI-AAR Tank Car Accident Database
• Resultant database provides a robust source of information for quantitative l i f t k f t d ianalysis of tank car safety design
• Treichel et al (2006) developed a logistic regression model to estimate tank car conditional probability of release
PRi|A = 0.533 e L(i) / (1 + eL(i))• The calculated regression equations for the four release sources are:
L(HEAD) = - 0.4492 - 1.1672 HST - 1.9863 HMT - 0.9240 INS - 0.4176 SHELF-0.4905 YARD
L(SHELL) 0 4425 0 6427 INS 4 1101 STS 1 5119 YARDL(SHELL) = 0.4425 - 0.6427 INS - 4.1101 STS - 1.5119 YARD
L(TOP FITTINGS) = - 1.0483 - 0.8354 PRESS - 0.8388 INS + 0.1809 SHELF - 0.3439 YARD
L(BOTTOM FITTINGS) = - 1.4399 - 0.3758 INS - 0.5789 SHELF - 1.4168 YARD
9/11/2009
11
Slide 11ILLINOIS - RAILROAD ENGINEERING
IlliTank: Tank car weight & capacity program
Cap + LW ≤ GRL
where:GRL = gross rail loadCap = tank car maximum lading capacity in lbsLW = tank car empty weight
= tank head and shell assembly + head shields + insulation + jacket + top fittings protection + bottom fittings + non-tank componentsnon tank components
9/11/2009
12
Slide 12ILLINOIS - RAILROAD ENGINEERING
IlliTank list of variables
Variable Description Input Range UnitMaximum Gross Rail Load Numeric Value, typically 263,000 lbsMaximum Gross Rail Load Numeric Value, typically 263,000 lbsProduct Density Numeric Value lbs/gallonTank Outage Numeric Value, typically 2 or 5 %Tank Inside Diameter Numeric Value in.Tank Head Thickness Numeric Value in.Tank Shell Thickness Numeric Value in.Ceramic Fiber Insulation Thickness Numeric Value in.Fiberglass Insulation Thickness Numeric Value inFiberglass Insulation Thickness Numeric Value in.Tank Jacket Constant None or Jacketed -Head Shield Constant None, Half-height or Full-height -Bottom Fittings Constant None or Equipped -Top Fittings Protection Constant None or Equipped -Additional Weight Increase/Reduction Numeric Value lbs
9/11/2009
13
Slide 13ILLINOIS - RAILROAD ENGINEERING
Enumeration of all tank car RROs The safety performance and weight for all possibleThe safety performance and weight for all possible
RRO combinations are enumerated
No…
No…
No…
No…Typical…Yes
Half-Height…Full-Height 1/16” increment
from 0.4375”
…
…Yes
Enhanced
Bottom Fittings Removal BFR
Top Fittings Protection TFP
to 1.5” 1/16” increment from 0.4375”
to 1.5”Jacket JKT
e o a
Head ShieldHHP/FHP Head Thickness
H Shell ThicknessS
Jacket JKT
2 x 3 x 2 x 3 x 18 x 18 = 11,664 combinations
9/11/2009
14
Slide 14ILLINOIS - RAILROAD ENGINEERING
Enumeration of the conditional probability of release• The conditional probability of release were enumerated with 1/16” incremental
head and shell thicknesses, up to 1.5”
Head Thickness (inch)
0.4375 0.5000 0.5625 0.6250 0.6875 . . .
0.4375
( )
0 3090 0 3051 0 3014 0 2981 0 29510.4375
0.5000
0 5625Shell
Thickness
0.3090 0.3051 0.3014 0.2981 0.2951
0.2981 0.2940 0.2901 0.2867 0.2835
0 2889 0 2846 0 2806 0 2770 0 27370.5625
0.6250
0 6875
Thickness (inch)
0.2889 0.2846 0.2806 0.2770 0.2737
0.2813 0.2769 0.2728 0.2690 0.2657
0.6875...
0.2751 0.2705 0.2663 0.2625 0.2590
9/11/2009
15
Slide 15ILLINOIS - RAILROAD ENGINEERING
Pareto optimization
• Once all possible RRO combinations have been considered, the decision space is searched for a set of Pareto optimal solutions, from which the final design will be chosen from
• A set of solutions is called Pareto optimal if there is no other feasible solution that would improve some objective function without causing asolution that would improve some objective function without causing a simultaneous decline in at least one other objective function
• This approach has its roots in mathematical consumer economics as considered by Pareto (1896)considered by Pareto (1896)
• Pareto optimization has been used extensively in vehicle safety and crashworthiness design problems (Kasprzak et al 1998, Andersson & Redhe 2003, Hamza & Saitou 2005, Lee et al 2006, Cristello & Kim 2007, Sinha 2007, Sinha et al 2007)
9/11/2009
16
Slide 16ILLINOIS - RAILROAD ENGINEERING
Stepwise algorithm used to identify the Pareto-optimal (non-dominated) solutionsPareto optimal (non dominated) solutions
lity
of
Baseline
l Pro
babi
ase,
PR
|A
ondi
tiona
Rel
e
Efficient Frontier
Weight, W
Co Efficient Frontier
1) Compute W, PR|A and ΔW for all RROi; set i = 0 (base case); 2) From RROi, find RRO with the smallest ΔW and lower PR|A than current PR|Ai
3) Insert solution RROi+1 that has the minimum PR|A among RRO identified in step 23) Insert solution RROi+1 that has the minimum PR|A among RRO identified in step 2 to the set of Pareto-optimal solutions, S
4) Repeat steps 2 and 3 for all feasible solutions
9/11/2009
17
Slide 17ILLINOIS - RAILROAD ENGINEERING
Trade-off between PR|A and tank car weightE t d ll ibl RRO bi ti d i l tiEnumerated all possible RRO combination design solutions
(263,000-lb maximum GRL for 20,000-gallon baseline tank car)
0.30
ease
, PR
|A Dominated Non-Dominated
0.20
abili
ty o
f Rel
0.10
tiona
l Pro
ba
0.00-20 0 20 40 60 80
Con
dit
20 0 20 40 60 80
% Change in Light Weight
9/11/2009
18
Slide 18ILLINOIS - RAILROAD ENGINEERING
Expected quantity lost• Account for source-specific conditional
probability of release and average release size0.125
0 08
0.12
0.16
Prob
abili
ty
leas
e
where:
= expected gallon capacity lost given a tank i d il d i id t
0.059 0.063
0.018
0.00
0.04
0.08
H d Sh ll T B tt
Con
ditio
nal
of R
el
car is derailed in an accident
Cap = tank car gallon capacity
= mutually-exclusive and collectively-
Head Shell Top Fittings
Bottom Fittings
62.2 61.780
e of
t (
%)
y yexhaustive conditional probability of release from source i given a tank car is derailed in an accident 20.1
44.2
20
40
60
age
Perc
enta
gC
apac
ity L
ost
Qi = average percent tank capacity lost from source i
i = tank head (H), tank shell (S), top fittings (T),
0Head Shell Top
FittingsBottom Fittings
Aver
aTa
nk C
bottom fittings (B), and multiple causes (M)(DOT-111 Non-Insulated Tank Car)
9/11/2009
19
Slide 19ILLINOIS - RAILROAD ENGINEERING
Trade-off between expected quantity lost and tank car weight
Enumerated all possible RRO combination design solutions (263,000-lb maximum GRL for 20,000-gallon baseline tank car)
3,000
lon)
Dominated Non-Dominated
2,000
ty L
ost (
Gal
l
1,000cted
Qua
ntit
0
Expe
c
-20 0 20 40 60 80
% Change in Light Weight
9/11/2009
20
Slide 20ILLINOIS - RAILROAD ENGINEERING
Identification of a compromise solution using the utopia point method
80
100
y Lo
st
utopia point method
60
80
nk C
apac
ity
40
enta
ge T
an
Compromise Solution
where:
N(x) = Euclidean distance
F( ) bj ti f ti t
0
20
Perc
e
Utopia PointMinimum distance
F(x) = objective functions vector
Foa,b = utopia point vector
x = feasible design space
Efficient Frontier
00 20 40 60 80 100
Percentage Increase in Light WeightPercentage Increase in Light Weight
References:Vincent, T.L. 1983, Game theory as a design tool. J. Mech. Transm. Autom. Des. 105, 165–170
Kämpke, T. 2007, The Reference Point Method in Primer Design, Methods in Molecular Biology, Yuryev, A., ed.
9/11/2009
21
Slide 21ILLINOIS - RAILROAD ENGINEERING
Examples of the applications of Phase I conceptual approach • Risk Analysis of Toxic Inhalation Hazard (TIH) Materials’ Transportation on U S• Risk Analysis of Toxic Inhalation Hazard (TIH) Materials Transportation on U.S.
Railroad Mainlines
• The utopia point method was used to select among the Pareto-optimal set of combinations to identify candidate designs for enhanced tank cars for TIHscombinations to identify candidate designs for enhanced tank cars for TIHs
• Safety and transportation efficiency were assumed equally weighted, in part because that is what the Association of American Railroads (AAR) specified, but largely because no explicit information on how to differentially assign thebut largely because no explicit information on how to differentially assign the preference level or weight on safety performance versus railcar capacity or cost was available
• Barkan C P L 2008 Improving the design of higher-capacity railway tank carsBarkan, C.P.L. 2008, Improving the design of higher capacity railway tank cars for hazardous materials transport: Optimizing the trade-off between weight and safety, Journal of hazardous materials, vol. 160, no. 1, pp. 122-134.
• A goal programming approach was used to identify the optimal safety designA goal programming approach was used to identify the optimal safety design combinations for higher GRL (total weight) tank cars for the AAR
• The industries had agreed a-priori that one third of the incremental weight would go toward enhanced safety and the remaining two thirds to extra g y gcapacity
9/11/2009
22
Slide 22ILLINOIS - RAILROAD ENGINEERING
Limitations in the Phase I approachLimitations in the Phase I approach
• The bicriteria tank car safety design optimization model enables identification of a specific, Pareto-optimal set that represents the most efficient combinations of tank car safety design options
• However, that model does not provide a means of determining what the optimal level of safety or performance is for any particular product
• The utopia point method or a goal programming formulation, can provide an objective approach to identify the optimal solution
• However the underlying assumption of equal preference in the utopia pointHowever, the underlying assumption of equal preference in the utopia point method, or a decision maker’s specification to allocate a specific weight increment for safety, leaves an element of subjectivity in the process of identifying the final decision for individual car designs
• Phase II of the tank car safety design optimization process gives an advancement to the work that has already been done by considering explicit chemical-specific hazard and the consequent benefits and costs to identify the
ti l l ti i th t t l hoptimal solution using the net present value approach
9/11/2009
23
Slide 23ILLINOIS - RAILROAD ENGINEERING
N t t l (NPV) l iNet present value (NPV) analysis
• Benefit and cost streams generally extend into the future from some decision point
• The NPV method accounts for the future benefits and costs and the• The NPV method accounts for the future benefits and costs, and the time value of money within a specific analysis period
• Provides an objective means for decision makers to compare the cost-effectiveness of different feasible alternatives
9/11/2009
24
Slide 24ILLINOIS - RAILROAD ENGINEERING
Fleet replacement schedule
• An important variable in the NPV analysis that determines how fast the full benefit and cost would be realized
• Immediate Replacement• Immediate Replacement• Chemicals with extremely high hazard may justify an immediate fleet replacement with
enhanced-design tank cars
With thi i th f ll b fit d t ld b d i di t l• With this scenario, the full benefit and cost would be accrued immediately
• Attrition-Based Replacement• Tank cars are replaced with enhanced designs at the end of their normal service life, p g
typically between 30 to 40 years
• With this scenario, the full benefit and cost are accrued proportionally over the life-span of a tank car
• Accelerated Replacement• 1/n of the fleet is replaced annually
• The benefit and cost would be accrued proportionally over the n-year period after whichThe benefit and cost would be accrued proportionally over the n year period after which the benefit and cost would be fully realized
9/11/2009
25
Slide 25ILLINOIS - RAILROAD ENGINEERING
Risk-based tank car safety design optimization• I develop a quantitative model that combines the bicriteria optimization method
with a benefit-cost approach based on maximizing the NPV
• Enables chemical-specific hazard and risk to be used with the NPV approach to objectively determine the optimal tank car safety design for each material
• The risk-based tank car safety design optimization concept will be illustrated by using idealized benefit and cost curves
• The first step involved is to define a set of Pareto-optimal solutions, then consider how chemicals with different hazard levels affect the optimality
4,000
2 000
3,000
uant
ity L
ost
on)
1,000
2,000
Expe
cted
Qu
(Gal
l
00 20 40 60 80
E
Percentage Change in Light Weight (%)
9/11/2009
26
Slide 26ILLINOIS - RAILROAD ENGINEERING
Risk analysis framework
• For a set of Pareto-optimal solutions identified, the accident-caused release risk can be p ,estimated as follows:
where:= accident-caused risk for transporting chemical j= accident-caused release rate for a tank car transporting chemical j
b bilit f l i i i t k l d it t t= probability of release size i given a tank car released its contentQi = average release quantity
= average percentage tank capacity lost for release size i × tank car capacityC h i l j lCj = chemical j release consequence
9/11/2009
27
Slide 27ILLINOIS - RAILROAD ENGINEERING
Risk per ton-mileRisk per ton mile
Assuming CL < CM <CH, where CM = 5 CL and CH = 10 CL
6
8
e (¢
)
4
per T
on-M
il
Chemical H
0
2
Ris
k p Chemical M
Chemical L
0 20 40 60 80Percentage Change in Light Weight (%)
• The higher the hazard level, the higher the risk for all weight increments
9/11/2009
28
Slide 28ILLINOIS - RAILROAD ENGINEERING
Estimating risk reduction or benefit
where:where:Benefitt = risk reduction or benefit at year t
= accident-caused risk when baseline tank car design is usedRR = accident-caused risk when enhanced tank car design is usedRR accident caused risk when enhanced tank car design is usedρt = proportion of total tank car fleet replaced at year t
= (t + 1)/θ if (t + 1) ≤ θ, else ρt = 1θ = phase in period based on tank car fleet replacement scheduleθ = phase-in period based on tank car fleet replacement schedule
• Net benefit over a certain present-value analysis period can be estimated as follows:
where:PVBenefit = present-value benefit or risk reductionBenefit pY = present-value analysis periodd = discount rate
9/11/2009
29
Slide 29ILLINOIS - RAILROAD ENGINEERING
Benefit per ton-mileBenefit per ton mile
Assuming CL < CM <CH, where CM = 5 CL and CH = 10 CL
8
10
Mile
(¢)
Chemical H
4
6
ft pe
r Ton
-
Chemical M
0
2
PV B
enef
Chemical L
0 20 40 60 80Percentage Change in Light Weight (%)
• At any specific weight increment the benefits are higher for chemicals with higher hazards
9/11/2009
30
Slide 30ILLINOIS - RAILROAD ENGINEERING
Estimating costs related to fleet replacementst at g costs e ated to eet ep ace e t
• Tank car replacement incurs incremental increases in both capital and operating costsand operating costs
• Capital includes tank car life-cycle cost, i.e. the cost of buying a new car, maintenance costs, salvage value and other expenses throughout the life span of a carthroughout the life-span of a car
• It must also account for the total number of tank cars required to replace a fleet, and the replacement schedule
• Operating cost accounts for the total number of shipments and the cost per trip
Total cost is the sum of capital and operating costs for any• Total cost is the sum of capital and operating costs for any particular design
9/11/2009
31
Slide 31ILLINOIS - RAILROAD ENGINEERING
Identifying minimum tank car fleet size
where:where:N = minimum total enhanced-design tank cars in a fleet
S = annual number of shipments with baseline tank cars
Cap = nominal gallon capacity of a baseline tank car
Cap’ = nominal gallon capacity of an enhanced-design tank car
T = tank car utilization rate (annual trips per car)T = tank car utilization rate (annual trips per car)
9/11/2009
32
Slide 32ILLINOIS - RAILROAD ENGINEERING
Tank car fleet replacement or pcapital cost estimation
where:PV present al e of total fleet replacement costPVFleet = present value of total fleet replacement cost LCTankCar = life-cycle cost of a tank carmt = total number of enhanced-design tank cars entering
the fleet in year t= N/θ
9/11/2009
33
Slide 33ILLINOIS - RAILROAD ENGINEERING
Operating cost estimationOperating cost estimation
hwhere:PVOpr = present value of total fleet operating cost
M = number of car milesM = number of car miles
COpr = operating cost per mile
9/11/2009
34
Slide 34ILLINOIS - RAILROAD ENGINEERING
Incremental cost estimationIncremental cost estimation
where:PVIncremental Cost = present value of total incremental cost PVFleet = present value of fleet replacement cost with
enhanced-design tank carsPVOpr = present value of operating cost with enhanced-design
tank cars= present value of fleet replacement cost with
baseline tank cars = present value of operating cost with baseline tank cars
9/11/2009
35
Slide 35ILLINOIS - RAILROAD ENGINEERING
Incremental present-value capital, operating & total costs per ton-mile
8
Incremental present value capital, operating & total costs per ton mile
Total6
8
Cos
t (¢
)
2
4
ncre
men
tal
r Ton
-Mile
Capital
0
2
0 20 40 60 80
PV In pe Operating
0 20 40 60 80
Percentage Change in Light Weight (%)
9/11/2009
36
Slide 36ILLINOIS - RAILROAD ENGINEERING
Cost effectiveness evaluation –PV benefits & costPV benefits & cost
7.5
9.0B
enef
it
Total Incremental
Chemical H Benefit
4.5
6.0
al C
ost &
Bn-
Mile
(¢) Incremental
Cost
Chemical MBenefit
1.5
3.0
Incr
emen
tape
r To
Chemical L Benefit
0.00 20 40 60 80PV
Percentage Change in Light Weight (%)
e e t
9/11/2009
37
Slide 37ILLINOIS - RAILROAD ENGINEERING
Cost effectiveness evaluation - NPV
4
6 (¢
) Ax
0
2
er T
on-M
ile Chemical H NPV
x
BxC
xDx
-2
00 20 40 60 80
NPV
pe
Chemical M NPV
x x
-6
-4
Percentage Change in Light Weight (%)
Chemical LNPV
9/11/2009
38
Slide 38ILLINOIS - RAILROAD ENGINEERING
Application of the risk-based tank car safety design optimization modeldesign optimization model
• Summarize a risk analysis of rail transportation y pinvolving a group of chemicals that pose hazard to the environment
• Use the risk analysis results to evaluate cost-effectiveness of tank car safety design enhancementsenhancements
9/11/2009
39
Slide 39ILLINOIS - RAILROAD ENGINEERING
Decision & risk analyses framework
Remote
0.1-4.9Environmental Cleanup Cost
Evacuation Cost
Train Delay Cost
0-5%
Clay
10
20
Rural
Suburban
5-9.9
10-19.9
20-39.9
BaselineDesign
Yes
0 5%
5-20%
Silt
Sand
20
50
100
Urban
High
E t l
40-59.9
60-99.9
AlternativeDesign
No 20-50%
50-80%
80-100%
200 Extremely High
≥ 100
Tank Car Design Alternatives
Accident-Caused Release
Release Quantity as Percentage
of Tank Car Capacity
Soil Type Depth toGroundwater,
(ft)
PopulationClass
Traffic Density Category (MGTM)
Decision Probability Consequences
9/11/2009
40
Slide 40ILLINOIS - RAILROAD ENGINEERING
Probability analysis• Accident-caused release rate metric was used to estimate the rate of a
l trelease event:
PR = PA x PR|A x M x Cap/Cap’hwhere:
PA = tank car derailment rate
P = tank car conditional probability of releasePR|A = tank car conditional probability of release
M = total number of car miles
Cap = nominal gallon capacity of a baseline tank carCap nominal gallon capacity of a baseline tank car
Cap’ = nominal gallon capacity of an alternate-design tank car
9/11/2009
41
Slide 41ILLINOIS - RAILROAD ENGINEERING
Accident-caused release rate summary
0 647Methanol
(The “probability” or frequency term in the risk definition)
0.1560.199
0.3650.647
Methyl MethacrylateVinyl Acetate
XylenesMethanol
0 0810.1030.119
0.152
C l hTolueneEthanolStyrene
y y
0 0380.0500.0720.081
BenzeneAcrylonitrile
Butyl AcrylatesCyclohexane
0.0260.0350.038
0 0 0 2 0 4 0 6 0 8
Ethyl AcrylateEthyl Acetate
Benzene
0.0 0.2 0.4 0.6 0.8Accident-Caused Release Rate
9/11/2009
42
Slide 42ILLINOIS - RAILROAD ENGINEERING
Consequence analysis• Impacts to Soil and Groundwaterp
• Hazardous Materials Transportation Environmental Consequence Model (HMTECM) was used to estimate soil and groundwater cleanup cost
• Accounts for physicochemical properties, soil type and depth to groundwater
• Population Exposure
Soil Type
• US Emergency Response Guidebook (ERG) was used to determine hazard area
• Impact in terms of evacuation cost was estimated
• Train Delay• Estimate impact due to additional costs related to
locomotives, railcars, fuel and labor
Depth to Groundwater
• Accounts for traffic density to estimate total number of trains delayed
References:
Yoon H Werth C J Barkan C P L Schaeffer D J & Anand P 2009 "An environmental screening model
Population Density
Yoon, H., Werth, C.J., Barkan, C.P.L., Schaeffer, D.J. & Anand, P. 2009, An environmental screening model to assess the consequences to soil and groundwater from railroad-tank-car spills of light non-aqueous phase liquids", Journal of hazardous materials, vol. 165, no. 1-3, pp. 332-344.
Schafer, D.H. & Barkan, C.P.L. 2008, "A prediction model for broken rails and an analysis of their economic impact", Proceedings of the AREMA Annual Conference, Salt Lake City, UT.
9/11/2009
43
Slide 43ILLINOIS - RAILROAD ENGINEERING
Total expected consequence costE t d Cl C t E ti C t T i D l C t
(The consequence term in the risk definition)• Expected Cleanup Cost + Evacuation Cost + Train Delay Cost
898,507907,833
1,069,5831,239,038
AcrylonitrileTolueneXylenes
Cyclohexane
815 172844,454882,007859,578898,507
BenzeneMethyl Methacrylate
Ethyl AcetateEthyl Acrylate
y
643,117775,925795,799815,172
Butyl AcrylatesStyrene
MethanolBenzene
559,041627,185
0 500,000 1,000,000 1,500,000
EthanolVinyl Acetate
Total Consequence Cost ($)
9/11/2009
44
Slide 44ILLINOIS - RAILROAD ENGINEERING
Risk estimationA id t C d R l R t T t l E t d C C t• Accident-Caused Release Rate x Total Expected Consequence Cost
390,711515,051
XylenesMethanol
0.0420.048
XylenesCyclohexane
99,801118,081124,871131,756
CyclohexaneStyrene
Vinyl AcetateMethyl Methacrylate
0.0320.0330.0340.035
Methyl MethacrylateEthyl AcrylateEthyl Acetate
Toluene
45,16846,187
66,64193,545
AcrylonitrileButyl Acrylates
EthanolToluene
0 0200.0220.024
0.0310.032
BenzeneEthanol
Vinyl AcetateMethanol
Acrylonitrile
22,07530,74430,998
0 200,000 400,000 600,000
Ethyl AcrylateEthyl Acetate
Benzene
0.0160.0200.020
0.00 0.01 0.02 0.03 0.04 0.05 0.06
Butyl AcrylatesStyrene
Benzene
Annual Risk Risk per Ton-Mile
Annual Release Risk ($) Risk per Ton-Mile (¢)
9/11/2009
45
Slide 45ILLINOIS - RAILROAD ENGINEERING
Considering tank car safety design enhancements
No
No…
No
H lf H i ht
…
No…
Enhanced
Typical…
Half-Height…Full-Height 1/16” increment
from 0.4375” to 1 5”
…
1/16” increment
…Yes
Enhanced
Top Fittings Protection TFP
Head Shield
to 1.5 1/16 increment from 0.4375”
to 1.5”Jacket JKT
HHP/FHP Head ThicknessH Shell Thickness
S
9/11/2009
46
Slide 46ILLINOIS - RAILROAD ENGINEERING
Identifying Pareto-optimal solutions
5,000n)
3 000
4,000
ost (
Gal
lon
2,000
3,000
Qua
ntity
Lo
0
1,000
Expe
cted
Q
Cyclohexane (H)Acrylonitrile (M)
Butyl Acrylates (L)
0 20 40 60 80 100
E
% Change in Light Weight
9/11/2009
47
Slide 47ILLINOIS - RAILROAD ENGINEERING
Estimating the benefit for Pareto-optimal solutions
0.20
0 12
0.16
-Mile
(¢) Cyclohexane (H)
Acrylonitrile (M)
0.08
0.12
fit p
er T
on-
0 00
0.04
Ben
ef Butyl Acrylates (L)
0.000 20 40 60 80
% Change in Light Weight
9/11/2009
48
Slide 48ILLINOIS - RAILROAD ENGINEERING
Estimating the incremental cost for Pareto-optimal solutions
16e Acrylonitrile
Pareto-optimal solutions
10
12
14
er T
on-M
ileAcrylonitrile
6
8
10
tal C
ost p
e(¢
)
Cyclohexane
Butyl Acrylates
0
2
4
Incr
emen
Cyclohexane
00 20 40 60 80
% Change in Light Weight
9/11/2009
49
Slide 49ILLINOIS - RAILROAD ENGINEERING
Estimating the NPV for Pareto-optimal solutions
00 20 40 60 80
-4
0 20 40 60 80ile
(¢)
Cyclohexane (H)
-8
per T
on-M
Acrylonitrile (M)
Butyl Acrylates (L)
-16
-12
NPV
16% Change in Light Weight
On the basis of the NPV, it is not cost justified to replace the fleets of any of the chemicals of interest with enhanced-design tank cars
9/11/2009
50
Slide 50ILLINOIS - RAILROAD ENGINEERING
Minimum risk-cost multiplier to attain positive NPV solutionsNPV solutions
3234
BenzeneStyrene
2123
2630
32
Ethyl AcetateButyl Acrylates
Methyl MethacrylateEthanol
Benzene
18212121
Vinyl AcetateToluene
Ethyl AcrylateEthyl Acetate
89
1517
XylenesCyclohexane
MethanolAcrylonitrile
0 10 20 30 40
Risk-Cost Multiplier, μ
9/11/2009
51
Slide 51ILLINOIS - RAILROAD ENGINEERING
Future researchFuture research• Addressing Constraints in Existing Work
• Considering Multiple Car Derailments and Multiple Car Releases• Considering Multiple-Car Derailments and Multiple-Car Releases
• Considering Varying Transportation Demand (i.e. shipments over 10-40 years)
• Improving Chemical-Specific GIS Route Creation Process
• Considering Other Decision Making Techniques
• Developing a More Detailed Uncertainty Analysis
• New Research Directions
• Evaluating Unconventional Tank Car Designs’ Performance
• Considering Multiple Hazards and Risk ImpactsConsidering Multiple Hazards and Risk Impacts
• Considering Transportation Security
• Considering Other Strategies to Reduce Hazardous Materials T t ti Ri kTransportation Risk
9/11/2009
52
Slide 52ILLINOIS - RAILROAD ENGINEERING
Summaryy
• Tank car safety design optimization is presented in my dissertation as a two-phase processp
• The first phase addresses the tradeoff between safety and transportation efficiency by using Pareto optimization to identify the most efficient non-dominated design combinations of safety performance and weight
• The second phase involves incorporating chemical-specific hazard level and the consequent benefit and cost to determine the optimal level of protection for different hazardous materials
• My dissertation research provides decision tools and parametric models to assess hazardous materials transportation risk, identify optimal tank car safety design, and estimate potential risk reduction options and their associated benefit and cost
• The framework presented in this research is intended to assist industry and government policy makers to make better-informed decisions for safer transportation of hazardous materials
9/11/2009
53
Slide 53ILLINOIS - RAILROAD ENGINEERING
Acknowledgements• RSI-AAR Railroad Tank Car Safety Project
• CN Graduate Research Fellowship
• Dow Chemical
9/11/2009
54
Slide 54ILLINOIS - RAILROAD ENGINEERING
QUESTIONS?QUESTIONS?