TRB2005 MRS2

Slide1: 

RELEASE RISK AS METRIC FOR EVALUATING TANK CAR SAFETY PERFORMANCE University of Illinois at Urbana-Champaign Railroad Engineering Program M. Rapik Saat Christopher P.L. Barkan

Outline of Presentation: 

Outline of Presentation Railroad safety trends Introduction to tank cars Development of tank car release risk model Utilizing release risk for tank thickness optimization model

Railroad Mainline Accident Rate 1980-2003: 

Railroad Mainline Accident Rate 1980-2003

Railroad Accident Caused Hazardous Material Release Rate 1982-2002: 

Railroad Accident Caused Hazardous Material Release Rate 1982-2002

Railroad Tank Car: 

Railroad Tank Car UTLX 662281 - example of a commonly used tank car design Class DOT 111A100W1 permitted to transport certain hazardous materials Non-pressure car used for transporting liquids Various top fittings for loading, pressure relief, gauging, etc. Bottom outlet for gravity unloading 7/16” A-516 steel tank

Tank Car Sources of Release in Accidents: 

Tank Car Sources of Release in Accidents Sources of release can be divided into two groups: Tank (= Head and Shell) Non-Tank (=Top and Bottom Fittings) These two groups perform differently in accidents and require different approaches to improving accident performance Bottom Fittings Top Fittings Shell Head

Probability of Release vs. Quantity Lost : 

Probability of Release vs. Quantity Lost The average quantity lost in an accident varies depending on the part of the car that is damaged The quantity of contents released affects the severity of a release incident Probability of Release Quantity Lost

Release Risk Model Development: 

Release Risk Model Development Risk = frequency x consequence Frequency = probability of release for a car damaged or derailed in an accident Consequence = quantity lost expressed as the percentage of the tank’s total volumetric capacity Release Risk = expected value of the quantity lost from a tank car given that it is in an accident

Tank Car Release Risk: 

Tank Car Release Risk where:   RR = release risk for a tank car in percentage of tank capacity lost per mile traveled m = number of release sources considered (RRTR,RRNR) n = number of release sizes considered RRi,j = risk for release size i from release source j

Tank Car Release Sizes: 

Tank Car Release Sizes Data from RSI-AAR Tank Car Safety Project database

RRTR|A is a function of tank thickness: 

RRTR|A is a function of tank thickness Data from RSI-AAR Tank Car Safety Project database Generic regression model: RRTR|A (t) = v + we(-y t +z) RRTR|A (t) = 0.40951 + 4.72098 e(-6.35515 t +3.22174) R2 = 0.884

Tank thickness versus “Light Weight” and Capacity of tank cars: 

Tank thickness versus “Light Weight” and Capacity of tank cars Capacity + Light Weight = Gross Rail Load (GRL) = 263,000 lbs Different size cars (gallon capacity)

Tank Car Release Risk: 

Tank Car Release Risk where: RRj = release risk from source j P Ri |A = conditional probability of release size i given car derails P Rj |A = conditional probability of release from source j given car derails PA = probability a car will derail per mile Vi,j = average percentage of tank capacity lost for release size i from source j K(t - t’) = percentage increase in shipments where: t’ = base case tank thickness t = tank thickness K = tank car specific constant (K = coefficient of a linear equation for the relationship between thickness and percentage increase in shipments required) PA(t) = PA [1 + K(t - t’)] = increased probability of an accident involving shipments in tank cars of thickness t

Tank car release risk as a function of tank thickness: 

Tank car release risk as a function of tank thickness RR (t) = RRTR (t) + RRNR (t) where: RRTR (t) = [v + we(-y t +z) ]PA[1+K(t-t’)] = tank-caused release risk as a function of thickness RRNR (t) = 32.125 • PNR|APA[1+K(t-t’)] = non-tank-caused release risk as a function of thickness PNR|A = conditional probability of non-tank-caused release given car derails (unique for each tank car safety design) = 0.207 (DOT-111-general-purpose tank car)

Optimality Model: 

Optimality Model Minimize [[1+K(t-t’)] • [v + we(-y t +z) + 32.125 • P(NR|A)]] objective function is solved numerically to find t such that RR(t) is minimized, t* t* = optimal thickness

Graphical Model: 

Graphical Model Non-insulated tank car with 20,000-gallon capacity, K = 0.237, baseline car-miles M’ = 1 million RR (t) RRNR (t) RRTR (t) t*= 0.9747 RR(t*)= 1.05

Tank car sizes are optimized for different density ladings : 

Tank car sizes are optimized for different density ladings Sulfuric Acid Alcohol Density = 14.26 lbs./gallon ca. 13,000 gallon tank Density = 6.58 lbs./gallon ca. 29,000 gallon tank

Different size tank cars have different values of K: 

Different size tank cars have different values of K 6 lbs/gallon 9 lbs/gallon 12 lbs/gallon 15 lbs/gallon (K = coefficient of a linear equation for the relationship between thickness and percentage increase in shipments required)

t* varies with tank cars for different product densities: 

t* varies with tank cars for different product densities Inverse relationship between product density and tank size Inverse relationship between tank size and t* 6 lbs/gallon 9 lbs/gallon 12 lbs/gallon 15 lbs/gallon t*= 0.911 t*= 0.960 t*= 0.996 t*= 1.025

Conclusion: 

Conclusion Release risk is a useful metric for assessing the benefit from changes in tank car safety design because it takes into account both the probability and severity of a release incident Tank thickness can be optimized to minimize release risk Constructing thicker tanks only reduces risk up to a point There is no single optimum Further work differentiate the tank and fittings elements in the release risk model analyze the effects of implementing other safety measures e.g head shields, top fittings protection, bottom fittings removal incorporate hazard levels of specific chemicals/products incorporate the risk tolerance by adapting formal decision analysis techniques