FREQUENCY MODEL RESULTS

The same nodes of the train derailment causes and their contributing factors were used in the frequency and severity causal models.

4.5.1.1 POINTS MACHINE PROBABILITY DISTRIBUTION

RESULTS

The probability distribution of the defective points machine, undetected defect in points machine and unrepaired defect in points machine nodes of the train derailment causal model is shown in figure 4.3. The probability distributions indicate that there is a 23.19% probability that a points machine is defective and results in a derailment under the following circumstances:

• an 84% probability that the defect was not detected during visual inspection; • a 5% probability that the defect was detected and was scheduled to be repaired but

caused a derailment before the day in which it would be repaired; and

• An 11% probability that the defect was detected and inefficiently repaired resulting in a derailment.

Considering that undetected points machine defects cause the greatest amount of derailments compared to detected defects, it is imperative that track masters are trained to improve their defect detecting skills.

Figure 4.3 Probability distributions of nodes concerning defective points machine.

4.5.1.2 RAIL GAUGE PROBABILITY DISTRIBUTION RESULTS

Figure 4.4 shows the probability distributions of incorrect rail gauges, undetected incorrect rail gauges and unrepaired incorrect rail gauges nodes of the train derailment causal model. The probability distributions indicate that there is a 5.96% probability that an incorrect rail gauge results in a derailment under the following circumstances:

• an 89% probability that the defect was not detected during visual inspection; • a 0% probability that the defect was detected and was scheduled to be repaired but

caused a derailment before the day in which it would be repaired; and

• An 11% probability that the defect was detected and inefficiently repaired resulting in a derailment.

Considering that undetected incorrect rail gauges cause the greatest amount of derailments compared to detected incorrect rail gauges, it is important that correct and well calibrated equipment is used to measure the rail gauges. Additionally, the rail gauge measuring skills of the track masters must be improved by training.

OP ERATI ONAL RIS K ANA LYS IS FOR THE M ANAGE MENT OF R AI LWAY INFR AS TRUC TURE MAINTE NAN CE

P HUMZILE DHLAM INI MAY 2010 56

Figure 4.4 Probability distributions of nodes concerning incorrect rail gauge

4.5.1.3 RETARDER/ADVANCER PROBABILITY DISTRIBUTION

RESULTS

The probability distributions of retarder/advancer defects, undetected retarder/advancer defects and unrepaired retarder/advancer nodes of the train derailment causal model are presented in figure 4.5. The probability distributions indicate that there is a 30% probability that a retarder/advancer defect may result in a derailment when the defect is undetected during inspection. The probability of retarder/advancer defects can therefore be substantially reduced by improving the defect detection skills of the track masters.

Figure 4.5 Probability distributions of nodes concerning defective retarder/advancers

4.5.1.4 BROKEN RAIL PROBABILITY DISTRIBUTION RESULTS

The probability distributions of broken rail, undetected broken rail and unrepaired broken rail nodes of the train derailment causal model are shown in figure 4.6. The probability distributions indicate that there is a 5.8% probability that an incorrect rail gauge results in a derailment under the following circumstances:

• an 87.5% probability that the defect was not detected during visual inspection; and

• a 12.5% probability that the defect was detected and was scheduled to be repaired but caused a derailment before the day in which it would be repaired.

Considering that undetected broken rail cause the greatest amount of derailments compared to detected incorrect broken rail, it is imperative the broken rail detection skills of track masters are improved. Additionally, the equipment that detects hidden rail defects such as ultra sonic measuring systems should be used at a greater frequency.

OP ERATI ONAL RIS K ANA LYS IS FOR THE M ANAGE MENT OF R AI LWAY INFR AS TRUC TURE MAINTE NAN CE

P HUMZILE DHLAM INI MAY 2010 58

Figure 4.6 Probability distributions of nodes concerning broken rail

4.5.1.5 SLACK PROBABILITY DISTRIBUTION RESULTS

The probability distributions of slack, undetected slack and unrepaired slack nodes of the train derailment causal model are displayed in figure 4.7. The probability distributions indicate that there is a 3% probability that a retarder/advancer defect may result in a derailment when the defect is undetected during inspection. The probability of retarder/advancer defects can therefore be substantially reduced by improving the slack defect detection skills of the track masters.

4.5.1.6 TRAIN DERAILMENT FREQUENCY PROBABILITY

DISTRIBUTION RESULTS

The probability distributions of train derailments, defective points machine, incorrect rail gauge, defective retarder/advancer, broken rail and slack nodes of the train derailment causal model are shown in figure 4.8. The probability distributions indicate that there is a 4.02% probability that a train derailment can occur due to the following reasons:

• a 23.19% probability of a defective points machine; • a 5.96% probability of an incorrect rail gauge; • a 30% probability of a defective retarder/advancer; • a 5.8% probability of a broken rail; and

• a 3% probability of slack.

It is estimated that the average amount of trains that move along the Johannesburg region per year is two thousand (2000); the number of train derailments that are likely to occur equals the product of the frequency probability and the number of trips that trains make in a year. Thus, the forecasted frequency is eighty (80) train derailments per year when there is an average of two thousand (2000) trains that pass the Johannesburg region railway infrastructure a year.

OP ERATI ONAL RIS K ANA LYS IS FOR THE M ANAGE MENT OF R AI LWAY INFR AS TRUC TURE MAINTE NAN CE

P HUMZILE DHLAM INI MAY 2010 60

Figure 4.8. Probability distributions of nodes concerning train derailments

4.5.1.7 THE EFFECT OF POINTS MACHINE DEFECT DETECTION

ON THE TRAIN DERAILMENT FREQUENCY PROBABILITY

DISTRIBUTION

The main contributing factor to all of the above causes of train derailments is the lack of detection of defects during inspection. This problem can be minimised using the following suggested solutions:

• the current defect detection techniques for visual inspection of rail infrastructure should be reviewed and ways should be found that can make defect detection more accurate;

• a program should be implemented in which trackmasters are given practical training by engineers, technicians and more experienced trackmasters that will empower them to improve their defects detecting skills; and

• the rate of use of equipment that detects underlying rail infrastructure defects that cannot be seen visually should be increased.

The use of causal modelling allows managers to predict the effect of any strategies that are made. Thus, the effect of decreasing the probability of any of the contributing factors of operational risks can be forecasted using causal modelling. In table xvii, the initially forecasted probability of undetected points machine defects is decreased by various percentages. This table indicates that a decrease in the probability of undetected points machine defects results in a decrease in the points machine defect and train derailment probability. Undetected defect in points machine 84% 75.6% 67.2% 58.8% 50.4% 42% 33.6% 25.2 % 16. 8% 8.4% 0% Points machine defect 23% 21.6% 19.9% 18.3% 16.7% 15% 13.4% 11.8 % 10. 1% 6.% 6.9% Train derailment frequency 4.02% 3.8% 3.6% 3.4% 3.2% 3% 2.83% 2.6% 2.4 % 2.2% 2%

Table xx: The forecasted effect of decreases in the probability of undetected points machine defects