Railway Investigation Report SR9401

1.1 Introduction

This report presents the results of a study on Canadian main track derailments. It includes:

• a statistical analysis of main track derailment trends since 1980;
• an evaluation of more recent occurrences;
• a review of safety actions recommended and taken; and
• consideration of the potential safety actions that may further reduce the frequency of main track derailments.

The Transportation Safety Board of Canada (TSB) has been concerned about main track derailments for several years. The Board initiated this study following apparent increases in such occurrences in late 1992. In addition, the Board received, in January 1993, a request to examine the issue from a member of the House of Commons Standing Committee on Transport^{Footnote 1}.

1.2 General Conclusions

From a study of both long-term trends and recent main track derailments, it was concluded that:

1. Canadian main track derailments have declined by almost a factor of three between 1980 and 1988. Adjusted for changes in reporting requirements, the data indicate that the annual number of derailments has remained basically unchanged since 1988.
2. The reason for the 1980 to 1988 decline in derailments is most probably a combined result of many different factors. These include:
   • improved installation and repair procedures for welded rail; increased use of automatic rail defect detection and track geometry measurement technology;
   • increasing proportion of the car fleet equipped with roller bearings; gradual elimination of straight-plate wheels;
   • improved marshalling requirements;
   • increased number of hot box detectors (HBDs); and
   • rigorous government safety regulatory enforcement and inspection programs.
3. The reasons for the levelling-off of the derailment trend post-1988 are probably many. Changes in technology and methods of operation affected some types of derailments both here and in the U.S., leading to a very similar pattern of decline followed by level trend in both countries. The appearance of a single point in time when the trend changed may, in fact, be coincidental. As to the possible impact of legislative and regulatory change, there is no direct evidence from the occurrence record to confirm or refute such a possibility.
4. No passenger or member of the general public has been fatally injured because of a main track derailment in Canada since 1983. Apart from two operating employees being fatally injured because of a 1992 derailment caused by a beaver dam failure, no employee has been fatally injured because of a main track derailment since 1984. Considering the annual toll of around 50 to 70 fatalities that occur at railway grade crossings, the hazards to persons posed by train derailments have proven to be relatively low. However, the hazards associated with derailments of trains carrying dangerous goods remain potentially serious. Actual releases of dangerous goods following derailments have declined from the early eighties, and are at a low level.
5. The rate of main track derailments in Canada is and consistently has been less than in the United States.
6. Persistent themes in the factors underlying many of the main track derailments since 1991 include: broken rail with previously undetected internal defects; inadequate track geometry maintenance; undetected roller bearing failures; worn truck components; and straight-plate wheel failures.
7. Although rail defect detection technology contributed significantly to reductions in broken rail derailments in the eighties, failure to detect internal defects has contributed to recent main track derailments attributed to rail failure.
8. Track geometry problems continue to cause main track derailments; some of these problems could be ameliorated by improved preventative maintenance.
9. Straight-plate wheel failures continue to lead to derailments. Action has been initiated by Transport Canada and the railway industry to eliminate this wheel design from the revenue car fleet; however, there is some debate as to the rate at which wheel replacement should proceed.
10. Derailments caused by undetected roller bearing failures continue and, while some improvements are possible in making the best use of available data, there appears to be no immediate solution without a major capital investment in HBDs.
11. Human performance is a major concern in both operations-related derailments, and in occurrences where the output of automatic detection systems was inappropriately actioned. At present, the TSB has little human performance data concerning these situations. The TSB and, increasingly the railway industry, recognize the importance of developing a better understanding of those human behaviours which can create or exacerbate an accident situation.

Towards the end of 1992, the TSB was concerned about the increase in the number of reported main track derailments in 1992 (122) compared with 1991 (106). There were also three major derailments in December 1992. One of these involved the release of dangerous goods, which caused a 22-day evacuation of the residents of Oakville, Manitoba, over the Christmas holidays and attracted considerable public attention. At about the same time, concern was also expressed over the possible impact of large reductions in the numbers of railway employees.

In January 1993, Mr. Iain Angus, then MP for Thunder Bay/Atikokan and New Democratic Party Transport Critic, requested the TSB to conduct a special investigation into Canadian railway safety. That request cited "...a 20-percent increase in mainline derailments since transport deregulation was implemented".

After conducting a preliminary analysis of the statistical trends, the TSB decided that there was no indication of a single systemic safety problem. However, considering the trends and the perceived potential impact that changes taking place in the industry have on safety, the Board directed, in February 1993, that main track derailments would be subject to increased attention during the following year, with specific emphasis on:

• Increased depth of investigation of derailment occurrences with respect to equipment inspection, maintenance, operating practices, and human factors.
• Statistical analysis of occurrence data at a disaggregated level.
• Review of recent safety initiatives that would reduce the frequency and/or severity of main track derailments. This would include Board safety initiatives, along with actions taken by Transport Canada (TC) and the industry.

This report presents the results of that study. The analysis is based primarily on the results of the Board's investigations of reported occurrences, either individually or in aggregate, and the safety deficiencies revealed by those occurrences. (For analysis purposes, occurrences previously reported to the Canadian Transport Commission (CTC) and to the National Transportation Agency of Canada (NTA) are included, as are some NTA safety actions that bear upon issues still of concern.)

It should also be noted that the Board's normal ongoing investigation and assessment of rail safety occurrences continued while this study was under way. All investigated derailments are the subject of full TSB investigation reports in the normal fashion. Similarly the identification of safety deficiencies revealed by occurrences has been proceeding, and a number of safety recommendations and safety advisories related to main track derailments were issued during 1993. Other safety actions may be proposed resulting from investigations yet to be publicly reported on by the Board.

The approach taken in this report is to review historical trends in derailment rates, severity, consequences, and causes. Recent occurrence patterns are examined in more detail, including an examination of those safety issues that have been the cause of concern to both the NTA and the TSB. Finally, the question of whether new safety actions are needed to improve the safety performance with respect to main track derailments is considered.

3.1 Data Sources

The data analysis presented in this report is based primarily on occurrence data reported to the TSB and previously to the NTA and the CTC. During the period covered by this analysis, there have been changes to regulations governing the reporting of main track derailments. Some of these changes did have an impact on the number of occurrences reported and could, therefore, affect the interpretation of trends.

In the light of these changes, some data series have been adjusted to reflect a consistent definition (see Appendix A for details). For comparison purposes, the derailments as reported are also shown. This is only done for the overall trend analysis. Detailed breakdowns of derailments are based on the number reported, since estimations of the impact of reporting changes within subcategories of derailments could be highly inaccurate.

3.2 General Trends

From 1982 to 1988, there were significant and continuing reductions in the rate of main track derailments^{Footnote 2} (Figure 1). Since 1988, the rate has remained essentially constant (after adjustment for changes in reporting criteria). While there was an increase in 1992, the rate declined again in 1993 to return to about the same level as in 1988.

The decline in the main track derailment rate from 1982 to 1988 indicates a significant improvement during that period. While this improvement is probably due to many factors, it is notable that the Grange Inquiry into the derailment at Mississauga (10 November 1979) submitted its findings in January 1981. Subsequently, the Railway Transport Committee of the CTC addressed a number of the factors influencing main track derailments, their early detection, and the reduction in risk of consequent release of dangerous commodities. Over the ensuing years, the majority of these decisions were implemented along with an ongoing program of investigation of occurrences and research into factors influencing derailments. All of these activities clearly had an effect on derailment rates during the eighties.

Given this strong improvement during the eighties, what changed in 1988 and subsequently? Before addressing that issue, it is appropriate to examine the causes and consequences of derailments.

3.3 Trends in Main Track Derailment Rates by Cause

To evaluate the reasons for historical trends in main track derailments, the variations in the causes of those derailments can be examined.

The majority of derailment causes can be grouped into three major categories which collectively account for over 80 per cent of the total:

• track-related: track buckle, broken rail and track geometry problems;
• equipment-related: broken wheels, bearing and axle failures, truck component related failures; and
• train operations-related: operating rule violations, train handling problems, etc.

Figure 2 shows the proportional distribution of derailments by these broad groups from 1980 to present^{Footnote 3}. Although there are some year-to-year changes, it is perhaps remarkable that, during a period when the total number of derailments reduced by more than half, the split by cause category has remained so stable. This suggests that the gains of the last decade were not made by a major leap forward in one area alone, but as a result of many small successes.

This pattern of many factors at play is illustrated in Figure 3 which shows the trends in main track derailments for a number of specific causes^{Footnote 4}. The details of each of these trends can be linked, to some extent, to changes taking place during the period. For example:

• the decline in derailments as a result of track buckling probably relates to improved continuous welded rail (CWR) installation and repair procedures;
• the decline in derailments related to broken rail probably relates to improved rail metallurgy and increased usage of improved rail defect detection technology;
• the decline in derailments as a result of track geometry problems likely relates to increased use of automated measurement of track geometry;
• the general decline in derailments due to roller bearing failure up to 1991 could reflect removal from service of a certain type of roller bearing that was found to be prone to failure; and
• the decline in main track derailments as a result of friction bearing failures is clearly a result of a major reduction of friction bearings in the revenue car fleet.

Over and above these specific impacts, however, it is interesting to look at the generally consistent patterns for these different causes. Most of them exhibit declines followed by some degree of levelling in the trend. The time of the levelling is not the same. Unlike the trend in total derailments where the change from before 1988 to after is quite distinct, such is not the case for individual causes. In other words, the 1988 level-off is probably the result of several different causal factors working in combination.

3.4 Trends in Derailment Severity and Consequences

In considering main track derailment rate trends, consideration has to be given to the severity and consequences of main track derailments. These aspects of the outcome of derailments provide an indication of both the potential and actual risk associated with derailments.

Severity, in this discussion, implies a measure of the scale of the derailment event itself. In other words, is this derailment a multi-car and/or high-energy occurrence with the potential for significant damage? Two measures are used the number of cars derailed, and the speed of the train at the time of derailment. Consequences, or actual results, are indicated by a number of measures, including injuries, damage levels and the release of dangerous commodities.

Figure 4 depicts trends of the average number of cars derailed per reported main track derailment. These data indicate that:

• the average number of cars derailed per reported main track derailment has stayed relatively constant up to the mid-1980s and gradually increased since that time;
• the most significant increase in the number of cars derailed per reported derailment is for track-related derailments post 1990; and
• operations and equipment-related derailments had an increase in the number of cars derailed per derailment since 1985.

Figure 5 provides more detail on the number of cars derailed per derailment. It can be seen that there has been a considerable increase in the number of derailments involving one car in the last two years (note that these data are as reported to the TSB). This reflects an increase (since the latter months of 1992) in the number of one-car low-severity derailments that are reportable under current regulations which were not reported previously.

The number of derailments with more than five cars derailed declined from over 100 in 1980 to about 50 in 1985, and has remained essentially unchanged since then. Meanwhile, the number of derailments with two to five cars derailed declined from 1981 to 1988 and has remained steady thereafter.

Another indicator of the severity of derailments is speed. Recognizing the potential effect of speed on the seriousness of derailment consequences, the CTC adopted measures following the Mississauga derailment to limit the speed of trains carrying dangerous goods in metropolitan areas. Figure 6 indicates that the average speed of trains involved in derailments has remained almost unchanged since 1983 (data on speed were not consistently recorded prior to 1983).

In general, while the rate and number of derailments has declined since 1980, the severity of those derailments has remained about the same.

Along with initiatives to reduce the incidence of derailments since 1980, many steps have been taken to limit the consequences. These included speed restrictions and gateway inspections in metropolitan areas for special dangerous goods trains, tank car improvements, including double-shelf couplers,head shields and thermal protection and bottom fitting protection requirements.

The consequences of main track derailments can be measured by such elements as:

• passenger injury;
• train operating crew injury;
• general public injury;
• release of dangerous goods;
• cost of property damage;
• disruption of rail service; and
• evacuation of local residents.

Figure 7 depicts the trends in fatalities and injuries involving employees, passengers and the general public, as a result of main track derailments. Following a significant decline in injuries from 1982 to 1984, there has been little change in the total number of injuries. A review of the number of fatalities and injuries by category of person shows that:

• no passenger or member of the public has been fatally injured as a result of a main track derailment since 1983;
• two employees were fatally injured in 1992 as a result of a derailment at Nakina, Ontario, but no other employee has been fatally injured as a result of a main track derailment since 1984; and
• one member of the public has been seriously injured as a result of a main track derailment since 1983.

A major safety concern associated with main track derailments is the post-derailment release of dangerous goods. The percentage of derailments involving dangerous goods cars has increased from 15 per cent in 1980 to 35 per cent in 1993. This reflects, in part, changes to the definition of dangerous goods to include substantially more goods. It may also reflect an increase in the proportion of trains carrying dangerous goods. In contrast, Figure 8 shows that the percentage of derailments involving dangerous goods cars which resulted in a release of product has declined from 15 per cent in 1980 to less than 5 per cent today.

Property damage resulting from main track derailments has increased in constant dollars from $150,000 per derailment in 1981 to about $200,000 per derailment in 1991 (data on property damage caused by derailments are not available for 1992 and 1993; however, from other data available, it would appear that there has been little change from 1991). This is because the reporting threshold for derailments was set at a low level of damage ($750) in the early part of the decade. This threshold was later increased to $7,350. Reported derailments in the earlier years with low damage levels would tend to lower the average.

Broken rails lead to serious consequences more frequently than other causes of derailments. Figure 9 presents a comparison of the number of derailments with serious^{Footnote 5} consequences for each cause category. Derailments for which the proximate cause is a broken rail account for more than twice the number of serious consequence derailments than the next most common category (roller bearing failure). In part, this is a result of the relatively high number of broken rail derailments, but is primarily because such derailments are more frequently associated with more serious consequences.

3.5 Personnel Reductions and Main Track Derailment Rates

The number of employees in the railway industry has declined fairly steadily since 1980. Recent highly publicised staff cut-backs by the railways, and the expectation of continued reductions in jobs have led some observers to speculate on the impact of these reductions on railway safety.

If the number of people available to do visual and manual inspection is reduced, but the use of technology to detect problems is improved and applied, overall safety may not be reduced or may even be improved. A comparison of the trends in railway maintenance staff for both equipment and roadway, and corresponding derailment rates for the same period is instructive (Figure 10). From 1981 to 1992, the reductions in staff levels have been significant. However, over that same period, derailment rates did not increase; in fact, they declined significantly. While this observation does not prove that staff reductions never lead to safety declines, it demonstrates that the relationship between the two is not direct and simple.

3.6 International Comparison of Main Track Derailment Rate Trends

Trends in main track derailment rates in Canada are compared with the United States as railway operations in the two countries closely match^{Footnote 6}. To facilitate comparison, Canadian data have been adjusted as far as possible to reflect the same definitions used by the Federal Railroad Administration (FRA) in the U.S. (Figure 11). Because of differences in reporting practices, this adjustment has been done for the years 1984 to 1991 only.

A number of observations can be made from this comparison:

• Canadian railways have consistently operated with a lower main track derailment rate than their U.S. counterparts.
• The relative difference in rates has remained constant throughout this period.
• Both countries have experienced a levelling of the decline in derailment rates taking effect in the 1986 to 1988 period. Because of the different reporting requirements of the FRA, the exact timing of the change in trend is different from that noted earlier. However, the pattern of decline followed by a level trend is apparent. The parallel tracking of the two countries in this respect is notable.

3.7 Analysis of Post-1988 Trends in the Derailment Rate

As noted earlier, a number of factors may have had a bearing on the trend in main track derailments following 1988. In particular:

• The series of initiatives during the eighties may now have "run their course". For example, the removal of friction bearings from the revenue fleet was mainly completed by 1987.
• Some cause groups declined fairly steadily from the early eighties to today, others levelled off before 1988, and others after. This suggests that a fairly complex set of factors may have led, perhaps coincidentally, to the pattern of decline until 1988, and level trend thereafter.
• The parallel pattern of main track derailments in the U.S. and in Canada (Figure 11) suggests that industry-wide changes in North America may have played some part in the levelling of the trend.

However, four other events that altered the Canadian railway environment roughly coincided with the beginning of the period where there was no further reduction in main track derailments. These four events were:

• introduction of the "freedom to move" transportation policy with the attendant proclamation of a new National Transportation Act;
• proclamation of the Railway Safety Act which replaced and modernized the railway safety aspects of the Railway Act;
• introduction of cabooseless train operations; and
• introduction of new railway operating rules.

3.7.1 Change in the National Transportation Act

The new National Transportation Act reflected what has been labelled "transportation deregulation". Although "freedom to move" was an economic deregulatory policy, transportation safety was not to suffer. Although there were arguments made by industry that there should be a parallel "technical" deregulation (i.e. cease or reduce government intervention that involves expense to be incurred by industry as a result of government direction), this was not part of the transportation deregulation policy initiative.

There are differences of opinion amongst those in the railway industry as to the impact of the new legislation on such things as the profitability of the railways. However, there appears to be unanimity, including both major railways and TC, that the introduction of this new legislation has not resulted in reduced railway safety and cannot be linked to the levelling-off of main track derailment rates

post-1988.

Some have hypothesised that firms under reducing profitability may reduce expenditures which could eventually reduce safety. However, assessing the impact of economic changes on safety is extremely difficult if not impossible.

Attempts to identify such a pattern in the U.S. airline industry following the 1978 deregulation have found little evidence to either confirm or refute the hypothesis. The same seems to be true in this case. In fact it may be even more difficult. There were many changes already unfolding in the railway industry long before 1988. The impact of the Staggers Act in the U.S. in the late seventies triggered industry-wide changes including cost-reduction activities. Similar changes were under way in Canada. For example, it has already been noted that personnel reductions were taking place from the early eighties and cannot be directly linked to changes in main track derailments.

Therefore, in considering the evidence presented by the investigation of individual occurrences, it is very difficult to identify the possible effect of economic changes on safety. In sum, the TSB occurrence record offers no direct evidence that economic deregulation has adversely affected main track derailment rates. However, one senior rail company official noted that:

...the effects of deregulation have hampered the ability of railways to generate or protect revenue growth due to heavy pricing and competitive pressures under the Freedom to Move philosophy. This situation, along with the severe economic downturn, has had an impact on the rail industry's ability to maintain capital programs for new rail installation and equipment acquisitions.

3.7.2 Change in the Railway Act and Introduction of the Railway Safety Act

The Railway Safety Act (introduced shortly after the new National Transportation Act) involved the modernization of the technical and safety provisions of the Railway Act. Railway safety regulatory responsibilities and staff from the Canadian Transport Commission/National Transportation Agency were transferred to Transport Canada.

In discussing the impact of changes related to the Railway Safety Act one railway official pointed to a "...change in philosophy and relationship between the regulator and the regulated...which simply put, places the emphasis for running a safe operation on the railways' themselves...."

Both the major railways and TC, agree that there is no evidence which supports a proposition that the changes brought about by the Railway Safety Act resulted in a decrease in railway safety, or a levelling-off of the main track derailment rates post-1988. Another railway official noted that it cannot "be demonstrated that somehow relaxed regulations have contributed to stopping the main line derailment rate decline".

Again, the occurrence record offers little illumination on this question. These factors, including the introduction of the Railway Safety Act, took place at about the same time. Therefore, the time series of occurrences cannot clarify the relative impact of any one factor.

3.7.3 Introduction of Cabooseless Train Operations

Both major railways and TC agree that there is no evidence to indicate that cabooseless train operations are less safe than operations with an occupied caboose. From the occurrence record, the TSB has not identified any evidence that cabooseless trains are more likely to be involved in derailments.

3.7.4 Introduction of new Railway Operating Rules

Both major railways and TC agree that there is no evidence to suggest that the introduction of the current Canadian Rail Operating Rules has any relationship with the levelling-off of the main track derailment rates post-1988.

3.7.5 Summary of Post-1988 Analysis

The reasons for the levelling-off of the decline in main track derailments are probably many. Changes in technology and methods of operation affected some types of derailments both here and in the U.S. The appearance of a single point in time when the trend changed may, in fact, be coincidental. As to the possible impact of legislative and regulatory change, there is no direct evidence from the occurrence record to confirm or refute such a possibility.

Nevertheless, it is undeniable that the derailment rate has remained virtually constant from 1988 through to today. It is therefore appropriate to turn to the major reasons for recent derailments, and the safety actions aimed at removing them.

3.8 Summary

The key observations to be made from this review of historical trends are:

• Main track derailment rates declined significantly from the early eighties until 1988 and subsequently levelled off.
• The consequences of main track derailments, such as injuries, property damage and dangerous goods release, have, in general, declined in magnitude since 1980.
• Trends in the individual types of derailments suggest a complex set of factors rather than a single factor.
• There is no obvious link between derailment rate trends and trends in the numbers of maintenance staff employed by the railways.
• Canadian derailment rates are consistently lower than U.S. experience.
• The trend in derailments since 1988 reflects the impact of many different factors. There is no direct evidence from the occurrence record to support the identification of legislative change leading to the change in trend.

This chapter examines recent main track derailments to identify the safety issues which underlie those derailments, the actions that have been and are being taken to correct the safety deficiencies, and the effectiveness of those actions. For the purposes of this examination, the main focus is on derailments since 1991.

The discussion begins by looking at the causes of recent derailments. The safety issues are then considered within the categories of track-, equipment- and operations-related causes.

4.1 Causes of Recent Derailments

The overall distribution of proximate causes of derailments since 1991 were presented earlier in Section 3.3 and Figures 2 and 3. As noted in that discussion, while there are year-to-year differences in the proportion of derailments linked to cause categories, there is little to distinguish recent years.

When the factors involved in recent derailments are examined in more detail, the recurrent themes which emerge are:

• broken rails with unidentified internal defects;
• inadequate track geometry maintenance;
• roller bearing failures (including a significant proportion where the HBD system failed to provide sufficient advance warning of failure);
• wheel failures (primarily involving straight-plate wheels); and
• operational violations (particularly involving switch handling).

These themes have been the subject of significant efforts by the industry over a number of years. However, they remain persistent elements in the derailment picture.

4.2 Track-related Safety Issues

Derailments related to track can be grouped into two main categories: rail failures and track maintenance problems. Rail failures can be further broken down into rail defect identification and rail replacement. These three subjects are discussed below.

As noted in the discussion of derailment consequences (Section 3.4), broken rail derailments more frequently have serious consequences than any other type. In terms of risk management, the value of better prevention of the underlying factors in broken rail derailments is substantial.

4.2.1 Internal Rail Defect Detection

A relatively frequent reason for broken rail derailments is the inability of internal rail defect detection equipment and practices to discover all internal rail defects ^{Footnote 7}. The problem is not new. The NTA issued five related recommendations in 1989 and one in 1990. The TSB made one recommendation in 1992 and five recommendations in 1993, and issued one safety advisory in 1993 concerning rail defect detection.

For example, in 1989, the NTA concluded that rail of certain manufacture had a higher propensity for internal rail defects. The NTA therefore recommended that TC collect information regarding the frequency of defects and failures specific to the manufacturer and the date of manufacture to verify this conclusion.

TC responded by indicating that the Canadian National Railway Company (CN) was providing data on defects, and that frequency of testing on the subject subdivision had been increased. TC stated that it was examining the issue of rail testing in conjunction with the establishment of track safety standards for Canadian railways, and incorporated a complete section on the subject in the 1992 Track Safety Rules.

In 1993, the TSB made a number of recommendations with respect to the equipment and procedures for rail testing. TC was urged to reassess the adequacy of the requirements for testing, taking into account the age of the rail and the nature of the traffic. Research to improve the effectiveness of testing methods was recommended. In addition, the TSB noted some aspects of testing which required more attention, specifically the ability of rail testing to identify defects at crossings, on curved track, and to identify vertical split head defects.

The responses to these TSB recommendations by TC indicated that TC inspectors had reviewed rail testing procedures and inspection programs and had determined that they are consistent with other U.S. and industry standards. More specific information regarding TC's plan to address the safety deficiencies identified by the TSB was not indicated.

More recently, CN indicated to the TSB that increased frequency of ultrasonic rail testing is underway. CN also indicated its intention to improve ultrasonic detection by improving the operator-qualification procedures and training. A new system of computerized warnings to the rail testing vehicle operator is intended to use pattern recognition as a method of reducing the number of undetected defects.

Between 1989 and 1993, both the NTA and the TSB have identified a number of safety deficiencies regarding the detection of internal rail defects. Even though safety actions have been recommended to TC and the railway industry, and some of them have been implemented, main track derailments resulting from internal rail defects are still occurring.

4.2.2 Rail Replacement

The prevention of track-related main track derailments, in part, depends on the rail replacement practices. The railways strive to maximize the length of time rail is in use, consistent with minimizing the number of rail failures which can lead to derailments. In the past, rail replacement was primarily based on manual checking and local decision making. Now, mechanised measurements, computer tracking and centralized management of the rail inventory are the norm.

In follow-up to two recent main track derailments (04 February 1993 on CN's Caramat Subdivision, and 17 February 1993 on the Canadian Pacific Limited (CP) Nelson Subdivision), consideration was given to rail testing, inspection, wear measurement, and replacement practices.

The main observations with respect to CN's rail replacement practices were:

• CN's standards and practices are consistent with good North American standards.
• CN monitors rail wear primarily based on test car data using an optical measurement system. This information is combined with rail failure rates to determine if replacement is required.
• Rail wear limits were established in 1981 (Standard Practice Circular 3200), and have not been revised nor are there plans to change them. Once wear limits are reached, rail is either relocated or replaced. There is no provision for limiting speed over rail which has passed the wear limit.
• Rail failure rates are based on the sum of the number of defects identified during rail testing and in-service defects found during inspections or as a result of actual failure of the rail. For example, for the five years 1988 to 1992, there were 270 rail failures on the Caramat Subdivision (i.e. 0.5 failures per mile per year), 201 of which were detected by rail testing and 69 were in-service failures. CN uses a criterion of 2.75 failures per track-mile per year on main track for replacement.
• Rail defect testing is not totally effective. As a result of a combination of the limitations of the technology in certain circumstances (e.g. curves) and the interaction with human operators, some defects are not detected. In the specific occurrence on the Caramat Subdivision, the rail which failed was tested eight days prior to the derailment and, although there were recorded anomalies, the rail was not identified as being defective.
• There are no formal qualifications required for operators of internal defect testing equipment. CN requires 500 hours' experience for an operator to become a chief operator.
• Work force numbers for the section of track related to this occurrence have not changed since 1989 when the work force was mechanised, allowing some reduction of staff.
• Track inspections had been carried out as required by TC.

The main observations with respect to CP identified many similarities with CN's practices. The significant differences are:

• CP's rail wear limits are currently based on joint bar clearances. CP has conducted extensive testing and research in conjunction with external research organisations to better identify rail wear limits related to yield strength and carrying capacity. This work identified that wear limits can be extended beyond those dictated by joint bar clearances. It is expected that new limits will be adopted by CP shortly except in jointed rail territories or in curves experiencing specific defects.
• In the case of the rail failure which led to the derailment on the Nelson Subdivision, an optical wear measuring technology was used, which did not correctly identify that the rail was excessively worn. CP has since replaced that technology with a more accurate system using lasers.
• In July 1993, track maintenance forces were reduced by 25 per cent as a result of increased mechanisation. Teams are now equipped with Hi-rail equipment and power tools which allow smaller crews to inspect and maintain the same amount of track.
• CP uses a criterion of 2 rail failures (excluding bolt hole failures) per mile per year for replacement in main track with tonnages exceeding three million gross tons per year.

The rail replacement approach used by the railways is critically dependent on two elements: the ability of the technology used to identify both rail wear and internal defects reliably; and the trigger levels built into the rail-replacement algorithms which essentially determine the level of risk of failure. As noted above, there are some areas of weakness in the defect-detecting technology and the interface with the human operator. In terms of the trigger levels for rail replacement, there is no indication that these have been changed in response to economic pressures. There are, however, differences between the two major railways. These differences do not necessarily imply differences in failure risk since many other factors are at play.

4.2.3 Track Maintenance

Main track derailments related to track maintenance are usually associated with track geometry that is inadequate to handle the passage of a train. Three recent main track derailments illustrate the importance of sound track maintenance.

On 25 March 1993, a CN freight train derailed 12 cars within the city of Moncton, New Brunswick. The derailment was caused by soft grade and cross-level variations. The track had not been recently tested by the track geometry car and poor track conditions were apparently not identified during track inspections.

On 09 April 1993, a VIA Rail Inc. (VIA) passenger train derailed two locomotives and one baggage car near Rapide Blanc, Quebec. The derailment was caused by a washout resulting from a blocked culvert which was unable to handle heavy run-off water.

On 13 May 1993, a CN freight train derailed six cars near Park Gate, Alberta. This was partly because concrete ties, damaged from a previous derailment, were unable to maintain correct gauge. A third derailment has since occurred in the same area.

Washouts and slides are of particular concern because they can cause deterioration of the roadbed and track in a very short time. For example, in July 1992, a CN freight train derailed near Nakina, Ontario. The subgrade underneath the track at the derailment site had collapsed into an adjoining pond prior to the arrival of the train. The locomotive derailed into the pond when it travelled onto the portion of the track that was suspended in mid-air. Two crew members lost their lives and the third sustained serious injuries. The track bed failure was caused by a sudden draw down of water which resulted from a failed beaver dam.

As a result of the Nakina investigation, the TSB made a number of recommendations to TC relating to the potential for similar rapid draw downs of water elsewhere in Canada. In response, TC indicated that CN was undertaking aerial surveys of beaver dam locations along the railway right-of-way and employing trappers to remove beavers at strategic points. CN has been requested to carry out a geotechnical analysis to determine if there are other locations which are susceptible to collapse under similar conditions. TC has communicated the safety issue to other railway companies. However, the focus of the response appears to be on simply communicating the existence of a problem and there is no indication of specific action by TC to institute a program to address the systemic issue as proposed in the recommendation.

4.2.4 Track-related Main Track Derailments Summary

The safety issues associated with track-related derailments are most critical in respect of rail failure occurrences. A review of rail replacement practices indicates that CN and CP are using similar approaches. They use data on rail wear and failure rates to maximize the use of rail while limiting the potential for catastrophic failure. At the same time, railways are shifting the emphasis from labour-intensive manual inspection approaches to more mechanised methods.

There is no indication from the occurrence data that these approaches are leading to a deterioration in safety performance. However, a rail replacement approach which is based in large measure on rail internal defect detection highlights the problems of undetected failures. In particular, the TSB has expressed concerns with the performance of the technology; concerns which are echoed in the industry itself. Improved displays for test car operators may help to alleviate the problem.

Inadequate roadbed maintenance continues to lead to main track derailments. In the main, this is related to seasonal factors. However, the investigation of a 1992 occurrence near Nakina, Ontario, raised systemic issues associated with potential for subsidence under main track. While action has been taken to address the proximate cause of that occurrence, the more systemic issues have yet to be addressed.

4.3 Equipment-related Safety Issues

While derailments related to equipment deficiencies involve a number of different factors, three issues are of continuing concern: wheel failures, roller bearing failures, and truck instability as a result of worn truck components.

4.3.1 Wheel Failures

Straight-plate wheels are known to be more susceptible to failure than are curved-plate wheels. In recognition of their potential for failure, the Association of American Railroads (AAR) established a process by which straight-plate wheels are to be phased out of service. Nevertheless, there have been at least five derailments in Canada resulting from straight-plate wheel failures on main track in the past 18 months.

In 1992, the TSB was concerned that the issue of straight-plate wheels was not being addressed quickly enough to ensure the safety of the transportation system. The Board recommended that straight-plate wheels be phased out in an accelerated manner. It was recommended that a single rim thickness condemning criterion be established which would accelerate the elimination of the straight-plate wheel.

TC replied that it was working with the Railway Association of Canada, the Federal Railroad Association and the AAR to reassess the issue of accelerating the phase-out of straight-plate wheels. The TSB also recommended that cars with straight-plate wheels which do not meet the Canadian condemning criterion be returned to the owner by the shortest feasible route.

The TSB was advised that, if the outcome of TC's review revealed that straight-plate wheels were not being eliminated in an acceptable time, then action would be taken under the Railway Safety Act. On 03 November 1993, TC advised that the Railway Freight Car Inspection and Safety Rules to come into effect in 1994 prohibit placing or continuing in service any car equipped with a straight-plate wheel which fails to meet new tougher standards. In addition, any foreign-owned car with condemned straight-plate wheels is to be removed at the first available maintenance facility.

The industry has reacted strongly to these measures, arguing that the costs involved will far outweigh the safety benefits, and that the program already in place to replace straight-plate wheels is effective and will quickly reduce the problem to minor levels.

Although all parties concerned are addressing the issue, derailments on main track continue to occur as a result of the continued use of straight-plate wheels.

4.3.2 Roller Bearing Failures

As a result of the internationally publicized 1979 Mississauga derailment, there was considerable emphasis on the replacement of friction bearings with roller bearings on the Canadian railway fleet and on increasing the number of HBDs. As discussed in Chapter 3, the effect of this change was a reduced main track derailment rate. There are now very few friction bearings causing main track derailments in Canada because there are relatively few in the revenue car fleet.

However, at the time of the CTC hearings following Mississauga, the CTC expressed concern that "... at least a possibility exists that a hot box detector program based on spacings of 25 to 30 miles may be inadequate to detect overheating of roller bearings." This concern remains valid today as roller bearings, once in distress, can overheat and fail in much less than the standard HBD spacing^{Footnote 8}.

HBDs measure the absolute bearing temperature, the relative temperature differential between the bearings on either end of an axle, and the average temperature on both sides of a car and the train. A differential above a specified value requires that the train be stopped and the suspect bearings be inspected. If the inspection reveals a hot bearing, the car is set out for repair and the train continues. This detection system and its spacing were designed for detecting failing friction bearings which take a relatively long distance to fail once they become overheated.

In 1992, the TSB recommended that the spacing of HBD devices be reviewed, and that the criteria for the temperature differential be reduced. The TSB also recommended that the development of new technologies be supported to improve the sensitivity in detecting roller bearings in distress. It was further recommended that more use be made of the information given by the changes in temperature differentials over time. In other words, rather than relying on one simple criterion, decisions should be based on all available information on bearing performance.

TC responded that, in conjunction with the AAR, it was then undertaking a review of the criteria for the HBD spacing and the temperature differential for stopping trains. In addition, CN now monitors data from the Journal Advance Warning System (JAWS) to identify when the same bearing on a car records a high temperature differential at more than one HBD, where that differential was insufficient to require the car to be inspected immediately. Such suspect cars are then left at a convenient yard for bearing inspection.

Safety deficiencies regarding HBDs have been identified by both the NTA and the TSB. However, in 1993, there were at least six main track derailments in Canada caused by burnt-off journals which were not previously detected by HBDs prior to the point of derailment.

The cost of adding a sufficient number of HBDs to the system to ensure that no failing roller bearing would go undetected would require a significant investment. There are, however, ways to detect potential roller bearing failures other than by heat differentials. Research and development including field testing has been underway for about five years on acoustic wayside detectors which can evaluate the condition of roller bearings in distress that are not to the point of overheating. Similarly, testing has been carried out on what is referred to as a "Smart Bolt" which detects overheating and radios the defect notification to the locomotive cab. Neither system is ready for full-scale installation; in fact, recent research casts doubt of the potential of the acoustic system, and some industry representatives see little benefit in further pursuing the option.

The effectiveness of HBD systems has also been compromised by human error (e.g. HBD remote operator error, improper identification of suspected overheated bearings by train crew). Following a 1989 occurrence in which a train continued on after two detectors revealed high readings to the HBD operator, the NTA recommended to TC that CN be required to improve its response management with respect to HBDs. In 1993, the following occurrences reveal similar human performance problems. Specifically, they indicate failure in communication, interpretation, and follow-up action.

On 07 February 1993, a CP freight train derailed six cars near Tempest, Alberta, as a result of a burnt-off journal on the 16th car behind the engine. A HBD (with radio notification capabilities) registered an alarm which would have prevented the derailment; however, the detector's broadcasting ability had been disabled by a technician who had mistakenly left the audio transmit switch in the off position. This error was compounded by the fact that 17 trains had passed the detector in the interim and only three of those crews had reported an audio failure.

On 15 February 1993, a CP freight train derailed one car near White River, Ontario. The car was an empty boxcar with severely skidded wheels. The train had been stopped by a prior hot box and hot wheel detector because of sticking brakes on that car. A crew member cut out the air brakes on the car but failed to make a pull-by inspection as required by company instructions; thus, he did not realize that the wheels were not turning. The railway has issued instructions requiring that in response to hot wheel alarms the suspect car wheels must be seen to move properly before the train proceeds.

On 15 April 1993, a CN freight train derailed one car near International Boundary, Manitoba. The train was stopped after a HBD indication but the burnt-off journal dropped. An alarm had been received from the previous HBD (49 miles before) but inspection by the crew did not identify a hot bearing.

On 03 August 1990, a train derailed as a result of a failed roller bearing causing axle burn-off. The hot box and dragging equipment detector indicated an abnormally high reading on the 43rd car from the head-end of the train. The train was operating with a caboose, and the conductor calculated the position of the car from the tail-end by using the train journal. The conductor inspected the calculated car plus three cars on either side as per company instructions. The inspection revealed no abnormalities. Before the train left Calder Yard, four cars had been added to the train consist but were not indicated on the conductor's train journal. Therefore, by counting from the tail-end, the wrong cars were inspected.

Such human performance failures are fundamentally similar to human factors in many other aspects of the rail transportation system. The root causes of these failures are not well understood in this context, although there is a growing body of work on this subject in many other fields. Since human factors play a large role in operations-related derailments, this question will be addressed more fully in Section 4.4. It is, however, important to note that human factors can lead to accidents even when automatic detection systems are in use and working properly.

4.3.3 Truck Component Wear

Equipment maintenance departments perform various types of car inspections to ensure that there is no excessive wear on any of the components of the trucks on a car. Each component has a limit specified by the AAR, determining the maximum amount of acceptable wear. Car inspectors measure each component separately to determine if it has exceeded the maximum limit for that component.

Until recently, combination truck component wear has not been addressed. In some instances, even though the wear of individual components does not exceed the limit, the condition of truck components in combination can result in main track derailments.

In 1990, the NTA recommended that TC improve its ongoing inspection program to ensure that rail cars are not placed in service or continue in service with truck components collectively worn to the point where the ride control of the trucks is adversely affected. The NTA also recommended that the Railway Freight Car Minimum Inspection and Safety Standards be amended to outline precisely the maximum allowable wear on truck suspension components. However, recently approved Railway Freight Car Inspection and Safety Rules do not address this matter.

In March 1992, the TSB made recommendations on truck component wear specifically pertaining to leased tank cars. The 01 January 1992 AAR standard requires tank car owners to ensure that truck component wear does not exceed specific criteria when a car is released to interchange service. The TSB was concerned that this standard is not applicable to cars that already exceed the condemning limit and are currently in interchange service as well as cars whose truck components deteriorate beyond the condemning limits while in service.

The TSB recommended that TC prescribe condemning limits for combination truck component wear for all leased cars in service. This would ensure that cars with various worn components be taken out of service even though the wear on individual parts might not exceed the maximum for that component. In its response, TC noted that wear limits were being reviewed, but did not explicitly indicate any proposed action with respect to the issue of combination wear. Recent communication from TC indicates that consultation is underway with the industry on the combination wear question.

Car maintenance work is performed usually only in a car's home yard. For example, if the car is not owned by the company whose yard it is in, there is no obligation for the company to make any repairs to that car. However, if a component is worn beyond the acceptable limit, the component is replaced and the costs for the repair is billed to the company which owns the car. One of the many reasons for a company not servicing a car is that the costs for the repairs are not always recoverable. Therefore, the TSB made a recommendation to TC to effect cost recovery by carrying railways which take corrective action to replace components exceeding specified condemning limits for combination truck component wear.

In response, TC stated that cost recovery is "the responsibility of the railway industry and is not a factor related to safe railway operation". It is considered unreasonable to expect Canadian railways to effect necessary maintenance on foreign cars without some means of reducing their financial risk. Short of refusing to accept cars beyond condemnable combination wear limits, the carriers are likely to postpone the maintenance, thus increasing the risk of equipment failure. Cost recovery would ensure that all necessary car repairs are identified and completed before returning the potentially defective car to the transportation system for movement.

The TSB has also issued a safety advisory indicating a concern regarding the poor quality of car inspections. There had been derailments which were caused by side bearing clearances exceeding the limits set by the regulations. It was suggested that TC confirm the adequacy of training and supervision of car inspectors.

In spite of all the safety action proposed regarding truck component wear, there have been nine reported main track derailments attributable to this condition since the TSB recommendations of March 1992 .

4.3.4 Equipment-related Main Track Derailments Summary

TC appears to be taking strong action to address the straight-plate wheel issue. However, the railways have expressed concern that the safety benefits will be small compared to the cost to industry of implementing the new standards.

The roller bearing failure problem, which is related to the current HBD spacing criteria, is still unresolved. Current technology may be too expensive to eliminate all roller bearing failures. However, research into other methods of detecting potential roller bearing failure is underway and may identify feasible alternatives.

Problems of truck component wear and inspection have not been adequately resolved, and continue to cause main track derailments.

4.4 Operations-related Safety Issues

The major concern in operations-related derailments is that of human performance, especially of those who either operate the locomotive or are responsible for actions which could result in a derailment (e.g. improper switch handling).

Virtually all operations-related derailments in which the actions or inactions of operating crews were a contributing factor involved violation of at least one of the Canadian Rail Operating Rules (CROR). In that the CROR are specifically designed to govern safe rail operations, such violations invariably constitute unsafe behaviour by some member of the operating crew.

While there is a considerable body of knowledge concerning the factors adversely affecting expected human performance, the TSB has thus far accumulated little specific human performance data concerning operations-related main track derailments. The TSB recognizes the need to investigate the underlying human factors in rail occurrences. Similarly, the railway industry generally appreciates the importance of developing a better understanding of those human behaviours which can create or exacerbate an accident situation. To this end, increased emphasis is being placed on the importance of such potential contributing factors as fitness for duty, supervision and training, work scheduling, team work, etc.

As the industry's understanding of the impact of such factors on safe operations increases, significant improvement in railway safety generally can be expected. There should be a concomitant reduction in main track derailments attributable to operations-related safety issues.

From a study of both long-term trends and recent main track derailments, this study has concluded that:

1. Canadian main track derailments have declined by almost a factor of three between 1980 and 1988. Adjusted for changes in reporting requirements, the data indicate that the annual number of derailments has remained basically unchanged since 1988.
2. The reason for the 1980 to 1988 decline in derailments is most probably a combined result of many different factors. These include:
   • improved installation and repair procedures for welded rail; increased use of automatic rail defect detection and track geometry measurement technology;
   • increasing proportion of the car fleet equipped with roller bearings; gradual elimination of straight-plate wheels; improved marshalling requirements;
   • increased number of HBDs; and
   • rigorous government safety regulatory enforcement and inspection programs.
3. The reasons for the levelling-off of the derailment trend post-1988 are probably many. Changes in technology and methods of operation affected some types of derailments both here and in the U.S., leading to a very similar pattern of decline followed by level trend in both countries. The appearance of a single point in time when the trend changed may, in fact, be coincidental. As to the possible impact of legislative and regulatory change, there is no direct evidence from the occurrence record to confirm or refute such a possibility.
4. No passenger or member of the general public has been fatally injured because of a main track derailment in Canada since 1983. Apart from two operating employees being fatally injured because of a 1992 derailment caused by a beaver dam failure, no employee has been fatally injured because of a main track derailment since 1984. Considering the annual toll of around 50 to 70 fatalities that occur at railway grade crossings, the hazards to persons posed by train derailments have proven to be relatively low. However, the hazards associated with derailments of trains carrying dangerous goods remain potentially serious. Actual releases of dangerous goods following derailments have declined from the early eighties, and are at a low level.
5. The rate of main track derailments in Canada is and consistently has been less than in the United States.
6. Persistent themes in the factors underlying many of the main track derailments since 1991 include: broken rail with previously undetected internal defects; inadequate track geometry maintenance; undetected roller bearing failures; worn truck components; and straight-plate wheel failures.
7. Although rail defect detection technology contributed significantly to reductions in broken rail derailments in the eighties, failure to detect internal defects has contributed to recent main track derailments attributed to rail failure.
8. Track geometry problems continue to cause main track derailments; some of these problems could be ameliorated by improved preventative maintenance.
9. Straight-plate wheel failures continue to lead to derailments. Action has been initiated by Transport Canada and the railway industry to eliminate this wheel design from the revenue car fleet; however, there is some debate as to the rate at which wheel replacement should proceed.
10. Derailments caused by undetected roller bearing failures continue and, while some improvements are possible in making the best use of available data, there appears to be no immediate solution without a major capital investment in HBDs.
11. Human performance is a major concern in both operations-related derailments, and in occurrences where the output of automatic detection systems was inappropriately actioned. At present, the TSB has little human performance data concerning these situations. The TSB and, increasingly the railway industry, is placing particular emphasis on the importance of developing a better understanding of those human behaviours which can create or exacerbate an accident situation.

This study has identified a plateau in the trend in main track derailments, following a number of years of significant declines. If further advances are to be made, continuing efforts will be required to eliminate safety deficiencies identified by the NTA, the TSB and TC. Specifically, outstanding recommendations in the following areas require full implementation:

• internal rail defect detection;
• straight-plate wheel replacement;
• HBD spacing;
• potential for subsidence of roadbed under main track; and
• combination wear in truck components, and the related issue of cost recovery for maintenance on foreign cars.

Finally, the TSB, TC and the railway industry recognize the importance of developing a better understanding of those human behaviours which can create or exacerbate an accident situation. Increased emphasis needs to be placed on the identification of safety deficiencies in such areas as fitness for duty, supervision and training, work scheduling, and team work.

POSTSCRIPT

While the focus of this study is on long-term trends and persistent safety issues, a very recent increase in main track derailments deserves mention. In the first three months of 1994, the number of reported main track derailments in Canada increased to 55, compared to 42 for the same period in 1993. This increase is especially notable given the fact that in February 1993 an exceptionally high number of derailments were reported.

Such short-term changes, especially involving relatively few occurrences (statistically speaking at least) should not be over-interpreted. However, while the causes of these derailments will not be known for some time, this recent increase does serve to underline the fact that the trend in main track derailments is now, at best, flat.

Appendix A Adjustments To Data

Because of changes in the reporting requirements for main track derailments during the time period covered by this analysis, an adjusted data series has been calculated. This adjusted series can then allow the analysis of changes over time unaffected by changing definitions.

Three main changes to reporting requirements during the period from 1980 to date are expected to materially have an impact on the number of main track derailments recorded. These are:

• Under the CTC and the NTA, main track derailments were reported if the estimated cost of the damage caused exceeded a specified dollar value. (In addition, all derailments involving dangerous commodities and/or where serious or fatal injuries resulted were reportable irrespective of damage caused). This value was set at $750 up to 01 November 1987. It was then adjusted to $7,000, and then subsequently to $7,350 on 01 January 1988. This value remained in force until the NTA regulations were superseded by the TSB Regulations in 1992.
• During 1991, a number of changes were made to the definition of dangerous commodities, for the purpose of reporting of transportation occurrences. These changes added additional commodities to the existing list. This would have resulted in additional derailments being reported following the changes.
• In July 1992, the TSB issued new reporting regulations for all occurrences reported under the CTAISB Act. Under these regulations, the dollar damage threshold was removed and instead a requirement to report if "...rolling stock sustains damage that affects its safe operation" was added. Other conditions related to injuries and dangerous commodities remain the same.

In general terms, the impact of these changes was to reduce reported derailments after November 1987, increase them somewhat after 1991, and then to further increase the number reported as the TSB Regulations were fully implemented in the latter half of 1992.

The adjustment of the data series proceeded in three steps. First, derailments prior to 01 January 1988 were reduced to reflect a damage reporting threshold of $7,350 adjusted for inflation. In other words, derailments prior to 1988 were taken out of the series if the damage caused fell below the threshold in constant dollars. Second, derailments from 1991 on which were only reported because of dangerous commodity involvement (i.e. did not involve injuries or damage above the threshold) were removed if they involved commodities not previously reported. Third, using data from prior to 1992, a regression equation was derived which estimated the dollar value of damage caused based on train speed, cars derailed, consist, and whether the derailment was caused by rail failure. This regression equation was then used to identify occurrences after July 1992 which would have probably not met the damage reporting threshold used prior to that date.

It should be stressed that this adjustment was only performed on the full data series. Attempts to apply such a method to individual subsets of derailments would be subject to potentially high errors.