Rail Safety Advisory 617-06/20
Place du Centre
200 Promenade du Portage, 4th floor
Gatineau QC K1A 1K8
11 September 2020
Letter addressed to:
Director General, Rail Safety
14th floor, Enterprise Building
427 Laurier Avenue
Ottawa, ON K1A 0N5
Rail Safety Advisory letter 617-06/20
Managing in-train forces
On 28 June 2019, westbound Canadian National Railway (CN) freight train M38331-27 (the train) was a key train operating on the CN Strathroy Subdivision, which is a key route. The train consisted of 2 head-end locomotives, 1 mid-train distributed power (DP) remote locomotive (situated between line 81Footnote 1 and line 82) and a total of 140 cars, including 125 loaded cars, 12 empty cars, and 3 residue cars. The train was 9541 feet long and weighed 15 674 tons. It contained a block of lighter, loaded, autorack cars equipped with long-travel end-of-car cushioning devices (EOCCDs) located ahead of and behind the DP remote locomotive (line 70 to line 97) followed by primarily heavily loaded cars on the tail-end (line 98 to line 140).
At about 0402 Eastern Daylight Time (EDT), the train, operated by a 3-person crew (i.e., locomotive engineer, conductor and brakeman) departed from Sarnia, Ontario, en route to Port Huron, Michigan. The train proceeded through CN’s Paul M. Tellier Tunnel (the tunnel) under the St. Clair River, which connects Sarnia to Port Huron and traverses the international border between Canada and the United States at Mile 60.63 of the CN Strathroy Subdivision. The track alignment on the approach and through the tunnel is mostly tangent Class 4 single main track. The track has an approximate 2.00% descending grade from Mile 59.32 to near the international border (Mile 60.63), where it levels out slightly, followed by an ascending grade of up to 2.10% to just past the west end of the Strathroy Subdivision (Mile 61.7) in Port Huron.
At about 0420 EDT, while travelling at 44 mph in the tunnel, a train-initiated emergency brake application occurred while the head-end locomotive was at Mile 61.19. Subsequent examination determined that a total of 46 rolling stock between lines 51 and 98 had derailed and came to rest on both sides of the international border inside of the tunnel. The derailed cars included dangerous goods (DG) tank car UTLX 95205 (line 68), which was loaded with sulphuric acid (UN 1830, Class 8, PG II). During the derailment, this tank car was breached and released most of its load in the tunnel (about 12 000 U.S. gallons).
The head end of the train came to rest outside the tunnel at Mile 61.46, west of the tunnel portal in Port Huron. The trailing end of the 51st car, flat-bottomed gondola DJJX 19371 loaded with scrap steel, and all wheels of the 52nd car, flat-bottomed gondola DJTX 30049 loaded with scrap steel, were derailed. There was no visible impact damage on the trailing end of DJTX 30049, which had come to rest at Mile 60.85. Behind (east of) DJTX 30049, the south rail had rolled and there was a separation of 696 feet leading up to the leading A-end of the 53rd car, bathtub gondola DJJX 30478 loaded with scrap steel, at Mile 60.72 (Figure 1). The adjacent tunnel reinforcement walls showed no visible signs of impact and there were no obvious track defects.
The A-end of DJJX 30478 (line 53) was extensively damaged and appeared to have collapsed. The truck had skewed diagonally and the south rail had rolled to the south side of the tunnel. DJJX 30478 was likely the first car to derail as the A-end of the car sustained structural failure when subjected to elevated in-train buff (compressive) forces while travelling within the tunnel. At the time of the accident, DJJX 30478 also exhibited a number of pre-existing defects that may have contributed to its reduced structural integrity.Footnote 2
In order to determine the magnitude of the longitudinal buff forces acting on the leading A-end of the car, the TSB Laboratory conducted a series of train dynamics simulations using the Train Energy Dynamic Simulation (TEDS) software. The baseline TEDS simulation estimated the forces acting on the train as it traversed the track profile through the tunnel. The train handling script was created from the LER train handling commands in time. The train handling commands were synchronously applied to the head-end locomotive consist and the DP remote locomotive while the train make-up and tonnage profile was as listed on the train journal (Figure 2).
In addition to the baseline simulation, 2 additional simulations were performed to assess the in-train forces associated with alternate train configuration options. The 3 simulations provided the following results:
- For the occurrence train, the predicted longitudinal buff force occurring at the leading A-end of car DJJX 30478 (line 53) immediately prior to the train-initiated emergency brake application, which was considered to be the derailment and separation moment as the end of the car collapsed, was approximately 388 kips.Footnote 3
- Moving the DP remote locomotive from its actual position in the train to line 117, in accordance with CN marshalling requirements, had minimal effect on the simulated maximum in-train buff force at the leading A-end of car DJJX 30478 (line 53).
- The autorack cars equipped EOCCDs were then remarshalled to the tail end of the train from their original location ahead of and behind the DP remote locomotive. Using the same track profile and with train handling commands and train make-up otherwise remaining consistent with the LER and train consist, the predicted maximum in-train buff force at the leading A-end of car DJJX 30478 (line 53) would be significantly reduced to about 235 kips.
Train operations have changed significantly in recent years. The widespread use of DP remote locomotives in train operations has facilitated the operation of longer and heavier trains. Before the mid-1990s, an average mixed merchandise train in main-track service was about 5000 feet long and weighed 6000 to 7000 tons. In contrast, trains in today’s operating environment are often over 12 000 feet long and weigh sometimes as much as 18 000 tons or more.
With the increase in average train length and weight, there have been increases in the associated in-train forces. Longer trains in particular can generate significant longitudinal draft/buff forces due to the slack action of the train. To minimize these draft/buff forces requires more critical management of freight car placement (train marshalling) within the trains in order to reduce in-train forces and maintain safe operations. To accommodate this change in operations, Canadian railways have implemented various train marshalling strategies.
In the early 2000s, Canadian Pacific Railway (CP) developed a train marshalling system (TrAM) that requires the strategic placement of cars and DP remote locomotives in a train in order to maintain in-train forces below 200 kips of steady-state buff or draft force. The system also considers and assists with the optimal placement of DP remote locomotives and cars equipped with long-travel end-of-car cushioning devices (EOCCDs) such as autorack cars. TrAM requires that freight trains be made up, to the extent practicable, with the loads located closest to the locomotives. The marshalling of heavy blocks of cars at the rear of the train is prohibited unless blocks ahead are equally as heavy. Light cars (empties), lightly loaded cars or blocks of these cars are usually required to be placed as close as possible to the rear of a train.
In 2009, the TSB investigated a CN main-track train derailment on the CN Kingston Subdivision near Brighton, Ontario.Footnote 4 The Brighton investigation identified that other than compliance with the Transportation of Dangerous Goods Act and Transportation of Dangerous Goods Regulations, railways were left to marshal their trains to suit their operations. Regulatory overview consisted primarily of monitoring for compliance to the Transportation of Dangerous Goods Act and Regulations because there were no other regulatory guidelines, regulations or rules in place that required railways to manage and minimize in-train forces.
The Brighton investigation also identified that, from January 2007 to June 2009, CN had an average of 151 failed knuckles/drawbars per year that resulted in train separations just on the Kingston Subdivision. In contrast, CP experienced fewer than 20 failed knuckles and drawbars per year across its entire system. The TSB determined that, while the train was marshalled in accordance with CN’s General Operating Instructions (GOI) and regulatory requirements, it was not configured in a manner that effectively managed in-train forces. The investigation further identified that the absence of effective regulatory overview of train marshalling presents a risk that some railways will continue to marshal trains for operational efficiencies without consideration for effectively managing and minimizing in-train forces. In conjunction with the release of the Brighton report on 29 September 2010, the Board issued a safety concern stating that Transport Canada may not have adequate tools in place to effectively monitor train marshalling practices and a railway’s ability to manage in-train forces.Footnote 5
In response to the safety concern, CN developed a series of train marshalling business rules, primarily related to train weight distribution, in an effort to more effectively manage in-train forces. Since then, CN train marshalling initiatives have evolved, with some elements moved into other CN documents such as its Train Marshalling Job Aid (July 2018) and GOI. However, there continued to be no regulatory guidelines, regulations or rules in place that required railways to manage and minimize in-train forces.
With regards to managing in-train forces, CN’s optimal placement of a single DP remote locomotive is based on weight distribution and generally results in the DP remote locomotive being placed at about 2/3 of the total train length. In this occurrence, the placement of the DP remote locomotive between lines 81 and 82 of the occurrence train consist did not comply with CN requirements. Given the make-up of the train departing Sarnia, the DP remote locomotive should have been placed between lines 114 and 115 (line 117 when the 2 head-end locomotives are included) of the train consist. However, TSB dynamic simulation results showed that, in this case, CN’s general approach to DP remote locomotive placement had little effect on the in-train forces.
CN train marshalling criteria does not specifically require the placement of empty and/or lighter loaded cars, such as autorack cars equipped with EOCCDs, at the tail end of a train. Instead, CN relies on a generalized train weight distribution rule (Rule 1) designed to prevent a train from having excessive weight on the tail end, a condition generally referred to as a “tail-end heavy” train. CN’s train marshalling Rule 1, in effect at the time of the accident, stated that no more than 33% of the train weight was to be placed in the rear 25% of the train’s length. The occurrence train had 31.1% of its weight in the rear 25% of its length. Although the occurrence train was compliant with the CN Rule 1 marshalling requirement, it was borderline “tail-end heavy” and the significant trailing tonnage contributed to the severity of the accident.
TSB dynamic simulations identified that re-marshalling the lighter-weight autorack cars equipped with EOCCDs to the tail end of the train was critical to reducing in-train forces. The simulations also confirmed that there were safer train marshalling alternatives available to CN that would have reduced the risk of structural failure of DJJX 30478. However, in the absence of any regulatory requirements for managing in-train forces, safer alternatives were not implemented for this particularly challenging tunnel track profile.
As demonstrated by this accident, the consequences of a freight car sustaining structural failure during train operations can be significant. Although bathtub gondola DJJX 30478 was in a deteriorated state at the time of the accident, elevated in-train forces also had to be present for this accident to occur.
While failed knuckles/drawbars can occur for various reasons, such failures are a leading indicator of a railway’s effectiveness in managing in-train forces. In 2009, CN had an average of 151 failed knuckles/drawbars per year on the Kingston Subdivision and the Board was concerned that Transport Canada was unable to effectively monitor a railway’s ability to manage in-train forces. Despite CN train marshalling initiatives since 2009, between 01 January 2016 and 31 December 2019, CN had an average of 157 failed knuckles/drawbars per year on the Kingston Subdivision.
With the evolution of longer and heavier trains, effective management of in-train longitudinal forces has become essential to safe operations. Although CN has implemented train marshalling business rules, it continues to experience challenges in its ability to consistently safely manage in-train forces. In order to reduce risk to the public, property and the environment, Transport Canada may wish to ensure that all railways have adequate practices in place to effectively manage in-train forces.
The TSB would appreciate being advised of Transport Canada’s position on this matter, and what action, if any, will be taken in this regard. Upon completion of the investigation into occurrence R19T0107, the Board will release its investigation report.
Original signed by
- Associate Administrator for Railroad Safety
Chief Safety Officer
Federal Railroad Administration (FRA)
- Assistant Vice President, Safety
Canadian National Railway
- Senior Counsel, Regulatory Affairs
Canadian National Railway
- Assistant Vice President, Safety and Sustainability
Canadian Pacific Railway
- Director, Regulatory Affairs
Railway Association of Canada
- General Manager – Mechanical Services
The David J. Joseph Company – Rail Equipment Group