Ghost Train: The 2009 Washington D.C. (USA) Metro Collision
Lamond Riggs is a neighborhood of 4238 people (as of 2022) in the northeast of Washington D.C. on the border with the federal state of Maryland, 7.5km/4.5mi north-northeast of downtown Washington D.C. and 44km/27mi west of Annapolis (Maryland) in the far east of the United States of America (both measurements in linear distance).
The neighborhood lies on the Red Line of the Washington Metro System, the oldest and busiest of the system’s five lines. The dual-tracked line is electrified via a third rail on the ground outside the railway track as it serves 27 stations on a 51.3km/31.9mi route. The first section opened in 1976 while the current status wasn’t achieved until 2004. The line serves a large chunk of the entire system’s 211500 daily riders (as of 2022).
The trains involved
Metro Service 112 from Glenmont to Shady Grove was provided by a six-car electric multiple unit entirely consisting of RI 1000-series train cars. The 1000-series trains made by Rohr Industries was introduced in 1976, with each car measuring 22,86m/75ft in length and offering space for 82 seated and 93 standing passengers. They always operate in pairs of two with a “mate car” placing a control desk at each outer end, meaning longer trains are always even-numbered (2, 4, 6 or 8 cars). The trains are engineered with strong acceleration in mind but can still reach 121kph/75mph.
Running in the same direction was Service number 214, consisting of 4 Breda series 3000 cars followed by 2 CAF series 5000 cars. The series 3000 was introduced in 1987 and is largely identical to the series 1000 but only offer 68 seats due to a different interior layout. The same applies for the series 5000 introduced in 2001, which can carry the same number of passengers as the series 3000 but returned to the series 1000’s cheaper wheelset-design. All Washington Metro trains can operate in 3 modes:
- Mode 1 (Automatic): The train operates essentially autonomously, with the Automatic Train Control (ATC) and Automatic Train Protection (ATP) controlling all the train’s functions from the doors to acceleration and braking. In this mode the driver has the role of an observer for both the track and possible error-messages from the onboard system. This is the standard mode all trains are expected to operate in.
- Mode 2 (Manual, with speed protection): The driver manually dictates acceleration and braking, while the ATP ensures adherence to the speed limit at all times.
- Mode 3 (Fully manual): ATP is disabled, the trains entire operation is at the hands of the driver with no backup protection of any kind. This mode is only intended to maintain operations if equipment malfunctions disable Mode 1 or 2.
On the 22nd of June 2009 Metro Service #112 departs Takoma station at approximately 4:54pm. It’s a Monday afternoon at the beginning of rush hour, and while exact numbers aren’t known it can be assumed that the train isn’t exactly empty. Right after leaving the station the train came to a stop on the open track, with the driver announcing that the system had registered another train (#214) up ahead and that they would resume driving as soon as the distance between the two trains increased. Witnesses would later say that, after the stop, the train quickly reached “full speed”, which would be 89kph/55mph on that section of the line.
Coming around a long bend on approach to the overpass of New Hampshire Avenue the driver of train #112 suddenly spots the rear of another metro train a short distance ahead, stopped. The driver immediately triggered an emergency stop, but as so often, it was too little too late. At 4:58pm train #112 slams into the back of the stationary train #214 at 79kph/49mph.
The leading car of the rear train digs beneath the stationary train car’s frame as the driver’s cab is obliterated, causing the rear car of #214 to destroy the interior of the leading car on a length of 63ft before coming to a halt, leaving some survival space in the rear section of the leading car. The driver of #112 doesn’t stand a chance as the two trains sleeve over one another, dying along with 8 of her passengers. At least 80 passengers are injured, 57 of which severely.
The first call to emergency services was placed by a passenger of the stationary train at 5:03pm, soon followed by more calls from survivors. A surviving passenger in the leading car of the rear car told the police officer on the line that he had been in the car with around 15 other people, had seen the seats and floor “coming towards me” and was now trapped a few centimeters from the ceiling. Responders arrived at the site a few minutes after the accident, being warned by survivors to steer clear of the third rail which had yet to be deactivated, posing a lethal risk to responders and survivors. The tracks at the site lie between two CSX-rail tracks used by regular trains, responders cut an opening into the fencing on each side of the tracks and carefully accessed the wreckage, it took until 5:20pm for CSX to confirm that their traffic had been shut down and the tracks posed no risk to responders accessing or leaving the site. Some of the first responders on site walked the tracks from the adjacent stations, allowing precise location of the wreckage and improved information for following responders.
The loud bang of the impact followed by the interior of the trains filling with smoke and debris caused a panic among the passengers, worsened by the doors not opening right away. Within two hours over 200 firefighters were on site, along with a similar number of police officers and medical personnel. Survivors were initially instructed to stay on the trains if they hadn’t left them already, allowing more efficient triage and rescue of more severely injured survivors through clear paths. The first car of the rear train held a number of trapped survivors, finding no other options firefighters quickly resorted to saws and the “jaws of life” (hydraulic spreaders) to pick apart the doors and the destroyed interior, allowing them to access the passengers. Within 1.5 hours all survivors were rescued from the wreckage.
Investigators already started examining the trains while the rescue was still underway, being tasked with how an automated, supposedly fail-safe system had had a fatal collision between two operational trains. Something like this had never happened before, the last accident had been almost 5 years prior when an empty train inadvertently rolled backwards on a sloped section, striking a train parked at the platform.
Reviewing the recorded data from both trains they found that the rear train had been operating in Mode 1 as it was supposed to, showing no sign of any malfunction right up until the driver triggered an emergency stop, which was initiated but came too late to avoid a collision. As such, the control systems should have detected train 214 up ahead and automatically maintain the programmed minimum distance, which the system had done shortly before the accident when it temporarily stopped the train right after departing Takoma station.
Reviewing data from train 214 they found that it had been operating in Mode 2, with the driver explaining that he used the expanded control to stop at the platform marker for 8-car trains in stations to ensure that his 6-car train was safely and properly inside the station even if he stopped a little short. This was considered a valid justification, with investigators concluding that the mostly-manual operation had no influence on the events that had unfolded. It didn’t answer why the train had stopped out on the open track, and why train 112 had seemingly randomly failed to detect it.
When operating in Mode 1 the ATC-system sends speed-commands to the trains at regular intervals, instructing the trains what speed to maintain for the section of track at hand. The investigation found that the stopped train had last received a 0 mph speed command, which, simply put, made no sense. The system had essentially behaved as if, from one sensor to the next, train 214 had ceased to exist, and a train that doesn’t exist obviously doesn’t get a speed command. Upon interviewing drivers who had passed the site ahead of the accident it was found that they, too, had briefly received a 0 mph speed command, but their trains had “coasted” through the section and soon picked up a regular speed command again as they entered the next section, barely losing speed. It has to be noted that a 0 mph-command is not the same as initiating an emergency stop, it is simply the absence of a command. However, while most trains had had no problem passing the 0mph-sectiontrain 214 had been operating in manual mode, travelling at a lower speed to begin with, and thus came to a stop within the 0 mph-section.
It transpired that, for some reason, the system failed to register train 214’s location past a certain point, erroneously reporting a clear path ahead for train 112. Even worse, the system still reported an empty track-section while the wreckage was at the site.
The track occupancy for the Washington Metro is detected by a system of two “bods” per section, one at each end. A transmitter-module at an offsite equipment room generates a specific audio frequency which is supplied into the track via the transmitting impedance-bod at one end of a track-section. If no train occupies the section in question the signal travels through the rails and is picked up by the receiving bod at the far end of the section. The receiving bod relays the signal it received back to the equipment room, where it is automatically checked against the sent signal, if it matches the programmed requirements a track relay in the tech-room is energized, which in turn is picked up by the ATC-system as confirmation of the section being unoccupied. If a train enters the section the trains wheels redirect and change the signal, which leads to the track relay not being energized which means to the ATC-system that the track is occupied.
Upon having the wreckage removed from the site, in the process of which the last victims could be recovered from beneath the wreckage, and getting the tracks repaired investigators acquired a six-car train and parked it right where train 214 had been standing ahead of the collision. As feared and expected, the system claimed that there was no train on the line. Investigators then proceeded to manually check 4 sections adjacent to the site of the accident, simulating a train’s axle by means of an added 0.06 ohm current. Each section got checked 3 times, at the transmitting end, in the middle and at the receiving end. At last they repeated the test at section B2–304, where the accident had occurred. The system detected the “axle” at the transmitter-end, but failed to detect it in the center of the section and only intermittently recognized occupancy at the receiving end, with the relay momentarily energizing before switching to the non-energized “occupied” setting.
Investigators found that the track-circuit for section B2–304 had failed to detect ANY train for the past five days ahead of the accident, at which point a component of the output transistors had been replaced. Through further experimentation investigators found that faulty output transistors were generating an undesired, so called parasitic oscillation, a frequency that was not intended to be generated at all. This vibration was tracked as travelling on a previously unknown path by using the metal equipment racks on which the components were mounted. It eventually reached the receiver-module, causing the module to pick up an unsteady vibration on the same frequency as the intended signal. This false signal, varying in strength, could get strong enough at times to be picked up by the receiver as a “track free” signal, energizing the relay and reporting a free track to the ATC-system regardless of the actual signal coming from the tracks.
Workers who had installed the, as it turned out, faulty impedance bod testified that they had had to raise the power-level of the transmitter from 30 to 55 percent before the relay would pick up the system. This increase in power was within official limits, but was enough of an increase to allow the parasitic oscillations to override the actual signal.
With these findings the investigators concluded that undesired oscillations had overridden the affected section’s true signal, reporting to the rear train’s control system that the path up ahead was clear as it no longer received the “occupied”-signal from the forward train. Train 214’s driver was cleared of wrongdoing as they had followed the instructions given to them by the control system, as they were supposed to. They couldn’t possibly have known that they were meant to disregard the speed command as erroneously. The report notes that the second generation GRS track circuit modules widely used by the Washington Metro are significantly susceptible to the kind of parasitic oscillation that led to the accident, recommending that their removal from the system receives the utmost priority to ensure the safety of the Metro’s passengers. Alstom, manufacturer of the second generation module blamed for the accident, stated that third and fourth generation versions of the component feature significant design changes and that they failed to detect any parasitic oscillations when testing those units. In the meantime the company recommended the installation of so-called ferrite chokes on the modules’ power-supply, which would reduce frequency interference, or installing insulating blocks between the transistor heat sinks which would cause vibrations to be unable to pass through them as they had ahead of the accident.
The investigation also recommended retirement of the 1000-series trains as they showed poor crash protection, especially regarding telescoping/sleeving, compared to the other models in use. In response the Washington Metro banned them from running at either end of trains, only using them in the middle for the remaining eight years ahead of their retirement in 2017. This was the second time the recommendation had been made, with the NTSB (National Transportation Safety Board) previously recommending the retirement or severe structural reinforcement of the series after a 2004 collision. Additionally, the series 1000 lacks onboard data-loggers, making it needlessly difficult to reconstruct the events ahead of an accident.
The series 5000-cars which made up the rear section of the forward train fared better regarding survival-space post-collision, but were still stripped for parts and scrapped after the investigation concluded. The last series 5000 cars were retired from public service in 2018. Of the trains involved in the accident only the series 3000 cars remain in service as of 2022, being scheduled to be withdrawn in 2024.
After the accident a memoritative plaque was unveiled at nearby Fort Totten Station, listing the victims and thanking the responders, civilian and professional. Survivors and relatives of victims also maintained a makeshift memorial on the overpass just a few feet from the site of the accident, which was “officialized” three years after the accident when Washington’s Mayor and the NTSB’s chairwoman unveiled an official memorial plaque at the site.
Lastly, six years after the accident, the “Legacy Memorial park” was opened a short distance from the site of the accident. The small memorial garden features nine stone pillars symbolizing the nine victims of the collision. Victims’ families and government officials attended the opening while Metro officials were notably absent. In a coincidence not missed by the media the plaque was dedicated the same week as an NTSB-hearing on a fatal incident the Washington Metro suffered in 2015 when a passenger died from smoke inhalation after an electrical malfunction. The 2009 collision is considered the Metro’s darkest day, but survivors and victims’ relatives didn’t hide the fact that they considered the Metro’s safety still questionable at best even after the collision’s cause had been fixed.
A kind reader has started posting the installments on reddit for me, I cannot interact with you there but I will read the feedback and corrections. I will link the post as soon as I can. Please give it some love, I’m very thankful for the posts being published there.