Sight-Distance Testing
On July 18, 2009, investigators conducted sight-distance tests in the accident area. The tests were conducted at the same time of day as the accident under similar weather conditions. The two trains used for the tests were of the same configuration as the accident trains and used the same type of equipment.
The simulated struck train was positioned in the right-hand curve on track B2 at the same location as train 214 on the day of the accident. The second train, simulating the striking train, was used to determine at what point the rear of train 214 would have been visible to the operator of train 112.
The tests revealed that the train 112 operator would have had a partial view of the rear of the stopped train from about 1,180 feet away. A full view of the rear of train 214, including both red marker lights, would not have been available to the operator of train 112 until the trains were 470 feet apart.
Speed and Braking Calculations
Because of the absence of event data recorders on train 112, investigators did not have speed and braking information for the striking train. However, the train control system did record the time the train entered and left each track segment. Train 112 was being operated in automatic mode at the time of the accident. Using WMATA-supplied train braking and acceleration data as well as known train response to speed inputs from the signal system, the NTSB was able to develop a computer program to simulate the position and speed of train 112 up to the point that manual braking had been applied. As a speed target, the simulation used the speed commands from the signal system based on the train’s position on the various track segments and the time elapsed on these segments. Acceleration or braking was applied as necessary. The simulation then integrated these accelerations to determine a new speed and integrated that speed to determine a new position for the next time step. The signal system had recorded the time the train entered and left each signal circuit segment, and the simulation matched those times.
Because the exact time of the collision was uncertain, an actual braking profile could not be developed. The NTSB therefore simulated several braking scenarios by entering various emergency braking locations into the simulation program. The decelerations taken from the emergency braking schedule table provided by WMATA were adjusted for the track slope for the location of the accident.69 Use of a collision time that was early in the range of possible times for the collision produced a terminal speed for train 112 that best matched the collision speed of 49 mph that was calculated based on damage to the railcars.
On the day of the accident, recorded train control system data indicated that at 4:56:41 train 112 began to receive a 0 mph speed command while the train was in block B2-336. While the train slowed and stopped, the train moved into block B2-328 while the 0 mph speed command continued. Based on calculated speed and position history and assuming normal service braking, train 112 would have come to a stop about 4:57:03. According to the recorded signal data, about 17 seconds later, at 4:57:20, the ATC system changed the speed commands transmitted to train 112 from 0 mph back to 55 mph, which would have caused the train to begin to accelerate to that speed automatically even though train 214 remained stopped ahead.
Although the emergency brake was found in the activated position, the operator’s reaction time could not be determined. The position and speed calculations indicated that, had the train 112 operator immediately applied the train’s emergency brakes (by depressing the emergency brake mushroom) when the rear of train 214 was fully visible (when the trains were separated by 470 feet), the train would not have been able to stop and would have struck the rear 69
of train 214 while traveling about 24 mph. Activation of emergency braking 3 seconds after train 214 had become fully visible would have resulted in train 112 striking train 214 while traveling about 44 mph.
Insulation Resistance Testing
Following the accident, investigators asked WMATA representatives to provide copies of maintenance records that included results of insulation resistance tests. The representatives responded that no insulation resistance testing had been conducted.
The Metrorail ATC technical procedure T031 Cable Insulation Resistance Testing, dated November 25, 2008, requires the periodic testing of all cables installed in conduits, in ducts, along tunnel walls, or buried along the right-of-way, and of all wires and cables entering and leaving train control rooms, equipment cases, and junction boxes. The cables and wires must be tested for insulation resistance from each conductor to all other conductors in the cable or wire and from each conductor to ground. Insulation resistance for conductors used in a power source of less than 600 volts must exceed 1 megaohm. Resistance in wires and cables used in a power source greater than 600 volts must exceed 10 megaohms. The tests must be conducted every 10 years or after any new installations. At the time of the accident, the T031 procedure was new, it was not superseding another procedure, and it was still undergoing review before final approval.
Investigators performed postaccident insulation resistance tests on bond cables and telephone lines installed from the Fort Totten train control room to both main tracks in the vicinity of the two impedance bonds for track circuit B2-304. The cables were routed together from the train control room to the field locations through conduits and manholes along an underground duct bank between the two main tracks. For several days before the accident, the Washington, D.C., area experienced heavy rains. Following the accident, the manholes were found partially filled with water, and the bond cables and telephone lines were submerged. After several days of dry weather, the bond cables were tested and found to have insulation resistance of less than 500 kiloohms to ground. As a result of the postaccident insulation resistance tests, the absolute block on track B2 through the accident area (which had been implemented after the accident) was extended to track B1 between Fort Totten and Takoma.
Investigators used a spectrum analyzer to test for stray audio frequencies from outside sources that could possibly affect the track circuits. Harmonics from the 720-hertz traction power return frequency and the adjacent track circuit frequency were detected on the bond cables for track circuit B2-304. Identified harmonics of adjacent track circuit frequencies were also found on the bond cables, but at very low levels that did not pass through the track circuit module receiver filter.
Postaccident testing revealed that maintenance communication telephone lines were faulted to ground inside the track junction boxes along the wayside. The insides of the track junction boxes were heavily rusted, and the wire terminals were grounded to the metal case of the track junction boxes through the rust. A spectrum analyzer detected the frequency for track
circuit B2-304 on the communication lines, but the signal was determined to be of insufficient strength to be recognized as a valid signal by the track circuit module receiver.
Corrugated Rail
The investigation identified corrugated rail as a condition that can possibly lead to track circuit failure. Rail-head corrugations can cause or intensify electrical arcing between the train wheels and the rail, and this arcing can create harmonics that have been shown to effectively mimic a valid track circuit signal that can be accepted by the receiver via the normal signal path. For such harmonics to lead to a track-circuit failure, trains must be traveling at a certain speed, and the corrugations on the rail must occur at certain intervals. Because these signals only occur when a train is moving and they cease to be generated when a train is stopped, WMATA ATC engineers considered the harmonics caused by corrugated rail to be benign.