Tuesday, January 14, 2014

Improper signal maintenance actions led to October 2012 Amtrak derailment

In October/November, the National Transportation Safety Board finalized and released their report on the derailment of an Amtrak ''Wolverine'' train in Niles, Michigan back in October 2012. I felt that this was an extremely important investigation because the incident occurred on a segment of Amtrak-owned "high-speed" track, the first segment of modern 110-mph running in the United States outside of the Northeast. I've covered the event previously (here and here). My initial article was a bit too speculative about what happened—it quickly became clear that the train engineer responded appropriately to the signal indications, and that the positive train control (PTC) overlay (specifically Incremental Train Control System, or ITCS, in this case) behaved as designed.

As seen in the left-hand photo above, the train engineer had a "clear" signal (green over red), permitting the train to operate at the line's maximum speed of 110 mph. However, the switch was "reversed", aligned to take the train into a yard where speed is limited to just 15 mph (right). The in-cab signals had also presented a "clear" indication to the locomotive engineer, but he waited until the physical signals were visible before beginning to accelerate. The train hit the misaligned switch at 61 miles per hour. Fortunately there weren't any fatalities in the incident, though 13 people were injured (8 were taken to the hospital, but none of the injuries were life-threatening).

There had been maintenance work on the track shortly before the Amtrak train went through. A maintenance-of-way crew had been using a tamping machine, which uses vibrating paddles to pack ballast rocks around the rail ties/sleepers. Somehow this action had made it so the switch to the Niles yard could not move between positions properly. At the request of the maintenance crew (who were finishing up their work), the remote train director (dispatcher) tried aligning the switch to allow them to move equipment into the yard, but at 9:17 AM, the switch reported that it could not properly detect the position of the switch points.

The dispatcher called an Amtrak signal supervisor at 9:18, who attempted to forward the request to one of the regular maintainers. After being unable to get a response, he headed out to the site himself and arrived at 9:55 AM. The Wolverine was due to stop at the station in Niles at 10:07 and then go through the switch a few moments later, so the supervisor had less than fifteen minutes to find and fix the problem for the train to continue on time.

From the NTSB report:
He said that he first attempted to correct the problem at the power-operated switch machine but was unsuccessful. He said that he entered the signal bungalow and removed two cartridge fuses, opened two terminal nuts on the terminal board, and applied local battery power using two jumper wires. The signal supervisor stated that when the battery power was applied, the local control panel indication lights showed that the #2 switch was aligned and locked normal. The signal supervisor stated that he did not verify the physical position of the #2 switch before applying the jumper wire.

The train director contacted the signal supervisor on the radio and informed him that the #2 switch was now indicating normal on the display and asked if everything was safe for train 350 to proceed eastward. The signal supervisor told the train director that the switch was good for the normal movement. The conversation concluded at 10:10 a.m. as Amtrak train 350 approached CP 190. The signal supervisor said that he observed train 350 approaching CP 190. He said that as the train entered the yard tracks, he realized what had occurred, and he then removed the jumper wires and reinstalled the cartridge fuses. The signal supervisor did not notify anyone at this time that he had used jumper wires just before the derailment, and he did not leave the signal bungalow to aid the passengers and crew on the derailed train.

This wasn't the correct way to do maintenance on the signals. According to the rest of the report, jumper wires are known to cause false indications, and should only be used after other methods of diagnosing the problem have been exhausted. In addition, jumpers are only to be used after getting appropriate approval in the management chain, notifying the dispatcher, and "implement alternative means of protection".  Presumably some flags or signs placed along the right-of-way near the signals to indicate they're undergoing maintenance.  The PTC system along the tracks should also have a method for transmitting information about work zones to locomotive engineers.

However, because those precautions weren't taken, the PTC system relayed what it was told by the legacy signal equipment. The crash likely would have been worse if not for the train engineer's distrust of the in-cab signals, as the train would have accelerated sooner and hit the switch at a higher speed.  Needless to say, I'm really curious how best-of-breed fully-integrated high-speed signaling handles this type of situation, as opposed to the overlay systems we're expecting to see deployed across the country as PTC is implemented.


  1. I have my doubts about PTC being online in 2015. I have even more difficulty accepting the safety of a system layered over legacy equipment that's been cobbled together and spliced to function between four or five railroads that have themselves been cobbled together out of dozens of pre-merger systems that were run since time immemorial. PTC is supposed to prevent accidents, so why do I feel like this is a step in the wrong direction?