Tuesday, October 23, 2012

Amtrak derailment in Niles, MI appears to show weakness in Positive Train Control

This past Sunday, October 21st, an eastbound Amtrak Wolverine train from Chicago to Detroit/Pontiac derailed just outside Niles, Michigan, shortly after leaving the station there. While most news outlets have simply reported the derailment and the modest number of minor injuries that resulted, the Detroit Free Press noted in its reporting that the train ended up on the wrong track.

View Amtrak derailment, Niles, MI 2012-10-21 in a larger map

The mainline here only has a single track, so the "wrong track" appears to be a spur into an old yard area adjacent to the mainline. There is a siding right in Niles so that trains can pass each other at the station. Heading east, the line narrows back down to a single track about 1.3 miles out. Immediately after becoming single-tracked, there's another switch that leads off to the aforementioned yard area. According to Google Maps imagery, this is the only spot where it's possible to leave the mainline until another siding appears in Dowagiac, 12 miles away from Niles.

The track in this area is able to carry trains at speeds up to 110 mph in part because it has a Positive Train Control (PTC) system installed (specifically GE Transportation Systems's Incremental Train Control System, or ITCS) which is supposed to improve safety by preventing collisions between trains. Unfortunately, it looks like that system failed in some way, and disaster was probably only averted through the use of a very simple device actually known as a derail—a wedge-like hunk of metal that is locked onto a rail with the express purpose of derailing anything that rolls over it. This is generally used to prevent parked rail cars from accidentally rolling out of a yard or spur track and onto the mainline, but (as happened in this case) can also prevent a train inadvertently coming off the mainline from slamming into parked rail cars at full speed.

In fact, a derail device can be seen on another track in this shot of the derailment from the Associated Press. The mainline appears to be the track way off on the left, partially obscured by the orangish freight cars, while the derail appears below them (in the foreground):

The National Transportation Safety Board (NTSB) is investigating this incident. Hopefully they will have an initial report coming on it soon, though their full investigations typically take a year to 18 months. The NTSB has been recommending PTC be installed on rail lines for many years now, so it will be interesting to see how they react to an incident in a zone where it is apparently in use. Perhaps this specific spot didn't have it. My gut is leaning toward the idea that the switch that led off to this spur was intentionally not included in the PTC signaling system. Is it a manually-thrown switch, or is it controlled remotely? Amtrak actually owns the mainline here, and dispatching is their responsibility. If the switch is remotely controlled, then the dispatcher out on the east coast probably screwed up.

It'll also be interesting to learn how fast the train was going when it entered the spur and when it (presumably) hit the derail, and when or whether the brakes were being applied. If the signals were all working correctly and the switch off into this yard area is actually wired into the signaling system, the train probably should have had a "stop" signal just west of the misaligned switch—either the engineer should have stopped there because of the signals he/she saw in the cab and/or in the signal heads adjacent to the track, or the PTC system should have automatically stopped the train.

However, there's a decent possibility that the expected outcome here would have been for a train to get a signal telling the engineer and PTC system that it could proceed at a slow speed into the yard (probably 30 mph at most, and likely less than that). But, looking at pictures of the derailment, it seems to me that the train was going much faster than 30 mph when it hit the either derail device or a warped-enough piece of track to cause it to start sliding along the ground. (That said, there's an incredible mass there, and my imagination may not handle it properly. I also don't know exactly where the derailment began.)

One complexity thrown into the mix here is that a train traveling between two points has a very different method of operation from a train working in a yard. Out on a mainline, two trains should never really occupy the same stretch of track, but yards, spurs, and sidings often involve partially-assembled trains shoving things around, attaching and detaching themselves to or from other train segments. In theory, the PTC system probably should have known that the Amtrak train is a passenger train and had no business going into a freight spur.

There's a possibility that the switch leading off the mainline actually broke either as the Amtrak train went through it or shortly before, but that seems pretty unlikely based on where the train ended up: The train probably would have derailed as it passed over the mainline switch, rather than managing to continue a couple train lengths beyond. Some of the train—particularly the lead locomotive—probably also would have continued forward along the mainline in that event. Stranger things have happened, though.

This line is really one of the primary test beds for PTC implementation around the country: Amtrak first installed ITCS in 2002 to test it out, and speeds have been raised progressively over the last decade. Starting out at 79 mph in 2002, the limit went up to 95 mph in 2005, and Amtrak trains finally began operating regularly at 110 mph in February 2012.

The derailment came just two days after Amtrak made their first 110-mph test run along the Chicago to St. Louis corridor, also enabled through the use of ITCS PTC. That line is owned by Union Pacific and has historically had a cab signaling system in place to prevent train collisions. The cab signaling system might have acted as a fallback if a similar situation occurred in Illinois, but it's not clear to me if the newer PTC system would override that or not.

Speaking as someone who wants to see passenger trains deployed much more broadly across the U.S. and operating at higher speeds to be more competitive with car and air travel, this is extremely frustrating. It really looks like someone did a half-assed job either in designing the underlying signaling technology, or at least in their implementation of it it on this line. For all that has been written about the problems leading up to that high-speed train crash in China last year, our signaling systems in the U.S. are generally far less able to prevent accidents than the system in use there. PTC is supposed to make things much safer, but it failed to do its job on Sunday.

[Edit: Please take a look at my update to this story.]


  1. I took the Empire Builder to CHI that day, and we heard about the derailment... not much about the cause, but a lot of people were stuck at Union Station and they were getting coach buses ready to run the route eastward that evening.

  2. I was on 350. To me it felt as though the train was doing 60 when it went through the first switch (it went through at least 2 and maybe 3 or 4). There was an amtrak workmen's pickup truck (with people in it) at the scene when the event happened; I wonder if they might not have thrown the switch by hand for some reason? In any case, they drove away immediately (maybe 5 minutes) after it happened. It certainly felt as though the brakes were applied.