Metro Train Derailment

[Awesome!]

http://news.yahoo.com/s/ap/20090623/ap_on_re_us/us_dc_metro_train_derailment

Who's at fault?

[BillBlom]

Great question. Isn't the Metro pretty much fully automated?  Hope it isn't another texting moment...

[tomikawaTT]

Awesome!

Who's at fault?

An engineer (civil, mechanical or electrical) asks, "What went wrong?"

A lawyer, politician or blame-assigner asks, "Who's at fault?"

If this is, in fact, a fully automated system, the answer is, "Nobody is at fault.  There was an electrical or mechanical failure."  Hopefully, the engineers among the accident investigators will find out what failed and how such failures can be prevented in the future, and the lawyers, politicians and blame-assigners will stifle themselves until the final report is submitted.

If things could be made perfect, there would be no need for erasers on pencils...

Chuck (Former Air Force QC inspector)

[oltmannd]

From the hints and glimmers of facts that have filtered their way through the news media, it sounds like a a track circuit failed to indicate the presence of the train ahead. Track circuits are "fail safe" devices - when they get out of whack, they show a train is present when one really isn't. Why this one behaved as it did hasn't surfaced yet. Metro's automation seems to have played no part.

[blue streak 1]

Now the Washington Post is reporting multiple track failures system wide. If you were the General Manager would you have wanted to admit such?  Congress will have a field day.  Want to bet his answers will be  " I don't recall"

[tomikawaTT]

If the GM is half as smart as a pet rock, he'll say, "Look, folks, I'm a political appointee/MBA.  You really need to ask the technical types who are supposed to:

1. Keep the system working.
2. Let me know when it isn't, and what needs to be done to correct it."

Asking a typical government-organization 'manager' about technical problems is like asking the check-out clerk at the local Wal-Mart about brain surgery.

Of course, the problem is that the system is in Washington DC.  How deeply would Congress dig into a similar mishap at BART?  (Hint:  They would actually have to go beyond the beltway and out of the glare of the various media 'Washington Bureaus.')

Chuck

[HarveyK400]

Good news, bad news, and the lack of perspective.

The bad is that system-wide tests and statement that there was no problem a couple months ago are belied by inspectors finding anomalies in 6 track circuits on the system as of today.

The good is that a huge gulf exists between "system-wide" failure reported by the media and the 6 failed block circuits that have been found around a system the size of Metro.

[aegrotatio]

 The General Manager of WMATA disputes the Washington Post's yellow journalism story today that claimed the failures are "system-wide."  They found 6 examples of possible anomalies that potentially cause the intermittent failure experienced in the accident out of 3,000 track circuits.

  They aren't system-wide and the Post should be ashamed of themselves.. 

[Awesome!]

Now, they are saying there is a problem in the system. I guess everyone is pointing fingers.

[blue streak 1]

HarveyK400

The good is that a huge gulf exists between "system-wide" failure reported by the media and the 6 failed block circuits that have been found around a system the size of Metro.

RWM: Hypothetical question. How would the FRA react if a calss 1 had 6 track circuits not detecting and the level of traffic carried (i know not possible)?

[HarveyK400]

blue streak 1

HarveyK400

The good is that a huge gulf exists between "system-wide" failure reported by the media and the 6 failed block circuits that have been found around a system the size of Metro.

RWM: Hypothetical question. How would the FRA react if a calss 1 had 6 track circuits not detecting and the level of traffic carried (i know not possible)?

Probably order proceeding at restricted speed on dispatcher authority for the failed blocks; and no more than medium speed under track warrant authority for parts of the system with similar signaling.

[aegrotatio]

 I was searching high and low for WMATA Metrorail control system image and this video just JUST published.

Important details reveal that central control maintains a near-realtime view of all trains in the system even on dead circuits.

There can be 2 or 3 trains in EACH block during normal operation.  A block is 400 to 1000 feet long.

In the case of a circuit anomaly, an Absolute Block is imposed, which means 1 train per 400-1000ft block.  Trains enter the block through communication with the Metro control center which grants a Permissive block at which trains move at restricted speed.

At no time are trains "lost" in the system as I jhave been saying before.

Here is the video:

 http://www.wtop.com/?sid=1724694&nid=25

 

[DMUinCT]

It's how you look at it!    Only 6 dangerous failures on a very large transit system.

If only 6 aircraft engines had fallen off (or wings had fallen off, or landing gear had failed) on there  Airliners in flight within a large airline system, and the Operator claimed only one resulted in a fatal accident.

Would the press let the Operator and Manager of the offending airline get away with that statement?   This would be headlines for weeks!

[oltmannd]

blue streak 1

RWM: Hypothetical question. How would the FRA react if a calss 1 had 6 track circuits not detecting and the level of traffic carried (i know not possible)?

That's not what Metro said.  They said they had 6 circuits that were defective.  They didn't say they weren't detecting trains.  More likely, they were shunting too easily - indicating trains where there were none.  This is a very common defect.  TOLs are everywhere!

[aegrotatio]

 That math doesn't add up.

6 anamalous signals out of 3,000 sensors that protect about 20 trains, I'm not sure the math is working for you on your statement.

[oltmannd]

aegrotatio

I was searching high and low for WMATA Metrorail control system image and this video just JUST published.

Important details reveal that central control maintains a near-realtime view of all trains in the system even on dead circuits.

There can be 2 or 3 trains in EACH block during normal operation.  A block is 400 to 1000 feet long.

In the case of a circuit anomaly, an Absolute Block is imposed, which means 1 train per 400-1000ft block.  Trains enter the block through communication with the Metro control center which grants a Permissive block at which trains move at restricted speed.

At no time are trains "lost" in the system as I jhave been saying before.

Here is the video:

 http://www.wtop.com/?sid=1724694&nid=25

Interesting. Thanks for posting that.

You are confusing the term "block" as they are using it here, with a single, track circuit "block".

You are also confusing the train dispatching system's ability to track trains with the signal safety system that keeps them apart.

There can be only one train allowed in a track circuit block at a time.  In addition, no train is allowed in the track circuit block immediately behind the train, either.  That's so there is room for reactive braking should the following train exceed authority.

The "permissive block" set up to allow operation in what can be viewed as "dark" territory had no direct relation to the track circuit blocks.  It's a big as it needs be to accomodate the manual operation and to protect against following trains.  Think manual block or track warrants as an analogous.

The dispatching control center tracks trains for supervisory control.  It is not part of the safety system (except when they give permission to move manually i.e. permissive blocks).  That the control center doesn't lose track of trains means nothing to the signal safety system.

A "turned off" track circuit is the same as one that show as occupied.  The problem with the crash was there was one that showed the track as clear when there was a train in it.

The simplest way to think about a track circuit is a circuit with a battery on one end and a relay on the other.  The two rails are the "wires" that connect the battery to the relay.  When no train is present, the circuit is complete and the relay picks up.  When a train is present, the wheelsets create a short  circuit and the relay drops. 

[HarveyK400]

While blame has been assessed to a relay failure, the logic escapes me.  The struck train had braked to a stop in compliance with a restrictive signal in the block approaching the occupied block at the station.   In both cases, the occupied station block (train A) was not cleared to re-energize the relay for the approach block.  The question is how did the track circuit for the approach block become energized and transmit a clear code behind the struck train?

• Even if the relay behind the struck train (B) failed to drop, the following train (C) would get a restrictive code or no code and begin braking in ATO mode upon entering the block before the operator had to put the train into an emergency stop.
  
• If, as reported, the struck train (B) failed to shunt the approach block track circuit, the following train (C) would get the same restrictive signal to brake as the struck train.

I can only speculate that a stray clear signal leaking from the adjacent track, being stronger than no signal from the shunted circuit, was picked up by the train C on-board antennae; and, as a result of post-collision testing, energized the track circuit relay as well.

[HarveyK400]

This isn't about six planes losing engines and crashing; but more analogous to finding missing bolts or cracks during inspections, grounding all planes for inspections, and releasing them for service if no defect is found.

[oltmannd]

HarveyK400

While blame has been assessed to a relay failure, the logic escapes me.  The struck train had braked to a stop in compliance with a restrictive signal in the block approaching the occupied block at the station.   In both cases, the occupied station block (train A) was not cleared to re-energize the relay for the approach block.  The question is how did the track circuit for the approach block become energized and transmit a clear code behind the struck train?

• Even if the relay behind the struck train (B) failed to drop, the following train (C) would get a restrictive code or no code and begin braking in ATO mode upon entering the block before the operator had to put the train into an emergency stop.
  
• If, as reported, the struck train (B) failed to shunt the approach block track circuit, the following train (C) would get the same restrictive signal to brake as the struck train.

I can only speculate that a stray clear signal leaking from the adjacent track, being stronger than no signal from the shunted circuit, was picked up by the train C on-board antennae; and, as a result of post-collision testing, energized the track circuit relay as well.

The failure at the "back end" of the bad track circuit could have been sending out "clear"regardless of the signal it was receiving (in this case, no code).... the equivalent of a stuck track circuit relay. So the following train would have seen a "clear" all the way until it entered the block with the stopped train.

[HarveyK400]

With bi-directional signaling, that seems possible but highly improbable since the third train would shunt, or at least weaken the signal from behind, the "back end."  Furthermore, aren't the coded signals polarized by direction; and are not the reverse signals red by a direction of travel relay?

If tests showed a train was lost, ie, did not adequately shunt the circuit from the front end, a weak stop ("restrictive" since a dispatcher could give permission to move) command should be picked up to protect the train at the station.

[aegrotatio]

 Okay, thanks for the correction on the terminology, but the fact remains that we still do not know why both the administrative system and the safety system, and the operator of the train had all three failed to prevent the wreck.

We're over a month out and they are still very far away from describing what actually happened.  They're out there every night and every weekend running trains back and forth scratching their heads.

 

[oltmannd]

HarveyK400

With bi-directional signaling, that seems possible but highly improbable since the third train would shunt, or at least weaken the signal from behind, the "back end."  Furthermore, aren't the coded signals polarized by direction; and are not the reverse signals red by a direction of travel relay?

If tests showed a train was lost, ie, did not adequately shunt the circuit from the front end, a weak stop ("restrictive" since a dispatcher could give permission to move) command should be picked up to protect the train at the station.

I think the direction of travel determines which end of the track circuit is fed (and sets all opposing signals to stop). In this case there was something very awry (either an extremely rare failure mode or human error on installation/maintenance) at the back end of the circuit where the train was stopped such that it had the following track circuit sending out a "clear". Crosstalk from an adjacent track would have to be ruled out unless the onboard amplifier failed at just the right moment that increased the sensitivity a huge amount. Otherwise, the cab signals onboard would have been flipping like crazy all trip.

[DMUinCT]

Looks like they are looking at the 35 year old PC Boards in the control system.

NTSB Report #4 below:

http://www.ntsb.gov/pressrel/2009/090729.html

[HarveyK400]

DMUinCT

Looks like they are looking at the 35 year old PC Boards in the control system.

http://www.ntsb.gov/pressrel/2009/090729.html

 

While there may be a problem with either or both signal strength or circuit boards, time is an issue with these separate tests.  I still suspect a leaking signal from the other track that may have something to do with the WZ bond and electrical grounding and soil resistance conditions at different times.  These factors may be exacerbated by the unusually long, 1,700-ft blocks for this type of signaling.

[henry6]

If printed circuit boards are full of micro and mini circuits, wouldn't it make sense that no matter how "clean and pure" the environment, that micro an mini amounts of moisture, dust or other corosive action, undetectable contanments at least to the naked eye for sure, become a factor over any given amount of time.  In other words how long can the integrity of a circuit board be guaranteed in and given environment at any given time and be deemed safe from contamination or invasion?

[DMUinCT]

Electronics, 

The late 1940s, the Transistor is invented, the radio tube is doomed.

Late 1950s TI comes up with the Intergrated Circuit Chip.

The 1970s, the Micro-Processor, a full computer on one chip, starts to find uses.

By 1974 PC Boards made up of IC Chips for switching with Transistors or SCR for input/output drivers. How long the board will last was not known, even if Military Spec.tested components that will work at high and low temperatures.

In 1981, IBM introduces the Personal Computer, the PC.

By the mid-1990s most boards still work, but for those that fail, repair parts are unavailable.  Been their, done that.  Replacement PC Boards to cover maintenance contracts must be designed to do the same job but with new components.  The mix of old and new PC boards MUST work together.

Do you have a 35 year old TV set ?   Will it still work ?   Maybe, maybe not.

[aegrotatio]

If this turns out to be tin whiskers I'm moving to Montana.

 

[DMUinCT]

aegrotatio

If this turns out to be tin whiskers I'm moving to Montana.

 

That would be on the newer boards made without lead.  It is the computer problem of the future.

[oltmannd]

HarveyK400

These factors may be exacerbated by the unusually long, 1,700-ft blocks for this type of signaling.

PRR coded track circuits/cab signal blocks were typically 2 miles long but generally were split into two track circuit "cut sections".  1700 feet would be a very short track circuit in the PRR world.

Still think it highly unlikely that leakage could have enough oomph to get over the on-board amplifier sensitivity.  The SD60MAC demo units running at 70 mph were feeding 100Hz power to the traction motors, but even with the cables only a few feet away from the cab signal receiver bars, there wasn't enough signal to get the 100Hz cab signal system to pick up.

[HarveyK400]

oltmannd

HarveyK400

These factors may be exacerbated by the unusually long, 1,700-ft blocks for this type of signaling.

PRR coded track circuits/cab signal blocks were typically 2 miles long but generally were split into two track circuit "cut sections".  1700 feet would be a very short track circuit in the PRR world.

Still think it highly unlikely that leakage could have enough oomph to get over the on-board amplifier sensitivity.  The SD60MAC demo units running at 70 mph were feeding 100Hz power to the traction motors, but even with the cables only a few feet away from the cab signal receiver bars, there wasn't enough signal to get the 100Hz cab signal system to pick up.

First, one comment was that most Metro blocks were 400-1,000 feet long.  Another writer mentioned a nominal 1,500-foot limit.  I recall the latter figure for the CTA's original cab signal system as well.  When it was quiet, you could hear the code pinging along the Lake-Dan Ryan and Jefferson Park extension.

As for the PRR, the reason the continuous 100-Hz traction frequency was not "picked up" was that code bursts at 75, 120, and 180 cps were transmitted, activating the respective tuned vane relay in the locomotive/emu receiver unit for 30, 45, & up to 125 mph commands.  As you pointed out, blocks could be up to two miles long; but also lengthened by cut sections amplifying and repeating the weakened block circuit signal.