RIO Tinto unmanned train collides with stopped train.

[BaltACD]

Euclid

Several news sources say this: 

 

"At this stage these [investigation questions] are focused on the operation of and adherence to signaling systems in the area."

 

Does anyone here know what that means?

 

There seems to be hardly anything in any of the news reporting on this wreck that conveys a clear picture of the points it intends to make. 

If you can't convince them of your brilliance - baffle them with bull and double talk.

I haven't been impressed with anything that has been posted to date.

[Euclid]

Several news sources say this: 

 

"At this stage these [investigation questions] are focused on the operation of and adherence to signaling systems in the area."

 

Does anyone here know what that means?

 

There seems to be hardly anything in any of the news reporting on this wreck that conveys a clear picture of the points it intends to make. 

[jeffhergert]

AN AUTONOMOUS TRAIN SHOULD NOT REQUIRE A COMMAND FROM ANY PERSON TO STOP SHORT OF ANY OBSTRUCTION. 

Jeff

[BaltACD]

Euclid

The problem was not an inability to stop an Autonomous Train.  The problem was a mistake in sending the stop command to the wrong train.  Presumably, the wrong train dutifully stopped (if it was moving).  But the sender of the command failed to realize his/her mistake in sending the vital command to the wrong train.

 

What the news report refers to as the “wrong train” was not the disabled train.  The news article says that it was this train:

 

“Rio Tinto train controllers initiated the ‘on-site’ feature [the stop command], [by] transmitting it to the autonomous train to the south of the 222 car disabled train.”

 

So that train was the “wrong train;” and it was not the disabled train. 

Simple!

What we have here is a failure to communicate!

I might state that the Cardinal Rule of Train Dispatching is KNOW WHO YOU ARE COMMUNICATING WITH!

[Euclid]

The problem was not an inability to stop an Autonomous Train.  The problem was a mistake in sending the stop command to the wrong train.  Presumably, the wrong train dutifully stopped (if it was moving).  But the sender of the command failed to realize his/her mistake in sending the vital command to the wrong train.

 

What the news report refers to as the “wrong train” was not the disabled train.  The news article says that it was this train:

 

“Rio Tinto train controllers initiated the ‘on-site’ feature [the stop command], [by] transmitting it to the autonomous train to the south of the 222 car disabled train.”

 

So that train was the “wrong train;” and it was not the disabled train. 

[blue streak 1]

One item about ACSES is that a mobile inductor can be placed on a track to stop any train.

[jeffhergert]

Positive Train Control, the Energy Management Systems used in the US use GPS as well as inputs from legacy signal systems.  An enhanced version ofNYAB/Knorr's LEADER system is part of Rio Tinto's automation package.  It would follow that Rio Tinto's trains use GPS, with or without inputs from signals or other wired in hardware. 

Unless they only run, and intend to run, one train at a time on a track, any autonomous system is going to need to know the locations of obstructions/occupancies and be able to react to them.  That is, if trains ahead are slowing down, a following train would also need to slow down.  If trains started to stop ahead, a following train would also need to stop.  If a train needs a command from a central office to stop because of traffic, or a command from local personnel, I wouldn't call it autonomous.

Jeff 

[tree68]

Euclid

They say the stop signal was sent to the wrong train.  If so, it would have stopped that wrong train.  If so, wouldn’t that sudden, un-programed stopping of that train have gotten a lot of attention from probably several people? 

If it was sent to a stopped train, who's to tell?

Euclid

...full-time system that would have all trains being able to detect another train that is fouling their route. 

I wonder if the rear end marker talks to the system (with GPS information) or if they just work with the known length of a given train.

[Euclid]

I was wondering about that.  I would think there would be some type of full-time system that would have all trains being able to detect another train that is fouling their route. 

 

I have read that they do have sensors on the locomotives that will detect any fouling at grade crossings.  But even with that, I don’t see how that alone could prevent crossing collisions since the sensor would be only line of sight, and that would not account for whether the train could stop in time to avoid a crossing collision.  And this problem would also have that limitation, and more so, if it was expected to stop a train short of another fouling train. 

 

I think doing that would require a track activated block system rather than just a sensor on the locomotive.  But if they do have such a block system, why not just rely on that to stop the loaded train before it reached the site of the stalled train?

 

They say the stop signal was sent to the wrong train.  If so, it would have stopped that wrong train.  If so, wouldn’t that sudden, un-programed stopping of that train have gotten a lot of attention from probably several people?  How long would it take to realize that correct train did not receive its intended stop signal?

 

One would think there would have been plenty of time to discover the error and stop the correct train.

 

But in thinking more about it, I suspect that although it was immediately clear that a stop signal had been sent to a train unintentionally.  It may be that nobody realized that another train had been intended to receive that stop signal, and did not receive it. 

[jeffhergert]

First, I would think a train would be stopped  before changing from autonomous mode to some kind of local/remote control. Stopping being done automatically in autonomous mode.  Then people on the scene making the change

Second, if the change command was done while the train was moving, but sent to the wrong train, the train still moving in autonomous mode should have still stopped itself.  The system should have still recognized that there was an obstruction, the disabled train, and the automation stopped the train short of the obstruction.  That the command was sent to the wrong train seems like trying to focus blame on humans rather then a possible flaw in their automation.

Jeff

[Euclid]

Overmod

Don't say "the plan failed" -- say WHY it failed.

According to the mining union, a command to transfer control of the loaded train from the autonomous system to local control was sent to "the wrong train" despite that command having to be checked by supervisory personnel before transmission.

This does not explain why the automated system on the loaded train (which would not have received an order for local control) did not stop properly short of the standing train.  Nor does it explain who the 'six workers' presumably attempting to get the stopped train to move, which the union hinted had to scramble to get away from the impending wreck in time to save themselves, were, and how they got where they were.

This is not going to make sense until an impartial agency reviews and reports on it.  It may not make sense even then.

 

When I said the plan failed I was referring to the plan to stop the approaching autonomous train.  I am comfortable in asserting that the plan failed because it is obvious.  In previous post on this page, I had already mentioned that the news reports the operators sent the stop signal/command to the wrong train. 

 

As to an explanation as to why the message was sent to the wrong train, or other details as to why the plan failed, I see no certain conclusion yet.  I agree that it does not prove what caused the collision.  But the article I posted does seem to conclude what it considers to be a definite cause for the wreck

 

That conclusion seems to be disputed between the Company and the Union.  In that dispute the Union seems to conclude that because this error was developed by more than one person, it cannot be called “human error.”

 

They say that the collision was not the result of the error of a single train controller.  Instead they say it was a systematic failure of Rio Tinto’s implementation of safety procedures surrounding automation.

 

In this sense, I think it boils down to whether the cause of the wreck is the fault of human failure or the fault of automation failure.

 

This is why I said in my previous post, “I will leave it to others to decide what that answer is.” 

 

Was the wreck caused by an employee or was it caused by management?

[BaltACD]

Overmod

Don't say "the plan failed" -- say WHY it failed.

According to the mining union, a command to transfer control of the loaded train from the autonomous system to local control was sent to "the wrong train" despite that command having to be checked by supervisory personnel before transmission.

This does not explain why the automated system on the loaded train (which would not have received an order for local control) did not stop properly short of the standing train.  Nor does it explain who the 'six workers' presumably attempting to get the stopped train to move, which the union hinted had to scramble to get away from the impending wreck in time to save themselves, were, and how they got where they were.

This is not going to make sense until an impartial agency reviews and reports on it.  It may not make sense even then.

So far the 'pronouncements' of the local authorities have paralleled those about the MV Dali destroying the FSK Bridge in Baltimore.

Bad things happened and the wreck then happened.  No further info as to WHY the bad things happened, which are the heart of the causes.  In the case of the MV Dali we have been told circuit breakers tripped - we have not been told what it was the CAUSED the circuit breakers to trip.

[Overmod]

Don't say "the plan failed" -- say WHY it failed.

According to the mining union, a command to transfer control of the loaded train from the autonomous system to local control was sent to "the wrong train" despite that command having to be checked by supervisory personnel before transmission.

This does not explain why the automated system on the loaded train (which would not have received an order for local control) did not stop properly short of the standing train.  Nor does it explain who the 'six workers' presumably attempting to get the stopped train to move, which the union hinted had to scramble to get away from the impending wreck in time to save themselves, were, and how they got where they were.

This is not going to make sense until an impartial agency reviews and reports on it.  It may not make sense even then.

[Euclid]

New information from yesterday, 5/20/24

 

In my previous post, I asked if the failure to send the command to stop the loaded ore train that was approaching the site of the stalled train was due to a technical malfunction of the automatic train control system, or was the result of a human operator sending the stop command to the wrong train by mistake. 

 

I may be mistaken, but it is my understanding from news reports that there was a train that had been stopped on the mainline due to some type of malfunction, and it was being approached by a loaded ore train running by autonomous operation, apparently at normal track speed.

 

The intention was to use the approaching train to move the stopped train into the clear because it was not able to move on its own.  So there was an intention to stop the approaching train once it was near the stopped train.  Then the operable train would move the inoperable train.

 

The intention to stop the approaching train was to be executed by using the autonomous operating system to send a stop signal/command to the approaching train by using a manual control actuated by a human operator, to send the stop signal to the approaching train, to be executed by the autonomous system that operated the train. 

 

This plan failed and the approaching train thus maintained normal clear track speed rather than slowing to stop short of the stopped train that was to be assisted to move.  So the approaching train then collided with the stopped train.

 

So why did the plan to stop the train fail? Was it because the human operator failed to manually operate some type of manual switch/control to select and execute the stop command?  ---Or was it because the operator made the proper manual execution, but the automatic/electronic/digital system itself malfunctioned, and thus failed to execute the operator’s proper execution?

 

This above news report seems to answer that question, but I will leave it to others to decide what that answer is.

 

The latest report from yesterday:

 

https://www.amsj.com.au/mining-giants-safety-systems-catastrophically-failed-says-industry/

[Euclid]

Euclid

 

 

Overmod

I've been boycotting this thread (and will probably go back to boycotting this thread) to help keep my blood pressure in a sensible range.

For the last time: the empty train hit the loaded one, not the other way around.  Is that a clear enough statement for Ron to finally, finally get it?  THE EMPTY TRAIN IS THE ONE THAT WAS DISPATCHED.  THE LOADED TRAIN WAS THE ONE THAT STALLED AND NEEDED HELP.  THE EMPTY TRAIN HIT IT.

It seems that multiple people have told him this, and he keeps bouncing back 'confirming' what they said by repeating the same inanity about the loaded train hitting an empty one.   

 

 

 

 

It has been reported both ways. 

 

Here is what Rio Tinto says:

 

“Rio Tinto (ASX, LON: RIO) said on Monday that one of its fully-loaded autonomous iron ore trains had crashed with a set of stationary wagons in Western Australia’s Pilbara region.

 

It was the fully loaded train that approached at speed and collided with the stationary train that needed to be recovered.”

 

 

Here is another news report:

 

“A laden Rio Tinto train has come off the tracks in the Pilbara after smashing into stationary wagons, the mining giant’s third driverless train to derail in the Pilbara within the space of a year.”

 

 

The plot thickens:

 

Earlier here, I speculated that stop signal/command sent to the loaded train that was approaching the site of the stopped, empty train, was not sent in time by a human operator from the remote location of control origination.  

 

Now apparently we learn that the stop signal/command was sent in time to stop the approaching loaded train, but the stop signal/command was sent to the wrong train.

 

That raises this question:  Did the stop signal go to the wrong train because it was mistakenly sent there by a human operator?  Or did the signal go to the wrong train due to some technical glitch in the control system? 

 

 

https://www.trains.com/trn/news-reviews/news-wire/australian-mining-railroad-reopens-after-autonomous-train-collision/

[Euclid]

Overmod

I've been boycotting this thread (and will probably go back to boycotting this thread) to help keep my blood pressure in a sensible range.

For the last time: the empty train hit the loaded one, not the other way around.  Is that a clear enough statement for Ron to finally, finally get it?  THE EMPTY TRAIN IS THE ONE THAT WAS DISPATCHED.  THE LOADED TRAIN WAS THE ONE THAT STALLED AND NEEDED HELP.  THE EMPTY TRAIN HIT IT.

It seems that multiple people have told him this, and he keeps bouncing back 'confirming' what they said by repeating the same inanity about the loaded train hitting an empty one.   

 

It has been reported both ways. 

 

Here is what Rio Tinto says:

 

“Rio Tinto (ASX, LON: RIO) said on Monday that one of its fully-loaded autonomous iron ore trains had crashed with a set of stationary wagons in Western Australia’s Pilbara region.

 

It was the fully loaded train that approached at speed and collided with the stationary train that needed to be recovered.”

 

 

Here is another news report:

 

“A laden Rio Tinto train has come off the tracks in the Pilbara after smashing into stationary wagons, the mining giant’s third driverless train to derail in the Pilbara within the space of a year.”

[jeffhergert]

Pure speculation, but it's so fun to speculate.

We seem to focus that the moving train somehow lost communication in some manner.  What if it was the disabled train that got lost?

I don't know if Rio Tinto's lines are signaled, or if any signal system has been discontinued.  There's a lot of proponents out their who propose to eliminate signals entirely and go to a CBTC (Communications Based Train Control) system.  (Some have said that the signal departments only want signals to keep their jobs.  Sems that also could apply to those proponents so they then have work.)

In CBTC, the main computer at the control center will receive inputs from everything affecting movement: trains, MOW, switch positions, etc.  The main computer will transmit to each train's onboard computer how far it can go until it has to stop.  A specific train won't necessarily know the location of other trains, just that it has to stop at milepost X.  What if the disabled train stopped communicating to the main office?  Then the main computer no longer knows there is an obstruction (disabled train), so then gives instruction to the moving train to continue to a point beyond where the disabled train is.  The human dispatcher (Probably not the correct term they use, but is what most of us would be familiar with.), for whatever reason, doesn't realize in time that a train is missing.  The realization is made with only enough time to warn those working on or around the disabled train.

Just some (junk) food for thought.

Jeff 

[Overmod]

I've been boycotting this thread (and will probably go back to boycotting this thread) to help keep my blood pressure in a sensible range.

For the last time: the empty train hit the loaded one, not the other way around.  Is that a clear enough statement for Ron to finally, finally get it?  THE EMPTY TRAIN IS THE ONE THAT WAS DISPATCHED.  THE LOADED TRAIN WAS THE ONE THAT STALLED AND NEEDED HELP.  THE EMPTY TRAIN HIT IT.

It seems that multiple people have told him this, and he keeps bouncing back 'confirming' what they said by repeating the same inanity about the loaded train hitting an empty one.  

We need to wait for the report to hear precisely how the system failed.  (Or whether there were 'employees working on the stalled train' that got early radio warning that a collision was coming...)

It was not a simple failure: the speed of impact was obviously substantial, locomotives first, so any sensors or any systems that consider a point on the locomotive to be the 'front' of a train were obviously not doing their collective job.  I have seen no evidence at all that the empty train was under RCO guidance of the 'usual' kind (which is I believe technically limited to short range in the United States by specific intent) and it might be well to review what the response of a RCO system to any loss of connectivity is (as it specifically has been designed to relate to what may be a very heavy consist being moved at the time). 

[blue streak 1]

Suspect that no  matter who hit whom that procedures for this kind of problem will be changed significantly.  Surely this is not the first instance and what failed this time?

[BaltACD]

zugmann

 

BaltACD

  Two COMPLETE trains are difficult to handle in the confined quarters necessitated by their coupling. 

I mean, we do it in yards all the time.   

This soudns like a failure of the automation system at first glance?  If it was, it will be buried quickly. 

I haven't seen any yards where you are handling 30K, 40K tons or more downgrade to a coupling.

Suspect it was a failure of automation caused by incompetence of design and implemention of failed logic.

[Euclid]

Actually even though the news says the plan was use the loaded train to move the empty train, they may not have meant that they intended to use the whole loaded train with engines and cars  The intention may have been to just use the power from the loaded train to move the whole empty train. It would not be a leap to say that the loaded train was going to rescue the stranded train even though they meant that they would cut off the cars and only use the power of the loaded train. It is just that it was the “loaded train” that was coming to the rescue.

 

But they also said that the loaded train was operating in autonomous mode, which would offer no means of slowing the train down to tie onto the stalled train. But here again, they may not have meant there was an intention to stay in autonomous mode when taking on the role of helper power. 

 

So they would either stop the approaching loaded train short of the stalled train by remote control, or let the automatic block system stop the train.  Next they would have had one or more of those 5-7 workers at the stalled train board the loaded train after it was stopped and taken out the autonomous mode.  Then the human operator would come ahead with the loaded train and tie onto the stalled train.  So there was a failure to stop the loaded train that was operating in the autonomous mode.  The failure may have been in the autonomous system, or it may have been in the block signal system----or it may have been human error in failing to initiate a remote control stop command soon enough to get the loaded train stopped in time. 

 

But even if such a human error occurred, I would think the automatic block system would a have stopped the loaded train short of the fouling empty train ahead.  The news does sound like a failure of the block system is where they are focusing to find the cause of the collision. 

[zugmann]

BaltACD

  Two COMPLETE trains are difficult to handle in the confined quarters necessitated by their coupling.

I mean, we do it in yards all the time.   

This soudns like a failure of the automation system at first glance?  If it was, it will be buried quickly. 

[SD70Dude]

BaltACD

This accident is why you don't have complete trains assist each other - no matter if the trains are manned or autonomous.  Two COMPLETE trains are difficult to handle in the confined quarters necessitated by their coupling.

I beg to differ, provided that you have a good engineer and the combined power is enough to move both trains. 

This happens sometimes on one of our hills in the mountains west of Jasper.  Why make the conductor tie down your train and clog things up for even longer when you can keep it and still push the stalled guy ahead of you?

https://www.youtube.com/watch?v=VuOl5kvxZxs&t=384s

[Paul of Covington]

   Actually, if I understand what they intended to do, it might have worked.  They just made the coupling a little too fast.

   I assume they don't use EOT devices.

[Paul of Covington]

      I think we need to go back to the story in the original post:

 

"The Office of the National Rail Safety Regulator (ONRSR) said it received a report about a collision between a loaded ore train and a recovery train.

"The recovery train is reported to have collided with the ore train it was sent to recover after it was disabled by a mechanical failure," a spokesman said in a statement."

 

       "loaded ore train"   and   "recovery train"

[BaltACD]

Euclid

Here is what I think happened: 

The loaded train was some considerable distance from the empty train when the empty train malfunctioned and stopped.  They had workers onsite after the empty train malfunction. 

 

So, intending to use the loaded train to move the empty train, they allowed the loaded train to approach at track speed because there was no reason to slow it down until it got close. 

 

I assume they have some type of block system that either includes wayside signals or at least keeps trains properly separated for the autonomous operation.

 

So the plan was to let the loaded train move in to the area and be stopped by the block system due to the presence of the stationary malfunctioning train.  Then to move it ahead to couple to the stopped train, they would have put a human operator onto the loaded train to bring it ahead and make the joint.

 

For some reason, the automatic block system failed to perform its function, so the train headed right into the stalled train at track speed.

This accident is why you don't have complete trains assist each other - no matter if the trains are manned or autonomous.  Two COMPLETE trains are difficult to handle in the confined quarters necessitated by their coupling.

[Euclid]

Here is what I think happened:

 

The loaded train was some considerable distance from the empty train when the empty train malfunctioned and stopped.  They had workers onsite after the empty train malfunction. 

 

So, intending to use the loaded train to move the empty train, they allowed the loaded train to approach at track speed because there was no reason to slow it down until it got close. 

 

I assume they have some type of block system that either includes wayside signals or at least keeps trains properly separated for the autonomous operation.

 

So the plan was to let the loaded train move in to the area and be stopped by the block system due to the presence of the stationary malfunctioning train.  Then to move it ahead to couple to the stopped train, they would have put a human operator onto the loaded train to bring it ahead and make the joint.

 

For some reason, the automatic block system failed to perform its function, so the train headed right into the stalled train at track speed.

[BaltACD]

If the empty train was the 'stalled' train - it would have been on an ascending grade area - otherwise it could have used gravity to drift down a descending grade.

The Rio Tinto loaded trains are normally 30K, 40K or 50K tons?  If the stalled empty train is on a ascending grade - that means the loaded train coming to rescue it is thereby on a DESCENDING grade with the load of its train making it even harder to control.

DUMB DUMB DUMB

My experience (pre PSR, pre DPU) has no relation to Rio Tinto.  In the territories I have worked - loaded and empty coal trains normally had the same power complement.  Two units for each train.  Normally engine failures happened on LOADED trains and the fix was to have an empty train give up one of their units at a meeting point.  If a empty train had an issue with one engine, then other engine would be sufficient to take the train to destination.

I have no idea how Rio Tinto operates their autonomous operations.  I do know that trying to safely handle tonnage moving downgrade is a difficult undertaking.

Grades are pesky things - both up and down.

[Euclid]

The loaded train that was following the empty train was intended to rescue the empty train which was standing still. The loaded train was in autonomous mode when it struck the stationary empty train. 

 

News reports that because the loaded train was in autonomous mode, there was nobody onboard to see the collision course with the empty, stationary train. 

 

So the plan was for the loaded train to approach the stationary, empty train while reducing speed to the point of being able to successfully couple with the stationary train. 

 

Here is my question:

 

Was it possible for this slowing approach of the loaded train to be controlled by the autonomous mode?

 

If not, why would they attempt such a move?  If they intended to make this move for slowing and coupling to the empty train, how could they have operated the loaded train during that move?