Longer Trains Cause More Derailments

[tree68]

Euclid

STAGE ONE:  Perform testing to gather data of in-train forces.  

STAGE TWO:  Build a model of those forces showing every detail of their potential.   

STAGE THREE:  Write a program to evaluate train make-up and all of the in-train forces that it can produce.

As Balt points out, and to my knowledge - this has already been done.  Hence the existing rules for train make-up.

The prime reason for block-swapping is to prevent handling a given car at every single terminal.  This is why humps are disappearing.  Not to mention that terminal dwell is a major productivity killer.

Keeping entire 200 car trains in the "proper order" would bring the railroad to a halt.  

Keep in mind, too, that a 50' boxcar loaded with balsa may well weigh less than an empty LP gas car (partly conjecture on my part, but...).

And even if you do get the train in the "proper order," a dragging brake changes the whole dynamic...

[BaltACD]

Euclid

 

BaltACD 

oltmannd 

BaltACD 

oltmannd 

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment.  

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen.  

Euc wants all that modeling done for each train that departs its originating terminal - for each and every train, every trip.  Probably ending in discipline for somebody if the train derails for any reason anywhere. 

Yup.  And what are you going to use for train handling inputs?  You put that exact train out on the road for 10,000 days in a row and you'll have 10,000 unique sets of train handling data (brake, throttle, DB, etc.) 

Which is a concept Euc can't or won't comprehend.

My favorite interest is automobile racing - virtually all forms from F1 through IndyCar, NASCAR and on down to Club racing and the local bull ring dirt tracks.  In many cases there are 'spec classes' where the intent of the specifications are to make the cars as identical as humanly possible to that the driver is the variable to the success or failure in any competition.  The drivers is the 'master sensor' and is the one that inputs throttle, brake and steering inputs to make the car go - good drivers are faster than less than good driver.  I won't hazard any guess to what a 'computer' driver's performance would be.

I have not been on any trains that are being operated with the various locomotive operation computer application, as such I am not in any position to comment on how well those programs peform, what their sensory inputs are and how they react to those sensory inputs.  Jeff Hergert would be a much better source on their operation and quirks. 

I do not understand why you and Don assume that the “modeling” I refer to will be done for every train before it is allowed to depart from its terminal.

 

My reference to “modeling” refers only to the second stage of a 3-stage process as follows:

 

STAGE ONE:  Perform testing to gather data of in-train forces.

 

STAGE TWO:  Build a model of those forces showing every detail of their potential. 

 

STAGE THREE:  Write a program to evaluate train make-up and all of the in-train forces that it can produce.

 

 

After the data is gathered by testing to learn the entire scope of those forces, computer models of those forces are built.  The final model will show every detail of the buildup and dissipation of those forces as they move through the train car by car. 

 

This would be a big task costing many millions of dollars.  It has not been done yet.  It would take considerable time.  It will be revised occasionally to account for new types of rolling stock.

 

The testing to gather data alone will also be a lot of work, yet I expect the data might show things that have not been known.  It might show force multiplication that is sometimes greater than has been determined by practice.    

 

And I don't think that everything developed for the model will have to come from actual measuring of the in-train forces. There can be a mix of approaches with the model making.  For one thing, they could model a 250-car train to learn about the in-train foreces.   

 

I think this is what the FRA has in mind.

 

I understand that the only similar project that has been done is extensive modeling of tank cars following the Lac Megantic wreck.  The purpose was to study how tank cars collide in high speed derailments in order to build them with more crashworthiness. 

 

In any case, the modeling for in-train forces would be an enormous task.  The cost will certainly be objectionable, as will the prospect of the resulting knowledge that might lead to mandates limiting train length. 

 

HOWEVER--  Once the modeling has been done, the occurrences of all in-train forces will be understood.  At present, they are not fully understood.  But when they are understood, a program will be written.  It is the program that will be used to evaluate every train to determine whether it is safe to run.  So the “modeling” will be done for the purpose of writing the program.  Then the program will evaluate each train to show all of the potential in-train forces and warn against operation of some the force potential is high enough to show a derailment hazard.

 

So you put the train car-list into the program, and it gives you the answer.  I don’t know how long that would take, but maybe only seconds.  So the data gathering, modeling, and developing a program is a massive effort needed to build the system, but it certainly does NOT have to be done for each train.

 

All three stages of development will have been completed before the working system will be used in actual service of evaluateing trains for excess force potential.  

You continually demonstrate you have no idea how trains come into being or why they end up in the form they do.

Railroads service customers - customers get and release cars - cars that customers get or release can be in one's and two's or twenty's and thirty's or even more.  Some trains are fully built at origin, many more are not.  Those that aren't fully built will be involved in pick ups and/or set offs as the train moves from its origin to its destination.  Such a train might depart origin with light power and end up with 200 or more cars when it gets to its destination - or it may arrive destination with light power after being 200 or more cars for periods during its trek.

A function of PSR among most of the carriers is a reduction in the number of yards and expanding Yardmaster responsibilities to cover multiple yards.  Another function of PSR  is the act of 'block swapping' between trains that have diffeent origins and different destinations at some point where both trains share their routes.  The block can be 1 car or 150 cars and it will change on a daily basis.

One thing modeling has done from my experience up until the time I retired was to create Train Handling Rules and Timetable Special Instructions concerning permissible train make up.  Some of the simple rulings are 1. blocks of 30 more more empties can have no more that 10 loads in whatever cars follow the empty block.  2. long (longer than 60 foot) empty cars are restricted to how much tonnage can trail behind them.  3. placement restrictions between various types of HAZMAT. 4. Placement restrictions between shiftable loads and HAZMAT.  There are probably a dozen or more other placement restrictions that will crop up from time to time - like clearance implicated cars being handled within five cars of the locomotives.  

[Euclid]

BaltACD

 

 

oltmannd

 

BaltACD 

oltmannd 

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment.  

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen.  

Euc wants all that modeling done for each train that departs its originating terminal - for each and every train, every trip.  Probably ending in discipline for somebody if the train derails for any reason anywhere. 

Yup.  And what are you going to use for train handling inputs?  You put that exact train out on the road for 10,000 days in a row and you'll have 10,000 unique sets of train handling data (brake, throttle, DB, etc.)

 

 

Which is a concept Euc can't or won't comprehend.

My favorite interest is automobile racing - virtually all forms from F1 through IndyCar, NASCAR and on down to Club racing and the local bull ring dirt tracks.  In many cases there are 'spec classes' where the intent of the specifications are to make the cars as identical as humanly possible to that the driver is the variable to the success or failure in any competition.  The drivers is the 'master sensor' and is the one that inputs throttle, brake and steering inputs to make the car go - good drivers are faster than less than good driver.  I won't hazard any guess to what a 'computer' driver's performance would be.

I have not been on any trains that are being operated with the various locomotive operation computer application, as such I am not in any position to comment on how well those programs peform, what their sensory inputs are and how they react to those sensory inputs.  Jeff Hergert would be a much better source on their operation and quirks.

 

 

I do not understand why you and Don assume that the “modeling” I refer to will be done for every train before it is allowed to depart from its terminal.

 

My reference to “modeling” refers only to the second stage of a 3-stage process as follows:

 

STAGE ONE:  Perform testing to gather data of in-train forces.

 

STAGE TWO:  Build a model of those forces showing every detail of their potential. 

 

STAGE THREE:  Write a program to evaluate train make-up and all of the in-train forces that it can produce.

 

 

After the data is gathered by testing to learn the entire scope of those forces, computer models of those forces are built.  The final model will show every detail of the buildup and dissipation of those forces as they move through the train car by car. 

 

This would be a big task costing many millions of dollars.  It has not been done yet.  It would take considerable time.  It will be revised occasionally to account for new types of rolling stock.

 

The testing to gather data alone will also be a lot of work, yet I expect the data might show things that have not been known.  It might show force multiplication that is sometimes greater than has been determined by practice.    

 

And I don't think that everything developed for the model will have to come from actual measuring of the in-train forces. There can be a mix of approaches with the model making.  For one thing, they could model a 250-car train to learn about the in-train foreces.   

 

I think this is what the FRA has in mind.

 

I understand that the only similar project that has been done is extensive modeling of tank cars following the Lac Megantic wreck.  The purpose was to study how tank cars collide in high speed derailments in order to build them with more crashworthiness. 

 

In any case, the modeling for in-train forces would be an enormous task.  The cost will certainly be objectionable, as will the prospect of the resulting knowledge that might lead to mandates limiting train length. 

 

HOWEVER--  Once the modeling has been done, the occurrences of all in-train forces will be understood.  At present, they are not fully understood.  But when they are understood, a program will be written.  It is the program that will be used to evaluate every train to determine whether it is safe to run.  So the “modeling” will be done for the purpose of writing the program.  Then the program will evaluate each train to show all of the potential in-train forces and warn against operation of some the force potential is high enough to show a derailment hazard.

 

So you put the train car-list into the program, and it gives you the answer.  I don’t know how long that would take, but maybe only seconds.  So the data gathering, modeling, and developing a program is a massive effort needed to build the system, but it certainly does NOT have to be done for each train.

 

All three stages of development will have been completed before the working system will be used in actual service of evaluateing trains for excess force potential.  

[BaltACD]

oltmannd

 

BaltACD 

oltmannd 

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment.  

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen.  

Euc wants all that modeling done for each train that departs its originating terminal - for each and every train, every trip.  Probably ending in discipline for somebody if the train derails for any reason anywhere. 

Yup.  And what are you going to use for train handling inputs?  You put that exact train out on the road for 10,000 days in a row and you'll have 10,000 unique sets of train handling data (brake, throttle, DB, etc.)

Which is a concept Euc can't or won't comprehend.

My favorite interest is automobile racing - virtually all forms from F1 through IndyCar, NASCAR and on down to Club racing and the local bull ring dirt tracks.  In many cases there are 'spec classes' where the intent of the specifications are to make the cars as identical as humanly possible to that the driver is the variable to the success or failure in any competition.  The drivers is the 'master sensor' and is the one that inputs throttle, brake and steering inputs to make the car go - good drivers are faster than less than good driver.  I won't hazard any guess to what a 'computer' driver's performance would be.

I have not been on any trains that are being operated with the various locomotive operation computer application, as such I am not in any position to comment on how well those programs peform, what their sensory inputs are and how they react to those sensory inputs.  Jeff Hergert would be a much better source on their operation and quirks.

 

[tree68]

It's those darned shades of gray, again...

[oltmannd]

BaltACD

 

 

oltmannd

 

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment.  

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen. 

 

 

Euc wants all that modeling done for each train that departs its originating terminal - for each and every train, every trip.  Probably ending in discipline for somebody if the train derails for any reason anywhere.

 

 

Yup.  And what are you going to use for train handling inputs?  You put that exact train out on the road for 10,000 days in a row and you'll have 10,000 unique sets of train handling data (brake, throttle, DB, etc.)

[BaltACD]

oltmannd

 

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment.  

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen. 

Euc wants all that modeling done for each train that departs its originating terminal - for each and every train, every trip.  Probably ending in discipline for somebody if the train derails for any reason anywhere.

[oltmannd]

Euclid

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment. 

 

Validation work had long been done.  Models work very well.

You can't practically model every train for every operational scenario before you dispatch. Train forces are the results of train handling which depend largely on what's happening where on the railroad plus unintended stuff.

You gonna model the train going into emergency every 50 feet?  For a separated air hose at every one of 200 locations on the train?  Or even every combination of DB and service brake application for every spot the train might have to slow down?  

Not happening.  Can't happen. 

[Euclid]

oltmannd

 

 

Euclid

 

 

BaltACD

 

 

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

...

 

 

Your lack of knowledge of railroad operations and supervision is SHOUTING.

 

 

 

 

Well if you have the knowledge; show me the use of computers to model the in-train forces in a 250-car train with suitable distributed power and best train handling practice in moderately hilly terrain, at typical road speeds.

 

Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring? 

 

 

 

 

They do have modeling tools.  They do use them to set rules for train make-up and territory. 

They are often used retrospectively to analyze derailments. 

They are generally difficult to use and tough to deploy as tools for train consist management.  Car geometry plays a big role as does ROW geometry and in-train forces.

 

Regarding my comment on the previous page:

“Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring?” 

 

++++++++++++++++++++++++++++++++++

 

This issue is not just a matter of having computers and doing modeling.  Everybody is doing that every day.  The point is that the computer modeling needed for solving the problem of excess in-train forces has got to be directed at that problem.  So the first step is to find examples of that problem by measuring in-train forces.  Then use the most applicable computer analysis to see if it verifies that same problem.  If it does not verify the problem, then try to figure out why a real problem is not detected by the computer program.  In other words, you need to test real trains to learn how to build the program that can flag that problem before it causes a derailment. 

 

I do understand that the derailments keep happening for good reasons.  But my question goes to why those reasons have not been acted on to eliminate them.  When I suggested that a new wave of testing is needed to find and eliminate causes of excess in-train forces that are derailing ultra-long trains, I am told that such testing has already been done.  Did the testing not find a problem? If it did not find a problem, does that mean there is no problem, or does it mean that the problem is in the software failing to find the problem that is derailing trains?  If so, does that mean there is no problem? 

 

If there is no problem, why are these derailments occurring???

 

The FRA believes there is a problem that is causing these derailments, and they believe the problem is related to ultra-long consists, issues with DP, and operator technique related to braking.  They say the problem is in-train forces that are too high.  They also say these forces are extremely complex to understand, predict.  Sometimes they are caused by train make-up, and other times they are caused by defective train handling technique. 

 

The FRA says that locomotive engineers are not able to diagnose the potential of these high in-train forces because they have no training in the matter.  And rules/training have not been provided because the industry itself does not understand the train dynamics that are causing the problem. 

 

The FRA also says that the problem has not been diagnosed because adequate testing has never been done, so they say they are starting to conduct such testing.

 

There is no point in developing new train handling techniques and rules based on theories and speculation about the behavior of excess in-train forces.  What is needed is actual proof of the behavior and its causes.  For that, new testing is required for what the FRA refers to as a very complex behavior.  That is why the FRA and I say new, comprehensive testing is required. 

 

I suspect that the state of PSR railroad management will not welcome the comprehensive testing because it might lead to restrictions on their long trains.  But if this apparent relationship between long trains and increased derailments continues to develop, that road leads to government mandates.  So I would think the industry should welcome the new testing to actually solve the problem rather than insist there is no problem. 

 

The issues with train make-up are also part of the problem, and for that issue, software has been developed.  But again, the question comes down to whether cause of the make-up problem has been adequately diagnosed, or whether it relies on theories.  Is there 100% agreement on what this software advisees for train make-up?  Do all railroads use the same program?  Does the program provide consistent answers, or does it is sometimes offer choices?  If these programs do the job except for factoring in rail car geometry differences; then factor those differences into the program so it can do its job.

 

Train simulators give feedback as training for operator techniques.  They are programed for the accepted assumptions based on empirical evidence from actual train handling.  But this problem of excess in-train forces is new and not fully understood.  So I doubt that existing simulators are a useful tool to unravel the new mysteries of ultra-complex in-train dynamics of ultra-long trains.

 

With all of this uncertainty, rather than using computer programs to look for problems that might occur,  I think it makes more sense to look for problems by testing real world trains for buff force that is great enough to be of concern.  Then work backwards from there, and see if the computer program can also predict that same buff force condition and show the circumstances that bring it about.  Otherwise you are just writing computer programs to look for what is already expected by speculation and past derailments. 

[M636C]

I was involved in investigations to produce computer models of train handling on the then Hamersley Iron and Mt Newman Mining railways, now Rio Tinto and BHP respectively. This was between 1976 and 1978. I believe this work was used in the work leading up to Rio Tinto's present remote controlled trains.

Back then the electronics required were quite big and clumsy. We could only instrument a pair of cars, which we moved to different locations in the train.

Of course these trains of 144 to 200 cars were always either completely empty of fully loaded, and the cars were effectively identical in size and weight.

But the tests were carried out and the models developed. These days much more could be done at a much lower cost. We actually had a black and white video camera and a reel to reel video recorder which may have been powered by car batteries, just to give an idea of what was available.

On the Hamersley trains the cars at each end were always the newest cars available, so that the trailing cars were always fitted with the best air brake valves to reduce the chances of dragging brakes on the trailing cars.These cars were all pairs with only a single air brake valve for the two cars, which substantially reduced the propagation time of the air brake signals.

200 cars with individual brake valves would be subject to greater delays in application and release of the brakes.

Of course, all these trains now have ECP brakes, whether the trains are remotely controlled or conventionally driven. The significant reduction in in train forces allowed by ECP brakes is regarded as a neccesity for remotely operated trains.

Peter

[tree68]

oltmannd

Or, why haven't these derailments stopped happening. 

As I noted earlier, there's rules, and there's following them.  

Too, consider the problem of combining proper placement with blocking.  

[oltmannd]

Euclid

 

 

BaltACD

 

 

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

...

 

 

Your lack of knowledge of railroad operations and supervision is SHOUTING.

 

 

 

 

Well if you have the knowledge; show me the use of computers to model the in-train forces in a 250-car train with suitable distributed power and best train handling practice in moderately hilly terrain, at typical road speeds.

 

Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring? 

 

Or, why haven't these derailments stopped happening. 

RRs have, for a very long time, built trains the would derail if they went into emergency in just the right spot. Once in a blue moon, they would and did!

[oltmannd]

Euclid

 

 

BaltACD

 

 

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

...

 

 

Your lack of knowledge of railroad operations and supervision is SHOUTING.

 

 

 

 

Well if you have the knowledge; show me the use of computers to model the in-train forces in a 250-car train with suitable distributed power and best train handling practice in moderately hilly terrain, at typical road speeds.

 

Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring? 

 

They do have modeling tools.  They do use them to set rules for train make-up and territory. 

They are often used retrospectively to analyze derailments. 

They are generally difficult to use and tough to deploy as tools for train consist management.  Car geometry plays a big role as does ROW geometry and in-train forces.

[zugmann]

tree68

There's rules, then there's compliance with the rules...

It's not like a RFE has to pull physical tapes anymore.  All the stuff has instant triggers, instant downloads, instant reviews. 

[BaltACD]

tree68

There's rules, then there's compliance with the rules...

And there is Supervision to ensure compliance.

Without proper Supervision, you don't have much.  Without a rules compliant culture you have even less.

[tree68]

There's rules, then there's compliance with the rules...

[BaltACD]

Euclid

 

BaltACD 

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation. 

... 

Your lack of knowledge of railroad operations and supervision is SHOUTING. 

Well if you have the knowledge; show me the use of computers to model the in-train forces in a 250-car train with suitable distributed power and best train handling practice in moderately hilly terrain, at typical road speeds.

 

Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring?  

All I can say is that such modeling was taking place when I was still employed.  That being said I have been retired since December 2016 and subsequently PSR attacked my former employer in the form of a dying EHH and virtually all Operating Management was replaced by EHH cronies - with all that being the case I have no idea what form of modeling is taking place.

I do know that when I was working the CSX Train Handling Rule and divisional employee timetable special instructions had a number of restrictions on how to build trains and how much trailing tonnage was allowed before a helper or DPU was required.

Personal observations from some local incidents AFTER EHH and PSR were thrust into operations, indicated to me, that some of the restrictions that were in effect when I was working did not appear to be in effect from the incidents (derailments) that happened.

[Euclid]

BaltACD

 

 

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

...

 

 

Your lack of knowledge of railroad operations and supervision is SHOUTING.

 

Well if you have the knowledge; show me the use of computers to model the in-train forces in a 250-car train with suitable distributed power and best train handling practice in moderately hilly terrain, at typical road speeds.

 

Also, if the industry has this capability, why aren’t they using it to predict derailments caused by in-train forces?   If they are doing that, why are these derailments occurring? 

[BaltACD]

Euclid

... 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

...

Your lack of knowledge of railroad operations and supervision is SHOUTING.

[Euclid]

Overmod

 

 

Euclid

I suggest a means to actually measure buff and draft force on each coupler joint in a test train that duplicates an actual revenue train.

 

 

Probably be easier to equip them with magnetic or clamp-on battery-powered accelerometers, and store the results with GPS location and timestamp, then stream all the results wirelessly.  Assign an IoT address to each one to keep them definitively apart.

 

Where the effort ought to be placed is better control of the 'fence' activity in Locotrol so that any trailing power modulates its dynamic correctly for the part of its train between nodes.  Without that, any long train on an irregular profile might indeed be another Springfield cocked and unlocked...

You'd use exactly the same set of accelerometers, feeding into one of the train management programs, to figure out what that ought to be, and what permanent methods should be implemented (probably as an AAR standard) to deal with it.

 

I agree that what is needed is an easy, quick way to equip a long test train with sensors that would record drawbar forces in relation to variable conditions such as train length, train makeup, track grades and curves, train speed, brake operation, and throttle settings.  I don’t think it would be practical to attempt to simulate all of the combinations of conflicting in-train forces that need to be evaluated.  If they have all of that gathered and evaluated now, as some may claim, then why are they not finding that trains are destined to derail?

 

I doubt the railroads would program simulators that show that derailments can be caused by excess buff forces in their ultra-long trains.  But I suspect that if they had real world data, it would show that in-train forces associated with these trains will derail them under certain conditions.  Indeed, that is the conclusion of the FRA and unions.  And there is mounting empirical evidence of the correlation.

 

Not only is real world testing needed, but it also needs to be done by independent technicians and experts.  It has to be done with real world trains, and it has to be presented in terms that are objective enough that the results cannot be spun. 

 

The real world is showing strong evidence that ultra-long trains are developing in-train forces that are causing derailments.  It has to be determined whether these accidents are being caused by operator error in train handling or by erroneous conclusions about the risks resulting from excess in-train forces. 

 

If the cause is operator error, new training is necessary.  If the cause is excess in-train forces, a combination of remedies must be called for and mandated if necessary.   These remedies would include limiting train length, the use of train make-up programs, and ECP brakes. 

[BaltACD]

tree68

It's probably already been done, virtually - ie, on a computer.

Acceleration, deceleration, and buff forces can easily be simulated, and I'm pretty sure most profiles have been recorded as well.

And probably for less money than it would take to equip some 200 cars.

In the early 1990's CSX had a Simulator for Engineers that married the calculation of in train forces by using the data of actual trains that had been operated and the ability to operate that train of data over any of the territory CSX was operating.

Of course in the 1990's Distributed Power was not on CSX's horizon.

I am certain that in the 21st Century, the CSX Simulator does replicate that various operating modes with DPU in one or more locations within specified trains; with those location(s) able to be manipulated on instructor demand.

In my use of the simulator what was missing was the actual physical impacts from slack action - the buff and draft kicks in the pants.

[tree68]

It's probably already been done, virtually - ie, on a computer.

Acceleration, deceleration, and buff forces can easily be simulated, and I'm pretty sure most profiles have been recorded as well.

And probably for less money than it would take to equip some 200 cars.

 

[Overmod]

Euclid

I suggest a means to actually measure buff and draft force on each coupler joint in a test train that duplicates an actual revenue train.

Probably be easier to equip them with magnetic or clamp-on battery-powered accelerometers, and store the results with GPS location and timestamp, then stream all the results wirelessly.  Assign an IoT address to each one to keep them definitively apart.

Where the effort ought to be placed is better control of the 'fence' activity in Locotrol so that any trailing power modulates its dynamic correctly for the part of its train between nodes.  Without that, any long train on an irregular profile might indeed be another Springfield cocked and unlocked...

You'd use exactly the same set of accelerometers, feeding into one of the train management programs, to figure out what that ought to be, and what permanent methods should be implemented (probably as an AAR standard) to deal with it.

[Euclid]

With the current trend of ultra-long trains, I understand that the preferred method of braking is not air brakes, but rather dynamic brakes.  Dynamic brakes apply only on the locomotive units.  When the units used to be all on the head end, the use of dynamic brakes limited train the length (and was limited by train length) because their application caused all of the trailing tonnage to run in against the locomotive units, whereas air braking restrained that run-in force by applying brakes to all the cars in addition to the locomotive units.   

 

However, with the advent of locomotive units being distributed throughout the train, this distributed power also distributed the dynamic braking.  It still had the disadvantage of slack run-in resulting from no braking on the cars, but only on the locomotives.   However it did offer the advantage of at least spreading out the dynamic braking even though it left no direct braking on the individual cars.  

 

Distributed power was a new development that eliminated the need to place all the power on the head end.  So by spreading out the power in the train, distributed power allowed longer trains without the danger of pulling them in two; a danger that is greatest with all of the power pulling from the head end.

 

DP also allowed increasing the use of dynamic brakes in lieu of air brakes because DP also spread out the braking force of dynamic brakes in addition to spreading out the motive power.  Distributing dynamic braking allowed longer trains without the danger of buckling them; a danger that is greatest with all of the power applying dynamic braking from the head end.   

 

HOWVER, although spreading out dynamic braking was a step in the right direction for braking longer trains without the danger of buckling, it still applied braking only to the locomotives, which resulted in the trailing un-braked cars running in to the locomotive when dynamic braking is applied.  All it does is limit the number of cars running into a locomotive power unit by dividing the train up into distributed power sections.  So each section has fewer cars running into its locomotive than would be the case will all of the power on the head end.  Nevertheless, the run-in of un-braked cars is more likely to cause train bucking than run-in of cars during braking with air brakes.  Ideally, when the train is divided into distributed sections of power and dynamic braking, it would also be divided into distributed sections of air braking. 

 

Distributing the air braking would prevent the brakes from applying serially from the head end to the hind end, which takes considerable time in the case of a “Service” application, which is used for routine train handling.  If this improvement were possible, it would reduce the need for dynamic braking with its fundamental disadvantage of applying only on the locomotives.   However, this improvement for air braking is not possible with conventional pneumatic controlled pneumatic braking (PCP).  However it is a function feature of ECP braking.  In any case, there is no practical way to implement it with current practice.  So with each individual locomotive power set dividing the long train into sections, each section acts like one train being braked by only the locomotive. 

 

I suspect the presence of excess in-train forces in ultra-long trains arises from excess reliance on dynamic braking even though it is made safer by the distribution of dynamic braking into blocks throughout the train.  However, despite that benefit, dynamic braking still suffers from the disadvantage of being confined to only the locomotive wheels and not to the railcar wheels. 

 

Then when you add the changing and opposing grades, which naturally occur in greater number under longer trains, the variation in car type and weight, and the random and complex patterns of slack surge that result; you get an overall train behavior that is impossible to predict by existing means of evaluation.    

 

References that I find concur with my conclusion on that point.   That is why I suggest a means to actually measure buff and draft force on each coupler joint in a test train that duplicates an actual revenue train.

 

[oltmannd]

Euclid

Here is some news about this question.  Does anybody actually know the answer?  Railroad labor advocacy seems to frequently cite this increase of danger with monster trains.  However, that view is understandable because everyone agrees that monster trains reduce crew costs, and thus cause a loss of jobs.   The media obviously jumped onto that bandwagon in response to the East Palestine wreck, which was then affirmed by the Springfield, OH wreck.   

 

This latest news suggests that engineers may need more training to know how to operate the longer trains. 

 

 

https://fortune.com/2023/04/28/railroad-derailment-long-trains-federal-regulators-problem-contributor/

 

 

The feds are warning railroads that their love of long trains is leading to horrible accidents and derailments—but they’re not doing anything about it yet

 

 

 

 

It's gonna be more territory specific than train consist specific.  FEC could run just about anything, anywhere, anytime no sweat.  Horseshoe curve?  CSX Boston Line?  We'd have to talk...

[Euclid]

I conclude that ultra-long trains made possible by distributed power, have disproportionably more complex and more violent train dynamics than relatively shorter trains such as less than 100 cars.  With the longer trains comes a large variety of random slack force patterns that can be difficult to understand and control.  These conditions found in the ultra-long trains cause derailments due mostly to hard slack run-ins. 

 

The core issue is not that locomotive engineers fail to understand what train handling is needed to offset these slack issues.  The real problem is that the core issues of long train slack dynamics are not sufficiently understood by the ones who decide to run the ultra-long trains.  The most attention to the problem has been focused on train makeup, which has always been recognized as being critical even before the advent of ultra-long trains.  But train makeup alone does not seem adequate to solve the greater tendency for derailments that comes with ultra-long trains.  Therefore, I don’t think that optimum train makeup will solve the problem.  It will help, but it does not go far enough.  

 

So, to the question of whether ultra-long trains cause derailments, I conclude that the answer is YES.   There is also a growing movement that concludes likewise.  The FRA says the solution is for locomotive engineers to solve the problem by keeping up with the demands of increasing train length practice.  That seems like a predictable response.  The FRA does not have a solution, but they are sure there is one.  So they conclude that the only challenge will be for operating people to apply the solution. 

 

Oddly enough, there is a solution right in plain sight.  The basic problem is train braking.  And this circles back to our old friend, ECP brakes.  Back when ECP first emerged, it came with several valuable new features such graduated release, lower in-train forces, and shorter stopping distance.  People naturally think of stopping distance reduction when they consider brake improvements.  Indeed, it is shorter stopping distance that always increases safety, and better safety always sells.  So the champion feature of ECP brakes was hailed to be shorter stopping distance. 

 

Yet this was extremely misleading and almost dishonest.  Trains basically have two different brake systems with two different names which are “Emergency” and “Service.”  While ECP impressively reduces stopping distance, it does so only with the Service application, where there is very little need for reduced stopping

distance.  It is only the Emergency application that has a profound need for shorter stopping distance, and for that, ECP offers almost nothing. 

 

The real champion of ECP features is:  LOWER IN-TRAIN FORCES. And this is also the most direct and meaningful benefit needed to prevent derailments with ultra-long freight trains.

 

ECP provides this feature by moving all brake shoes against all of the wheels simultaneously, and with equal pressure.  Fundamentally, this eliminates the reason for slack to run in or out.  However in a practical sense, even with simultaneous application, brakes will vary in stopping power due to the weight of the car.  So car load sensors must be included with ECP.

 

There is also the issue of brake and wheel wear, which can cause the braking response to vary.  This may require new regulations to improve maintenance on brake shoes and wheels for cars that will qualify for ultra-long train service. 

 

But of course this solution to the ultra-long train derailment problem will be fiercely opposed by the railroad companies because of their dread of ECP brakes.  It is only the social impact of train disasters written in blood that will push this forward.  But the industry would give up the long trains if it would avoid ECP.     

[Euclid]

Here is what the FRA has found in their preliminary investigation of the March 4, 2023, Springfield, OH derailment on the NS Railway:

 

Quoting from this link:

 

https://www.railwayage.com/wp-content/uploads/2023/04/Safety-Advisory-2023-03.pdf

 

“The train consisted of three head-end locomotives and two mid-train DPUs, with one head-end locomotive offline. The train was traveling on an ascending 0.6% grade with the heavier part of the consist (the back end) on a 0.7% downhill grade. The weight was mostly concentrated at the head and rear ends of the train.

 

During the incident, dynamic braking was applied only to the head-end locomotive consist, while the DPUs were idle, making the train function like a conventional train.

 

The derailment happened at the sag between ascending and descending grades, with short, empty rail cars designed to ship coiled steel being the first to derail.

 

Buff forces peaked as the downhill portion of the train ran-in, causing the derailment of cars 70 through 72 (the short coil cars) and the subsequent pile-up.

 

The train was classified as a key train,7 with 28 loaded hazardous materials (hazmat) cars distributed throughout. No hazmat cars derailed. FRA’s investigation into this incident is currently ongoing, but preliminary indications show excessive buff force due to train makeup and train handling are the primary causes of the incident.”

 

===================================

 

What about the loose wheels which NW found and concluded were the cause of the wreck?

[BaltACD]

jeffhergert

...

The FRA wants more crew training.  Maybe that's not where the problem is.  Crews run these things everyday.  They get to know what works and what doesn't.  Maybe railroad management needs more training in making up trains.

Jeff

On 'my' division of CSX, there were a number of TTSI's that related to train make-up that limited the positions of various cars/types when built into trains.  Many of these restrictions were related to the trailing tonnage that could be in the train behind such cars.  My division did include mountainous territory with a high degree of curvature in surmounting those grades.  These restrictions had been developed from the examination of causes of derailments over the years - learned in 'blood'.  While these restrictions were second nature to the yardmasters on the division who made every effort to comply with the restrictions - my divisions restrictions were not uniformly enacted across the company as a whole.  Flatlanders did not have the restrictions, even though they were building trains that would operate through the territory that occasioned the restictions.  A company failure in my mind.

[Euclid]

 

Generally, ECP brakes have been promoted on the basis that they provide much shorter stopping distance of freight trains.  This claim was always quantified by the percentage reduction of stopping distance.  However, to the best of my knowledge, this claim was never fully qualified in public promotion.  Such qualification would have stipulated that the claim only applied to brake “Service” applications because with brake “Emergency” applications, the stopping time/distance reduction was negligible.  

 

Here is an example of that claim from The Railway Technical Website:

 

http://www.railway-technical.com/trains/rolling-stock-index-l/train-equipment/brakes/electro-pneumatic-brakes-d.html

 

“With the new responsiveness of ECP braking, braking distances will be reduced. A range of 30 - 70% reduction has been quoted. This will allow shorter stopping distances and will, in turn, allow higher speeds.”

 

This specification makes no mention of the fact that it only applies to airbrake “Service” applications, while it is the “Emergency” application that demands relatively short stopping distance.  That is because that application is for situations that spontaneously pop up often when a train is too close to stop with a "Service" application,  or an “Emergency” application.  Fundamentally, emergency stopping requires the quickest stopping response. 

 

This fact took the wind out of the sails for the mandate for certain tank cars to be equipped with ECP brakes about 6 years ago.  In any case, the industry is vehemently opposed to ECP brakes because it requires that all locomotives and cars must be converted to remain compatible within the standardized interchange pool.

 

So the industry argued against ECP brakes only on the basis of the reduced stopping distance, but they did not bother to mention the misleading premise that had always surrounded the issue as I mention above.  Instead the railroads only based their objection on the fact that very few accidents are caused by brake failure.  They completely sidestepped the advantages of ECP mitigating in-train forces. 

 

And in the time since dodging that ECP mandate, the railroads have fallen in love with monster trains, which feature a centerpiece of potential, excessive in-train forces.  And the industry cannot dis-prove that excess in-train forces are causing more derailment with the longer trains because nobody has done enough testing to find the answer.

 

However, with ECP brakes, all in-train forces can be virtually eliminated.  Actually it is a far greater feature benefit than improved stopping distance.  So monster trains and ECP brakes is a marriage made in heaven.  Now, unless the high in-train forces issue can be disproven, the only argument opposing EPC is the cost of the added safety.  

[jeffhergert]

Within the last year or so, UP announced a computer program that can run a train's route and show where problems might happen.  I don't think it is used to reposition cars, just give them an advance heads up about where something might happen.

Just looking at a train list is sometimes enough because it's obvious that things should be changed.  A few times I know where a crew raised objections and were told to go with it.  Then the train derailed on that crew's portion of the run.  (Funny how even though the crew told the FRA investigators that fact, it didn't make it into the FRA accident report.)

DP is not the panacea it's made out to be.  Yes they do help.  All our trains over 10000 feet need a mid train DP.  Longer trains over a certain weight threshold need a mid and rear DP.  Some still get torn up, sometimes into three or four  pieces.  Sometimes it's a problem with the DP that causes the initial action that results in a train breaking into two.

Not all trains are the same.  One of our mixed manifests is often 60 or 70 % cushioned drawbars.  We have a couple that have few of the cushioned drawbars.  Both types can often be in the 12000 to 15000 foot range.  Guess which type is likely to have problems out on the road. 

We've had some intermodals up tp 18000 feet.  Many are often in the 12000 to 15000 foot range.  They normally don't have many problems.  (Other than open container doors.)  They are relatively light and don't have cushioned drawbars.  They aren't bad to run.

Regarding air brake usage,  Most trains I get I only have to use air when stopping.  Eastbound, except for light trains, like empty unit trains, I will need to use air once for sure to control speed.  That's coming down the short 1.25% grades coming down into the Missouri River valley.  There's two other areas eastbound where I may need to use air to maintain speed, but even in those places if I have enough good dynamics I won't need to use air. 

The FRA wants more crew training.  Maybe that's not where the problem is.  Crews run these things everyday.  They get to know what works and what doesn't.  Maybe railroad management needs more training in making up trains.

Jeff