Oil Train

[Euclid]

We have had a lot of discussion about derailment sensors on freight cars and how they can help get a train stopped in the derailed-dragging phase before a pileup begins.  Some have suggested that the derailed-dragging phase is too short to take any sort of preemptive action.  It has even been claimed that the instant the first wheelset leaves the rails, it digs in to the track bed and piles up the train.

My contention is that the derailed-dragging phase can be anywhere from nearly non-existent to several miles long.  A case in point was the CSX derailment in Maryville, TN last July.  A tank car full of acrylonitrile derailed, dragged some distance, and caught fire.  Eventually the train stopped before the derailment led to a pileup.  The derailed car remained coupled and in line with the track.

At the time, there was no report of how far the derailed car dragged, but one photo of a ripped out grade crossing indicated that the derailed-dragging event persisted for at least half a mile—plenty of time for a sensor to stop the train. 

But now, we learn that the derailed car actually dragged Nine Miles before the train stopped. 

http://www.thedailytimes.com/news/lawsuits-train-car-carrying-toxic-chemical-derailed-miles-before-reaching/article_a183b5db-6409-53ae-b749-195868ae8934.html

[Deggesty]

I do not remember all the details, but when I was going from Memphis to Chattanooga in 1964 on the Southern's day train, I noticed, as I was standing at the rear, that many ties on the stretch between Stevenson and Wauhatchee had been cut into in the middle. The flagman told me that a car had been derailed and was dragged along for some distance, cutting the ties.

[Euclid]

I have seen that sort of dragging event too.  One car on the ground can just be carried along, with the aid of the couplers, without disrupting the train, and without derailing it.  It does surely damage track, and it tears out grade crossings as though they were not there.  It is amazing that this can continue for miles without any realization that something is wrong.  How can wayside detectors eliminate most of this dragging event?

Certainly a derailed-dragging car should be able to set off any detectors, so what is the answer when it comes to wayside detectors?  If they were placed every mile, would that contain most derailment developments to one mile?

In the case of the one above (Marysville, TN) where the CSX dragged a derailed car nine miles, it appears that the car eventually caught fire from mechanical friction of an axle.  I am not sure what happened next.  The train eventually stopped with the car on fire.  I do not know whether the train stopped because of an intentional crew reaction to a warning of smoke; or whether the car catching fire simply burned up the air hose.

Some of those derailed-dragging events do not progress beyond the one derailed car and a long stretch of damaged track.  Others end in catastrophic pileup of the train.  The wreck in Marysville was in between those extremes.  There, no big pileup ensued; but the burning carload of acrylonitrile was an added element of danger to the base line of just tearing up a few miles of track in the minimum derailed-dragging event.  

[Jim611]

In the summer of 1973, I worked on a N&W section gang on the ex Wabash in Fort Wayne.  I got to work one morning and the foreman and assistant were not there.  Nobody knew where they were.  They showed up out an hour later and advised there was a derailment west of town. As we approached a crossing near the derailment, the gates went down and a train went by at max track speed. 

 

During the night, a wheel had derailed and tore up 4 miles of track before re-railing on an old farm crossing. The train speeding by had not received the slow order yet. They speculated it was the rear wheel of an auto rack car.

 

The track was slightly out of line in a few spots and we spent the next week replacing damaged bolts, spikes and creepers.

  

[Euclid]

Jim611,

Like the one that you describe, the derailed dragging events that somehow manage to rerail and keep going are an interesting variety.  Here is one from the monthly tabulation of train accidents in the U.S. published in the Railroad Gazette.  This one did enough damage to derail the next train: 

 

January 1892

 

15^th, on Atchison, Topeka & Santa Fe, near Brenham, Tex., a passenger train of 6 cars was derailed, and although there were a large number of passengers aboard, only 7 of them were injured.  It appears that the rails had been loosened for a considerable distance by the cutting out of the spikes by a derailed freight car in a train which had passed a short time before.  This derailed car, after running about 2 miles, was rerailed at a road crossing, so that the injury done by it to the track was not promptly discovered.

 

Here is an amazing derailed-dragging event involving multiple phases.  In one phase, they left a truck alongside of the track and the train continued on:

 

November 1896

 

9^th, on Cleveland, Cincinnati, Chicago & St. Louis, near Larue, O., a brake beam of an empty platform car, in the middle of a long westbound freight train, fell upon the track and tore up a crossing plank.  This plank was carried along about nine miles when it dropped off and derailed one truck; this truck ran on the sleepers for about half a mile and then ran off clear of the track, the train going on.  At the same time that this truck jumped the track one truck of the next following car also went off, and this one ran on the sleepers about 2,000 ft. farther than the other; then it got back on the track and behaved all right.  After running about six miles farther the trainmen noticed that they had lost a truck and stopped the train.

[Euclid]

 

The final report on the cause of the Lynchburg, VA oil train wreck has been released.

http://www.roanoke.com/news/virginia/ntsb-says-broken-rail-likely-caused-lynchburg-train-derailment/article_8aba3f76-cb8d-543f-8799-3f6e5e4b48cf.html

 

NTSB Report:

http://www.ntsb.gov/investigations/AccidentReports/Reports/RAB1601.pdf

[Euclid]

 

The news reporting in this article and in the earlier preliminary release from the NTSB last summer has been very difficult to comprehend.  There were parts of that news last summer about the crew of a passing train hearing strange sounds in passing the rail defect.  I never saw any clarification of that rather provocative information.  Also included with that news release was a long and highly detailed interview relating to those in charge of repairing the rail defect.   

The article I linked above is the most incomprehensible explanation I have ever heard.  The way to explain something complex is to start with the overview, and then give the details.  That way, someone hearing the details can tell where they are headed. 

The writer of this article starts by piling up details in all directions with no indication of how they fit together or where they lead.  Then it ends there, and the writer starts describing other train wrecks.  After that, he comes back and starts linking all the tangle of details of the Lynchburg wreck which he left piled up at the beginning. 

The NTSB report is clearer to follow through the rather complex set of factors, but it ends before answering this question:

If the type and size of rail defect was known to be not in danger of breaking, why did it break?

Also, there was news at some point that assured us that the NTSB would examine this wreck in full detail with complete 3D scanning of all of the wreckage accompanied by detailed explanation of how it all happened. 

I do not see anything in the final report approaching that level of information.  Interestingly, only one tank car ruptured.

[dehusman]

Euclid

If the type and size of rail defect was known to be not in danger of breaking, why did it break?

That's not what the report said, what the report said was that those types of defects were not considered serious enough to fail.  That means in most cases they don't progress to failure, this one did.  There doesn't appear to be any obvious reason documented why it failed.  The anwer probably has to do with a combination of the specific metalurgy of the rail at that point, combined with the specific stresses on the rail.  Replicating that specific set of circumstances would be very difficult.

Bottom line is that the train being an oil train had nothing specific to do with the cause of the derailment.  It was a broken rail.

[Wizlish]

You're both missing the key point here -- interestingly, although the report goes into some discussion of it, it's pretty evident the reporter doesn't really understand what's involved, and goes into the detail that in part is confusing Euclid (some of which appears to me to have some political aspects, perhaps talking up Administration policies and actions that turn out to have little if anything to do either with the proximate cause of the derailment or any practical measures to prevent a repetition, but I digress slightly)*.

This was a REVERSE DETAIL FRACTURE, originating on the underside of the railhead and with little, if any likelihood of being caused by any aspect of properly-defined rolling contact fatigue.  There is specific mention of how difficult detection of this kind of defect is (relative to things like gauge-corner cracking, shelling/flaking, and other aspects that can be scanned or monitored easily by devices and techniques for the actual 'rolling contact' stressed areas.  There is a mention -- only as a reference to something I haven't found yet -- that a "20% defect" turned up in the ultrasound records, but this seems to have transmogrified into only 5% in some later parts of the discussion (one being the detail about 25-mph train speed not constituting excessive for a defect of the claimed magnitude).  The report also mentions something about RDF tending to cause a more likely catastrophic progression to breakage, although I saw no source reference -- I'm not really concerned because buslist and perhaps others can easily provide the references or summarize them for us.

What I am concerned about here is that the discussion about 'harmonic loading' or special stresses set up, or supposedly set up, by unit oil train consists will be re-opened with the excuse that such factors mitigate some 'greater likelihood of RDF' (I am not sure what physical argument would be used, and I'm not saying there can't be one, just that it is MUCH less likely than the stated probable cause, a fabrication or handling defect in the lower 'corner' of the railhead that propagated and then proceeded quickly to failure).

The great red flag here is that it opens up a whole new can of worms to people who are concerned about track integrity, particularly insofar as there appears to be no effective testing for RDF defects or crack progression other than interpretation of data whose methodology, processing, and interpretation is largely skewed in favor of RCC analysis.  God help us if Schanoes' 'Schumenthal' gets hold of this as a place to enforce new regulations, or FRA decides more oversight procedures to detect RDF should be mandated as emergency orders or whatever.  On the other hand... yes, I think this MIGHT need to be more carefully detected and tracked.

*And before any of the usual suspects start whining about 'experts' having worded everything in this report, consider the last part, which says

For more details about this accident, visit www.ntsb.gov/investigations/dms.html [this is all one line in the original, of course] and search for NTSB accident ID DCA14FR001.

 

Follow the link given ... and it is broken.  Follow the cited accident ID in their search ... and it maps to a BART incident, almost as far away as you could get and still be in the Continental United States... that is not much indication of 'expert' presentation or oversight, let alone editing before release or even elementary checking.

 

EDIT -- Pulling a few teeth, it appears the correct accident ID for the docket material ends in 8, not 1.: Here is a direct link to the materials:

 

http://dms.ntsb.gov/pubdms/search/hitlist.cfm?docketID=57646

[Euclid]

Wizlish

You're both missing the key point here -- interestingly, although the report goes into some discussion of it, it's pretty evident the reporter doesn't really understand what's involved, and goes into the detail that in part is confusing Euclid (some of which appears to me to have some political aspects, perhaps talking up Administration policies and actions that turn out to have little if anything to do either with the proximate cause of the derailment or any practical measures to prevent a repetition, but I digress slightly)*. 

Wizlish,

I do not understand what point you think I am missing, nor anything else in the above quoted portion of your post.  I simply asked this question:   “If the type and size of rail defect was known to be not in danger of breaking, why did it break?”

Here is the point of that question from the report:

 

“The size of a rail defect determines if mitigation is required by FRA regulation. The suspected rail defect that failed at the point of derailment was a 5 percent reverse detail fracture. Historically, regulations have not considered 5 percent reverse detail fractures to be a defect subject to complete failure prior to progressing to a larger size.”

 

I conclude that the part in red was the justification for not repairing the defect as soon as it was detected.  Obviously the justification was wrong. 

The report seems to confirm that the justification was wrong with this quote (with my emphasis in red):

 

“The RSAC adopted the Rail Failure Working Group recommendations on April 16, 2014. The final recommendations developed with industry and other stakeholders formed a consensus document of best practices or guidelines to manage the risks related to rail wear and rolling contact fatigue.  Before the guidelines were implemented by CSX, the Lynchburg accident occurred; if they had been implemented, this accident would likely have been prevented.”

[Wizlish]

You're picking the wrong nits again.

The first point:  Reverse detail fractures are difficult to detect.  An associated point, not entirely made clearly (and buslist is better qualified to address this specifically) is that disambiguating ultrasonic or other test detail to produce evidence of an early case of RDF is difficult.

The second point: it's 20%, not 5%, that constitutes the 'threshold' for the usual kinds of transverse failure, the point at which the railroad would have (promptly!) dispatched a crew to put on just the sort of 'remediation' they did 30" west.  Now, your point about the potential importance of a 5% reverse detail fracture is well taken; the report does note that this kind of failure is notable for quick catastrophic progression from even slight extent (or so I understood the report to say).  Your first 'rubrics' are addressing the perception in the industry that RDFs were not considered more critical at very early stages than more common kinds of transverse failure, most notably those more directly attributable to RCF.

How the second point concerns this is not quite clear.  I do not know, although I presume buslist and others do, whether the 'recommendations' you are describing in the second red passage specifically involve a requirement to repair RDFs promptly whenever detected at 5% or greater level ... which necessarily presupposed a requirement to detect them and then characterize them accurately in the first place, with minimum latency.  Perhaps we'd better look at and discuss the specific passages and context of the 'recommendations' before addressing whether the conclusions are even at all germane to this specific failure ... let alone something that would have produced definitive preventation if applied.

[Wizlish]

Incidentally, there is something else valuable in this report that has not been discussed at all adequately: the specific mode of failure for the 1232 car that was so radically breached.  Note that this was not a 'puncture', but caused by something that 'scraped' across the car until it 'picked' a circumferential weld, and then apparently followed the weld (either by failure or preferential 'guiding') leaving what was described as curls of metal inside the tank.  Remember that this occurred from a speed below 25 mph.

[Euclid]

Wizlish,

Picking nits?

There was policy in place that allowed the defect repair to be delayed after discovery.  So they scheduled the repair for two days later.  The defect caused a train wreck one day later.   Is it not logical to conclude that there was something wrong with the policy?

As it turns out, the answer is apparently yes because the Rail Failure Working Group changed the policy shortly before the wreck.  And the NTSB report says that had the policy change been implemented before the wreck, it would have prevented the wreck. 

[Wizlish]

Euclid

There was policy in place that allowed the defect repair to be delayed after discovery. So they scheduled the repair for two days later. The defect caused a train wreck one day later. Is it not logical to conclude that there was something wrong with the policy?

I'm sure for you it's logical.  I don't have that good 20/20 foresight.

I am not altogether sure, having looked over a couple of the interview transcripts, that the thing is as 'open and shut' as finding defects and ordering them all fixed ASAP (which is what appears would have been required to 'prevent' what happened for this specific defect).  If I understand correctly, they made a decision to replace a full piece of rail rather than patch a defect for two days and then remove it.  An interesting thing is that they talk about imposing a slow order (10 mph) before fixing a defect 30" from a bad weld, but I don't get the impression that is that same defect elsewhere described as a reverse detail failure that was the supposed proximate cause of the derailment.

Reading between the lines, there may have been a bad batch of rail involved in this section, and it may be interesting to see how that 'angle' develops.  Look at this transcript of how the testing was done on this segment and compare to a couple of the other interviews.

I was particularly interested to see how the discussion of MGT wear on the double track developed between the questions asked in the first interviews and the information between Track 1 and Track 2 use that came up in this one.

EDIT - let me find exactly what the working group changes involve regarding reverse detail fractures and then see how this would affect systematic predictive repair.  I find I misunderstood what you were saying and will correct that before commenting further on this aspect.  Out of curiosity -- are you using Martz' article in the Richmond paper as source material, or have you looked at the actual working-group study?

FURTHER EDIT:  I came across this .pdf of an earlier state of the NTSB report, which has some highly interesting 'conclusions' in it.  Among other things, they apparently get the MGT of Track 1 wrong, they attribute the rail failure to rolling-contact fatigue damage wrongly, and they quote the findings of the RSAC rail failure working group in their Appendix B ... noting that the results of the study would have prevented the accident if its cause had been rolling-contact fatigue.  With that not being the case, I have to review the actual RSAC material to see what is different.

[Wizlish]

Apparently some of the dates of these 'reports' vary depending on the agency involved.  I managed to track down something potentially relevant, that the origin of the specific RSAC working group (in 2012) was related to issuance of a specific recommendation from an earlier accident (the ethanol train in New Brighton, PA).  The history of how this recommendation came to be 'closed by alternative means' is probably highly relevant to the current discussion.

http://www.ntsb.gov/_layouts/ntsb.recsearch/Recommendation.aspx?Rec=R-08-011

Note the association of defects with 'wear' -- I am still looking to see if there is a specific association between 'special treatment for reverse-detail fracture defects' and specific levels of observed wear or markers of fatigue.

I continue to think there is another aspect in here somewhere that has not been completely brought out.  This is supposition based on testimony, but y'all can fact-check it against the source material to some extent.  There were two tracks, on a line that essentially had 'single-track throughput'.  The two tracks in the section involved were not utilized equally, with Track 1 receiving far less MGT/year than Track 2, apparently at least partly through preferential dispatching of empties on that track.  It might follow that preferential dispatching of HHFT or hazmat might be done on this track because it would (presumably) have lower average wear or RCF defect probability.

Meanwhile, we also find that because of the lower loading, the maintenance cycle for rail inspection was much longer (123 days vs 31, was I think the figure given) and, at least prospectively, defects not immediately identified as critical would have a longer 'response time' (due to the lower expected average loading?).

I have the suspicion that the different priorities resulted, here, in a 'catch' situation where a more unsafe condition was produced by efforts to produce a safer one.  I do not think it would be fair either to tar CSX maintenance for making this mistake or to assume that rules for correcting RCF-related defects more quickly would necessarily have caught this particular one, and it would appear (from the CSX representative's comments in the News-Leader article) that CSX has voluntarily adopted a far more rigorous standard for repairing defects in this general situation than the Federal rule (that supposedly 'would have prevented this') calls for.  However, I wonder whether there should be a wider recognition of the general principle that routing 'key' trains on less-utilized parts of the track structure, in order to expose them to a lower probability of wear-associated track defects, needs to be matched with a higher level of detection and response corresponding not to average track-related measures but to the 'key' nature of the trains themselves.  If I am not mistaken, some of the recent safety recommendations and legislation do call for this, but I am still not sure that the specific RSAC rail integrity group's report specifically addresses how the reverse-detail fracture was to be detected and remediated in a short timeframe (vs. the circumstantial observation that it would have been removed in track scheduled for remediation 'by other means' for other defects).  And I would repeat that the discussion involves only reverse-detail fracture in this context when discussing anything that 'would have prevented this accident' in any meaningful context that is a useful predictor for future prevention of similar problems.

[Euclid]

Wizlish

 

Euclid

There was policy in place that allowed the defect repair to be delayed after discovery. So they scheduled the repair for two days later. The defect caused a train wreck one day later. Is it not logical to conclude that there was something wrong with the policy?

 

 

I'm sure for you it's logical.  I don't have that good 20/20 foresight.

Wizlish,

I am not claiming to have any foresight let alone 20/20 foresight.  I am only basing my conclusion on this sentence from the NTSB:

“…if they [the revised regulations] had been implemented, this accident would likely have been prevented.”

One point I noticed was the development of a 5% fracture not being a problem requiring an immediate repair, but a 20% fracture requires an immediate repair before running more trains on it.

But then the stipulation is added that a 5% may turn into a 20% almost instantly. Of course this raises the problem of such sudden conversion happening while a train is running over it.  Then, the train may have begun to cross the defect when it was safe, but failed to complete the crossing before it became unsafe.   

Therefore, unless there is some way of predicting when a 5% fracture will turn into something greater requiring immediate action; I don’t see how it can be concluded that a 5% fracture does not require immediate action. 

[tree68]

Euclid

I don’t see how it can be concluded that a 5% fracture does not require immediate action. 

I would opine that perhaps 5% fractures have been found to last some period of time before getting worse.  In fact, I'd imagine that if railroads had found that a 5% fracture was a significant problem, then they would have policies in place for immediate replacement.

It could also be that a change in the type of traffic might, indeed, exacerbate the issue.  And that result may not have become apparent in the MOW crystal ball.

[Euclid]

 

 

The rail defect is discussed in detail on page 8-11 in this NTSB report:

http://www.ntsb.gov/investigations/AccidentReports/Reports/RAB1601.pdf

I have read this several times, and find it to be incredibly confusing. 

Here is the text quoted from the report.  I have revised the formatting and removed the foot notes.

The following is quoted from the report:

Reverse Detail Fracture

The derailment occurred at a sudden break of a rail originating from a reverse detail fracture on the gage corner of the railhead of the high rail in the curve.

A CSXT contractor performed ultrasonic testing in the area of the derailment the day before the accident. Investigators reviewed the ultrasonic test data for the failure location. The data confirm that the test equipment functioned properly and responded to known rail features that would normally be detected by the ultrasonic test probes within the area of the failed rail.

The data showed defects discovered during the ultrasonic testing including a 20 percent transverse detail fracture; this was noted as number 151 on Sperry report number 119A, dated April 29, 2014.

CSXT Petroleum Crude Oil Train Derailment and Hazardous Materials Release

NTSB/RAB-16-01 9

The size of a rail defect determines if mitigation is required by FRA regulation. The suspected rail defect that failed at the point of derailment was a 5 percent reverse detail fracture. Historically, regulations have not considered 5 percent reverse detail fractures to be a defect subject to complete failure prior to progressing to a larger size. These types of defects cause a stress concentration on the surface of the rail and may cause a complete rail failure at a much smaller size than typical detail fractures.

Rolling Contact Fatigue

Rolling contact fatigue results from the cumulative effects of railhead wear and rail surface conditions, such as shelling, head checks, or flaking. The detrimental effects of rolling contact fatigue can occur before a worn railhead profile or side wear is noted. Rail wear on the gage corner and side of the rail are easier to find and manage; however, the detection of fatigue in the lower corner of the gage face of the rail is more difficult. That part of the railhead it not easily scanned by ultrasonic equipment, and a regulatory remedial action was not mandated to address these flaws.

A stress concentration, often called a stress riser, is a location in an object where stress is concentrated.

Shelling is a progressive horizontal separation that may crack at any level on the gage side, generally at the upper gage corner. It extends longitudinally—not as a true horizontal or vertical crack, but at an angle related to the amount of wear.

Flaking is a progressive horizontal separation on the running surface of the rail near the gage corner, with scaling or chipping of small slivers. Flaking should not be confused with shelling, as flaking takes place only on the running surface near the gage corner of the rail and is not as deep as shelling.

Head checks are transverse surface cracks on the gage corner of rails resulting from cold working of surface metal. These are sometimes referred to as gage cracks.

The FRA Track Safety Standards do not address this type of rail defect at the size that failed in the Lynchburg accident (5percentreverse detail fracture). The FRA remedial action chart addresses transverse detail fractures, but does not mandate remedial action until the defect is 20 percent or four times the size of the defect that caused this derailment. At that time, the railroad owner would be required to reduce speed to no more than 30 mph and apply joint bars within 20 days to the defective rail condition.

Post Accident Actions

CXST

Prior to the Lynchburg accident, if a transverse detail fracture had been20percentof the cross-section of the rail head, CSXT engineering standards required that the defective rail be changed out within 5days or joint bars be installed to the rail at the site of the defect. Since the ultrasonic testing data indicated a transverse detail fracture near the location of the derailment, CSXT planned to replace the rail on May 1, 2014. Since the operating speed for that area was 25 mph, the rail defect did not require a speed restriction in accordance with CSXT maintenance procedures or FRA regulations.

FRA Rail Failure Working Group Recommendations

The NTSB derailment investigations in New Brighton, Pennsylvania; Columbus, Ohio; and Ellicott City, Maryland, led the FRA to determine that each of the accidents resulted from rail failures. In September 2012, the FRA established a Rail Safety Advisory Committee (RSAC) Rail Failure Working Group to address rail wear issues such as rolling contact fatigue. The working group studied the effects of railhead wear and resulting rail surface conditions (better known as rolling contact fatigue) and how such rail conditions can adversely affect the results of ultrasonic rail testing. The Rail Failure Working Group met four times beginning in January 2013, and completed its task on July 31, 2013.

The group proposed new performance-based recommendations for determining rail wear and internal rail inspection criteria. These criteria ensured the FRA’s ability to effectively monitor rail integrity programs that require track owners to quickly identify and remediate areas that could lead to a derailment. The FRA’s efforts and industry’s acceptance of these best practices should significantly reduce rail accidents caused by broken rails resulting from rolling contact fatigue and improve the industry’s rail risk management programs.

The RSAC adopted the Rail Failure Working Group recommendations on April 16, 2014. The final recommendations developed with industry and other stakeholders formed a consensus document of best practices or guidelines to manage the risks related to rail wear and rolling contact fatigue. Before the guidelines were implemented by CSX, the Lynchburg accident occurred; if they had been implemented, this accident would likely have been prevented.

 

[End of quote]

[BaltACD]

Euclid

 

The rail defect is discussed in detail on page 8-11 in this NTSB report:

http://www.ntsb.gov/investigations/AccidentReports/Reports/RAB1601.pdf

I have read this several times, and find it to be incredibly confusing. 

Here is the text quoted from the report.  I have revised the formatting and removed the foot notes.

The following is quoted from the report:

Reverse Detail Fracture

The derailment occurred at a sudden break of a rail originating from a reverse detail fracture on the gage corner of the railhead of the high rail in the curve.

A CSXT contractor performed ultrasonic testing in the area of the derailment the day before the accident. Investigators reviewed the ultrasonic test data for the failure location. The data confirm that the test equipment functioned properly and responded to known rail features that would normally be detected by the ultrasonic test probes within the area of the failed rail.

The data showed defects discovered during the ultrasonic testing including a 20 percent transverse detail fracture; this was noted as number 151 on Sperry report number 119A, dated April 29, 2014.

CSXT Petroleum Crude Oil Train Derailment and Hazardous Materials Release

NTSB/RAB-16-01 9

The size of a rail defect determines if mitigation is required by FRA regulation. The suspected rail defect that failed at the point of derailment was a 5 percent reverse detail fracture. Historically, regulations have not considered 5 percent reverse detail fractures to be a defect subject to complete failure prior to progressing to a larger size. These types of defects cause a stress concentration on the surface of the rail and may cause a complete rail failure at a much smaller size than typical detail fractures.

Rolling Contact Fatigue

Rolling contact fatigue results from the cumulative effects of railhead wear and rail surface conditions, such as shelling, head checks, or flaking. The detrimental effects of rolling contact fatigue can occur before a worn railhead profile or side wear is noted. Rail wear on the gage corner and side of the rail are easier to find and manage; however, the detection of fatigue in the lower corner of the gage face of the rail is more difficult. That part of the railhead it not easily scanned by ultrasonic equipment, and a regulatory remedial action was not mandated to address these flaws.

A stress concentration, often called a stress riser, is a location in an object where stress is concentrated.

Shelling is a progressive horizontal separation that may crack at any level on the gage side, generally at the upper gage corner. It extends longitudinally—not as a true horizontal or vertical crack, but at an angle related to the amount of wear.

Flaking is a progressive horizontal separation on the running surface of the rail near the gage corner, with scaling or chipping of small slivers. Flaking should not be confused with shelling, as flaking takes place only on the running surface near the gage corner of the rail and is not as deep as shelling.

Head checks are transverse surface cracks on the gage corner of rails resulting from cold working of surface metal. These are sometimes referred to as gage cracks.

The FRA Track Safety Standards do not address this type of rail defect at the size that failed in the Lynchburg accident (5percentreverse detail fracture). The FRA remedial action chart addresses transverse detail fractures, but does not mandate remedial action until the defect is 20 percent or four times the size of the defect that caused this derailment. At that time, the railroad owner would be required to reduce speed to no more than 30 mph and apply joint bars within 20 days to the defective rail condition.

Post Accident Actions

CXST

Prior to the Lynchburg accident, if a transverse detail fracture had been20percentof the cross-section of the rail head, CSXT engineering standards required that the defective rail be changed out within 5days or joint bars be installed to the rail at the site of the defect. Since the ultrasonic testing data indicated a transverse detail fracture near the location of the derailment, CSXT planned to replace the rail on May 1, 2014. Since the operating speed for that area was 25 mph, the rail defect did not require a speed restriction in accordance with CSXT maintenance procedures or FRA regulations.

FRA Rail Failure Working Group Recommendations

The NTSB derailment investigations in New Brighton, Pennsylvania; Columbus, Ohio; and Ellicott City, Maryland, led the FRA to determine that each of the accidents resulted from rail failures. In September 2012, the FRA established a Rail Safety Advisory Committee (RSAC) Rail Failure Working Group to address rail wear issues such as rolling contact fatigue. The working group studied the effects of railhead wear and resulting rail surface conditions (better known as rolling contact fatigue) and how such rail conditions can adversely affect the results of ultrasonic rail testing. The Rail Failure Working Group met four times beginning in January 2013, and completed its task on July 31, 2013.

The group proposed new performance-based recommendations for determining rail wear and internal rail inspection criteria. These criteria ensured the FRA’s ability to effectively monitor rail integrity programs that require track owners to quickly identify and remediate areas that could lead to a derailment. The FRA’s efforts and industry’s acceptance of these best practices should significantly reduce rail accidents caused by broken rails resulting from rolling contact fatigue and improve the industry’s rail risk management programs.

The RSAC adopted the Rail Failure Working Group recommendations on April 16, 2014. The final recommendations developed with industry and other stakeholders formed a consensus document of best practices or guidelines to manage the risks related to rail wear and rolling contact fatigue. Before the guidelines were implemented by CSX, the Lynchburg accident occurred; if they had been implemented, this accident would likely have been prevented.

 

[End of quote]

Hindsight is always 20/20 - foresight is generally 20/400 or worse.

[Euclid]

Wizlish was able to track down the supplementary information for the NTSB report.  Apparently the link in the report was incorrect.  So he found this correct link:

http://dms.ntsb.gov/pubdms/search/hitlist.cfm?docketID=57646

 

The supplementary information is extensive, and I have not looked at a large portion of it.  But I did look at documents #1-3, #5, and #47, and found them all to be quite interesting. 

#47 is the aerial survey of the wreck showing the location and identity of the tank cars.

Car #34 was the last car still on the rails in the string of cars ahead of the derailment.  Interestingly, while car #34 did not derail, its wheels were damaged by striking the rail break that began derailing the cars behind car #34. 

It would be interesting to see what that rail defect looked like when the train began passing over it.  Regarding that question, the report says something that seems to be related to the answer, but again, it does not resolve the matter.  It is in document #2 which shows some video still shots.  One shot shows the track ahead of the locomotive where the defective rail was waiting about 50 feet beyond the locomotive.  The caption says this:

 

• At about 13:52:56 EDT, BNSF 7485 had just passed the aforementioned bridge and was rounding a right hand turn, travelling at about 24 mph. Figure 5 shows a rail plug2 (annotated) to the right of the track. As BNSF 7485 crossed this section of track, the train oscillated side-to-side and made a rattling sound.

Figure 5. BNSF 7485 showing rail plug to right of track.  

 

Whatever was going on with the defect, it was felt and heard by the crew on the engine as they passed over it.  I wonder if the previous train noticed any sound or oscillating movement when passing this defect.  I wonder if the investigators asked that question.  It would be interesting to ask the engineers of all trains that passed since the day before when the defect was scheduled for repair two days later. 

I also wonder if they asked the engineer of the derailed train what he thought when passing over the defect and noticing side to side oscillation and a rattling sound. 

And I wonder what caused the rattling sound.  The only thing I can think of would be the truck wheel flanges snagging on the misaligned nature of the broken rail as if trying to climb the rail, but falling back down into proper position before being able to fully climb the rail. 

[BaltACD]

Euclid

Whatever was going on with the defect, it was felt and heard by the crew on the engine as they passed over it.  I wonder if the previous train noticed any sound or oscillating movement when passing this defect.  I wonder if the investigators asked that question.  It would be interesting to ask the engineers of all trains that passed since the day before when the defect was scheduled for repair two days later. 

I also wonder if they asked the engineer of the derailed train what he thought when passing over the defect and noticing side to side oscillation and a rattling sound. 

And I wonder what caused the rattling sound.  The only thing I can think of would be the truck wheel flanges snagging on the misaligned nature of the broken rail as if trying to climb the rail, but falling back down into proper position before being able to fully climb the rail. 

If T&E crews feel something about the track structure - they report it.  If there was no report made to the Dispatcher about 'rough track', no rough track was felt.  Since all radio communications between Train Dispatchers and field personnel are recorded - any period of time can be searched, and I am sure the NTSB in their investigation did review all radio communications on the territory.

[Euclid]

BaltACD

 

Euclid

Whatever was going on with the defect, it was felt and heard by the crew on the engine as they passed over it.  I wonder if the previous train noticed any sound or oscillating movement when passing this defect.  I wonder if the investigators asked that question.  It would be interesting to ask the engineers of all trains that passed since the day before when the defect was scheduled for repair two days later. 

I also wonder if they asked the engineer of the derailed train what he thought when passing over the defect and noticing side to side oscillation and a rattling sound. 

And I wonder what caused the rattling sound.  The only thing I can think of would be the truck wheel flanges snagging on the misaligned nature of the broken rail as if trying to climb the rail, but falling back down into proper position before being able to fully climb the rail. 

 

 

If T&E crews feel something about the track structure - they report it.  If there was no report made to the Dispatcher about 'rough track', no rough track was felt.  Since all radio communications between Train Dispatchers and field personnel are recorded - any period of time can be searched, and I am sure the NTSB in their investigation did review all radio communications on the territory.

 

I don’t doubt what you say about the policies.  However, I would expect the report to follow through on the matter of the crew experiencing side to side oscillation and a rattling sound rather than to just make that pregnant point and move on.

It forces me to consider the prospect that had somebody actually looked the rail defect prior to the arrival of the train; they would have found the defect to have progressed to a point where the passage of a train would not be permitted.  I am not saying that such an inspection was necessary if it was not actually required by the rules.  It is just an interesting point that suggests the need for a change in the rules. 

I think it would be an interesting challenge to rewrite this whole sequence about the development of the defect in a way that it could be understood.  Certainly, the facts are available.  They just have to be rearranged and presented without the smokescreen. 

A clear explanation should be the mission of the NTSB, but that is difficult because they also have a mission of not finding blame. 

[Wizlish]

Euclid

I think it would be an interesting challenge to rewrite this whole sequence about the development of the defect in a way that it could be understood. Certainly, the facts are available. They just have to be rearranged and presented without the smokescreen.

I second this motion.  As part of that discussion, it seems to me that the noise and 'oscillation' reported had nothing to do with either the particular reverse-detail fracture or its timing as far as the "preventability" of this accident is concerned.  However, it would be a fine demonstration of an 'enhancement' of the policy of the RSAC group about prompt action in cases of actual RCF involvement (particularly transverse gauge-corner fractures) OR to bump up priority in the 'automatic defect tracking system' in as close to realtime as possible when a crew reports common-sense signs of track damage on a section with other known (but not yet 'critical enough to prioritize') defects -- as appears to have been the case for the rail on Track 1 in Lynchburg.

Much of the discussion is going to hinge around two things: the defect-detection timeline and the physical defect types and locations in that "40' section' of rail that was scheduled for changeout on what turned out to be not-quite-just-in-time scheduling...

[dehusman]

If they had dispatched somebody out to the scene how would they have :

a)  Found the 5% defect.

b)  Determined it had progressed to a larger state

The rail had broken prior to the passage of the train that derailed, why hadn't the signal system indicated a track occupancy?

I have a feeling that unless the rail was broken, any person inspecting the track would not be able to determine that there was a defect in the rail or its extent.  It is entirely possible that they may have "determined" the noise, etc. was caused by some other factor (probably related to the previous repairs) and, depending on what the rest of the track looked like, might have at most put some sort of additional slow order on the track (10 mph?) which would have meant that the derailment would have probably still happened but at a slower speed.  Unless the rail is broken, any noise or wobble would not be caused by the defect in the rail.

[Euclid]

dehusman

If they had dispatched somebody out to the scene how would they have :

a)  Found the 5% defect.

b)  Determined it had progressed to a larger state

The rail had broken prior to the passage of the train that derailed, why hadn't the signal system indicated a track occupancy?

I have a feeling that unless the rail was broken, any person inspecting the track would not be able to determine that there was a defect in the rail or its extent.  It is entirely possible that they may have "determined" the noise, etc. was caused by some other factor (probably related to the previous repairs) and, depending on what the rest of the track looked like, might have at most put some sort of additional slow order on the track (10 mph?) which would have meant that the derailment would have probably still happened but at a slower speed.  Unless the rail is broken, any noise or wobble would not be caused by the defect in the rail.

 

Dave,

My thought was that the oscillation and rattling noise that the crew experienced when passing over the defect was an indication that the defect had evolved past the 5% fracture range where it stood the day before.

I understand your point that a rail break that has separated would have caused a stop signal indication, so that indicates that the break had not evolved to a complete separation.  Perhaps part of the head and web was missing or out of position, but the electrical continuity was still maintained through the rest of the rail. 

I speculate that the fracture might have been obvious to an observer on the ground, but of course, that was not required since the rules allowed the assumption that the fracture would not cause a derailment in the given timeframe. 

I do not conclude that the rattling sound was necessarily due to the movement of a loose portion of the rail against the fixed portion.  It could have also been due to the wheel contact with an irregular shape of a rail break.

Such a rattling sound combined with unusual side to side oscillation suggests that the locomotive nearly derailed and then abruptly recovered its alignment, and the “rattling sound” may have been the impact of the three wheel flanges against parts of the broken out portion of the rail.    

In the NTSB interview that was conducted with the person in charge of repairing the rail defect, there was exhaustive questioning of every conceivable detail.  Given that stunning level of attention, I cannot imagine the NTSB not asking the engineer what he thought was happening when his locomotive suddenly oscillated and made a rattling sound when passing over the fracture that caused the derailment.  Leaving that out of the report is almost comical.    

[Wizlish]

Euclid

My thought was that the oscillation and rattling noise that the crew experienced when passing over the defect was an indication that the defect had evolved past the 5% fracture range where it stood the day before.

Where is it documented in the record that the oscillation and rattling were caused at the reverse-detail fracture site?

That's not to say it wasn't, but there is no smoking gun to indicate there wasn't in fact a catastrophic progression and break at the 5% reverse-detail fracture site somewhere back in the particular train, perhaps right at the point the evidence appears to indicate the first sign of wheel damage to a car (around the 33rd car, if I remember correctly). 

If I remember there was a 20% transverse fracture present right in the vicinity, and this was the thing that 'tipped the balance' over into removing the whole 40-foot piece rather than putting in another set of joint bars and field-welding things.  In the 2014 report, in fact, the accident was directly attributed to RCC-related failure, which would have to be something at the gauge corner or head/flank contact surfaces, and this assumption was the reason for the initial claim that the RSAC 'recommended procedure changes would have prevented the accident'.

 

As you noted, we need a careful timeline, and a carefully-done diagram showing all the defects organized east to west or vice versa, probably best in the direction of the derailed train's movement, within the context of how the track was aligned.   There won't be any sense out of this except bandying claims forward and backward until we have distinguished the failures, and the propensity of each to have caused the particular derailment observed.

Something else that needs to be covered if we're going to start a shoulda-woulda-coulda blamefest on the part of the engineer noting the 'oscillation and noise' is the procedure by which those observations would work themselves 'quickly enough' up the chain of notification and down the chain of command to get quick remediation.  Perhaps there needs to be some procedure (within CSX, or perhaps NORAC/GCOR more formally) for implementing the equivalent of a "911 call" that puts an automatic slow order out (via the PTC system, most probably, and presented as more a civil restriction than an operating one so it's understood clearly) - this would have some specific 'rapid response' checking with the database of track defects referred to in the testimony and ideally some feedback on this to the engine crew(s) concerned so they know what they might be experiencing.  (Do I think that's practical to implement, or that crews would pay much attention to it if it generates 'cry wolf' slow orders too often?  don't ask too often... but someone like Sarah might)

 

BTW, I did not find the level of detail in the parts of the interviews I read to be 'stunning'.  In fact I find it astounding that the NTSB people either did not pick up on, or deprecated for some reason known best to themselves, the observation that Track 1 was preferentially underloaded relative to Track 2 (they say explicitly in the 2014 report that both tracks were approximately equally used, which was in one of the testimonies but not from the person responsible for track assignment...)

Something I do not know is whether the 'friendliness' of the investigators to some of the people called in was genuine, or the sort of technique used to 'gentle' potentially hostile or dissembling witnesses into letting their guard down and 'volunteering helpful information'.  I like to think it was the former.

[BaltACD]

Wizlish

Something else that needs to be covered if we're going to start a shoulda-woulda-coulda blamefest on the part of the engineer noting the 'oscillation and noise' is the procedure by which those observations would work themselves 'quickly enough' up the chain of notification and down the chain of command to get quick remediation.  Perhaps there needs to be some procedure (within CSX, or perhaps NORAC/GCOR more formally) for implementing the equivalent of a "911 call" that puts an automatic slow order out (via the PTC system, most probably, and presented as more a civil restriction than an operating one so it's understood clearly) - this would have some specific 'rapid response' checking with the database of track defects referred to in the testimony and ideally some feedback on this to the engine crew(s) concerned so they know what they might be experiencing.  (Do I think that's practical to implement, or that crews would pay much attention to it if it generates 'cry wolf' slow orders too often?  don't ask too often... but someone like Sarah might)

Current procedures - Crew reports track anomally - Dispatcher puts out Restricted Speed Slow Order for the location of the anomally, automatically - The location of the anomally gets reported to MofW for inspection.  After inspection, MofW will change the slow order as they see fit, or take the track out of service.

[Wizlish]

BaltACD

Current procedures - Crew reports track anomally - Dispatcher puts out Restricted Speed Slow Order for the location of the anomally, automatically - The location of the anomally gets reported to MofW for inspection. After inspection, MofW will change the slow order as they see fit, or take the track out of service.

BaltACD

Current procedures - Crew reports track anomally - Dispatcher puts out Restricted Speed Slow Order for the location of the anomally, automatically - The location of the anomally gets reported to MofW for inspection. After inspection, MofW will change the slow order as they see fit, or take the track out of service.

Leads me to wonder -- tongue only partly in cheek -- whether there ought to be some apparatus on the engine that can mark the rail with a different color than that used for any of the Sperry cars, to identify the precise location of an issue and perhaps allow recognition as a complication or new problem among perhaps-already-recognized issues even before the MoW people can get there to check things.

Something I'm still not quite seeing: the discussion clearly indicated that a 10 mph restriction was discussed for the rail section we're concerned with, but was not imposed (the 25 mph constituting 'slow enough' for their procedures).  The problem is that too many 'foresight' 10 mph restrictions are going to cripple the railroad.  Does the dispatcher have some idea of how serious a 'track anomaly' is from the description he gets and the questions he subsequently asks?

Hey, another potential use for those octos!  Drop in and get some high-resolution pictures and scans...

[BaltACD]

Wizlish

 

BaltACD

Current procedures - Crew reports track anomally - Dispatcher puts out Restricted Speed Slow Order for the location of the anomally, automatically - The location of the anomally gets reported to MofW for inspection. After inspection, MofW will change the slow order as they see fit, or take the track out of service.

Does the dispatcher have some idea of how serious a 'track anomaly' is from the description he gets and the questions he subsequently asks?

When the crew mentions 'rough track' - that is all that is needed to issue a Restricted Speed train messages until the track is inspected.  Dispatchers are not qualified to make decisions on how serious a track situation is - that is what MofW personnel are trained for.  The ONLY thing the Dispatcher needs to know is that the train crew took exception to the track quality at a specific point - the rest is up to MofW.  Train crews don't necessarily know that track has been Sperry inspected or when the inspection occurred, nor do they really care - it isn't their job.

Remember the requirements of Restricted Speed - 'proceed at a speed that will permit stopping short of obstructions or track anomallys, within 1/2 the range of vision, not exceeding 15 MPH'.

[tree68]

Wizlish

Leads me to wonder -- tongue only partly in cheek -- whether there ought to be some apparatus on the engine that can mark the rail with a different color than that used for any of the Sperry cars, to identify the precise location of an issue and perhaps allow recognition as a complication or new problem among perhaps-already-recognized issues even before the MoW people can get there to check things.

It would be imprecise at best.  By the time the crew says to themselves, "boy, that was rough," and reacts by hitting the "mark this spot" button, they may well be a significant distance beyond the actual location, especially at higher speeds.

On the other hand, their knowledge of their territory can lead to a more precise location, both by milepost and perhaps a physical characteristic.  Odds are the restriction would be for the next nearest mileposts in both directions anyhow, at least until something more specific is found.

Then we get into maintenance of the marking devices...

NORAC defines restricted speed as already noted, but with a limit of 20 MPH.  We use 20 MPH on the main, but 10 MPH in "Rule 98" by timetable rule.

Remember, too, that a 5% crack may not "register" with the crew at all.  Even a completely broken rail, if still properly secured, may pass unnoticed.  And speed can make a difference.  Track that is downright nasty at 15 MPH may seem relatively smooth at 40 MPH.  Been there.

This is why track is regularly patrolled, and inspected with devices like the Sperry cars.