A Review of the Department of Transportation's Plan for Analyzing and Testing Electronically Controlled Pneumatic Brakes Letter Report (Phase 2) (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

26 Appendix G DOT Response to NAS Letter Report on Electronically Controlled Pneumatic Brakes This appendix contains briefing materials presented to the committee by DOT on July 6, 2017. Summary The NAS had three primary recommendations:  Consider specific brake response profiles  Confirm parameters through multivariate analysis  Additional model validation The DOT has responded to these recommendations as outlined here:  Conducted additional field and rack testing to capture brake response profiles - Included these in the simulations and analyses  Conducted regression analysis and interfaced with the RSI-AAR project on this effort - These confirmed that the variables included in the DOT’s analyses were appropriate, inclusive, and prudent  Reasonable validation has been completed NAS Recommendation # 1 Emergency Brake Propagation Profiles 1.1 Use test racks to simulate the performance of the different braking systems 1.2 Measure the time required for the emergency application of ECP brakes and DP and EOT systems from initial train separation to the point at which all of the brakes on the train are fully applied.  Conduct tests on in-service, full-scale locomotive and revenue car trains that are standing and moving slowly.  Collect EOT and DP radio-system operational delay times and then apply them in test rack simu- lations. DOT Test Actions – Brake Propagation Times Field Testing at the NS yard in Conway, PA on a 100-car oil train to measure:  Operational delays [radio latency] from DP and EOT trains  Brake cylinder pressure build-up curves from an operational train Rack testing at the NYAB rack in Watertown, NY to measure:  Brake propagation and cylinder build-up profiles for DP, EOT, and ECP systems from the point of hose separation

Field May 31, 2 Test for the FR Obse At variou  B  T  D  B A total of Note that brake pipe testing was 017. ing and prim A. Additiona rvers include s locations th rake pipe wa ime between ifference betw rake cylinder 32 tests were , although EO in response conducted o ary data coll l data collect d representa roughout trai s vented to in BP venting a een these tim build-up pro DO performed, 2 T transmits to this signal average std dev number of t Overview o n a stationar ection was p ion was perfo tives from AA Te n, the followi itiate emerge nd loss of BP es is the rad files at multip T Test Resu 6 of which p emergency m . front t back 1.83 0.17 ests 12 DP Ra Ti   27 f Conway, P y revenue tra erformed by rmed by GE R, NS, BNS st Process ng actions w ncy pressure at e io system tra le cars were lts – Conwa roduced radi essage to th o  back to  front 2.29 0.31 7 dio Latency  me, sec A Tests in in Norfol NYAB, Wab . F, CN, TRB ere performe ach end of tr nsmission la also collecte y, PA Tests o latency dat e lead loco, front to  back b f 1.81 0.12 7 EOT Radio La Time, se k Southern’s tec, and Sha , FRA and NT d: ain was colle tency d. a. the locomoti ack to  ront N/A N/A tency  c Conway Ya rma & Asso SB. cted ve does not rd on ciates dump

28 Overview of Watertown, NY Tests Rack Testing was conducted on NYAB brake rack with:  Conventional air brakes in both DP and EOT configurations, and  ECP brakes in a standalone configuration Testing and primary data collection was performed by NYAB and Wabtec. Observers included representatives from AAR, NS, GE, BNSF, TRB, and FRA. Test Process Other configuration details:  The DP units were connected to each other through a wired network; no additional delay was added.  The EOT configuration was laboratory specific and thus, 0.5 seconds of delay was added. At various locations throughout train, the following actions were performed:  Brake pipe was vented to initiate emergency  Time between BP venting and loss of BP pressure at each end of train was measured  Time from hose separation to the last car reaching full brake cylinder pressure was measured  Brake cylinder build-up profiles at multiple cars were also collected Test Results – Watertown, NY Tests Average DP latency: 2.2 seconds Average EOT latency: 1.95 seconds Time for last car to achieve full brake pressure:  DP/EOT systems: 13.8 to 10.8 seconds, depending on location of hose break  ECP systems: 10 seconds (regardless of venting location) Time delta between hose separation and brake initiation on adjacent car:  DP/EOT/Conventional Systems: 0.17 seconds  ECP Systems: 0.67 seconds (for all cars) Updated Analysis Configurations Based on the test results from Conway, PA and Watertown, NY, the derailment simulations and puncture analyses were revised. Revisions included:  DP/EOT radio transmission delay of 2 seconds  Measured brake cylinder pressure profiles  Measured brake propagation times, adjusted for brake pipe length

 M – – The data fo the GE dat field condi [NOTE: T The Derailmen Analys Original Revised 2.1 Use on model 2.2 Crea accident r 2.3 Dev relates to but are kn easured time DP/EOT/C ECP Syste r the conventi a. This data in tions better th he slide repr likely numbe t) at 40 mph is Conv multivariate predictions. te a merged ecords and ta elop a multiv a set of speci own with ins delta betwee onventional ms: 0.67 seco Revi onal/DP/EOT cludes the eff an test rack d esents profile r of punctur . Likely entional 6.6 6.3 NAS Recom analysis to en database for nk car–specif ariate regress fic accident c ufficient accu n hose separ Systems: 0.1 nds (for all c sed Brake C system is a co ect of the actu ata. Propagati s only for the Revised es was recalc No. of Punctu DP/EOT 5.9 5.8 mendation sure DOT’s tank car dera ic informatio ion model fro ircumstance racy, and the   29 ation and bra 7 seconds ars) ylinder Pres mposite of the al brake riggin on time is no railroad car Analysis Res ulated for a res ECP/O 4.3 4.7 #2 DOT Ana plan is focuse ilment accide n as found in m the merge s. Seek insigh refore are wo ke initiation o sure Profiles measured bra g on the cars, t shown on th adjacent to ults 100 car train ECP over CL lysis and Te d on the fact nts from FRA the RSI-AA d database to ts into which rthy of testin n first adjace ke cylinder pr and is though is chart for co the point of d (100 cars b Advantage onventional 35% 25% st Plan ors that have ’s RAIRS t R tank car ac examine ho analysis pre g or addition nt car: essure profile t to represent nventional an erailment.] ehind the Po ECP Advant over DP/EOT 27% 19% significant im ank car derai cident databa w tank car sp dictors are st al measurem s from actual d DP. int of age pact lment se. illage rong, ents.

30 DOT Actions in Response – Multivariate Analysis  Collaborated with RSI-AAR [Railroad Tank Car Safety Research and Test Project] on the poten- tial for combining the databases and conducting additional analysis  Conducted multiple regression analysis of data from FRA’s RAIRS database  Both of the above actions confirmed that the variables considered in the DOT’s Analysis were appropriate, inclusive, and prudent Engagement with the RSI/AAR Data Group  The RSI-AAR data group graciously agreed to engage with the FRA on this effort  Several conference calls were held and the FRA shared a potential list of questions that could be answered by combining the data sets  In the end, both parties agreed that literally merging the FRA and RSI-AAR databases does not offer any value: - However, the RSI-AAR data group identified several key variables that should be considered. Key Variables From RSI/AAR Data Group All of these variables were included in the DOT’s original analysis Car features:  Thicknesses of tank head and shell  Steel specification for tank  Inside diameter of shell  Jacketed: yes or no  Head shield: none, half-height, full-height  Top fittings protection; yes or no [evaluated separately]  Bottom fittings: present, none [evaluated separately] Accident details:  Train speed at time of derailment  Number of freight cars derailed  Location of derailed tank car within derailed string DOT’s Multiple Regression Analysis  Rail Equipment Accident data was queried from FRA website (RAIRS database)  200,908 records reported between 1975 and 2016  Data was filtered and cleaned  Statistical analysis: multivariate regression model used to determine factors that have significant impact on derailment damage predictions Summary – Recommendation # 2 The DOT puncture analysis approach includes all of the variables identified as significant predictors of derailment severity as recommended by both:  The RSI-AAR Tank Car Safety Project  DOT’s internal multivariate analysis

Overall, th ed in the D 3.1 Vali ual derailm modeled s prediction Review o large leve  T co Ther  E  It Prior wor qualitative tured  T e DOT belie OT approac date the mod ent event w cenario shou is to be com f accident dat l of variabilit he level of sc mmittee for efore, simula specially, for is more critic k published b ly shown ex his work was ves that the s h are signific NAS R eling approac ith the actual ld be the sam pared. D a (in general) y in accident atter in derai consideration D tion of a spec a model that al to ensure y the DOT cellent agree done at an ag tatistical ana ant, prudent a ecommend h by compar outcome of e as that inv OT Respons , as well as, r severity due lments is sign in developin OT Respons ific derailme seeks to estim that the mode (and shared ment on key gregate leve   31 lysis conduct nd effective ation #3 Mod ing the mode that one even olved in the a e to Model V ecent multiv to random fa ificant [Note g its Phase 1 e to Model V nt may not b ate an ‘avera l captures the through lette parameters, s l and not at a ed so far con . el Validatio l’s predictio t. The type o ctual train de alidation ariate analysi ctors : The figure b report.] alidation e a good pred ge’ value inte mean and th r reports with uch as the nu n individual d firms that th n n of the outc f braking sy railment wit s have confir elow also w ictor of mod nded for an e e trends app the NAS a mber of cars erailment le e variables in ome of an ind stem include h which the m med that ther as provided t el performan conomic ana ropriately nd the public derailed or vel. clud- ivid- d in a odel e is a o the ce: lysis. ) has punc-

32 DOT Response to Model Validation In addition, the rear car distance traveled in a comparable set of LS-Dyna simulations was compared to the Aliceville locomotive’s event recorder data, and found to match with a difference of less than four percent.  This shows that in spite of all the potential variations, our derailment simulations closely matched reality, as evidenced by the event recorder download. Supplemental Materials Brake Force Graphs22 NOTE: Prior to ECP brake initiation, conventional braking occurs as is indicated in the plots for Cars 1, 5, and 11. The plot labeled “ECP” represents the profile for all cars following Car 11, after initiation of ECP braking. Data were obtained from testing conducted at the New York Air Brake facility in Water- town, New York. 22DOT’s graphs representing ECP Overlay Brake Cylinder Propagation and Conventional Brake Cylinder Propa- gation, received on July 7, 2017. 0 0.2 0.4 0.6 0.8 1 1.2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 N or m al ize d  Fo rc e Time, seconds ECP Overlay Brake Cylinder Propagation Car 1 Car 5 Car 11 ECP

NOTE: D York. ata were obtained from te sting conduct   33 ed at the New York Air Brake facility in Watertown , New