I recently discovered the results of the FRA's Positive Train Control Passenger Braking Algorithm Enhancement project dated September 2023. As regular readers are aware I have been a frequent critic of the overly conservative braking algorithms used by various PTC systems and vastly under perform the performance capabilities of the rail vehicles. While this FRA study was confined to the entirely wireless ETMS system used outside the northeast, it might still offer meaningful improvements to these systems in general. Long story short don't get your hopes up. In fact the results of this study might make things even worse.
So I encourage everyone to try and work their way through the linked paper (mirror here), but in my quest to add value I will provide a summary of the key points. The study examined four potential algorithm enhancements, Target Approach Management (TAM), specified consist length trains, tuned train types, and adaptive braking. TAM is for low speed (5-10mph) approach to stop signals while the other three are general purpose braking algorithms that better take into consideration various quirks of passenger train and EMU/DMU operation. The study was almost entirely carried out via computer simulation where real world variables like train consist, rail adhesion and brake pipe latency were tested in tens of thousands of combinations to estimate real world performance. There also appeared to be real world tests carried out on a test track.
The first important takeaway is that the performance baseline for defining an "undershoot", as in PTC stops the train too quickly is 500 feet for speeds under 30mph and 1200 feet for speed over 30mph. Under the baseline scenario these limits were exceeded 20-25% of the time. The PTC braking algorithms calculate a stop point probability distribution and then add a safety factor. Higher speeds mean more uncertainty so the typical stop point for 25mph might be a couple hundred feet short of the signal, the stop point for 90mph could be multiple thousands of feet short. This is why PTC forces so much aggressive braking at line speed.
As for the results of the study the good news was that the new TAM algorithm improved low speed undershoots (defined as more than 100 feet) from nearly 50% under the current proprietary EMTS implementation to a fraction of a %. Unfortunately that's all the good news you are going to get as the other "improved" general purpose algorithms blew up undershoots in the simulated runs. Specified Consist undershoots increased between 9 and 17% with some passenger train consists undershooting up to 40% of the time. Tuned Train Type saw mixed results with some equipment having less undershoots and some more. Adaptive showed no change for the commuter type equipment while undershoots for passenger type equipment went up from 20 to 30%. The gain for this drop in performance was an improvement in meeting the stop target from 98 to 99%. The real life tests resulted in a mixed bag of performance gains and losses, however unlike the costless simulations far fewer real life tests could be carried out and under much more limited conditions.
The problem with the study is that it ultimately treats a safety overlay system like autonomous operation. PTC isn't what has to stop a train before it hits something. That's the job of a skilled locomotive engineer. PTC is what should step in when the Engineer is clearly going to be doing something unsafe. Even unsafe conditions rarely lead to physical impacts at which point the crash safety system prevent potential injury or death. PTC preventable accidents of all types were already rare and PTC should eliminate at least 98% of those. Meanwhile poor train performance is pushing riders onto the roads where they'll die in car accidents. Celebrating a 1% improvement in PTC effectiveness at the cost of yet more performance isn't the win the FRA thinks it is. The infuriating part is that over on those highways self-professed self driving cars are blowing throw stop signs left and right and Federal regulators do nothing.
No comments:
Post a Comment