Ten90 Solutions LLC

I have, periodically and repeatedly, attempted to determine the “reliability” of the PTC systems deployed by the seven US Class 1 railroads.  Of course, the first thing I should do is define what I mean by reliability. 

Given the paucity of publicly available information concerning PTC problems experienced in day-to-day train operations, I use the broadest brush to define that reliability—and that would be the number of trains that begin and finish their assigned daily movement without an “unscheduled” misfunctioning, disabling, rebooting, reinitializing, cutting out, by-passing of the on-board PTC equipment including the visual display.  That’s about as broad as I could make a pass-fail criterion.  It might not be broad enough.

Because everything railroad has to do with performance over time, failure makes sense as a sample, a percentage, a frequency, a ratio of events or incidents to the total opportunities, or total train movements in a specific time period. 

I’ve tried to come up with a total number of trains operated by Class 1s, excluding passenger trains, work trains, yard trains, “yard-limit” train movements, light-engine movements, etc., each day.  I guessed 2500 per day based on…what?  Based on something I read somewhere that half the freight trains in the US operate at some point in their travel through Chicago, and some other thing I read some other where that the number of freight train movements in Chicago is 1200 each day.  Like I said I guessed. 

I had a guess for the denominator of the calculation and absolutely nothing for the numerator.  Expressed mathematically, that’s  0/? which isn’t even close enough for government work.  

However, speaking of the government, two different agencies have provided me with the first approximations of what might be real information that might be used to populate the fields of numerator and denominator, while at the same time not overwhelming my 5^th grade elementary school mathematical skills, bless you Ms. Schrager. 

In response to the alarm raised by numerous shippers over the recent decline in rail service performance and reliability, the US Surface Transportation Board has required Class 1 rails to provide weekly additional reports on their performance.  One of those supplementary requirements is a report of “weekly average number of train starts per day by [type of] train,” which, footnote (2) says “data augment the weekly average number of trains held per day data currently reported here under 49 C.F.R. § 1250.2(a)(5)), sorted by train type (intermodal, grain unit, coal unit, automotive unit, crude oil unit, ethanol unit, other unit, and manifest).” 

So we have a way to calculate a denominator. 

As for a numerator, NTSB issued its report, and opened its docket, of its investigation into the August 25, 2021 stop signal violation and subsequent collision of Canadian Pacific train 296-23 near St. Paul, Minnesota.  Included in the docket is the daily log of PTC trouble reports communicated to the CP’s PTC “help desk.”  The docket document covers the period from July 26 to August 25, 2021.  The log identifies 63 incidents in the 30-day period.  Five reports pertain to Amtrak trains, and we eliminate them from the total, leaving 58 incidents, or about 2 trains reporting PTC trouble per day.

During this time period, CP (identified in the STB data by the names of its once US subsidiary “Soo”) weekly average of daily train starts as reported to the STB was calculated (don’t go anywhere without the Sharp EL-S25 calculator its New York Yankees livery) at 25.38 trains per day, for our purposes, 25 trains per day. So our rate of failure per trains per day is 2/25 or eight (8) percent. 

Using the data reported to the STB by all seven Class 1s, the weekly average of daily train starts measured 1920.34 starts per day.  Now we assume CP is having no more or less problems with PTC than all railroads, so that our eight percent incident rate equates to (possibly) 154 trains per day with (possible) PTC failures. 

Yes, I know the sample is small.  I know the assumptions may be mistaken.  I know I’m leaving out Conrail Shared Assets and all the switching railroads that may be required to operate with PTC engaged over PTC-rules territory.  I know all that.  But this is the closest I get to even take a shot at estimating reliability given the paucity of available data.

 I sincerely hope my assumptions are unwarranted, that CP was having an extraordinarily bad reporting period, and that I’ve made errors of such severity that I should be forced to reup in the fifth grade of my former elementary school and sit at the front of the class so I might get the special help I so much require.  Pray for the teacher. 

I hope all that because an eight (8) percent failure rate of a train-control apparatus is unacceptable. 

Imagine that all 1920 daily train starts took place on a single commuter railroad providing service to a metropolitan area.  The equipment requirement to provide the scheduled service is 350 trainsets. The trainsets have locomotives/power cars on each end.  Your fleet includes reserves, a shop margin, of 35 trainsets, bringing the fleet size to 385 trainsets.

The 8 percent train-control failure rate removes 28 trainsets from service each day. Mandatory locomotive inspections claim another 4+ trainsets daily.  In a single day, the inspection requirement and the PTC failures alone have nearly consumed your shop margin. If any additional defects develop in the operating fleet, you’re in big trouble.  If any repairs, inspections, maintenance cannot be completed within 24 hours, you’re facing service reductions due to equipment shortages. Sure you can increase your shop margin, but engineering reliability is supposed to improve asset utilization, not impair asset utilization.  

Sure, you already have a big-ass shop. And sure you can work three shifts each day at the shop, but that is an extraordinarily expensive way to mitigate by operating practice engineering failures.  Engineering reliability into system-wide applications is supposed to reduce the unit costs of train operations, not increase them.

The problem is not contained or expressed solely, fully, or adequately by the minutes of growing train delays.  The problem is made manifest in two critical factors: 1) the inability to meet the service requirement and 2) the rate of train-control failures degrades and eventually nullifies the purpose for originally installing the apparatus.  

magine you’re in charge of this commuter railroad with this density of service, and during each hour of service you have to deal with 7 train-control failures.  The operation of so many trains under the special restrictions for failed train-control movement magnifies the risk of human failure to a point where the risk exceeds that risk which existed before the equipment was ever installed.

Now clearly US Class 1s do not operate with that density of movement.  Space, gaps in time and distance reduce the risk of any single train-control failure jeopardizing the safety of other trains  However, in that case, the reduced risk to safe train operation is dependent on luck, chance, rather than the application of the train-control system.  The value to your operation embodied in  the train-control system heads south of zero.

 So please, somebody with better data prove me wrong.  Make my New Year.

David Schanoes 

January 5, 2023