Framingham/Worcester Line Questions

[jaymac]

BandA » Thu Jul 21, 2016 6:40 pm
I don't know if people in West Suburban, Metrowest and Worcester area have less political pull, or if it is because the "T" purchased the perpetual easement more recently that riders on the Framingham/Worcester line seem to get less attention?

Used to be that the Worcester Line seemed first among equals, now it seems to be merely among equals. Lt. Gov. Tim Murray resigning all those years ago probably started the loss of primacy, but the Baker administration -- Stephanie Pollack to the contrary -- seems to take a more adversarial approach to most things T than did the previous administration.

[dbperry]

But there is a LOT of political pull for Framingham-Worcester, especially at the Worcester end. Tim Murray still has considerable influence and a platform / (loud) voice as president of Worcester Chamber of Commerce. Lt. Gov. Polito is from Shrewsbury. Rep. Hannah Kane is from Shrewsbury and gets involved with CR issues. Secretary Pollack is from Newton, and although she doesn't take the commuter rail very often, she does take it sometimes, so she has at least more first-hand / personal usage knowledge of our line than others. Even the Mayor of Worcester has some pull at the statehouse.

Lt. Gov. Polito was very influential in advocating for the 'bullet' / #HeartToHub train. And I've got some pretty cool documents to back that claim up...all I can say about that.

We can all argue about the usefulness of the bullet train, but it is purely a politically driven train. So that's 'attention' even if it is maybe misguided or misplaced (for most of us).

Hard to say if the line would have gotten the rail destressing and tie replacement work without political pull. The acquisition of the Framingham-Worcester segment from CSX might have been all that was needed to jump start the work. But a LOT of money flowed to those projects - and not all at once - the $$ got released in waves. I believe there was probably some political influence in pushing the work along and getting it all done essentially in one continuous mobilization (kind of). In the grander scheme of the MBTA and government, I actually think the rail destressing projects moved relatively quickly.

The recent culvert work in Natick and the project at Natick Center to mitigate flooding there is another example of $$ coming our way. Again, probably not politically driven and more of an operational or safety budget line item, but still $$ coming our way.

All the delays caused by the construction make it SEEM like we get the short end of the stick, but in fact we're getting a disproportionate amount of money and work.

And as I've mentioned on my blog, we've got one more season of construction pain to live through (tie replacement on track 2 this fall).

As far as I know, P508 and the other 8 double coach set from from Providence in the AM - along with their corresponding return trips (P521 on F-W) - are still the only "high priority" trains on the south side. Although I wonder if the bullet train gets added to that list. Supposedly that means more attention, better dispatching, & better equipment, but I'm not sure it makes any difference in the real world. For a while those 'high priority' trains ran with two locos as insurance against breakdowns.

[Komarovsky]

Today Keolis announced changes to the timetable for certain inbound and outbound trains to try and address the terrible OTP that have plagued the Worcester line for the past two months under the new schedule (see my analysis of the first month here: viewtopic.php?f=65&t=161520&start=330#p1392367). All the changes are an increase of 5-10 min to the running time of selected trains, all for trains that arrive/depart SS during the afternoon rush. Full details here http://www.keoliscs.com/news/worcester- ... justments/

Breakdown of increases below:

Inbound trains
588: 10 min increase
522: 10 min increase
526: 5 min increase

Outbound trains:
517: 5 min increase
519: 7 min increase
521: 7 min increase
593: 7 min increase
523: 10 min increase
525: 5 min increase

Given the speed of these adjustments, it's clear that track work wasn't the sole cause of delays and perhaps the schedule was based on some faulty assumptions (especially the 522/521/593/523 dependencies that kills the OTP of all those trains). Keolis also mentions changes come the fall, which should also be interesting to contemplate as more info comes out.

[Rockingham Racer]

Aren't there companies with simulators that can show how performance goes prior to putting out a schedule? Does anyone know if Keolis bothered to use one?

[MBTA F40PH-2C 1050]

No. Franklin line is not as bad as Worcester, but the afternoon gets ugly with some of the meets and delays trains anywhere between 5-20 minutes almost every night.

[dbperry]

Aren't there companies with simulators that can show how performance goes prior to putting out a schedule? Does anyone know if Keolis bothered to use one?

Remember, the MBTA set the new schedules. Keolis just has to operate them (from what I've heard, they had some input but not a lot). Maybe you knew that & I'm just being ticky tacky.

Anyway, yes, they hired a consultant and there were computer simulations run to model the runs. Some of the complaints about the modeling I've heard are:

- the modeling did not properly account for dwell times in stations - a generic average was used, which isn't appropriate for the high load busiest trains.
- the modeling did not properly account for the Delay In Block rule (see definition in my glossary: http://framwormbta.weebly.com/glossary" onclick="window.open(this.href);return false;
- the modeling did not properly account for the effective 15 mph speed through the CP 4 switch near the new Boston Landing station on the Framingham-Worcester

All of those reasons make the modeling (and therefore predicted transit times / trip durations) more optimistic (faster) than real world conditions.

A description of the modeling process is shown in this PowerPoint starting around page 26:
http://www.dbperry.net/MBTA/articles/mb ... iative.pdf" onclick="window.open(this.href);return false;

[BandA]

Today Keolis announced changes to the timetable for certain inbound and outbound trains to try and address the terrible OTP that have plagued the Worcester line for the past two months under the new schedule...

Pathetic. In the '80s and '90s you could set your watch by this train line.

[Komarovsky]

Dave, thanks for linking to that that provides a bit of detail on the modeling. I work on modeling of another sort and I'm really surprised to see how little uncertainty the output provides around the results. In my business there is a lot of sampling along distributions for variables in the model(amongst other techniques) to try and quantify that uncertainty(or at least make the model a bit less deterministic). If the RTC modeling doesn't include these uncertainties, I'm both surprised and disappointed in the transit modeling industry for doing such a bad job.

[Rockingham Racer]

+1. As we all know about computers: lousy input yields lousy output. Then there's the actual field testing of doing a run to see if the simulation is accurate.

Thanks, Dave, for the info. At least there was some attempt to model the real-life service.

[leviramsey]

Dave, thanks for linking to that that provides a bit of detail on the modeling. I work on modeling of another sort and I'm really surprised to see how little uncertainty the output provides around the results. In my business there is a lot of sampling along distributions for variables in the model(amongst other techniques) to try and quantify that uncertainty(or at least make the model a bit less deterministic). If the RTC modeling doesn't include these uncertainties, I'm both surprised and disappointed in the transit modeling industry for doing such a bad job.

Including the uncertainty pretty much implies increasing the padding (and limiting the capacity) at the points of contention along the line. If you had the model use sufficiently realistic distributions for the relevant variables (for reducing rush hour delays to any desired level less than recent schedules), I suspect with a very high degree of confidence, the model would conclude that no schedule exists for the Framingham-Worcester line that provides more than, say, two-thirds of the current count of rush-hour trains. If you want a schedule for the line that has the number of trains (especially the number of expresses and reverse commutes and without massive gaps in the schedule) and you're not interested in/able to rebuild the line almost from scratch (some combination of starting over with the signalling, adding three-track sections in places, adding several new crossovers, consolidating stations (Newton, Wellesley, and Natick, I'm looking at you), etc.), then you're going to have to live with a schedule where 15-30 minute late trains aren't really that rare; it wouldn't surprise me at all if the MBTA knowingly used unrealistic distributions in order to get a schedule that wasn't a major service cut. Cut a third of the trains from the rush hour schedule, or close a few stations, and the pain is concentrated enough that it's politically untenable. Consistently late trains spread out the pain enough that the politics allow for a status quo to be preserved, and I think it's an open question (given the anguish over things like destressing, which are the mere tip-of-the-iceberg of what the line needs) whether there's enough political will to allow for, say, a year of 7-day-a-week bustitution or multiple years of smaller-scale service disruptions to get the line into a state appropriate for its current and future level of service.

[Komarovsky]

I agree with you somewhat with your contention that there is probably no 100% delay proof scheduling for an average day on the line without a ground up rebuild but I disagree with the need to have 15-30 min delays, especially when it's the same 3-5 trains every day. If the modelers and planners were doing a good job they'd work on solving the optimization problem that exists between delays and the schedule as a whole, so you wouldn't have clustered and correlated delays but clearly they've not done that in this case. Given the fact that it's always 522, 593 and 523 that are late every day, I think that the assumptions used in scheduling those trains are totally wrong and wouldn't be solved by modeling uncertainty correctly.

[dbperry]

The pre-May schedule generally worked for on-time performance, especially when there were no outside disturbances (i.e. construction). And it had the same general overall traffic capacity. So I don't buy the argument that there is no valid solution.

I'm not sure about including uncertainty or using a Monte Carlo type multiple variable analysis. I guess you could include a range of assumptions for acceleration profiles, braking patterns, and dwell times, but I would guess the output of that model would only vary by a few minutes across an entire trip duration. And it's entirely possible that the modeling did that (at least for acceleration & braking). We don't have the full documentation of the modeling.

But before we go too far off topic and start debating computer modeling techniques ad nasueum, I think we can recognize that the MBTA is now willing to admit that the modeling was flawed and the new schedules don't work. Hence the addition of time to trip durations for this line and others.

Comparing present day performance to any time period from long ago is also inappropriate, given the expansion of service and the growth in ridership. I will agree with the idea that the current schedule is probably the maximum number of trains that can fit on the line for the AM and PM rush hours. You could get more capacity for riders by making every train an 8 double set, but that's the only way you could increase overall capacity.

[BandA]

The Newton and Wellesley stations are not the cause of schedule delays. They are "transit oriented development" for the past 180 years, and they've gotten no investment for the last 50 years. Bypassing them only gets you only a couple minutes per station in reduced runtime/dwell time. And many trains already bypass such that rush hour service is less than it used to be before the Worcester extension. Consolidating stops would require construction of very expensive garages instead of walk-up customers & spread out parking.

Looking at the May schedule, the closest together the trains run is 8 minutes. There is essentially no freight train interference on this line. The "Rabid-Transit" lines run as close as 4 minutes on double track I believe, though they have smaller blocks. This entire line with the exception of Beacon Park Yard is double track. So if running express and local trains and reverse-commute trains is too complex for the blocks & crossovers, then you could run all locals with right-railing in the AM & "wrong"/left-railing in the PM and should easily be able to accommodate the number of trains on the schedule. Use that as the base, then figure out how much "complexity" can be added back.

The big problem with schedule keeping is equipment failure and the delays accessing South Station.

[MBTA F40PH-2C 1050]

They are running too many trains on track (CP3-CP21) with out dated signal technology. Trains at Rush hour are "riding the Yellows" to their destinations. As B&A wrote, delays is what messes up the rush hours...too many trains running on long, 2 mile blocks outside of CP11

[Komarovsky]

They are running too many trains on track (CP3-CP21) with out dated signal technology. Trains at Rush hour are "riding the Yellows" to their destinations. As B&A wrote, delays is what messes up the rush hours...too many trains running on long, 2 mile blocks outside of CP11

Agree, many times I've noticed a sudden drop in track speed on an express and after going to Stephan's web page I see a behind schedule local as the culprit. 90% of the time we end up stuck at restricted speeds the entire way to Framingham because of the lack of crossovers as well.

That said I have a feeling that there is a more optimal schedule for the current number of trains in the schedule, it's just that this set of simulations(for whatever combination of reasons discussed above) didn't deliver it. The other half of computer models of course is that you still need an expert to look over the results and to do some sanity checking. If trains with the same departure time and number of stops are suddenly scheduled 5-10 minutes faster according to the new simulation, that should be a major red flag that something is wrong.