Tuesday, December 30, 2014

Improving LOSSAN-South Part 1: Surfliner rolling stock

This is the first of several posts looking at various ways to improve the performance of the Pacific Surfliner, Metrolink, and Coaster services upon the Los Angeles-San Diego-San Luis Obispo rail corridor as well as the cost-effectiveness of such improvements. As I lack an employee timetable for Union Pacific's Santa Barbara subdivision, this is necessarily limited to the San Diego to Los Angeles section rather than extending north to Santa Barbara. Additionally, for the sake of my own sanity in putting this together, it only looks at northbound travel. I do believe that any performance gains are likely to be substantially mirrored, but they will differ somewhat. This particular post will look at replacing the current Pacific Surfliner equipment with a variety of other options.

A few notes before continuing on:
1) No attempt was made to model the impact of inclines upon train performance.
2) An aggressive stance was taken towards acceleration and braking which may not be the actual practice with Amtrak.
3) The information that I am using to construct the timetable is several years out of date (and a newer Metrolink timetable has replaced it); there are track improvements that are not reflected in it.
4) This is not a professional level simulation (it is, in fact, using spreadsheets to solve it as a set of physics problems) and while I think that the output is generally going to be substantially accurate, it shouldn't be treated anything at all like it's gospel.
5) I've done this with a unidirectional track assumption, which meant going onto a siding with a couple of Coaster stations where the northbound platform is on a siding a bit than the mainline where the Surfliner (normally) doesn't stop, which does slow the train somewhat over a purely optimal run.

Lastly, I would like to thank the folks at the Railway Performance Society, especially David Stannard and John Heaton, for their assistance with British train performance.

To begin with, here's a profile of the Surfliner corridor up to Los Angeles. Note that while there is a maximum speed of 90 miles per hour, there are a number of slow downs which severely limit the amount of time spent at that speed and resulting in the current average speed of only 46.5 miles per hour.


For reference, the following is a list of stations on the corridor with normal Amtrak stations in bold.
San Diego Santa Fe 0.0 miles
San Diego Old Town 3.3 miles
Sorrento Valley 18.5 miles
Solana Beach 25.7 miles
Encinitas 29.8 miles
Poinsettia 34.2 miles
Carlsbad Village 38.3 miles
Oceanside 41.1 miles
San Clemente Pier 62.7 miles
San Juan Capistrano 70.3 miles
Laguna Niguel 73.8 miles
Irvine 82.5 miles
Tustin 88.0 miles
Santa Ana 92.3 miles
Orange 94.9 miles
Anaheim 96.9 miles
Fullerton 102.2 miles
Buena Park 107.3 miles
Santa Fe Springs 112.6 miles
Commerce 119.1 miles
Los Angeles 128.8 miles

Now let's look at that with the performance of Amtrak's current equipment overlaid upon it.
Acceleration was derived from the Fairmount DMU study (with some extrapolation) and the minimum breaking specifications for the Next Generation Bilevels (page 37). Both acceleration and braking may be understated, especially as the Fairmount study train set tops out at only 90mph and the Surfliner equipment is rated for 110mph. However, rounding up every arrival to the next minute (which is probably overly conservative), adding two minutes of dwell time at every station (except Old Town), and adding a 7% pad into the Los Angeles arrival time produces a run time of 2:49, comparing well with the actual best scheduled time of 2:45. On the other hand, the model appears to be somewhat faster than the actual trains on during the journey with an accumulation of 1-3 minute errors; compare with the first run of the day for example



Amtrak 763 Model Push-Pull Difference Intermediate Differences
San Diego
0h 0m
0h 0m
0h 0m
0h 0m
Old Town
0h 7m
0h 6m
-0h 1m
-0h 1m
Solana Beach
0h 38m
0h 35m
-0h 3m
-0h 2m
Oceanside
0h 57m
0h 52m
-0h 5m
-0h 2m
San Juan Capistrano
1h 29m
1h 21m
-0h 8m
-0h 3m
Irvine
1h 43m
1h 34m
-0h 9m
-0h 1m
Santa Ana
1h 54m
1h 46m
-0h 8m
0h 1m
Anaheim
2h 3m
1h 56m
-0h 7m
0h 1m
Fullerton
2h 12m
2h 5m
-0h 7m
0h 0m
Los Angeles
2h 45m
2h 49m
0h 4m
0h 11m

On the other hand, using Amtrak's best run between each city would show the model to be a bit slow, but not overly so except for Fullerton to Los Angeles. I take it to indicate that the model is grossly accurate with probable errors in dwell or padding assumptions.


Best runs Model Push-Pull Difference Intermediate Differences
San Diego
0h 0m
0h 0m
0h 0m
0h 0m
Old Town
0h 6m
0h 6m
0h 0m
0h 0m
Solana Beach
0h 35m
0h 35m
0h 0m
0h 0m
Oceanside
0h 49m
0h 52m
0h 3m
0h 3m
San Juan Capistrano
1h 19m
1h 21m
0h 2m
-0h 1m
Irvine
1h 33m
1h 34m
0h 1m
-0h 1m
Santa Ana
1h 43m
1h 46m
0h 3m
0h 2m
Anaheim
1h 51m
1h 56m
0h 5m
0h 2m
Fullerton
2h 0m
2h 5m
0h 5m
0h 0m
Los Angeles
2h 34m
2h 49m
0h 15m
0h 10m

When it comes to improving the run times, the most straightforward thing for LOSSAN to do would be to do as All Aboard Florida is doing and simply run each train with two locomotives. EMD was kind enough to provide acceleration curves in their protest letter and this was used to create the improved push-pull model, retaining the braking from the earlier one. Out of an abundance of caution, I used the curves for two EMD F125s hauling 1,360,000 pounds (eight cars) with 600kW of HEP rather than those that they produced for the competing Siemens Charger. As it is, Metrolink is purchasing F125s for its own use, which will make the use appropriate in the forthcoming commuter edition of this, and there are no plans to deploy Siemens Chargers on the Surfliner (second to last page). It should be noted that this model unfairly penalizes the proposal somewhat as the current planning is to lengthen the Surfliners to only 7 cars, not 8.


It's immediately obvious that adding another locomotive is superior, but dwell time and padding requirements prevent the improvement from being as great as it visually appears. Still, there's a consistent one minute reduction for nearly all stations, leading to a 13 minute (8%) reduction in total travel time.


Model Push-Pull Improved Push-Pull Difference Intermediate Differences
San Diego
0h 0m
0h 0m
0h 0m
0h 0m
Old Town
0h 6m
0h 5m
0h 1m
0h 1m
Solana Beach
0h 35m
0h 32m
0h 3m
0h 2m
Oceanside
0h 52m
0h 48m
0h 4m
0h 1m
San Juan Capistrano
1h 21m
1h 16m
0h 5m
0h 1m
Irvine
1h 34m
1h 28m
0h 6m
0h 1m
Santa Ana
1h 46m
1h 39m
0h 7m
0h 1m
Anaheim
1h 56m
1h 48m
0h 8m
0h 1m
Fullerton
2h 5m
1h 56m
0h 9m
0h 1m
Los Angeles
2h 49m
2h 36m
0h 13m
0h 4m

There's two different ways we can go about expensing this option. The first is simply to estimate the expense of purchasing two locomotives for each of the nine Surfliner train sets, an additional two for protection and maintenance reserve, and seven cars to replace the current single level Amfleet/Horizon set. With one cab, one café, and five coaches, the cars would come out to $19.3 million (see Table 9.2 on page 3 of attachment C for prices) while the locomotives, at $7 million each, would raise the total price to $159.3 million, approximately $12.2 million per minute saved. However, the Surfliner and Superliner cars currently making up the other 8 consists aren't exactly spring chickens themselves and will be needing replacement in the relatively near future; something that might as well be done while purchasing the new cars to lengthen them. In that case, using the existing cars for maintenance reserves, it would cost $219.7 million (accounting for the new bilevels already on order) to replace and lengthen every train set, for a cost of $16.9 million per minute cut from the schedule.

Of course, if we're going to replace everything, why not just go whole hog and go with a DMU? Lower weight and distributed traction allows us to have even faster acceleration and we don't waste any space with locomotives. There is the slight downside of passengers reporting increased engine noise and vibration however.

Using the more conservative lower bound level acceleration of the Class 221 is still more than enough to rapidly outperform a beefed up push-pull train with an overall 18 minute (11%) reduction in travel time (braking was a constant 2 mphps, the source for which I've lost at some point in the six months between creating the model and writing this up). Given how far off the Los Angeles padding appears to be however, the actual result may be even better in the real world.


Model Push-Pull Model DMU Difference Intermediate Differences
San Diego
0h 0m
0h 0m
0h 0m
0h 0m
Old Town
0h 6m
0h 5m
0h 1m
0h 1m
Solana Beach
0h 35m
0h 32m
0h 3m
0h 2m
Oceanside
0h 52m
0h 47m
0h 5m
0h 2m
San Juan Capistrano
1h 21m
1h 13m
0h 8m
0h 3m
Irvine
1h 34m
1h 25m
0h 9m
0h 1m
Santa Ana
1h 46m
1h 35m
0h 11m
0h 2m
Anaheim
1h 56m
1h 43m
0h 13m
0h 2m
Fullerton
2h 5m
1h 51m
0h 14m
0h 1m
Los Angeles
2h 49m
2h 31m
0h 18m
0h 4m

In 2002, HSBC Rail purchased 127 Class 222 Meridian DEMUs derived from the same family of vehicles for £154 million. At 2002 exchange rates that would be approximately $231.5 million or $303.9 million in 2014 dollars. This would make the price per car $2.4 million (alternately, the Class 395 cost $3.3 million per carriage with similar exchange and inflation though this may include some maintenance). Per Wikipedia, two conjoined 5-car sets would come closest to matching the current Surfliner capacity and ten such sets (9 for operations and one reserve) would cost $240 million (or $330 million with the alternative calculation). Simply purchasing new locomotives is more cost effective, but it isn't by much as the DMU option ranges $13.3-18.3 million per minute saved.

What about electrification however?  The short and simple answer on electrification is that it won't be terribly cost-effective, compared to the other options. In addition to the cost of rolling stock, catenary would need to be installed over the corridor and several multibillion dollar tunneling projects already roughly scheduled for ordinary capacity increases would get lumped in, both for pragmatism, NIMBY complaints, and environmental reasons. But let's hand wave this for a moment to see what might be; we'll pretend that California has awoken to a thunderous "Fiat electris!" and a bill for the newly installed catenary.


First, I must admit that I'm not overly thrilled with the modeling of the EMU. Acceleration (taken from page 17) doesn't quite match up with other sources and there's just something that feels off about it. Braking was 2.91 mphps in accordance with information from the TGV (page 128). For intercity rail, I am assuming that high speed vehicles are used as an extension of the California high speed rail system; if the Surfliner is to have its own separate electric trains then slower but faster accelerating trains would make more sense. As it is, electrification, with no other track improvements, only cuts an additional three minutes from the schedule, though it does now come below two and a half hours.


Model Push-Pull Model EMU Difference Intermediate Differences
San Diego
0h 0m
0h 0m
0h 0m
0h 0m
Old Town
0h 6m
0h 5m
0h 1m
0h 1m
Solana Beach
0h 35m
0h 30m
0h 5m
0h 4m
Oceanside
0h 52m
0h 45m
0h 7m
0h 2m
San Juan Capistrano
1h 21m
1h 11m
0h 10m
0h 3m
Irvine
1h 34m
1h 23m
0h 11m
0h 1m
Santa Ana
1h 46m
1h 33m
0h 13m
0h 2m
Anaheim
1h 56m
1h 41m
0h 15m
0h 2m
Fullerton
2h 5m
1h 49m
0h 16m
0h 1m
Los Angeles
2h 49m
2h 28m
0h 21m
0h 5m

Using the High Speed Rail Authority's estimate of $45 million per set and assuming that ten sets would remain necessary, we would have a price of $450 million just for the rolling stock, $21.4 million per minute saved. But at approximately $3.6 million per route-mile for electric traction (calculated from here), we have an additional $463.7 million to account for ($913.7 million total) raising it to $43.5 million per minute saved. Even worse, an improved push-pull or DMU is able to continue north of Los Angeles to Santa Barbara and San Luis Obispo without needing a diesel locomotive to be brought in to tow it and is able to make improvements on that segment as well; an electric train is not able to do so without spending a significant sum electrifying that portion as well.

In short, reequipping the Pacific Surfliner with a new fleet of diesel multiple unit trainsets provides the greatest reduction in travel time at reasonable cost and may provide the greatest reduction in travel time for the least expense, depending on the actual costs and how existing equipment is accounted for. The next post, hopefully in less than six months, will look at Coaster and Metrolink rolling stock while further posts will deal with track improvements.

Models may be found here. The Numbers files are the originals, Excel versions are provided as well but may have conversion errors. Additionally, I used Metrolink's Timetable No. 7 and BNSF's California and Los Angeles Divisions Timetable #3.

12 comments:

  1. Have you tried working out how the new timetable will affect meets between trains? Speeding things up doesn't do a lot of good if it means trains have to wait longer for opposing moves to clear the single track sections. Fortunately, there's been enough double tracking that it's not as huge of a problem now in San Diego County, though the San Clemente section is still the bottleneck of the whole corridor.
    On the topic of electrification, another possible alternative is replacing the diesel locomotives with ACS-64s. With just one, it shouldn't be too hard to switch to a diesel at Union Station, whereas with two, even an 8 car train would be quite fast, but you'd lose the through service to Santa Barbara (though maybe gain one to Newhall).

    ReplyDelete
    Replies
    1. Not as of yet. It's something I may end up doing once I get around to track improvements however.

      Delete
  2. What are the slowdowns at miles 15 to 17, and at 92 to 97? How much would eliminating those do to the travel time?

    ReplyDelete
    Replies
    1. 15-17 is the canoyon north of San Diego, which really should be bypassed by a tunnel (but that is expensive). 92-97 is a series of curves from Santa Ana north through Orange, where the ROW is relatively constrained and has to make a sharp turn just north of the Orange station anyway. They can be eased somewhat (I think the current situation might actually be a bit better than what's on the graph).

      Delete
    2. The tunnel is expected to be about half a billion dollars (though they made a typo in a recent planning document and put it down for around 2.3 billion). Not sure how much it would cost to improve things between Santa Ana and Anaheim, but I doubt it'd be all that cost effective thanks to the short distance; probably no more than one or two minutes saved by improvements.

      Delete
    3. Thanks for the explanations. So, if one could get rid of freight and use better cant, the speeds could be increased.

      It might also be worthwhile to look at tilting trains, and see how much that would get you better.

      Delete
    4. You don't need to get rid of freight entirely to improve speeds, just run the existing passenger trains at higher cant deficiency (I believe it's 6" on the NEC, instead of 3" everywhere else in the US). If the NEC can do it, there's no reason the Surf Line can't, though it may require higher maintenance standards.
      Also, the real killer of the line's capacity (and reliability) is the long stretch of 40 mph through San Clemente, since it's a long single track that's impossible to double. The long single track limits headways, and either end has relatively short passing sidings and more single track before you get to a long double track section, so you inevitably have meets on either end of the section that, if missed, result in delays to the train going the other way, more missed meets, and general unreliability. I was hoping that once they added some official crossings and fenced it in, they'd raise the speed, but it seems like they're not too interested in doing that. At least they're adding more double track in the vicinity, hopefully making the service a bit more reliable.

      Delete
    5. @Max getting rid of freight is a non-starter on that route. In addition to the half-dozen or so freight trains per night that regularly use the line, it's part of STRACNET, since it's the only rail connection between the domestic rail network and several major US military facilities.

      I've always thought it seemed like a natural for tilting trains (either electrified or similar to tilting DMUs used in Europe), but I'm unaware of any serious study of using them there.

      Delete
    6. I know there's a signaling law to FRA that makes the common 79 mph limit, what's the deal with 90 mph limit here? Enough equipment has top speed of 110 mph that it may make sense to get that limit raised, seems like it could offer the most significant end-end time savings.

      Also, could having platform doors or announcements (perhaps LED boarding/alighting countdown timers) allow for lower dwell times?
      In Japan, commuter trains dwell for no longer than 90 seconds.

      Delete
    7. The 90mph limit is thanks to a legacy ATS system allowing the trains to exceed 79mph (in this case, track quality is kept up to 90mph standards). As for dwell time, while I think there's probably some cheap ways of improving it (as simple as painting lines on the platforms), my experience is that dwell (measured as "train stop to train movement again") averages 90 seconds already and reductions wouldn't make a noticeable improvement in travel times.

      Delete
    8. How expensive is "legacy" ATS and why is it so uncommon?

      Delete
    9. Rose Canyon definitely needs to be bypassed by the tunnel. What's not obvious from the profile is that it's a massive detour; the tunnel would make the route significantly *shorter* as well as eliminating the speed restrictions. But it needs half a billion dollars...

      Delete