25 August 2018

Over-Promising on Electrification

Numerous recent Caltrain materials include the following quantitative claims (see slide at right) about the service benefits of the electrification project:
  1. A baby bullet train making 5-6 stops will make the SF - SJ trip in 45 minutes, down from 60 minutes today.
     
  2. A train making the SF - SJ trip in 60 minutes will be able to stop 13 times, up from 6 stops today.
Both of these claims are greatly inflated. They are easy to verify using a computer program known as a train performance calculator, which numerically integrates the differential equations of motion of a train based on the known characteristics of the track (vertical profile, curve, speed limits, station stops, etc.) and of the train (power, weight, tractive effort, drag, etc.) Physics and math can predict timetable performance quite accurately.

Myth #1: the 45-minute Baby Bullet express

Today's diesel performance
(pure run time, no padding)
Here is what a typical baby bullet run looks like today, with an MP-36 diesel locomotive, six Bombardier coaches, and a load of 600 passengers. There are five stops in this example, each lasting (very optimistically, as riders will attest) just 60 seconds. The pure run time from San Jose to San Francisco 4th and King is 52:22 under ideal conditions, without any margin or padding that is added to a real timetable; compare to the weekday northbound timetable at 64 to 67 minutes, or up to 25% longer (!) than the pure run time. Note that the weekday timetable has been extensively padded lately due to crowding; in 2012, the same run was timetabled at 59 minutes with 12% padding.

Tomorrow's EMU performance
(pure run time, no padding)
All other things being equal, let's substitute an EMU train for our slow diesel. The same run drops to 48:15, just four minutes quicker. This isn't surprising: baby bullet trains spend most of their time cruising near the speed limit, where the faster acceleration of EMUs doesn't provide a benefit. With all other things being equal (including crowding and long dwell times--why would electrification resolve these?) we can expect the timetable for our five-stop baby bullet to drop by the same four minutes, or 60 to 63 minutes. That is a full 15 to 18 minutes slower than claimed by Caltrain! Even if you remove the copious 5-8 minutes of extra padding present in today's timetable and compare to the 2012 timetable, we're still 10 minutes slower than claimed, at 55 minutes.

EMU performance at 110 mph
(pure run time, no padding)
How could you possibly get to 45 minutes? One approach is to raise the speed limit to 110 mph, which is planned in the long term but clearly outside of the scope of the electrification project. Changing only that variable, and slowing down as needed where curves limit the speed to below 110 mph, our EMU now makes the same San Jose to San Francisco run in 41:32, almost seven minutes faster. However, we're still 7 to 10 minutes slower than Caltrain's 45-minute claim, or 2 minutes slower when using 12% padding. Again, the reasons for having such enormous amounts of timetable padding will not suddenly disappear after electrification!

The best way to get there is with level boarding, which alleviates Caltrain's crippling dwell time problem. Level boarding has two benefits: the primary benefit is in the form of reduced dwell time during each stop, and the secondary benefit is in the smaller amount of timetable padding that is needed, thanks to the improved schedule adherence that is possible when the occasional wheelchair lift deployment no longer threatens to inject random three-minute delays. Padding could conceivably be cut to 7%, and dwell time to 30 seconds. No new simulation runs are required-- our five-stop 79 mph EMU makes it in (48:15 - 2:30)*1.07 = 49 minutes on the timetable; the 110 mph EMU makes it in (41:32 - 2:30)*1.07 = 42 minutes.

Caltrain's claim of a 45-minute baby bullet is readily attainable only after three major improvements are made. These are not included in the scope of the electrification project and are currently unfunded:
  1. Conversion of the baby bullet fleet from diesel to EMU
  2. Implementation of system-wide level boarding
  3. Curve realignment, track upgrades and grade crossing safety upgrades for 110 mph
To promise a 45-minute baby bullet run in the short term is at best misleading and at worst a flat-out lie. Once the electrification project is complete, we can expect approximately zero improvement in baby bullet performance, with timetabled runs in the range of 64 to 67 minutes. If the initial slight increase in capacity of the electrification project relieves crowding (but will it, enough to offset the performance loss from dragging a seventh Bombardier car?) then we could return to the 2012 timetable performance of 59 minutes.

Myth #2: the one-hour, 13-stop limited

Let us assume for the moment that padding returns to the 2012 level of about 12%. Assuming 60-second dwells and a 79 mph speed limit, how many intermediate stops can a limited train make between San Jose and San Francisco before the timetable hits one hour?  Subtracting 12% pad from one hour, we need to make a pure run time of 53:34.

With today's diesel bullet performance, Caltrain's claim of six stops in one hour checks out reasonably closely at 54:57 or just over one hour including padding, i.e. close enough. Let's change the assumptions, one by one:

Simulation CasePure Run TimeTimetable
Case A, Diesel, dwell 60, 6 stops, 12% pad0:54:571:01:33
Case B, EMU, dwell 60, 6 stops, 12% pad0:50:100:56:11
Case C, EMU, dwell 60, 7 stops, 12% pad0:52:040:58:19
Case D, EMU, dwell 60, 8 stops, 12% pad0:53:581:00:27
Case E, EMU, dwell 30, 8 stops, 7% pad (level boarding)0:49:580:53:28
Case F, EMU, dwell 30, 9 stops, 7% pad (level boarding)0:51:220:54:58
Case G, EMU, dwell 30, 10 stops, 7% pad (level boarding)0:52:460:56:28
Case H, EMU, dwell 30, 11 stops, 7% pad (level boarding)0:54:100:57:57
Case I, EMU, dwell 30, 12 stops, 7% pad (level boarding)0:55:340:59:27
Case J, EMU, dwell 30, 13 stops, 7% pad (level boarding)0:56:581:00:57
Case K, EMU, dwell 30, 13 stops, 7% pad (level boarding), 110 mph0:53:080:56:51

Simulation Case K
(pure run time, no padding)
Case D shows that the maximum number of stops permissible under post-electrification conditions is at most 8, just two more stops than today, and not 13 as claimed by Caltrain. Only after level boarding does the number of stops increase to 13 as shown by Case J, but once again, level boarding is not included in the scope of the basic electrification project. Case K illustrates the diminishing returns from increasing the speed limit to 110 mph; the more stops a train makes, the less benefit there is from the higher allowable speed. Case K (see diagram at right) shows the train almost constantly accelerating and braking, which is not how one would choose to operate given the cost of electricity in the real world.

The takeaway message to Caltrain is this: don't over-promise and under-deliver on the modernization project. Your electrification project reduces time in motion and establishes a foundation for further improvements, but is not sufficient by itself. To deliver the service benefits promised in your public presentations, you absolutely need level boarding to reduce time at rest.

(do I sound like a broken record?)

42 comments:

  1. Yes, level boarding can't come fast enough.

    But one obvious question is how were Caltrain's post-electrification stops vs. run-time claims developed (and by whom), and what simulation tool(s) and assumptions did they use? Caltrain didn't just make those post-electrification run-times up, did they?

    ReplyDelete
    Replies
    1. They have the tools and the analysts (or hire the consulting firms that do.) I think the marketing claims just got away from the analysis, and need to be reeled back in.

      Delete
    2. Is "got away from" a euphemism for "made up out of thin air"?

      Caltrain has been boasting quantitative (numeric) post-electrification run times for various stop counts — not just qualitative marketing stuff like "better", "faster" or "much faster." So the question remains: where did those hard numbers come from?

      Delete
    3. Where did those "hard numbers" come from? We won't know until Caltrain gives us more details. But that doesn't matter in the end, because there just won't be a 45-minute baby bullet in 2022.

      Delete
  2. A note on precision, for simulation nerds: when I quote times down to the second, I don’t mean to imply that there isn’t the possibility of error in these calculations. I could very well be off by a minute or two due to differences in parameters, or to the higher fidelity incorporated in a professional-grade train performance calculator.

    However, any such systematic errors are compensated by my padding assumptions, which are calibrated against the real timetable. For example, if my run times are systematically 2% too short and I derived a 12% padding factor against the real timetable, I will still predict the same relative changes in the timetable as I might have by using the “correct” timing with a 10% padding factor.

    Because of this calibration, small modeling errors wash out, and the predicted timetable times (after including padding) should be plenty accurate, likely to the one-minute level. Which is to say, I am quite confident in my analysis.

    ReplyDelete
  3. Sounds like the 1 hour, 13 stops is based on an optimistic 30 second average dwell time without level boarding. However, for a 13 stop run, the AVERAGE dwell is probably less than 60s even without level boarding since many of the stations are relatively low ridership. I'm still hopeful that limited trains will be able to make SF-SJ in about 1hr if service starts with 8-car EMUs and 7-car diesel bullets to reduce crowding.

    ReplyDelete
    Replies
    1. I sure hope that Caltrain won't veer into the same shenanigans as the high-speed rail people to make inflated performance claims.

      Delete
    2. Do you mean like their inflated capacity increase claims?
      http://www.caltrain.com/Assets/Caltrain+Modernization+Program/Electric+Multiple+Units/EMU+Capacity+Details_SeatsStandeesBikes+Feb+2018.pdf

      Delete
    3. Q: How do you tell when a Caltrain consultant/employee/perma-consultant is lying?

      A: His lips are moving.

      Good luck with that "exposing outright fraud and lies about billions of capital expense" business! The consequences of doing so for decades now have been ... deliciously delightfully delectably Trump-tastically profitable.

      Delete
    4. That’s a bit harsh. The advertised figures are achievable, but only after additional planned improvements well beyond the scope of the electrification project. That’s a softer sort of lie than making up numbers that are impossible to hit.

      At this point, Caltrain has no need to resort to exaggeration to keep the project sold. I would argue that overpromising now hurts the chances of follow-on funding for level boarding.

      As for capacity, at least they revealed their calculations, if a bit unwillingly. Their math checks out and is likely to be overcome by events anyway when options are exercised for more EMUs.

      Delete
    5. "Opening up our request to alternative designs will foster competition between more car builders, which potentially will result in better pricing, a more modern car and a better experience for our customers," said Metra Chief Executive Officer and Executive Director Jim Derwinski.

      The railroad will seek new rail car designs that boost seating capacity since yard space and platform length currently limit the number and length of trains that can operate.

      In addition, Metra will consider proposals for alternative car designs that provide lower, platform-level boarding for riders with disabilities and families with strollers."

      https://www.progressiverailroading.com/mechanical/article/Metra-to-reissue-rail-car-RFP--55147

      Delete
    6. Metra is where Caltrain was almost 20 years ago.

      Delete
    7. They are clearly 20 years ahead in terms of procurement. Please elaborate.

      Delete
    8. They are angling to purchase Bombardier bilevels to supplement and replace their gallery cars, which Caltrain did in 2002 or so. And they didn’t want to buy Chinese after Nippon Sharyo skipped town. What amazing shrewd move do you read into this?

      Delete
  4. What time does your simulator show for a standing-start mile on level track, MP36 with six Bombardier cars? Haven't checked lately, but today's actual trains probably take 105 seconds or a bit more. How long do you figure the last mile takes, to the stop? 80 to 85 seconds?

    If today's express schedules are annoyingly padded, that means ... 90% of them are arriving on time? Are they actually doing that well?

    ReplyDelete
    Replies
    1. @Tim, as per Tables A & B on page 22 in the August JPB meeting package, on-time performance hovers around 94.5%.

      As per this November 2017 On Time Performance & Delay Mitigation slide deck, any train arriving at its final stop less than 6 minutes (i.e. 5 min. 59 seconds or less) late, counts as "on-time." Also, on page 4 there is a slide showing 5 years' worth of monthly on-time performance stats.

      Delete
    2. My standing start mile happens in 92 seconds, with a "pedal to the metal" assumption making use of every Newton of available tractive effort. Real life operation is more gentle as the engine runs up, especially when operating close to the adhesion limit (below about 23 mph), so your figure of 105 seconds (14% slower) sounds pretty reasonable. On the braking side, my last mile takes 87 seconds.

      Small discrepancies like this wash out in the padding calculation against the real time table, as I pointed out above.

      Delete
  5. Just curious: Is Caltrain presently (diesel) or in the future (electric) limited in acceleration by adhesion?

    ReplyDelete
    Replies
    1. None of us knows just how quick today's diesels build up power at the start -- if they take, say, 30 seconds to reach full horsepower they've probably reached 15+ mph by that time (with 5-6 cars) so adhesion isn't much of a problem. Maybe not no problem at all, tho.

      The control system on AC-motor diesels may allow them to reach full power quicker. I didn't get a chance to study NJ Transit's 710-powered commute diesels, but suspect they do a little better.

      Delete
    2. Just a quick point of order: all trains are limited by adhesion in how fast they can accelerate. Full power is neither available (no matter how fancy your control system might be) nor needed at low speeds. The limit in that regime is how much tractive effort you can develop, keeping in mind that power = force * speed. (Well, power at the rail, which is less than rated power.) Once you hit the speed where you can use all the power you’ve got, the tractive effort begins to drop. For the MP36 consist we were discussing that’s north of 20 mph, i.e. plenty of time to rev up.

      Delete
    3. "all trains are limited by adhesion in how fast they can accelerate."

      Depends on what you mean by "limited". Does a Caltrain diesel reach that limit, accelerating with 6 cars on level track? I'm guessing usually not. As it accelerates thru 5 miles/hr the V-16 hasn't built up enough power to slip the wheels; ditto as it accelerates thru 10 miles/hr; I'm guessing ditto as it accelerates thru 15 miles/hr.

      Delete
    4. Who knows? This is a second-order effect and can safely be ignored. The art of simulation is to have enough fidelity, but no more.

      Delete
    5. I can not speak for the diesel-hauled trains, but for the KISSes, where full torque is available from speed 0 on, it is theoretically possible to get into micro-creep state (that's where maximum force can be transferred to the rails, just before slipping starts). On dry rails, this corresponds to a friction coefficient of around 0.4. Just a guesstimate (someone may do some calculations), this could lead to an acceleration of around 1.8 m/s2, which is uncomfortably strong. Maximum torque is controlled by the drive control software, and limited to lower values (also in order to take care of less optimal rail surfaces). Micro-creep can be recognized by a screeching sound.

      I personally have never heard a KISS (or a FLIRT) reaching the micro-creep state. But when the Zürich S-Bahn DPZ trains (a locomotive plus 3 bi-level cars) were new, they regularly got into that state, and the screeching was really noticeable. Some time later, after complaints by people living along the "gold coast" line, they reduced the maximum torque. Acceleration is still adequate, but no longer as strong as before. Of course, the higher torque also increases wear on the gears.

      Delete
  6. Clem,
    In the case of the Baby Bullet, why is 110mph operation outside the scope of electrification? Granted they could use more passing tracks, but since there are so few locals running anyway that shouldn't be much of an issue.

    ReplyDelete
    Replies
    1. 110 mph requires upgrade of track to FRA class 6, quad gates and intrusion detection at every crossing, minor curve realignment, and safety upgrades at all platforms. Not in the $2 billion tab for work now underway.

      Delete
    2. Unless I'm mistaken, quad gates and intrusion detection are only required for speeds in excess of 110mph. Not sure what you mean by platform safety upgrades...

      Delete
    3. If Michigan / Illinois are anything to go by, that’s what they had to do to satisfy FRA. Also the grade crossing status (gates down and clear of obstacles) had to be patched in to the PTC system with enough time to come to a full stop from 110 mph... that means 2-minute gate down time per train. Not something you’ll see on the peninsula rail corridor. I could see San Mateo / 9th to Redwood City / Whipple upped to 110 mph, and maybe some of the other long crossing-free stretches.

      Platform safety is an unresolved issue. It’s already kind of marginal at 79 mph. I don’t have an answer on that one.

      I don’t think Caltrain needs 110 mph; the time savings are marginal. There is far lower-hanging fruit to be picked by shortening time at rest (level boarding!!)

      Delete
    4. There is a primary rule in robotics or anywhere where you move things around, to make the slow movements as fast and short as possible. So, Clem is absolutely correct that Caltrain should go for the far lower-hanging fruit.
      Level (or just one step) boarding is definitely something. Making space around the doors in the vehicles the next (I am not familiar with the Bombardier cars, whether they have lot of space in the entrance area; S-Bahn Zürich KISSes do have (and are faster than the Siemens DTZ).
      If there are switches passed on deviation, there may be a lot of potential; making it possible to run with 60 km/h instead of 40 km/h over deviation saves about 10 seconds directly, plus deceleration/acceleration for a 300 m long train (unless I got my math wrong…}

      OTOH, SNCF in the Alsace, and DB between Hamburg and Hannover, for example, operate at 200 km/h with grade crossings and non-gated non-manned stations.

      Delete
    5. @Drunk: Quad gates and intrusion detection are required for speeds between 90 and 110mph. Speeds between 110 and 125 MPH additionally require Vehicle Arresting Barriers (VABs) IN THE USA ONLY.

      @Clem: There is nothing (other than pesky level boarding island platforms) stopping trains from exceeding 125 MPH "on long crossing-free stretches". The answer to platform safety issues was presented to the Transit-Land Use Committee in February 2017: http://hsr.ca.gov/docs/brdmeetings/2017/brdmtg_021417_TLU_Board_Meeting_SFO_pres.pdf (slide 8).

      Last but not least, it is unclear why a train capable of stopping from 125 MPH in less than 45 seconds would require a "2-minute gate down time per train" ("Il lui a fallu seulement 43 secondes sur 1,25km pour passer de 200km à l'arrêt complet.")
      http://www.newsbombardierfrance.com/2016/06/regio-v200-termine-ses-essais-velim.html

      Delete
    6. We’re not that flush with money or space, so I would anticipate the platform safety solution for 110-125 mph to look more like platform railing (see this video from Germany). Definitely not ideal, but it’s cheap.

      Gate down time requires a few seconds for detection and transmission to the signaling system, and significant extra time for emergency braking in the rain. I can lower my estimate to 90 seconds for a 110 mph interlocked grade crossing, up from ~30 sec today.

      Delete
    7. It's just awful the way the cameraman gets blown back into the parking lot as the train comes through in far off exotic Rhode Island. https://www.youtube.com/watch?v=l30Z1dV2SVU

      Delete
    8. I appreciate the sentiment, but San Francisco / Silicon Valley during rush hour is quite unlike rural Rhode Island.

      Delete
    9. Yes it is and if you had a four track system they could be out on the express tracks at 125ish. I'll see if I can find 90 mph Acelas in Secausus. Someday in your wildest dreams Palo Alto is Secaucus. Secaucus is designed so that, one of the speculative fantasies is this, the center island goes away and the side platforms become islands between the 5th and 6th tracks that are cantilevered off the sides that were overdesigned to do that. Or sumptin. there's fantasy where its expressy-est expresses flying over which is probably excessive except in scenarios where cars and airplanes have been banned. Enjoy your two tracks. at 110 you can keep your quaint grade crossings forever.

      Delete
    10. According to page 7-29 of the former (AKA now obsolete) Chapter 7 of the Caltrain engineering standards, "Caltrain does not have and does not allow at-grade crossings where there are four (4) tracks(passing tracks)."

      On a related note, thank you for another great shot explaining why the NEC continues to experience systemic catenary failures even on brand new sections: https://youtu.be/l30Z1dV2SVU?t=12

      Delete
  7. OT Breaking News: "The authority is looking to get that Valley-to-Valley service running by 2026 or 2027 using one track through the Pacheco Pass tunnel, and then have the second track ready for use in 2029, after the tunnel is fully built out."
    https://thebusinessjournal.com/california-texas-riding-different-tracks-to-high-speed-rail/

    ReplyDelete
    Replies
    1. I rather doubt that... twin bore is required for fire safety in long tunnels.

      Delete
    2. But does the second bore NEED to be fully built out?

      (I know, considering overall cost, it is cheaper to do everything at once.)

      Delete
    3. @Clem Did you miss the reduction in tunnel diameters in the 2016 Business Plan that was made possible by the elimination of pesky tunnel ventilation fans interfering with air flow?

      Delete
  8. 1st Law of Holes:
    http://www.caltrain.com/Assets/__Agendas+and+Minutes/JPB/2018/2018-09-06+SSF+presentation.pdf

    ReplyDelete
  9. Is anyone knows dwell time of BART? We will expect Caltrain to be "RAPID TRANSIT" not a electrified "COMMUTER RAIL". So, we should benchmark BART's non-SF stations (Fremont-Oakland or Richmond-Oakland) where ridership /population profile are similar to Caltrain Penninsula.

    ReplyDelete
    Replies
    1. "We will expect Caltrain to be "RAPID TRANSIT" not a electrified "COMMUTER RAIL""

      BART dwells on non-CBD stations are on the order of 10 to 20 seconds.

      Caltrain will never be RAPID TRANSIT, sadly.

      Level boarding will never happen, insane EMU procurement nothwithstanding.
      Frequent service will never happen.
      Regular interval service will never happen.
      Acceptable off-peak headways will never happen.
      Acceptable operating costs (one person operation, maintenance costs anything approaching those of industrialized democracies) will never happen.

      So BART is irrelevant, except the desginated eventual replacement of Caltrain -- sole-source whack-job low-capacity EMUs, insanely over-built super-cost electrification infrastructure (so so so many extraneous poles), electrification leading to higher operating costs, etc.

      Enjoy the next 20 years of long dwell times, infrequent service, insane skip-stop timetables, out-of-control ticket prices and tax subsidies, and effectively zero off-peak service. The next 20 years are going to be exactly the same as the last 20, including the mañana-mañana talk about "rapid transit" or "regional rail" or whatever. Hasn't happened, won't happen, because nobody involved with Caltrain in any capacity wants it to happen.

      The status quo is fucking awesome. Being paid to do less than nothing is a great number of peoples' idea of a perfect set-up, and they're not about to let that change.

      Delete
    2. But they will have two tracks, scenic grade crossings with nostalgic sound effects and won't have to cut down any invasive weed trees that are fire hazards!

      Delete