News:

The Forum Rules and Guidelines
Our forum has Rules and Guidelines. Please, be kind and read them ;).

Locomotive trials, 1901

Started by zook2, February 18, 2014, 03:31:35 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

zook2

I am curious about the choice of locomotives for various tasks - short and long passenger routes vs. cargo hauls, so I set up a race track with seven identical and parallel lines, no curves, no stops or signals between start and finish. There are 8 up-slopes on the track and 10 down-slopes. All are made of Steel Track, max. 160 km/h. Length is 49 km.



Each engine pulls 120 tons of identical wagons, which means total train weights between 201 and 213 tons. Most of my current trains are lighter than that, in the 180 ton range, but my heaviest cargo haulers weigh about 250 tons.
The race track is not at all typical for my network. Most stops or stations are only 10-15 km apart, with few slopes but many curving sections and a lot of traffic, so it's mostly stop-and-go for my "real" trains, and acceleration/braking power counts much more than top speed, which is rarely reached, even on the 49km race track. But the trials could help me get a feeling for what an engine is really capable of.

Also, I'll have to repeat the trials with different loads - a heavy 200-ton load for simulating cargo, and perhaps a 100-ton for light passenger routes.



Engine, engine weight + tender weight, power/traction, total train cost/km, official top speed, time


LBSCR B4 Class     (49+32t, 276/79, 22.50 c/km, 138 km/h)        - 0:34:59
Midland 2601     (50+11t, 263/66, 10.16 c/km, 133 km/h)         - 0:36:30
Claud Hamilton     (50+40t, 258/76, 21.85 c/km, 129 km/h)        - 0:37:45
HR Jones            (57+25t, 258/109, 6.78 c/km, 110 km/h)        - 0:38:11
Midland Belpaire     (56+37t, 253/89, 16.50 c/km, 127 km/h)        - 0:38:48
Midland Spinner     (47+11t, 238/68, 10.08 c/km, 125 km/h)         - 0:38:54
Alfred the Great    (55+35t, 235/136, 11.00 c/km, 122 km/h)        - 0:40:26

(Note that I don't use Total Trip Time for the results, as it is currently somewhat unreliable.)

The savegame is available here, if anyone's interested:
http://forum.simutrans.com/index.php?topic=13274.msg132136#msg132136

zook2

The photo finish:



Now for interpreting the results. Only a few first ideas here:

First: only one engine is somewhat outclassed. "Alfred the Great"... isn't.

Secondly: top speed is mostly irrelevant, even in racing. Claud Hamilton barely beat the more powerful HR Jones because it reached a slightly higher speed for a few minutes on a long stretch of perfectly even terrain, but the difference is marginal at best.

Third: Disqualify Alfred, which is a heap of junk, and the difference between first and last place is about 10%. Likely even less in real-life traffic. Costs, however, vary by 300%.

Fourth: I never paid attention to tenders, but their weight probably explains a lot of the differences in the results. The Spinner + tender weighs a total of 58t, while Alfred comes in at a whooping 90t.

Fifth: I shall do more testing with different train loads, but if it's only about building a cost-effective network, I'd buy only HR Jones' right now for all purposes.

Finally: tenders and steam engines. Some tenders weigh 11t, others 40t. I don't know whether that's supposed to be loaded or empty. Then add 10 or 20 tons of water in the boiler, which has to be filled up regularly, too, and you get huge weight differences between the start and finish of a journey. Presumably a train carried only as much coal and water as required for the trip, but a larger tender could mean longer journeys between refueling/refilling. Currently, a heavy tender is just a millstone around an engine's neck, as fuel and water are not simulated.

Sarlock

Interesting results... on mostly flat terrain the faster locomotives would have a more significant advantage, but with a few hills the HR Jones quickly closes the gap with its much higher traction, largely negating its clear disadvantage with a lower top speed.

Regardless, the HR Jones is much cheaper to operate.  With a heavy load I imagine it will perform even better.
Current projects: Pak128 Trees, blender graphics

Tazze

What I learned today: speed = power.  ;D

ӔO

Main thing to look at is power/weight ratio, as that is what dictates top speed.
Although, there is a hidden 'air resistance' number on some trains that will make them perform better despite their power.
My Sketchup open project sources
various projects rolled up: http://dl.dropbox.com/u/17111233/Roll_up.rar

Colour safe chart:

AP

The issue with the tenders being essentially "dead weight" in the game is a valid point however. If the advantages of larger capacity tenders are not simulated (e.g. shorter stop times at stations, i.e. no need to "take on water"), then why should the disadvantages be simulated. Make all tenders have exactly the same stats, just be graphically different.

Although longer station-stops for tank locomotives or those with small tenders would be an interesting setting to add to the game...

zook2

Quote from: AP on February 18, 2014, 08:06:34 PMIf the advantages of larger capacity tenders are not simulated (e.g. shorter stop times at stations, i.e. no need to "take on water"), then why should the disadvantages be simulated. Although longer station-stops for tank locomotives or those with small tenders would be an interesting setting to add to the game...

I think you have a point there. And you could do the same for airplanes.

ӔO

A better way to look at it, is that not everything was created equally.
Some locomotives are lemons, despite what the designer originally intended them to be.


IMO, this can be a good place to use the monthly maintenance value to offset their running costs, which has, thus far, not been used.
My Sketchup open project sources
various projects rolled up: http://dl.dropbox.com/u/17111233/Roll_up.rar

Colour safe chart:

jamespetts

A very interesting experiment and resulting discussion. A few points:

(1) the results will vary depending on the load and on the steepness of gradients (a locomotive such as the Jones Goods, which was designed for hilly routes, would be expected to do better when there are many hills);
(2) the next major release will have half heights, meaning
- normal (half height) gradients will be treated in physics terms as being slightly more than half as harsh as existing full height gradients now; and
- high (full height) gradients will be treated in physics terms as being slightly more harsh than existing full height gradients now;
(3) with the exception of the above change relating to gradients, physics should be balanced by now, but costs are not balanced yet (the plan is to have a close relationship between power and the per kilometre cost);
(4) there has been careful consideration given to adding range, although it was not popular when suggested a while ago in large part due to the extra demands that it will place on the player having to work out the distances between things (and is likely to take a very large amount of work to implement in the code in a satisfactory way, which effort could better be spent elsewhere);
(5) the locomotive and tender weights are based on actual weights (if I have made any errors with the actual weights, please let me know the correct data - data are not always available, so some have to be guessed based on contemporaneous types); and
(6) the relative efficacy of the locomotives will also vary with the distance between stops: 49km is a medium distance - a very long distance express train might have twice as long between its stops, whereas an urban train might have stops only one or two kilometres apart.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

zook2

More results, same conditions, but this time with a 192-ton-load of empty cars (120t in the first race). Results for the 2nd race and placing in first race are in the last columns.

1. LBSCR B4 Class    (49+32t, 276/79, 22.50 c/km)       - 0:34:59 - 0:42:01 (1st)
2. Midland 2601      (50+11t, 263/66, 10.16 c/km)        - 0:36:30 - 0:44:38 (2nd)
3. HR Jones             (57+25t, 258/109, 6.78 c/km)       - 0:38:11 - 0:45:03 (4th)
4. Claud Hamilton    (50+40t, 258/76, 21.85 c/km)       - 0:37:45 - 0:46:04 (3rd)
5. Midland Belpaire    (56+37t, 253/89, 16.50 c/km)      - 0:38:48 - 0:47:06 (5th)
6. Midland Spinner    (47+11t, 238/68, 10.08 c/km)       - 0:38:54 - 0:48:09 (6th)
7. Alfred the Great   (55+35t, 235/136, 11.00 c/km)      - 0:40:26 - 0:49:32 (7th)

So the only difference here is that the HR Jones did indeed do slightly better with a heavy load.

zook2

Quote from: jamespetts on February 18, 2014, 11:29:28 PM
(1) the results will vary depending on the load and on the steepness of gradients (a locomotive such as the Jones Goods, which was designed for hilly routes, would be expected to do better when there are many hills);
It does slightly better with heavier load, but the differences are marginal, at least in this test.

Quote(6) the relative efficacy of the locomotives will also vary with the distance between stops: 49km is a medium distance - a very long distance express train might have twice as long between its stops, whereas an urban train might have stops only one or two kilometres apart.
Ahhh, then it's map parameters (and play style) that decide the issue. I play on a map of 1920x1920 tiles, but I have lots of cities and not one line that is even close in length between stops. I guess I shouldn't expect real-life differences in an unrealistic setting.

Even my few express trains, on a four-track connection (two for each direction), are forced to stop regularly because of heavy traffic. There are probably a lot of deficiencies in my network and figuring out solutions is what makes the game fun.

Quote(4) there has been careful consideration given to adding range, although it was not popular when suggested a while ago in large part due to the extra demands that it will place on the player having to work out the distances between things (and is likely to take a very large amount of work to implement in the code in a satisfactory way, which effort could better be spent elsewhere);
(5) the locomotive and tender weights are based on actual weights (if I have made any errors with the actual weights, please let me know the correct data - data are not always available, so some have to be guessed based on contemporaneous types);
But the point was that realistic tender weight is irrelevant without considering fuel. Imagine two engines...

...

After some consideration, I'm beginning to see the huge can of worms this would open. Fuel simulation is pointless without refueling stations, a la Railroad Tycoon, among other things. It's not entirely satisfactory, but I guess we have to live with that.

zook2

The third test, this time with only 72 tons load (three 24-ton passenger cars). Results are in the last column.

1. LBSCR B4 Class    (49+32t, 276/79)      - 0:34:59 - 0:42:01 - 0:29:50
2. Midland 2601    (50+11t, 263/66)       - 0:36:30 - 0:44:38 - 0:31:12
3. Claud Hamilton    (50+40t, 258/76)      - 0:37:45 - 0:46:04 - 0:32:02
4. Midland Belpaire    (56+37t, 253/89)      - 0:38:48 - 0:47:06 - 0:32:44
4. Midland Spinner    (47+11t, 238/68)       - 0:38:54 - 0:48:09 - 0:32:44
5. HR Jones      (57+25t, 258/109)      - 0:38:11 - 0:45:03 - 0:34:11
5. Alfred the Great   (55+35t, 235/136)      - 0:40:26 - 0:49:32 - 0:34:11

(Note the two ties at 4th and 5th position)

Not surprisingly, with lighter load, top speed rules on a track like this; again the B4 winds hands down. The high-traction but low top-speed HR Jones loses badly. The Midland Spinner, despite being the lightest by far, is clearly outdated in 1901.

With a heavy load (192t, 2nd column), the difference between first and last place is 7 1/2 minutes, or 16.6% of the fastest time.
With a medium load (120t, 1st column), the difference between first and last place is 5 1/2 minutes, or 15.7% of the fastest time.
With a light load (72t, 3rd column), the difference between first and last place is 4 1/3 minutes, or 14.4% of the fastest time.

Next, I'll have to test the same trains with lots of stops between start and finish. Also hilly terrain.

Milko

Hello

Quote from: zook2 on February 18, 2014, 10:08:46 PM
I think you have a point there. And you could do the same for airplanes.

Yes, there is the same problem for aircraft.
When calculating the weight value using a tank half full.

The speech of the range had already been discussed but I think it was shelved due to the excessive complexity of realization.

Giuseppe

jamespetts

Interesting results! You might also want to try as an alternative a 100km stretch of straight, flat track.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

ӔO

Another test would be several 5km to 8km intervals strung together with no hills.
Not only does it test acceleration, it also tests brake strength.

I would expect the HR-Jones to win this from past games, but maybe Alfred has a chance here?
My Sketchup open project sources
various projects rolled up: http://dl.dropbox.com/u/17111233/Roll_up.rar

Colour safe chart:

zook2

Stop-and-Go

This is the same track, now with 15 stops between start and finish. On average one stop with a 16-second loading time every 3km. All trains pull the same load as in the last trial (72 tons each).





1. LBSCR B4 Class     (49+32t, 276/79)      - 1:06:30
2. Alfred the Great    (55+35t, 235/136)     - 1:08:21
3. HR Jones           (57+25t, 258/109)        - 1:08:31
4. Midland 2601     (50+11t, 263/66)         - 1:09:59
5. Midland Belpaire     (56+37t, 253/89)      - 1:10:16
6. Claud Hamilton     (50+40t, 258/76)        - 1:10:56
7. Midland Spinner     (47+11t, 238/68)       - 1:11:35

Again, the B4 wins, but this time poor old Alfred comes in second! 136kn tractive power finally pay off. HR Jones makes a respectable third place. But all in all, the differences between the locomotives almost disappear: 8% difference between first and last place (see the last screenshot, taken at the last stop before the end of the race).

Also, they all took about twice as long for the run as in the last trial with no stops. Over one hour, of which a total of four minutes was loading time at the stops.

zook2

Quote from: ӔO on February 18, 2014, 11:04:50 AM
Main thing to look at is power/weight ratio, as that is what dictates top speed.

Power-to-Weight ratio

Power : loco+tender, without cars / Power : loco + tender + 192t load

Midland 2601       (50+11t, 263/66)    4.3 / 1.03
Midland Spinner    (47+11t, 238/68)    4.1 / 0.95
LBSCR B4 Class    (49+32t, 276/79)   3.4 / 1.01
HR Jones             (57+25t, 258/109)   3.1 / 0.94
Claud Hamilton    (50+40t, 258/76)   2.8 / 0.91
Midland Belpaire    (56+37t, 253/89)   2.7 / 0.92
Alfred the Great   (55+35t, 235/136)   2.6 / 0.83

There are huge differences in P/W ratio, but as soon as you add a large load and consider the entire train, the picture becomes a little blurred. The heavier the train, the less important locomotive weight becomes. And in the end, simply the most powerful loco won all tests.

jamespetts

You should try some of the larger tank engines on your stop/go test, too.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

ӔO

Quote from: zook2 on February 20, 2014, 12:41:13 AM
Power-to-Weight ratio

Power : loco+tender, without cars / Power : loco + tender + 192t load

Midland 2601       (50+11t, 263/66)    4.3 / 1.03
Midland Spinner    (47+11t, 238/68)    4.1 / 0.95
LBSCR B4 Class    (49+32t, 276/79)   3.4 / 1.01
HR Jones             (57+25t, 258/109)   3.1 / 0.94
Claud Hamilton    (50+40t, 258/76)   2.8 / 0.91
Midland Belpaire    (56+37t, 253/89)   2.7 / 0.92
Alfred the Great   (55+35t, 235/136)   2.6 / 0.83

There are huge differences in P/W ratio, but as soon as you add a large load and consider the entire train, the picture becomes a little blurred. The heavier the train, the less important locomotive weight becomes. And in the end, simply the most powerful loco won all tests.

yeah, I meant the power/weight ratio for the entire train, not just the locomotive.
My Sketchup open project sources
various projects rolled up: http://dl.dropbox.com/u/17111233/Roll_up.rar

Colour safe chart:

zook2

I will try out a few others as well. But first, the mammoth race: 192 tons with 15 stops.

1. HR Jones      (57+25t, 258/109)   - 1:25:35
2. LBSCR B4 Class    (49+32t, 276/79)   - 1:27:08
3. Alfred the Great   (55+35t, 235/136)   - 1:29:19
4. Midland 2601    (50+11t, 263/66)    - 1:29:29
5. Midland Belpaire    (56+37t, 253/89)   - 1:31:29
6. Claud Hamilton    (50+40t, 258/76)   - 1:31:50
7. Midland Spinner    (47+11t, 238/68)    - 1:33:46

And for the first time in recorded history, the B4 does not win! Although she'd been in the lead for most of the race, but the final, steep hill, slowing everything down to a crawl, brought the HR Jones to the front and made her the clear winner. And Alfie comes in third - did I call her a piece of junk?

But again, the difference between first and last place is 9% and much less than I expected.

jamespetts

Do bear in mind that, in reality, railway companies would not be able to afford to use their best express locomotives on low profit suburban runs, nor in many cases afford the infrastructure necessary to allow locomotives of that weight. That reality is something that we are trying to simulate, although the pricing element of things is not yet complete. Nonetheless, the smallness of the advantage of the fast express locomotives (and the B4 was indeed considered the finest in its day) on such an intensive run is telling: they would be much more expensive for a very small advantage. They come into their own rather more on the longer runs for which they were designed.

It is also noteworthy that, in reality (again, which will hopefully be simulated when price balancing is completed) railway companies had to make locomotives last for decades to get a sensible return on their investment, so only a few lines could have the latest and greatest locomotives: others had to make do with cascaded locomotives.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

AP

Quote from: jamespetts on February 20, 2014, 11:47:32 AMso only a few lines could have the latest and greatest locomotives: others had to make do with cascaded locomotives.

Is there an efficient way to do this in simutrans? I mean if you have a network with dozens of lines, hundreds of vehicles, just sending everything to the depot isn't a fun prospect. The idea of buying new stock, shuffling it all around, and retiring an aging unit on a branch line somwhere, sounds good, but is it what players actually do?

There's an "Upgrade" function in game, does it retain or sell the stock when it upgrades each train? If so, I assume  once line A has finished its upgrade you could (with "show obsolete" on), upgrade line B, etc, is that the best way?

jamespetts

It is planned to make some changes to the vehicle replacer to allow for easier cascading: the plan at present is to unify the stock held in all depots and allow using of vehicles held in the depots to be specified in the replacer window. When this is done, the procedure to cascade will be to replace the top link stock with the newer items, leaving the replaced items in the depot, then replace the second link stock with the displaced stock from the depot.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

ӔO

Currently, the easiest way to cascade convoys is to...
1. "Promote to line" a convoy to be replaced and send it off to depot.
2. Create a new fleet of convoys, set it on the newly created line and send them out.
3. In schedule list, open up the line to be replaced and add a nearby depot to its schedule.
4. Wait for all old convoys to be collected, disassemble and reconfigure them to your liking.
5. Repeat

You can or should keep the duplicate line, since you can reuse it in the future, if nothing has changed.


Alternatively, if there are fewer convoys in the line
1. Create new convoy fleet, set them on a line and send them out.
2. Send all older convoys in line to depot manually with 'send to depot' or 'retire'.

Typically, you shouldn't need more than 8 convoys per line. If you do, you should look at upgrading the service with something that has more capacity.
My Sketchup open project sources
various projects rolled up: http://dl.dropbox.com/u/17111233/Roll_up.rar

Colour safe chart:

AP

#24
Quote from: ӔO on February 21, 2014, 03:30:47 AM
Currently, the easiest way to cascade convoys is to...
1. "Promote to line" a convoy to be replaced and send it off to depot.
2. Create a new fleet of convoys, set it on the newly created line and send them out.
3. In schedule list, open up the line to be replaced and add a nearby depot to its schedule.
4. Wait for all old convoys to be collected, disassemble and reconfigure them to your liking.
5. Repeat

Its that portion that makes it tedious. Especially pre 1830 when everything is moving at a snails pace. On a server game, where things happen when you're not playing / at work, it's bad economics to have large numbers of vehicles idle in depots.

Quote from: ӔO on February 21, 2014, 03:30:47 AM
Typically, you shouldn't need more than 8 convoys per line. If you do, you should look at upgrading the service with something that has more capacity.
It depends entirely on the length of the line and the speed of the vehicles / frequency of service. Fewer high capacity units can cause terrible peaks in supply/demand, which can be a problem both for freight networks (industries cease producing) and passenger networks (waiting times at interchanges skyrocket). Because passengers only interchange 7 times, on a large map you need long routes.

zook2

Stop-and-Go, Heavy Load (192 tons)
The year is 1902 now. As suggested, I'm testing some tender-less engines now. Also the newer models that have appeared since, some of which are quite powerful.

1. GWR Saint    (69+44t, 455/91)  - 1:15:08
2. Midland 1000 (62t, 317/97)     - 1:23:43
3. GWR City     (56+37t, 300/79)  - 1:26:32
4. LNWR B       (53+30t, 260/104) - 1:24:03
5. GNR Klondyke (61+43t, 286/70)  - 1:31:15
6. Watford Tank (71t, 168/73)     - 1:46:41
7. Midland 2441 (49t, 104/93)     - 2:03:46

Not surprisingly, the GWR Saint wins, due to its massive engine power. One of the three tender-less locos, the Watford Tank didn't do very well, but considering that it has only half the power of the top engines, the difference is smaller than I expected. The Midland 2441 is the cheapest loco in operating costs (1.99/km), and it shows. Finally, the GWR Saint has only an 8-minute lead over the Midland 1000, despite having almost 50% more power.