Are long consists impossible?

[GaryG]

Ok, unless you have locos with modified dynamics, RailWorks' 10% to 15% dynamics setting is probably close to a 100% equivalent prototype setting.

Let us look at what you did. You had about 5% dynamics (probably 40%-50% prototype equivalent) set when you released the brakes. After thirty seconds or so, the cars would have probably all been freely rolling with most of the slack bunched against the locos.

You now had five locos with dynamics suddenly changed from about 50% equivalent dynamic braking to 100% dynamic braking applied to a freely rolling, longish train with all slack bunched and no brakes applied in the cars. I would think that under these conditions (heavy braking at the front and heavy pushing from the rear) that having the train pop off the rails on curved track would actually be quite prototypical.

You should have reapplied the brakes first to allow the slack to run out first. Even better, the brakes should not have been released, instead the dynamics should have been released and perhaps if needed, a notch or two of power could have been applied if speed was dropping too much. Knowing the route profile and how a train will handle is the reason why handliung a train is a skilled task.

GaryG

[MikeK]

You are almost certainly right. Either way the game as delivered is very unrealistic.

Here is a question I hope someone can answer. Maybe it will shed more light on the issue.

When slack is "stretched" or "bunched", is it metal on metal at that point? By that I mean once the slack runs in or out a certain distance, is that as far as it will go?

In the game, the couplers don't seem to be modelled that way. It seems as though they are modelled as springs. In other words, when you apply throttle they stretch out a certain distance. If you apply more throttle they stretch further. Even more throttle and they stretch even further. Even worse, they will stretch to a certain point, then spring back. The same happens with bunching the slack with dynamic brakes.

This doesn't seem right to me. Once the train is stretched you should be able to add any amount of throttle and the whole train would accelerate as a single unit. Once it is bunched you should be able to add more retarding force at the head end and it won't bunch up any more.

Am I correct or am I misinformed?

If you cycle between the F4 and F5 views you get the speed at the head end of the train (F5) and the tail of the train (F4), and it is pretty clear that the couplers are all very bouncy springs. If you use dynamics the tail end of the train ends up at a higher speed than the head end as the slack bunches up. This is what I expect to see.

But after a few seconds when the speed at each end of the train equalise they don't stay there, the reverse happens. The head end starts going faster as the slack runs out again, even though dynamic brakes are still applied. If you watch the 2 speeds over time the slack runs in, then stretches out, then runs in and so on, like a big slinky.

These videos show what I mean. In reality there is an inch or two of slack, but the couplers should definitely not be modelled as a spring:

http://www.youtube.com/watch?v=SaF82LCBSyA
http://www.youtube.com/watch?v=75MVm7atiWM
http://www.youtube.com/watch?v=9QiyXLqItJc

[GaryG]

Hi

As near as I can find (Googling), there is no metal-to-metal contact.

It looks like the main slack movement of an inch or two in and out is controlled by compressable springs, the force of which is probably enough to keep the coupler's drawbar centered (front to back) in the pocket when uncoupled. After the slack action spring is fully compressed, there is another way to have some additional movement but that movement does require much higher forces. It looks like there is a somewhat compressable block, perhaps rubber, that allows the extra movement. I don't know how much additional movement but I suspect that it could be as much as four inches in the forward and reverse directions.

Perhaps an expert can add corrections or additional information.

GaryG

[MikeK]

Since slowing a train from the head end leads to derailments, I tried another test this morning. I built the same train at the summit of Cajon (with the locos right at the summit). I wanted to see how going down backwards and essentially slowing a train from the "rear" works. No issues at all, I could manage the train speed using throttle in the forward direction (unrealistic) or use up to 50% dynamic brakes to control the speed with no derailments at all.

So having a massive force stretching a train from the head end of a train causes no derailment risk, but having even a tiny force compressing a train causes huge derailments. Are real train this fragile when the slack is run in? I always assumed that in reality the big risk was from breaking a coupler when stretching the slack, not from derailing when compressing the slack.

Also, going up the hill 750 amps x 7 locos = the force to keep this train moving against friction and gravity.

Going down the hill 130 amps x 7 locos plus friction = the force to hold the speed against gravity.

To me this implies that the force due to friction is equal to 620 amps x 7 locos. How is this possible, a train would never move if that was the case!

I did want to see how much slack there was in the train, so I started the scenario again, released the brakes and used forward throttle to keep the speed at the front of the train as close to 0 as I could. While I did this I watched the rear of the train to see how much slack ran out. This is what I saw:

2011-11-21_00002 - Copy.jpg

12 cars x 64' each = 768 feet, spread over 100 connections = an average 77 feet per coupling! Of course, couplers at the rear of the car are only stretched a little, the ones at the front have most of the force on them.This is the view a few cars closer to the front of the train:

2011-11-21_00003 - Copy.jpg

So I think we have 3 problems when moving trains downhill:

- One is that the force generated by the dynamic brakes seem completely out of proportion to the amps being used
- The second is that there is waaaaaay to much slack in the train because they are modelled as simple springs. In reality, a coupler cannot stretch 77 feet.
- The third is that the freight car brakes are too powerful. A minimum brake application is too much in most cases.

Can anyone who knows how to edit these settings do me a favour? Can you pick a freight car (maybe that reefer for consistency), and try to massively up the spring rate and damping for the couplers, and reduce the effectiveness of the brakes as Kali did in another thread?

If we can get the couplers so that there is a few inches of slack per car instead of 77 feet of slack per car, a lot of these problems might go away.

[Griphos]

Now that's very clever and illuminating work, there, Mike. Very helful. I hope there is something we can do about these parameters.

[GaryG]

Excellent tests and observations Mike.

Why don't you submit this link with a covering note to the RailWorks support email. Perhaps they will address this excessive coupler movement in a update eventually. MSTS also had the same problem although it didn't show up until you tried to find settings that emulated real world physics.

Some of the braking settings are changeable in the blueprints. They are more or less based on a force controlled by a percentage of the weight of the car/loco. Dynamic brake settings are controlled partially by a file that sets up a force/speed curve and depends on a couple other settings as well, again probably blueprint accessable.

GaryG

[arizonachris]

Someone down in the BLLW section was noticing that excessive stretching on those heavy weight passenger cars. And as I've said before a few times, DB is way over powered, so I stopped using it. I use train brakes, period. I don't get the weird stretching or the derail due to consist tilt anymore. That one really has me baffled. Consist tilt. Hmmmm.

[MikeK]

I have been trying to work out why the consists don't derail when a huge stretching force is being applied, but the slightest compressing force and you get the consist tilt message.

I think that when the consist is stretched, the cars just move unrealistically far apart, as in the above pic. Unrealistic, yes, but doesn't trigger any rules that would cause the game to decide that the cars should derail.

When the consist is compressed however, it only takes a few inches before the hitboxes for each car are on top of each other, and it derails.

So couplers can stretch 7+ feet with no problem, but compress them a few inches and the consist derails.

If this is true, then this incorrect coupler modelling could explain all the derailment issues I have been seeing under braking.

[MikeK]

I should also point out that I was seeing an average stretch of 7 feet per coupler, not 77.

Sometimes the coffee just doesn't kick in early enough :)

[Kali]

Yeah, there's been funny couplers for quite some time - ever since people asked for coupling slack, I think. Probably about time someone experimented with them a bit. I would think the spring force needs a couple of 0s on the end at least, perhaps when the cars compress the couplers under braking the collision boxes are overlapping.

The default SD40s dyna brakes aren't terrible - you might want to try the ones I posted somewhere in the "dodgy physics" thread which have been reduced a bit again. Note I posted two lots, try the second one. And yes the car brakes are too strong, we discussed that there too.

Edit: default couplers live here:

Assets\Kuju\RailSimulatorUS\RailVehicles\Couplings\Buckeye\Type-E\buckeye_type_e_coupling.bin

and I'd imagine the lines to mess with would be these:

<SpringCoefficient d:type="sFloat32" d:alt_encoding="0000000000708740" d:precision="string">750.0000</SpringCoefficient>
<Damping d:type="sFloat32" d:alt_encoding="0000000000C08240" d:precision="string">600.0000</Damping>

[MikeK]

Nice find Kali!

This confirms at least that couplers are definitely modelled as damped springs. Based on that I am guessing that there is no way to introduce limits at either end of the coupler's travel without RS making changes to the code.

[Kali]

There are no limits, other than when you bend the knuckle back enough that it won't couple :) ( or pull the coupler out of the car anyway ) just different strength springs. But no, there is a strength option which is in a different section, but other than that it's just those two coefficients. I suspect if you go dig up some basic equations for springs they'll be obvious numbers. Try sticking a 0 on the end of each and redoing your test.

[arizonachris]

Mike, the "hit box" has to be it as far as the engine braking too hard. Someone else (here or at the Steam Forum) was having trouble with a consist he created derailing right at the start. Of course, that was cars overlapping.

I made a nice jacknife earlier with only 20 double stacks, two SD40-2's and only dynamics. The thing flew apart in a second, and I had only just hit the dynamics. Didn't even get to 5%. So I just won't use them anymore. And as far as editing the game code, I would mess it up somehow (I always do)

Seems "power braking" as was suggested in the VORA sessions I have attended is the way to go. I'm finding it works really well once you get the hang of it!

[Kali]

Well, that was interesting.

80 loaded coal cars ( and a near weightless logging carriage, not relevant ), three of my modified SD40s, 1.8-2% grade downhill on Cajun, and coupler spring forces modified *100.

* Test one, rake as above; once the sd40s had stopped slipping, got it up to 50mph or so and hit the dynamics... aaaand nothing much happened at all. Checked to see if they're working, and they were! ok, so I'd modified them and 80 coal cars at 50mph is a fair bit, so added some train brake. Then a bit more, and a bit more, and stuck it in full service. Wondered if the brakes were working. Eventually several miles later, stop. Don't think these particular cars are overbraked!

* Test two: stuck unmodified Dash9 on the front. somewhere at around 42mph accelerating, train derails with no reason given. Dash9s have their own couplers which I'd not touched.

* Test three: unmodified ES44 instead of the Dash9 ( by default it has modded couplers ). Waited till the train found a corner at about 55mph, threw the dynamics on... train starts slowing. Eventually the ES44 starts slipping, but it's losing speed and not falling off the track. It actually appears to be slowing down as fast as test 1, only without the autobrake ... admittedly the ES44 by default has 1/4 of the dyna brake of the Dash9.

* Test 4 will be me editing dynamics on various things to try and jackknife it.

Maybe someone else can try this? this is admittedly a somewhat modified consist. I've attached a copy of my default coupler:

* back up Assets\Kuju\RailSimulatorUS\RailVehicles\Couplings\Buckeye\Type-E\buckeye_type_e_coupling.bin
* put the attachment in there ( or set SpringCoefficient to 75000 and Damping to 60000 if you don't mind editing )
* go and abuse trains.

I've no idea what something that rigid will do for switching... I did not sit down and do any maths.

Edit: well, 6 SD40s with 450kN Dynamics and using a trick to getting full adhesion will jackknife my coal train if I use 100% dynamics instantly, but really you'd kinda expect that. If I set them to a saner 45kN then all is smooth. Now to go uphill & see what the stretching is like.

[Kali]

Something else to check is coupler heights too,I think.