Atomic Systems

DarwinS wrote: I will give you what information I can find today, but will probably need to wait about 2 weeks until I have time to get into the library/archive at the National Railway Museum or perhaps even the MR archive in Derby, to get any additional technical details.

Thanks for the feedback.

When you go to the NRM, would it be possible to get a photocopy of the following publication if they have it?

BR Test Bulletin #19 - English Electric 3,300 HP Diesel-electric DELTIC

DarwinS wrote: Now that I have the friction parameters correct my model does work as a compound with the correct parameters (need to test more):

Sounds good, keep us posted on your testing.

Can you share the performance log that you are using to set the "reality benchmark"?

DarwinS wrote: including cylinder cocks operating which does not function on present ORTS compound

Is that still the case, this was supposedly fixed in #3797

systema wrote: My idea is to have a spreadsheet similar to those in FCalc but modified to allow input in Imperial units for Weight, Weight on drivers where applicable and Frontal Area in Square feet. This would show the relevant conversion to US tons and Metric Tons plus Square meters. The calculated Davis Coefficients in both US units and Metric Units would be shown.

This appears to be duplicating FCalc to a certain extent. I will give it some thought.

Thanks

@Systema

Is this what you are looking for:


I could put this together, but not sure about protecting cells so that users can not mess it up.

Please check the metric values for A and B. I have used the conversion factors that you have posted, but my spreadsheet gives me lower values than the FCalc program... even though it gives the same values for US units.

@steamer1960 and systema

A question of accuracy and limits of error here. Looking at the Davis equations used in FCalc all the coefficients are convenient whole numbers (two significant figures). The Davis equation is an empirical one that works because it fits the data. If Davis did not see fit to determine his constants to more than 2 sf, then presumably the calculations from it were not intended to be any more accurate than that. I have in the past taken the output of FCalc to four significant figures assume it to be accurate, now I am of the opinion that any numbers we use in a simulator should have only two or at most three significant figures.

Looking at the page above, should there also be an alternative C value for streamlined steam locomotives - and if so how might this be calculated?

Is that still the case, this was supposedly fixed in #3797

I am running OR version 1.2 was this a subsequent fix?

The first reality check that I am using is this one:

Compound
(11) 329 / 350
St Pancras 0 00

Kentish Town 4 17
West Hampstead
Cricklewood
Hendon 12 23 57
Mill Hill
Elstree 18 42 45
Radlett 21 34 67
St Albans 26 09
Sandridge
Harpenden 31 36
Chiltern Green
Luton 37 01 67
Leagrave 39 25
Milepost 34
Harlington 43 18
Flitwick 80
Ampthill 46 46 75
Bedford 53 32 eased
Oakley 56 44
Sharnbrook 60 24 60
Milepost 59 ¾ 64 17 42

Irchester 72

Wellingborough 69 08 66
Finedon 72 29
Burton Latimer

Kettering 75 49 59
Glendon 47
Desborough North 83 33 45
Market Harborough 88 00 68/53
East Langton 91 32 62
Kibworth 95 10 47
Great Glen / 70
Wigston Magna 100 30
Leicester 106 10

I am expecting to approximate to this within about 10%, because as often the case with the MSTS world the Thames Trent MSTS route is about 10% shorter than it should be and some of the gradients are rather steeper than they ought to be.

The second is this one:

Compound
166 / 175
Girvan 0 00
Milepost 1 ¼ 5 15
Loco slipped severely on wet leaves at Milepost 2 and almost stalled.
Speed recovered to 17.5 mph at top of 1 in 54 increased to 21.5 on 1 in 80 and maintained 21 on 1 in 56.
Milepost 4 14 53
Pinmore* 16 50//
Pinwherry* 4 35//
Max speed 28 mph on gradient
Barrhill* 9 58//
25.5 mph maintained on 1 in 67
Milepost 15 ½ 8 35
Milepost 16 ½ 10 08
Glenwhilly* 14 45//
Milepost 22 2 52
pws
New Luce* 7 15//
Challoch Jc* 6 25
Dunragit 8 20//
Castle Kennedy 4 50//
Stranraer 5 30

I have used Big Vern's old Stranraer-Ayr route and added the extra stations. Although an old route the gradients are reasonably accurate and very severe.

I have just searched the various online catalogues for the NRM (Search Engine). I can not find a copy of

BR Test Bulletin #19 - English Electric 3,300 HP Diesel-electric DELTIC

I will try asking them (e-mail). The nearest I could find was:

637 TEST DIESEL 8 English Electric Co Locomotive Test Reports: Deltic Co-Co Prototype and Production 1956-1962
ALS3/74/C

This might perhaps include Bulletin 19.

I have been away at a Eurotrack modelling show for the weekend, and I see this thread has been quite busy. As one who comprehensively tested the code as it was developed, I have but one thing to say with regard to data input. Where the real world sizes are known adhere to them, do not make changes. |The code will take those figures and give an output that should be very close to the real deal in terms of steam production and usage, power output at the wheel and drawbar. The one thing that the code does not do is work very well with an Fcalc calculated Davis C value. What one needs there is an existing resistance curve for an item of rolling stock and using one of them behind an engine, record the values at 10 mph intervals, then see how that plots against the known curve. Adjust the C value if the curve is wrong and retest. Once that is done, all stock of a similar type can use the same Davis figures. Then with a known set of figures and a train to match set up the loco with the initial Fcalc figures and run the train to its maximum attainable speed. See if the complete train can match the published figures. If the figures achieved are lower than published adjust the loco Davis C downwards and try again. Keep going until the figures achieved match the known ones. The loco will then perform accurately in all circumstances, lighter trains will give better fuel figures, heavier ones will give higher fuel consumption for the same distance and run at a slower speed due to the extra resistance.

Using those methods, I have been able to Run the Royal Wessex activity on Dorset 6 and have on time arrivals all the way. The only problem is a lack of watering points at the station stops, to have to rely on the Shift+t immediate refill, but unfortunately this also fills the coal bunker.

I am away for a couple of days but will pick up the discussion when I get back. I would say however it seems odd to me that OR doesn't use the C coefficient properly. I understood that fcalc was meant to replicate closely the Davis Curve. I have no idea where to get real resistance curves. Copper pens process sounds very laborious.

Mick Clarke
MEP

The process may well be laborious, tedious even, but is the only way to get performance correct. As the Davis curve is an empirical set of figures made to fit a particular set of circumstances, and the formula can change, depending on what source you refer to, it is quite probable that OR is not giving an expected result "out of the box" where the C value is concerned.

As long as the baseline locomotive performance is correct, steam output and usage figures do not exceed the expected rates,the C data can be altered to match a known result which should then allow the model in OR to do the same as the real world performance. In the end, the single factor that will determine a locomotives maximum performance is the boiler/firebox/calorific value of fuel combination versus total train resistance.

There is still some work to do on the steam firing code, at present it acts like a continuous mechanical stoker, whereas it should be using the little but often principle for the most part.

As one who comprehensively tested the code as it was developed, I have but one thing to say with regard to data input. Where the real world sizes are known adhere to them, do not make changes. The code will take those figures and give an output that should be very close to the real deal in terms of steam production and usage, power output at the wheel and drawbar.

So far I have found some adjustments to be useful, representing the fact that some steam locomotives were less efficient than others.

(1) ORTSBoilerEvaporationRate ( ) - in some cases, boilers could not reach the maximum output calculated by ORTS because of factors such as draughting (improved by redesigned blast pipe, double chimney, giesel ejector..) or for other reasons related to length of tubes compared to diameter, free gas area, etc. Where boiler output has been measured I fell this adjustment is justified.

(2) ORTSCylinderInitialPressureDrop( ) - represents the valve gear setting.
I have used this in the Princess so that work achieved at a given cut off is realistic - otherwise I might have to haul trains with 5% cut off where in real life 20% was used for the same speed/power. I would also use this to represent short travel valves on older locomotives, where steam supply to the cylinders got choked off at higher speeds. [MR 2P 4-4-0 Mean Effective Pressure at 300 rpm was 30% of that available on starting compared to 60% for a rebuilt Royal Scot.]

(3) ORTSBurnRate( ) - should help to give a more accurate coal consumption, however it is not easy to get information on this. I tried to make an estimate for the Jubilee based on pounds of water evaporated per pound of coal in road test reports. When I later found that one of this class had been laboratory tested I found that my original estimate was not correct.

The one thing that the code does not do is work very well with an Fcalc calculated Davis C value. What one needs there is an existing resistance curve for an item of rolling stock and using one of them behind an engine, record the values at 10 mph intervals, then see how that plots against the known curve. Adjust the C value if the curve is wrong and retest. Once that is done, all stock of a similar type can use the same Davis figures. Then with a known set of figures and a train to match set up the loco with the initial Fcalc figures and run the train to its maximum attainable speed. See if the complete train can match the published figures. If the figures achieved are lower than published adjust the loco Davis C downwards and try again. Keep going until the figures achieved match the known ones. The loco will then perform accurately in all circumstances, lighter trains will give better fuel figures, heavier ones will give higher fuel consumption for the same distance and run at a slower speed due to the extra resistance.

This is where most of us will need help. Perhaps someone with a better understanding of railway physics and engineering could trawl relevant sources and post some guideline figures (in the correct units for ORTS!) that will get us started and help things move in the correct direction.

Since the majority of things that I model are late 19th and early 20th century I doubt that there is much data available so will need to know how to make reasonable estimates from data that is available.

FCalc is easy to use for any rolling stock - would it be possible to make a similar C value estimator that would be of more general value for ORTS?

As the C value for locomotives (or leading vehicles) is very much greater than that for other vehicles, might we accept FCalc values for other vehicles as being acceptable and then concentrate on adjusting the locomotive values?

On further testing (Girvan-Stranraer!) I can say that using the ORTS compound version does not seem to be giving me the required performance at low speeds.
Running on the Midland Mainline was not so bad as although uphill speeds were low - which I partly attributed to some of the hills being too steep, it was able to reach some good speeds down the hills.
On steeper grades it is immediately apparent that there is not enough hauling power at low speeds - having let speed drop to 3mph on the 1in54, the best I could recover to before the easing of gradient was 13mph. On the easier stretch we recovered to 18 and then dropped again to 17 on the final section of 1in56.

compare to:

Speed recovered to 17.5 mph at top of 1 in 54 increased to 21.5 on 1 in 80 and maintained 21 on 1 in 56.

Full gear, full regulator, plenty of steam that was going out of the safety valves rather than through the cylinders.

Using the simple version given yesterday then the speeds at those points were 20, 24 and 23 mph. With a slight tweak to the Initial Pressure Drop then that can be changed to an almost exact to the log 17 / 22 / 21 mph. The cylinders then read:

Comment ( *** Cylinders *** )
NumCylinders ( 2 )
CylinderStroke ( 26in )
CylinderDiameter ( 19in )
ORTSCylinderInitialPressureDrop( 0 0.94 50 0.73 100 0.66 150 0.65 200 0.64 250 0.62 300 0.54 350 0.39 )
SteamCylinderCocksOperation ( Manual )

Giving power outputs as follows:

5mph 241ihp 17.9klbf
10mph 420ihp 15.8klbf
15mph 576ihp 14.4klbf
20mph 739ihp 13.5klbf

At somewhat higher speeds then the outputs from this model are presumably much the same as using the "correct" compound settings. Anyway that is the best feedback I can provide until I can research further about the real thing. Hopefully I can find some technical details and some road test data, possibly with indicator charts and so on. (None of these locos were ever put on the rollers so there will be no lab report. (Incidentally all of the reports for locos that were put on the rollers at Rugby are available at the NRM.)

I have Verns' old route so am able to do some testing on it. I don't currently have any Midland/LMS carriages installed, still archived. If you can give me the make up of your test train, I will see what results I get for comparison. I have my compound set up with the 7ft drivers, but it is an easy change to 6ft 9in.

With regard to adjustments, the three items you mention are designed for that purpose. It is the dimensions that should not be altered.

DarwinS wrote: I will try asking them (e-mail). The nearest I could find was:

637 TEST DIESEL 8 English Electric Co Locomotive Test Reports: Deltic Co-Co Prototype and Production 1956-1962
ALS3/74/C

This might perhaps include Bulletin 19.

Thanks for that. This site (http://www.traintesting.com/steam.htm) maybe of assistance in correctly specifying it.

systema wrote:I would say however it seems odd to me that OR doesn't use the C coefficient properly. I understood that fcalc was meant to replicate closely the Davis Curve.

FCalc does replicate the Davis equations, but sometimes "more accurate" information is available with a bit of research. Train resistance is an important subject for railway engineers as it can significantly impact the performance of the train, and its efficiencies, so there is a fair amount of research and information around. The only challenge, like all railway topics, it could be a bit daunting.

systema wrote:I have no idea where to get real resistance curves.

Try doing a Google search (there are quite a few papers available on the Internet), alternatively a trip to the railway archives might turn up some curves.

For example, this site ( http://users.fini.net/~bersano/english- ... /BR-tests/) has a number of test reports with resistance information quoted. In particular you might be interested in the BR_P&E_No15_BR_Std8 report on Graphs 38 (coaching stock) and 37 (locomotive).

systema wrote:Copper pens process sounds very laborious.

I have developed a process with a spreadsheet tool, that hopefully should make the process faster and more accurate.

See - http://www.coalstonewcastle.com.au/phys ... nce-fcalc/

I still have a couple of conversion tools that I going to add to it.

@steamer1960
Thanks for the comparison spreadsheet. What this shows me is that the values I have from FCalc2 for the A and B constants must be too large. Assuming there would not be a large difference between LMS stock and BR Mk I stock then this means my resistance at low speeds is too great now. This could be a factor in the hill climbing test with the compound.

Anyway this means that I can now build an appropriate conversion factor into my spreadsheet that should hopefully be good for UK rolling stock.



Hopefully third time lucky and I will not have to redo the Davis A, B and C factors for my LMS carriages again now. Before I start mass replacement I will try and see what happens with the hill climbing test.

The next trick will be to try to look at the locomotive in a similar way. If I input the data for the BR 8P into FCalc and then work out how to adjust to get a best fit line, hopefully the same can then be applied to other locomotives.

Here is my attempt to apply the same to the locomotive. As the published data refers to locomotive including tender, I have entered both as a single unit into FCalc. The green curve is the standard FCalc steam loco curve, the blue curve for rotary cam gear (which I think applies to this particular loco!).



From that chart it seems that FCalc is good for low velocity, but that a much higher value of C would be needed.



So I think I have my working numbers that can go into the sim and can hopefully apply the same factor to other locomotives.

The changes in resistance did not make a significant difference to the performance of the "Compound" so I feel quite sure the model is not developing sufficient power at low speeds.

Anyway I must take a break for some real world work just now. I will follow the thread and contribute where possible, but will not have access to Open Rails again until next week.

Darwin,

Your spreadsheet is wrong. A UK ton is more than a metric ton and a Metric ton is more than US ton.

so for 32UK tons = 32x1.0160469088 = 32.514 Tonnes (metric) = 32x 1.12 =35.84 US tons.

For Brits your table should start with Imperial units on the left and show the calculated US and metric amounts. The appropriate correct Davis numbers can then be calculated from these.

Mick Clarke
MEP