IMLEC results

Trying to insert some xls charts into this thread, how do I get from .xls or .png to .jpg so I can paste in here?

Copy the chart to the cipboard, paste into an image program and save as a jpg and put in your album.

If you are good at flying excel, copy a chart, paste it as an image in word, then right click and save as image.

Neil

More likely that the people publishing the results aren't aware of, or don't care about, the convention.

Too many significant figures in this type of endeavour doesn't upset me at all. I know without the need for a formal assessment that there must be many inaccuracies in the methodology.

By the way, I hope no-one takes the statement that there must be inaccuracies as a criticism of IMLEC. It's an interesting event for Model Engineers rather than an experiment at CERN. As such I like it very much.

Now if we organised an "IMLEC for Geeks", things would be different. It would be possible to tighten up the experimental conditions to obtain more accurate results and better data.

The trouble is that "IMLEC for Geeks" would be time consuming and dull. Better numbers at the cost of human interest. Perhaps this sort of Model Engineering is an acquired taste. I suspect that relatively few forum members are interested in theory, alas! For example no-one has commented on Neil's use of r-squared, or his observation that there must be errors in the table.

Cheers,

Dave

PS I still haven't seen this year's IMLEC results. I'm one of the victims of the ME4542 catastrophe, apparently the worst thing to happen in publishing since the Wicked Bible hit the streets. I'm having counselling...

you have

r-squared is built into Excel without having to do any hard maths and gives a rough and ready idea of the extent of correlation.

I would be interested if others agree the table must have errors if efficiency is calculated on work done and coal used alone.

If anyone involved with Imlec reads this, do they have the accurate figure to share?

Neil

The data I used came from a photo of the display board at IMLEC. Here's a plot of DB load vs persons

It doesn't look too unreasonable, but the left hand 3 don't look part of the rest of the population, and there is a lot of scatter at the 18 persons mark

If I then plot drawbar load per person against persons, you'd expect a straight horizontal line

Again the 3 at the left stand out, and there is the same scatter. Plotting DBL per person vs speed gives

We have the same 3 outliers, but otherwise no pronounced change with speed. I would expect a train of 18 persons, at much the same speed, using presumably the same trucks, to have much the same DBL, unless the passengers differed significantly in weight. To explain the scatter at 18 persons the passengers on run 1 would have to weigh on average 34% more than those on run 8, eg if run 1 passengers weighed on average 12st, run 8 passengers weighed 16 st. This hardly seems likely.

By the way, it's bad practice to ignore outliers unless you have some good explanation

Edited By duncan webster on 10/09/2016 18:54:41

Edited By duncan webster on 10/09/2016 18:56:35

Hi Duncan,

I wouldnt anaylse these things too much based on one year's results.

There have been all sorts of surprising results since 1969. Very quickly the water consumption figure was dropped from the equation. Some years have been blighted by bad coal, and some years by greasy wet rails. Some years certain passenger trolleys were defective and kept derailing.

Large tender locos do very well overall. It does not take much on the driver's part to rest his arm on the tender inadvertantly and perhaps push the tender forwards so causing a complete misreading on the drawbar in the loco's favour. The late Phil Hains got disqualified from IMLEC for this in 1970 or 1971 with his 5"g LBSCR E1 tank loco where his edging on the drawbar was more noticeable than on a tender loco.

Certain individuals have regularly won. Percy Woods with bog standard LBSC locos, Bill Perrett with his much improved Speedy, and Lionel Flippance, Alan Crossfield and John Heslop to name but a few. Others whom I knew and might have been expected to do well with excellent locos came well down the list regularly.

I remember quite a bit of correspondence years ago that the dynomometer cars (in those days the Bristol and Birmingham cars) produced way off results for certain heavy loads, but I stand to be corrected on this.

The formula used for IMLEC is not exactly fool proof. The maths is beyond my abilities, but there are certain problems in the formula. This has (to the best of my knowledge) been acknowledged for many years. The canny drivers take advantage of this. The rest of us who have taken part are so scared stiff and stressed that just completing the 30 minutes is enough!

Cheers,

Julian

Work done is measured direct, no estimation (summation of drawbar pull * distance travelled). It is quoted by organisers to 6 significant figures, I make no apology for doing my analysis to 4 sig figs.

In 'sad person' mode I found it interesting this morning to analyse the published IMLEC results using other criteria to rank performance.

IMLEC rewards thermal efficiency, which resulted in first 'BR Proposed', second 'Duchess of Abercorn' and third 'LMS Patriot'

On the other hand:

• Most powerful: first 'Big Boy', second 'Hunslet Dholpur', third 'Duchess of Abercorn'
• Best Distance: first 'LMS Patriot', second 'Duchess of Abercorn', third 'LNER A4 Wild Swan'
• Fastest : first 'Isle of Man Peveril', second 'Duchess of Abercorn', third 'LMS Patriot'
• Best passenger distance: first 'Big Boy', second 'Duchess of Abercorn', third 'Hunslet Dholpur'
• Best passenger distance per pound of coal: first 'Duchess of Abercorn, second 'BR Proposed', third 'S&D 7F'

The 'Duchess of Abercorn' stands out as a top three performer as determined by six different performance criteria. This is a substantial achievement. 'LMS Patriot' does well too, being 'top three' in 3 categories. 'Big Boy' is a more useful engine than 20th Place in IMLEC might imply. If only it burned less coal!

'Duchess of Abercorn' came second in IMLEC due to a 0.03% inferiority in thermal efficiency compared with 'BR Proposed'. As the measurement errors must be larger than this I think it would have been fair for the Duchess to have been awarded joint first.

If the other performance criteria I tried were relevant to the competition (which they're not), then the qualities of the 'Duchess of Abercorn' are even more impressive.

Health Warning: All conclusions assume the numbers used are correct and that I did the calculations properly!

Well done to everyone who took part and many thanks to IMLEC for running the event and publishing the results.

Cheers,

Dave

Edit: can't spell...

Edited By SillyOldDuffer on 12/09/2016 11:19:11

And therein lies the answer.

In my view, if you really want to test locomotive efficiency, construct a test bed. If you want to demonstrate how to drive efficiently enter IMLEC (or IMLDEC?) and have a thoroughly enjoyable weekend.

I answer to Neil - yes there were errors in the table as published. Issue 4545 may include an amended table.

Diane

Thanks Diane.

Duncan - the 3 at left are with few passengers, part-full trolleys mean each of the (few) passengers is accompanied by an excessive share of trolley weight?

The first two (driver & observer) have a whole trolley to themselves. I suspect that the 'outliers' had one or two partly empty trolleys in fairly short trains.

As passenger trolleys are typically very heavy to ensure stability with joe public on board, I think you need to know the weight of a trolley as well as the number of persons for the calculations to be meaningful.

Neil

Fair point Neil, but if you just consider the runs which had 18/19 passengers (including driver/observer) there is a lot of scatter. I would imagine that they used the same number of trucks, in fact the same trucks. I can see that driving style affects coal consumption, but not the drawbar load needed to pull the same number of people and trucks round the same track at much the same speed. Of these 4 runs, the one with the highes DB load per person is the second slowest, so speed does not seem to be the answer

Hmm. The picture of Karl Midgley shows him with two cars, one empty, the other just him and the observer - which suggests that the 2 are already counted in the passenger count. Is he one of your outliers?

Other photos show cars with both 3 and 4 passengers per car (always 2 on the driving car although it is the same size).

As there is a speed limit, the fastest drivers might have been very steady, while slower drivers were cautious of maintaining momentum around curves or struggled to keep to the limit, both of which might waste energy braking?

A driver who speeds up and down and uses the brake would do more work, and this might (or might not) offset they reduced effciency while not affecting passenger-feet.

A cunning driver would brake the train on downhills allowing the loco to do extra work pulling against the brake while probably helping efficiency. Legitimate but in the class of 'gamesmanship' I would say. This would increase work done per passenger foot.

The need for the big loco to take three runs to climb the 'hill' must have affected its work done figure as well. I wonder if the distance travelled includes the retreat and the re-run up the hill? In other words is the distance from the dynamometer or from the number of completed laps?

It's a right nest of vipers if you ask me!

The real answer could be gotten on a test rig as Diane says, but where's the fun in that?

Neil

Absolutely no fun at all. Let's keep stirring the vipers ...

I've been pondering Big Boy's excessive fuel consumption, the engine's outlier being obvious top right on a graph of work done against coal consumed.

Neil's query about measurements during hill restarts got me thinking. Restarts, drivers leaning on the engine, non-standard loads, and dropping passengers mid-run etc will all distort the figures. When and how the measurements are taken in these circumstances could make big differences too. Perhaps that's why Big Boy appears to be so greedy.

Does anyone have details of the Big Boy used at IMLEC that might explain the fuel consumption? There are too many hits on the web for 'Big Boy' for me to wade though I did find Alberto Celot's model at *LINK* . This is is a Propane burner and wouldn't be burning coal at IMLEC. However Part 7 of the Build description lists a raft of energy gobbling devices: Steam Driven Pump, Electric Pump, Air Compressor, Mechanical Lubricators, Whistle and a steam turbo-generator! Perhaps they are the cause?

IMLEC calculations may be a nest of vipers but I enjoyed 'Raiders of the Lost Ark' and 'Snakes on a Plane' too.

Dave

Following on from Neil's suggestion about number of trucks I decided to try to make some allowance. This isn't shown on the results board, so I ran the You Tube video and counted them for some of the runs. Why I didn't do all will become apparent. Whilst I'm counting trucks I might as well count passengers. Just quoting 3 off. BR Proposed reported as 16 passengers, actually 17 on 5 trucks. Duchess reported as 17 passengers, actually 16 on 4 trucks. Wild Swan reported as 12 passengers, actually 13. Can't Mancunians count? No wonder there is scatter. Note I'm working from a photo of the results board, not ME table.

Our club trucks weigh 50 kg, so allow these at 3/4 person, then repeat the sums and the BR Proposed and Duchess come out within 1.5% on DBLoad per person. If I could raise the energy I'd sit through and check all the other runs, but it's pretty dull running at 1/4 speed to count passing feet.

To answer SOD's questions, only positive DB load is recorded, but distance is incremented forward and backward. Hence BR Proposed did 8.6 laps, Duchess did 10.

To raise another problem, a big loco might have a grate say 30 sq.in. 1" of coal spread across this will weigh 19 oz. When the reported coal use was of the order of 30 oz, how on earth do we allow for the difference in coal on the grate at start and finish of run? 1/4" difference in level would be 4.5 oz, which is more than the difference between 1st place and 2nd.

Edited By duncan webster on 15/09/2016 02:45:27

Thanks Duncan - fascinating stuff, it makes my head hurt.

Another curiosity: all the engines in the competition are 5" gauge apart from 'Russell' which is 3½". I would have expected a smaller gauge engine to be disadvantaged compared with a larger, yet 'Russell' did rather well achieving 6th place.

I spotted another mistake in the ME Table. In 5th place GWR Praire № 4587 is given a thermal efficiency of 1.255% which, if correct, would make her 9th. Presumably it's a typo for 1.5nn%. I'd assumed that the ME Table came straight from a spreadsheet: now I think some poor soul had to transcribe it.

Lots of devils in the detail for sure!

Dave

Hi Duncan,

A most interesting discussion.

The pure drawbar pull is not a factor of just the weight pulled (number of passengers plus weight of passenger trolleys, and the dynamometer car - which is really quite heavy and I reckon equal to carrying 1 or 2 extra passengers). The number of passengers, as has been commented previously, is really a red herring.

A canny driver will request the numbers of passenger trolleys and the number of passengers on each, and whether adults, children, and perhaps no 'large' adults.

The driver is not allowed to touch the brake unless in an emergency for obvious reasons. If he is going too fast, the Observer watching the dynometer readings will warn the driver. If the driver ignores, then disqualification can result.

I have mentioned previously that the best way to get a decent near constant drawbar pull is to pull on the uphill gradients slowly, and then keep pulling down the downhill sections. This also optimises the steaming rate as a constant, and is line with Sam Ell's work at Swindon, and also the best bit of advice re IMLEC that came from Laurie Lawrence.

So many factors affect drawbar pull - as others have mentioned. Drag on curves, quality of the passenger trolleys, behaviour of the passengers etc. A heavily loaded passenger trolley may cause the bogies not to swivel easily on curves creating a great deal of extra friction.

One has to take a view on grate size. A larger grate requires more coal used to get it into a fit state ready for the run. More coal left on the large grate at the end of the run therefore cancels this out. The 30 minute run should be sufficiently long for the coal consumption figures to be reasonably accurate. There is a proviso to this - most drivers will run the fire down very low on the last two laps. This requires a great deal of confidence and skill to avoid a disaster on the last laps.

There are lots of examples of the same design of loco producing wildly differing results in the hands of different drivers.

Cheers,

Julian

Fascinating stuff, and as a newcomer to the hobby, inspirational.

Can someone point me to the details of the 'dynomometer car'. I would like to understand the mechanism and the data logging used.

I am not unfamiliar with dynomometer testing (in the automotive sense) but would like to understand it applied in this way,

Thanks, Phil.