Steam speed records including City of Truro and Mallard

No - but if you are doing something for an academic thesis, then you need to accept that your assessors will want to peer review it. So it is better to get to the bottom of any underlying issues early rather than when you are sitting in a viva with a hostile external examiner ...

The fundamental question is about data, but also about what question you are trying to answer. To take two extremes: if you want to know "how fast did Mallard go?" then the only data you have are the timing marks on the dynamometer roll and their relationship back to a datum point: everything else is calculated from that. On the other hand, if you want to know "how did the LNER market itself on the back of a high speed record?" then the most interesting thing is not the actual speed recorded, but the technical / management process by which the LNER publicised that the the speed was 125mph (and not, say, 124 or 126); and also why that figure changed in 1948. At that point understanding the annotations is interesting to determine who first calculated the figure and how they did it, and what confidence they had in it.

Both are interesting questions, but they are not the same question; and in particular, if you really want to answer the first, then you have to go right back to the actual data that is relevant to a speed calculation - which is the relationship between distance and time.

Tom

 

I transported family in five Allegroes,three Maxis and a Princess between 1982 and 1991.
They were so ill reputed that I just bougth another cheap when terminally sick.
In 1991 we could buy a new Punto that was basicly a 127 underneath and scrapped it around 2005 with 666666km on clock.
My son blew top gasket first time out after getting drivers licence.Damned fool.

 

But, as I have pointed out repeatedly, if you take the quarter mile measurements you remove that error. The distortion he says exists, doesn't actually exist.

I don't believe there is a fault, exasperatingly I think it is a case of misinterpreting the data and not doing enough work on it.

I think this would be classed under a range of assumptions. The LNER considered it accurate and nobody at the time questioned that the Dyno pen for seconds was out in any way. The Dyno car was calibrated before and after runs, and against other dyno cars, in its known history. So I think we have to accept that it probably does record seconds accurately, together with distance.

Deep breath...!

There's not three rows of speed estimates.

You have the original LNER team which calculated certain average speeds from the roll, in an inconsistent manner.

You have Dr Andrews interpretation which takes the same points and declares - wrongly - that there is an issue with the recording device.

You have my attempt which is a different approach altogether. Time and speed are measured on the same graph roll. Each quarter mile is recorded and the time against it can be measured. You do a standard distance/time equation and you get the instantaneous speed at that point.

We are all looking at the same graph and data contained on it and we are doing the interpretation differently.

I agree with that Tom, but I think I had better send you a screenshot so you can see why I am describing what I have in front of me in the way I am describing it. You are right to say "there are no square sine waves" - but when you see what's on the roll, you will understand where I am coming from!

The question I - actually - want to ask is "how fast did Mallard accelerate up Stoke bank really" - that's why I am doing every single quarter mile recorded for the roll. Seeing the speed record result is a by product of doing the whole roll.

The really interesting thing for me is that we have the ability to reasonably calculate the instantaneous speeds over the whole of Mallard's run, not just a portion of it, which is what has happened before.

You are absolutely correct to say that I should be going right back to the actual data that is relevant to a speed calculation - the relationship between distance and time.

That is precisely what I am doing, by recognising that we have a consistent marking (every quarter mile) and the relationship with time on the graph means we know how far Mallard travelled (each quarter mile) and how long each quarter mile took, which will give us the instantaneous speed at that point.

Neither the LNER team nor Dr Andrews took every quarter mile measurement and calculated the speeds at that point. I am supremely confident that when you put all of this on a graph, the issues that Dr Andrews believes he has highlighted will in fact disappear - we will have more points on the graph than before and that will show a likely more natural set of curves for Mallard's speeds (showing its acceleration better).

And if it doesn't at the end of this, then we go back and ask if the dyno car was operating correctly. But I believe it was - I think the data has been misinterpreted.

Why haven't this been done this before? Frankly, because only a very few people have bothered to pay the NRM's staff to have the roll professionally photographed!

 

I believe I have done so, repeatedly. Please go back and re-read my comments and you will see what I am saying.

And with respect to Tom - he doesn't have the dyno roll in front of him, but I do.

With respect to Dr Andrews - I think he made an error in interpretation and that if he had thought about his approach more he could have done what I am attempting to do.

That's not a slight on him - I think his work is important and it has highlighted potential issues - but ultimately I think his interpretation is wrong. I am allowed to disagree, I hope?

 

As a humble non-engineer, I think that however many dots you have, the function you use to join them in the vicinity of a pretty flat maximum is going to be critical to one of the questions you are addressing. There are probably curve fitting choices here which are not resolved by appeals to 'what the data shows', but might be more or less consistent with the data plot.

On the spectrum of theory--model--data you are clearly at the data end of the spectrum, that's fine, but the discussion above is to my mind another example of the three needing to go hand in hand.

I suspect many PhD students would give their eye teeth for the level of interaction you are getting here. Just saying.

 

Absolutely agree, but I am still going through the data. Food for thought, of course.

Agree.

You are quite right, which is why I have apologised to Tom via PM separately, for coming across a bit exasperated and defensive. I do think though we need to start with the data and get a consistent interpretation throughout the roll.

 

I find Andrews' identification of a sinusoidal variation of apparent speed with a one mile period persuasive. There may or may not be an additional smaller-amplitude component with a quarter-mile period. If, Simon, you deny that any such variation exists, please explain why you so believe and where you think Andrews made a mistake.

Using the distances at one-second intervals will indeed provide more data points but also more noise. I am dubious whether it will reveal more useful information, but there's no harm in seeing what emerges. If the mile-period variation is genuine it should show up just the same.

I also strongly support the request for someone to look into the details of the gear train between the fifth wheel and the chart mechanism. Either there is or there is not one wheel in that train which revolves once per mile.

Whether Andrews is right (as I am at present inclined to believe) or wrong makes a big difference to the question of Mallard's peak speed but no difference to the subject of its overall performance up the bank and down the other side.

 

Because, as I have said repeatedly in this thread and elsewhere, the thing he is claiming makes the "noise" is a designed in feature that is part of the LNER's operating instructions for making calculations on the graph paper.

I questioned this when we originally looked at his results, went off and found a partial copy of the operating instructions, and I am satisfied that he's misinterpreted what is happening on the graph paper. The mark made by the LNER team show a consistency with the operating instructions.

The truthful answer is I have a fair idea of what it will show, but I have to finish my work on it.

The fifth wheel at the rail head will spin in relation to the speed of the train. Obviously it will make many rotations before achieving a mile. The chart recorded by the equipment in the coach on the dyno roll will change length between each quarter mile recorded due to the speed.

The dynamometer car thus records time as a consistent pattern of one (1) second intervals, distance recorded on the graph varies by length in relation to speed.

So in short:

• time is recorded as consistent 1 second intervals by way of chronograph
• distance is recorded and varies in length between the recording of each quarter mile, due to the speed of the train
• This means we can treat time as consistently recorded, and recorded that between each quarter mile measurement on the roll
• Knowing the distance as a quarter mile, we can calculate speed by measuring how many seconds it took to achieve that quarter mile
• we can use the standard distance/time equation to produce the instantaneous speed at each quarter mile achieved

The problem is there is an inconsistency in the recording and analysis of the same data. Both the LNER team and Dr Andrews have ignored what is actually on the roll.

In the case of the LNER team, they weren't tasked with doing the whole run, they were only interested in the top speed attained and being able to claim above 124mph (which I think everyone likely agrees, 125mph was certainly achieved).

Dr Andrews has identified something that he believes is a mechanical fault, which it isn't. The graph roll works in a very specific way and should have been interpreted in a very specific way.

 

Oh boy. A lot of this is starting to worry me. Yes, you can grab figures for instantaneous speed off the roll, but they won't be very accurate. That's intrinsic to an electro mechanical recording mechanism. And, to be quite honest, its very easy to assume the mechanism has a greater accuracy than it actually possesses. Its clear that the car was more than good enough to do the job it was designed for, but that job was not recording instantaneous speed measurements at very high resolution.

Andrews has highlighted the problems with the measurement of the 'square' wave, and yes, you can use the pen overrun to get it a little more precisely, and I expect a really talented programmer would be able to write a program that grabbed the arithmetic mean of each tick, and gave you the precise distance and the variation. Its decidedly a non trivial exercise though. I note from Andrews' graphic that the shape of each overrun tick is different, which is exactly what you would expect of ink onto paper in an unsteady environment, but still makes it difficult to get down to decimal points or further.

I too would be a lot happier with Andrews' evaluation if I could figure out what components might actually be creating that mile and quarter mile wave. I doubt there'll be a single wheel, I can't see that you would want to put a one revolution per mile axle in the drive train - if you did inaccuracies from gear backlash and the like would be very significant. More likely its an artifact of numerous inaccuracies in the mechanism reinforcing and cancelling to create the effect. After all, if you measure accurately enough every gear is off centre on its axle to some degree, every cog set is slightly irregular, that's just inevitable.

 

I agree, actually, Jim, but - and I think this is key - doing this exercise will still give us a better indication of the run than we have had to date. I know I can't claim a perfect interpretation - but I'm not trying to. I am trying to give the best picture possible using the data I have available.

That didn't stop the LNER team from applying two decimal points, mind.

 

Yes of course, the chart paper is intended to move at a speed proportional to the speed of the train. But that movement is produced by some kind of gear train between the fifth wheel and the rollers that drive the chart paper.

We're not disagreeing with any of that.
What Andrews claims (persuasively) to have noticed is a periodic variation in the calculated speeds above and below the smoothly rising curve (of speed, calculated as above), which could plausibly be caused by a slight eccentricity somewhere in the gear train. Are you denying the existence of that fluctuation?

 

I have tried my method on a significant amount of the roll and the points I have plotted don't show this same phenomenon currently. But I know (because I know how many plot points he has, and how many plot points I will end up with) that having more plot points is going to help us see more of what was happening throughout the whole run.

If I am consistent and apply my method to the whole roll, then we are likely to get an interesting comparison.

But I suspect - as I have said all along - that Dr Andrews' account is not consistent with how the dynamometer car actually works.

Can I just make one thing abundantly clear, if I hadn't made it clear earlier in the thread.

Milepost 96 marked on the graph =/= milepost 96 at the side of the track. The thing is we don't know absolutely that the dynamometer car's automatic recording of quarter miles matches every position of every quarter mile post by the side of the track. It very likely doesn't! So we have to split our thinking between machine that might have some form of defect (question mark) automatically recording, compared to reflexes of man in coach jotting things down on a moving roll and all the issues that presents.

The point being that if we remove those elements of thinking and start from the position of "The Dynamometer car started at 0 miles and then recorded its distance travelled every quarter mile" we can at least say we can calculate a value for every quarter mile recorded based on that recorded on the roll.

 

I mostly agree, but 1 mile and 1 quarter mile markers would have been useful, and they would have to be driven by the electro-mechanical system somehow. My main retirement interest apart from old railways is classic and vintage motorcycles, and the issues here are similar to those that caused the inaccuracy of their electro-mechanical ignition and valve timing.

 

Simon, sorry to be picky but are you confusing David Andrews with Dr.Henry Ivan Andrews of
the LMS latterly BR, who was very much involved in the LMS mobile dynamometer Tests and
post Nationalisation? His PhD was wrt stresses in locomotive coupling rods. His work
for BR wrt effects of wind speeds on rolling stock resistance was important.

I think David is Eur Ing David Andrews MSc CEng MiMechE. as per his article in the SLS Journal wrt Mallard.
David has experience both as an engineer and as an archaeologist.

Michael Rowe

 

Some thoughts, I believe the paper moved at a nominal 24inches per mile. The distance
between the second marks changes with speed. Does the roller that pulls the paper
have a 24inch circumference, and rotates once per mile ? The error was introduced
in the final part of the mechanism I.e. once per mile.

David Andrew’s observations suggest that the quarter mile marks were not effected by
the paper speed variation.

I understand that the time from measurements taken on the role from MP90.75
to 90 was 21.75 seconds, suggesting 124mph average over the 1320 yards ?

Michael Rowe

ps although Tom has reminded you that errors are cumulative I think when
considering successive quarter mile timings you minimise errors
if you simply take passing times.

 

If finished my popcorn and am trying to get my head around some of the statements being made in recent posts

So, the paper is moved by a mechanism driven by the fifth wheel (my teminilogy but it's actually the ninth wheel). My simple mind can accept that and even understand it. i'm not sure about your second statement. I suspect that it should say 'rotation of the wheels.' Do we know how the paper is actually driven? Is it a positive drive with sprockets or a friction drive, perhaps between rollers? I just hope it isn't a drive to a spool which winds the paper onto it. Is there a clutch in the drive to engage and disengage the drive? I would expect there to be but don't know and is that clutch a positive drive or a friction drive?

Now this simple lad is getting more confused! If the chart is driven by the fifth wheel its movement is recording distance and nothing else. How, then, can it change length between quarter mile markers? That must be consistent. We have bee ntold that 24" of chart movement equates to one mile so each ¼ mile must be 6" of paper movement. Unless the chart drive is not direct and liable to error in the 24" to the mile the only reason for the ¼ mile markers are for ease of interpretation of the chart. They might be a calibration check but, if both chat movement and ¼ mile markers are derived from the fifth wheel I can't see the logic in that.

On the assumption that the chart is moving relative to distance, all that has to be done to calculate speed is measure the chart movement between second markers and you have the answer. However, it is not as easy as that as, at 120 mph chart movement/second is only 0.8" and, as has also been said, the difference in chart movement between 120mph and 126mph is 0.04" or slightly more that 1/32". The difference between 125 and 126 mph is about 6 thou. Not impossible to measure but probably not with any accuracy on something like a chart . The more seconds you measure chart movement over the more accurate you will get but the more you are getting an average speed over that longer time and not the speed at the highest point.

In all this, admittedly very interesting discussion, one thing bugs me. We are not revisiting all this with the benefit of modern instrumentation that wasn't available to the test engineers in 1938. We are looking at exactly the same indo that they did. These people who interpreted the trace in 1938 were people who did this for a living and were well used to the dynamometer car and its idiosyncrasies. They weren't stupid people and probably knew more about the job than the Nat Pres collective and how to interpret the trace.

I've got fish and chips to look forward to now. I'll have another bag of popcorn ready for tonight's learned discussion. Might even enjoy a glass of red.

 

I’m not trying to cast aspersions on those who did it before. I’m simply doing my own analysis and am sharing it on this thread.

If that’s offensive to a few of you, I don’t know what to tell you, because most academics do this every day.

 

A good travelling microscope has a resolution of about 0.01mm, so say 4/10000in.

 

I don't think anyone has accused you of giving offence. In fact @Steve said he's enjoying the learned discussion, as many academics do whilst challenging one another's alternative theories.