On Sat, 1 Mar 2003 18:39:04 +0000 (UTC), Tim Woodall <
[email protected]> wrote:
>>>>The release of 70mph's worth of kinetic energy in a crash is almost certain to kill.
>> Are you not just a little concerned and interested that since in the vast majority of cases the
>> previous kinetic energy has gone before the smash, that there might be something worth observing,
>> discussing and understanding? People are dying out there and 1/2.m.v^2 does not come anywhere
>> close to explaining reality.
>Paul. 1/2mv^2 is reality. Together with F=d(mv)/dt
No. It's physics. On the road it's of great importance once a collision has started, and as far a
crashes are concerned, little interest at other times.
This is a massive problem with a conventional mechanistic approach to road casualties. We look at a
crash and ask: "How could we make that crash different?" We examine a series of coincidences of
chance events and ponder changing some of them. Meanwhile, and if we succeed, there's a risk that
the crash will simply "pop up somewhere else" with a different series of chance events.
I'm trying to take a broader view. Instead of looking at accidents and how they might be reduced, I
want to look at avoidances of accidents and see how they might be improved. There's no KE to worry
about. (yeah yeah brake heat... blah blah).
From the non-accident side there are still plenty of failures to examine, many more failures
actually. But by luck, or something, they don't develop into actual accidents.
>But think about it. I know you don't like the studies but IIRC the Austrialian 60kph study found
>that in 25% of the crashes they studied the driver didn't brake at all before the crash. Regardless
>of what you think about their conclusions and methods I think we can probably all agree that they
>probably got this about right.
I have that study on my HDD somewhere, but I can't seem to find it. It's not under "Australia",
where it should be. ISTR that it studied accident where an ambulance was called, which is a big
filter for big accidents. Non-braking accidents are likely to be bigger accidents (to an enormous
degree actually), so I think the study did a large filtering job.
>Now, I would probably be reasonable to assume that this proportion will transfer to other roads in
>other countries, maybe not exactly but say something between 10% and 50% of crashes on "fast" roads
>the driver(s) don't brake before collision.
At 1 to 5%, I'd have a chance of agreeing.
>So, take the lower limit, 10% of crashes on the motorway occur without braking. Lets assume 10% of
>those are 70mph crashes - the rest being people in queues who don't realise that the traffic has
>stopped again and have a minor bump. (I suspect both these 10% are on the low side)
>So we have 1% of crashes on the motorway occur at 70mph+. If V=dV then we would expect basically
>all of these to be fatal. But I think less that 1% of crashes on the motorway are fatal, therefore
>one or more of the assumptions must be wrong. I have picked what I think are probably low numbers
>for the percentages therefore it must be the V=dV assumption which is wrong.
No way! These motorway crashes have low average impact speeds. I've got some work in preparation
which models the safety system failures and matches the real results. It's essentially very simple.
But it might take a bit of explaining. I'll do a quick sketch here, but if it's insufficient we'll
have to wait until the new SafeSpeed "inattention" page is ready.
*Drivers always try to plan to be able to stop before they crash. They look out for obstructions and
slow when required. Most of the time they have a view of a clear braking space ahead. They are
following the safe speed rule.
*But they are imperfect. Their imperfections have the effect of making them SLOW TO REACT to changes
ahead. This slowness to react can be the result of inattention, poor observation, distraction, bad
planning, hazard perception failures.
*The profile of these failures in drivers is such that failures amounting to larger number of
seconds are very less frequent. Inattention lasting 1 second is extremely common (and manageable)
Inattention lasting 10 seconds is very rare and very dangerous.
*If you create a braking model with inattention and look at impact speeds CONSIDERING at the same
time that before the inattention the driver was able to stop within the distance he could see to be
clear you soon get a profile of crash speeds and fatality risk which matches reality.
End of sketch. The stuff is coming along and I should be able to put it online soonish.
>And this is where you are having problems - if I am driving along the motorway at 69mph and someone
>runs into the back of me at 70mph then, provided I don't end up in a spin and hit something else,
>expecially something sturdy and stationary like a bridge pillar, I am unlikely to be hurt.
Agreed.
>So this part of your mantra is right : "THE SPEED DOESN'T MATTER".
The speed matters a great deal when it is matched to the amount of clear space ahead. The speed on
the tin sign has little to offer, in fact it's a bit of a distraction.
>But "THE DELTA V DOES". and as a basic rule of thumb, dV(max) = 2V(max) for two way roads and
>dV(max) = V(max) for motorways and dual carriageways
But only after the safety system failure.
>But what you want is:
>dV(average) - which if I recall your 12mph page correctly you calculate[1] as 21mph = \alpha
>V(ave-accident) where V(ave-accident) bears some resemblence[2] to the speed of the drivers who do
>crash on the road in question _before_ you consider how much braking etc they actually do.
We're back to the mechanistic approach, and it's hard or impossible to make a match with
non-accident driving.
>Finally, once you have got an average value for \alpha you need to think about where that comes
>from. I would be very suspicious of any interpretation that doesn't end up with an expression for
>\alpha which isn't at least quadratic in V(accident) (Maybe linear but it is so far away from my
>fields of experience that I really don't know how all the various things like driver attention,
>braking ability of cars at varying speeds, risks of skidding etc might interact)
>This work is EXTREMELY difficult to do, even finding "rules of thumb" from real world crashes is
>difficult to do because separating out the random variables from the contributary variables is
>tricky and population sample sizes tend to be small compared to the enormous number of degrees of
>freedom there are.
All true. But what about the driver?
>There is a short essay by Isaac Asimov (I can't recall the title but I can probably find it for
>you if you would be interested) about the "Laws of Nature" and where they come from. To
>paraphrase (badly).
>I throw a green rock into the air...
[snip Azimov]
>I hope you can see that a simple rule is getting extremely complicated with all the conditions you
>need to consider.
Sure.
>F = d(mv)/dt is so remarkable because there really aren't many special cases. And if you do get
>into the realm where v->c there is a slightly more complicated equation to do the same job.
>And this very simple equation will work regardless of how complicated the crash is. It will even
>account for the fuel that is burned between the crash becoming inevitable and the crash happening
>if you want.
>Joksch (Hope that is right) has come up with the "Everything that is thrown up comes down again but
>not necessarily straight away and only if it is doing less than 7 miles per second when it leaves
>my hand" sort of rule of thumb. And then you are applying it to a Saturn V rocket on its way to the
>moon. It's not that your calculations are wrong: as far as I am aware there is nothing wrong with
>your maths, it's that they are meaningless. This seems to be the stumbling block you are having.
I don't agree that they are (were?) meaningless.
>The same thing occurs with your strange lines and predictions on the accident rate curves. I don't
>know how you have got them or why you are so sure they are right. I've played with the figures and
>I just don't know how to make any predictions from the data it all seems so random. I'm sure there
>are some people who post here who have sufficiently advanced stats knowledge that they know whether
>any sort of best fit has any meaning. I suspect that the answer would be that you can't make a fit
>to that sort of data and say something changed in 1991 (or whenever) and they could probably back
>it up with pages of maths about variances, regressions etc but they don't need to. They can take
>one look at the data and say "You can't do that, it won't work." The onus is then on you to prove
>that the noise etc isn't sufficient to invalidate your claims, not on them to prove you wrong.
>Maybe you are the new Galileo, Copernicus or whoever. Maybe you are going to prove the experts
>wrong. But Galileo and Copernicus knew their opponents fields well, and they could point out the
>problems with their opponents arguments. Now, maybe a stats professor should take the data, do the
>umpteen hours work, and give you the confidence figures for there being a sudden change in accident
>rate in 1991. But given that you have given no evidence that you would understand that work, why
>should they waste their time?
I've talked to a LOT of road safety researchers since I started Safe Speed. So far, and completely
without exception, they have known virtually nothing about expert road driving. The degree of
blindness to the road drivers point of view has been absolutely mind boggling.
There's something about driving that makes folk say "I know enough". Drivers do it, and the
researchers do it.
This is why I can illuminate road safety issues and approaches that the researchers have missed.
They universally ignore the specific role that the driver plays in placing his vehicle "at a speed"
and "with respect to hazards".
>I'm sorry if this seems a bit blunt. I think you mean well. You have made the effort to improve
>your driving and you want the roads to get better and safer. That is a laudible aim. But you are
>way out of your depth in the theoretical stuff you are doing and it just looks silly.
This characterisation is false. Not all my hats are a perfect fit, I agree. But I can illuminate and
highlight massively important aspects of the whole debate which are being ignored. I'm not afraid of
pushing back the boundaries, and I'm happy to make mistakes along the way.
>In addition, cyclists tend to come from the higher socio-economic groups and be better educated
>than the population at large. So you are far more likely to find a lot of people on u.r.c. who are
>going to look at your work and say "that can't be right" rather than "I don't like speed cameras
>and I don't understand this work so I think it is great because it 'proves' cameras are dangerous".
Sure. And like I said in another post, at the end of the day it just helps make my case stronger.
--
Paul Smith Scotland, UK
http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives