Slashdot Mirror
Simulations May Explain Loss of Beagle 2 Mars Probe
chrb writes "Researchers at Queensland University have used computer simulations to calculate that the loss of the US$80 million British Beagle 2 Mars probe was due to a bad choice of spin rate during atmospheric entry, resulting in the craft burning up within seconds. The chosen spin rate was calculated by using a bridging function to estimate the transitional forces between the upper and lower atmosphere, while the new research relies on simulation models. Beagle 2 team leader Professor Colin Pillinger has responded saying that the figures are far from conclusive, while another chief Beagle engineer has said 'We still think we got it right.'"
another chief Beagle engineer has said 'We still think we got it right.'
They got it right, yet the mission failed. What sort of weird mental block do these people have?
Something else might have gone wrong even tho the choice of spin rate was the correct one.
In Doctor's terms: The operation was a success, but the patient died.
Sleep your way to a whiter smile...date a dentist!
Yes, perplexingly, there exists the possibility that they selected the correct entry spin rate but the mission failed for another reason. It's one of those perverse physics concepts like things falling down and cats and dogs not living together. Frankly, if it turns out that their own spin-rate calculations were valid and correctly computed, yet they caused the mission to fail, we'll have learned something very important for future missions. They could just throw their hands up and say "yes, our spin rate calculations were guff" to satisfy you that they do not have a "weird mental block" but that's not how it works in grown-up land.
No kidding!!! What do you say at this point?
Furthermore, their choice of spin-rate may have been made with the best possible available knowledge and approximation, and still been fatal. In which case we will have learned something - perhaps something really important, given how many Mars missions go awry - about successfully landing probes on planets.
No kidding!!! What do you say at this point?
.... that The Transformers got it.
In fact, it broadcast 13 seconds of footage. It can be viewed here. http://au.youtube.com/watch?v=wFvUdt9BQhU
This space for rent
Name a one thing British ever made right.
Canada.
Now, this may be crazy talk, but shouldn't you do the computer simulations BEFORE sending the $80m craft on it's way?
...Had this been an actual emergency, we would have fled in terror, and you would not have been informed.
"Beagle 2 team leader Professor Colin Pillinger has responded saying that the figures are far from conclusive, while another chief Beagle engineer has said 'We still think we got it right.'"
-So they really *did* intend to burn the craft up on re-entry? If they did, what's all the hubbub about?
Knowing Google's lust for data collection, the Soviet Union is still alive and well inside the psyche of Sergey Brin....
well according to the current score, the game is about a 20 all tie -- although this doesn't count any points scored this year
http://www.bio.aps.anl.gov/~dgore/fun/PSL/marsscorecard.html
There is something wrong but interesting about the idea that a computer simulation can explain what happened in a real-life incident. In the normal usage of "explain", only causally-related events can explain other events.
There is undoubtedly something to the contention that a computer simulation does some explanatory work, but it must be in a roundabout way. Maybe this: the computer simulation provides evidence to the effect that some prior event was able to cause the known outcome; but then it is the prior event (the bad choice of spin rate in this case) that explains the loss of the Beagle 2, not the computer simulation.
Come on, Cider is french :-D
Herve S.
Are you talking to me?
At the bottom of the
Dji Ingliti langwij and haw it iz speld.
At the bottom of the
There has been a very detailed assessment of this failure by the European Space Agency at the time, and the guys were quite competent being the ones that built the (successful) Huygens probe to Titan embarked on Cassini.
:(
ESA found many issues, mostly due to way too severe cost constraints (a "british-only" program...).
Among them IIRC, the main parachute that was changed in extremis (when the unpaid earlier maker announced they wouldn't go up to offer the flight model too) resulted in a drag coefficient that was smaller than the drag of the front shield, this big "bottom" device that you drop immediately after the peak entry aerothermal flux. Having such a drag ratio means the front shield could perfectly have slammed back onto the descent module upon release, or just inside the parachute itself with the consequence you can imagine (all of this happens at around Mach 1).
And that one was just an issue among others...
I'm searching for that ESA document but I just can't track it back right now
Name a one thing british ever made right.
Railways. Television. Electric motor. Flushing toilet. Steam engine & locomotive. Computer. Seed drill. Tank. Custard. Cat flap. Jet engine. World wide web. Penicillin.
http://www.thespacereview.com/article/330/1
http://www.thespacereview.com/article/347/1
As for Colin Pillinger, note that the (initially secret) ESA report on the Beagle failure put much of the blame on project management failings and he's not been put in charge of any large project since.
It's one of these: http://www.sciencecartoonsplus.com/images/miracle3.gif
Name a one thing british ever made right.
Scotch Whisky.
Rolls Royce cars.
Aston Martin cars.
TVR cars.
Lotus cars.
Triumph motorbikes and cars (and bras and knickers...)
Marshall amps.
Trace Elliot amps.
Orange amps.
Vox amps, guitars and organs.
The AVOmeter.
Harrier V/STOL aircraft.
The Hillman Imp.
No, wait...
Squirrel!
So what is a "bridging function"? Definitely not something about Ethernet bridging... but what is it?
From the context, I would imagine that it is a function that interpolates between the behavior of and forces on the craft in orbit (well studied by previous orbiters) and the behavior expected in the lower atmosphere (well studied by previous probes). The intervening region is probably not that well covered by available data, so some sort of function must be guessed to fill-in the gap between the two datasets. There may be features in the upper Martian atmosphere that were not present in the bridging function that they used to model the probe landing.
If the spin rates were off afterall, then they could have lost two or three probes instead of one. It's always a gamble, and if a mistake with big consequences has been made, sending more probes might not give you more chance of success...
For something so mission critical, well... you'd think they would have more than one of them up there.
Beagle 2 was not mission critical - it was an underfunded bolt-on to Mars Express, which is doing quite nicely, thank you.
Best troll for a long time, but I cannot resist biting ... how about Monty Python?
How many different independent forces could have influenced Beagle? Represent each with a variable. Calculate how many emergent properties could have influenced the craft (those arising from interactions between main variables). Assign these a variable. Estimate the range of values for each variable. Calculate the dynamics of each variable (ie. linear, logarithmic, hyperbolic, etc., including estimation of those whose behavior does not fit a simple function, instead requiring complex functions). For each variable, estimate a reasonable granularity (they may be analog, but the resulting computation would include infinities, so digitizing is necessary). Calculate the matrix necessary to represent all the possible results. Determine whether the calculations could be completed in polynomial time. Almost certainly not, so estimate how many variables (and their dynamics) must be retained and drop the rest. Calculate the solutions matrix for this reduced set. Check for polynomial time solution. If no, reduce yet again. With each reduction estimate the error range introduced, and whether any of them are unacceptable and the prior value retained.
Estimate the amount of computational power/time necessary to complete the solutions matrix, including the cost of buying/building/renting/etc. and your available resources. Calculate how many orders of magnitude there are between what's necessary to solve the problem and what you have to work with. From that estimate how much you have to reduce the solutions matrix in order to be able to arrive at some solutions, as well as how inaccurate any results will be.
Once you have the calculation of the solution set down to polynomial time and within your budget, look at how inaccurate your results will be. If the accuracy is found to be acceptable, and the calculation therefore worth doing, chances are you've made a mistake in your estimations. Almost certainly the inaccuracy will become too great before your reductions result in a solvable problem. Also note that the minimal matrix dimension will probably not be an integer. Choosing the best number of variables would be trivial, as you simply choose the next highest integer. However just because the solution here is between N and N+1 does not mean that there is only one variable with a fractional influence; estimate how many and which variables are best characterized as non-integers and select the best set of variables to use in the model. Calculate how far back into non-polynomial time your solution estimate has drifted, or at best how far over your resource budget the calculations will require.
Take a dose of analgesic of your choice sufficient to eliminate your headache. Begin building a model using the minimum number of (integer) variables necessary to arrive at a variable/value set that produces a result matching the behavior of the phenomenon you wish to model. Ignore the probability calculations that would indicate how likely it is you're wrong, and how many such wrong solutions you'll arrive at before you happen on a possibly right solution. Instead of using probability estimates to calculate statistical significance of any calculated solution, use the fact that a solution can be found that results in the same behavior as the one to be modeled, and wrongly call that accidental similarity 'practical significance'. Publish a factually unsupportable assertion that your model describes what happened based only on the fact that your model achieves the same result and count on the fact that nobody else on your research team, or anyone for that matter, is capable of accomplishing the necessary calculations described here to conclusively state you're wrong, or at best that you can't say you're right.
Estimate the positive influence the number of publications, regardless of validity, has on the probability of receiving future funding and amount thereof. Conclude that minimal-guess "modeling" provides you with the ability to say something that sounds reasonable whereas attempting to achieve real validity would take too
"I may be synthetic, but I'm not stupid." -- Bishop 341-B
Name a one thing british ever made right.
Australia.
but that's not how it works in grown-up land.
...slashdot's not grown-up land.
This is clearly a matter of an unfortunate choice of names. Naming it "beagle" was clearly asking it to wander off and get lost. Any beagle owner will be able to attest to this.
It's no different than the "dark matter/energy" concept of
"Damn, your calculations are totally wrong, and your model is a piece of shit."
"Nooo... The universe lies to us! There is some dark stuff that we can't see, or measure, or anything ever, that has just the effect to fix our model, and nothing else!"
"How convenient!"
Any sufficiently advanced intelligence is indistinguishable from stupidity.
Okay, okay. Besides railways, television, the electric motor, the flushing toilet, steam engines, computers, seed drills, tanks, custards, cat flaps, jet engines, world wide web and penicillin; what have the British ever done for us?
How could you leave off Monty Python?
Well, thanks to the Internet, I'm now bored with sex.
b = guess(a)
Table-ized A.I.