First of all, simulation and modelling has been a very large part of my
career over the last 20 years, and I'm familiar with both the capabilities
and the pitfalls of modeling. I'm not as sanguine as Glenn is about the
ability of ariframe companies to use computer simuations to get aircraft
designs right, because I remember the problems Convair had getting the 990
design right. They did extensive modeling when they designed the 880, and
the prototype and production aircraft matched the predictions of the
simulations so well that they decided to cut back heavily on prototype
evaluations when they designed the 990, which was not a very different
aircraft from the 880. But the 990 had stability problems that they ended
up correcting with a very ugly strake on the tail stabilizer, probably
after considerable expenditures trying to identify the cause and cure of
the problem without redesigning the aircraft. Of course the 880 and 990
were designed in the 50's and 60's when computing power wasd miniscule
compared to what's available today. But no matter how much computing power
you have, you still cannot model every phenomenon perfectly, and some
judgment is required to determine what approximations are valid. My
understanding is that the airframe companies today can do very accurate
modeling of the behavior of new designs, but validation is still necessary.
What then is the value of using computers to model poorly understood, or
complex phenomena? I believe there is significant value, provided you
don't put unreasonable demands on yur model. After all, the Convair
engineers were able to model the 880 quite successfully. It was only when
they tried to push their model outside the region of parameter space where
it workd that they got into trouble. Approximations they made for the 880
were not valid for the 990, which was slightly different. The objective of
modeling poorly understood systems is to look for models which suitably
approximate some aspect of the behavior for some region in the parameter
space, with some degree of invariance due to ignored parameters. Glenn is
not trying to show exactly how the Cambrian explosion occurred, he's simply
trying to show one plausible model. Ideally, the next step would be to ask
what detectible consequences Glenn's model, if it is valid, would produce
in the geological record and look for them. That might be a tall order, of
course. But a model, once it is proposed can be critiqued and compared with
data and results from other studies and learning takes place. A model that
is not proposed is of no value whatever.
As to Ray's lack of computer and mathematical expertise, I sympathize. I
have very little understanding of biology, population genetics, or
biochemistry. I wouldn't dare propose a model of some biochemical system
without clearing it with someone like Terry Gray who understands the
biochemistry and has a good grasp of computer modeling. I'd also feel a
need to learn more about the biochemical system I'm trying to model. All
this to say that we can't know everything. We have to work together and
know our limitations.
Bill Hamilton | Vehicle Systems Research
GM R&D Center | Warren, MI 48090-9055
810 986 1474 (voice) | 810 986 3003 (FAX)