any nuclear physicists hanging out in here?

[LtPage1]

I'm taking a class on nuclear weapons (specifically, the involvement of the University of California in their research, development, etc.), and This paragraph jumped out at me from one of the readings. What the hell does it mean?

from p.58 of The New Nuclear Danger, by Helen Caldicott

The goal of the SS & M program is to construct a three-dimensional, full-physics, full-system computer simulation of the explosion of a nuclear weapon. The problem is that as much more new and sophisticated data is fed into these supercomputers, the old nuclear codes derived from nuclear testing become confused. In other words, new data erodes the predictive capability of the old nuclear weapon performance codes. And change in the proven design of a hydrogen bomb undermines its reliability. Therefore, the greater the number of changes induced by a massive array of experiments, the greater will be the incentive to resort to full-scale underground testing.

I think I can follow everything in there except the bits about "nuclear weapons performance codes."

 

[CycloWizard]

Basically, computers can only solve problems that you lay out for them in a language that they understand. This language is a program, and what you tell the computer in that language is your model. The model that you derive is based on existing data. Since the data is now different, due to the factors listed in your passage, the old models aren't really so good anymore.

The other alternative that might explain their usage of 'confused computers' is an autocorrelative approach in which the computer automatically calculates trends based on data that is fed and uses these trends in the model equations that you give it. In either case, the conclusions are pretty much the same.

 

[soccerballtux]

KJI, is that you?

 

[LtPage1]

Gotcha. I actually asked the professor about it today, and he said that the "codes" refers to old data collected from actual testing done before the test ban treaties. The new computer programs attempt to use that information to determine how modern weapons will react; as newer and more sophisticated weapons are conceived, especially designs using very different characteristics than the old warheads (like the proposed "bunker-busters"), the old data isn't as useful.
So pretty much the same answer you just gave me.

 

[f95toli]

One problem with many of these old "codes" is apparantly that they don't make a lot of sense. Essentially, a lot of the information gathered from tests were collected into empirical "rules". Now, the only problem is that some of these rules (which I guess in reality are just equations, e.g. best fits to various experimental curves etc) is that they tend to e.g. violate the laws of thermodynamics and a few other laws that most physicists tend to follow.
(Laughlin writes about this in his book "A Different Universe" which I read recently).

Hence, I suspect they keep running into all sorts of problems when they try to combine these old "codes" with modern computer simulations since the latter tend to be based upon "proper" physics.