Stuxnet News

Another piece of the puzzle:

New research, published late last week, has established that Stuxnet searches for frequency converter drives made by Fararo Paya of Iran and Vacon of Finland. In addition, Stuxnet is only interested in frequency converter drives that operate at very high speeds, between 807 Hz and 1210 Hz.

The malware is designed to change the output frequencies of drives, and therefore the speed of associated motors, for short intervals over periods of months. This would effectively sabotage the operation of infected devices while creating intermittent problems that are that much harder to diagnose.

Low-harmonic frequency converter drives that operate at over 600 Hz are regulated for export in the US by the Nuclear Regulatory Commission as they can be used for uranium enrichment. They may have other applications but would certainly not be needed to run a conveyor belt at a factory, for example.

The threat of Stuxnet variants is being used to scare senators.

Me on Stuxnet.

Posted on November 22, 2010 at 6:19 AM • 41 Comments

Comments

gregNovember 22, 2010 6:33 AM

I find it odd that frequency converter drives that operate at over 600 Hz are regulated for export. I mean its really not hard to make yourself. I have done it for a power rating of just 800watts. But the whole thing should have been fine for 1.6kW with the cooling and component choices I made. Moving up to the next level of power transistor and ~10kW is no problem.

Seriously, i would expect any EE grad to be able to do this with "unregulated" components from the local radio shack.

guillemNovember 22, 2010 6:43 AM

Stuxnet is a Very Fine combination of software and intelligence :O Regardless of its real origins or purpose I admire its creators, intellectually. Does it have an official logo or something? You know, for t-shirts :D

WinterNovember 22, 2010 6:47 AM

Stuxnet interferes with the operation of devices needed to uranium enrichment in Iran.

Why am I not surprised?

@greg:
I think the relevant word is "Low-harmonic".

I suspect you need high precision, low distortion frequency conversion. Maybe they are chained to passive converters for driving ultra-fast centrifuges. The motors powering these centrifuges must be very precise. A centrifuge powering up and down during operation will mess up the separation process.

singuNovember 22, 2010 7:13 AM

10 kW is still peanuts in heavy industry. Now factor in that all that stuff need to be controllable and of course - failing got very, very nasty consequences ( think Chernobil )

AnonNovember 22, 2010 7:24 AM

I find it especially annoying that Stuxnet is being used to scare senators - when the most likely author of it is the US or one of its allies.

The notion that it may be reused or repurposed is silly. It's now one of the most well-studied and well-known pieces of software out there. Threats *like* it are of course reasonable - which is why we should invest in defense of these vulnerable systems... Most other tactics (like attacking countries or passing laws to punish people) are silly.

Clive RobinsonNovember 22, 2010 8:22 AM

@ Winter,

'I think the relevant word is "Low-harmonic"'

Not that much of a problem you can do it with Class D amplifiers summing their outputs into a toroidal transformer.

You put the first few "Walsh sequency terms" in and you get a quite nice looking sinewave output with very very low spectral content. The efficiency can be very high (up in the 90% range) and with the right choice of ferrite quite broadband.

You can using very small ferrites get a 400KHz to 2.5MHz span (AM modulated) at upto 10KW quite easily suitable for connection to an antenna.

Juha-Matti LaurioNovember 22, 2010 9:20 AM

Finnish vendor Vacon has stated that Stuxnet doesn't directly impact to its AC drives.
Quoting their posting at http://www.vacon.com/Default.aspx?Id=480277
"According to experts, Stuxnet seems to be capable of infecting one manufacturer’s certain programmable logic controllers (PLCs) which in turn can be used to control the operation of Vacon AC drives."

SteveNovember 22, 2010 9:20 AM

@greg: "Seriously, i would expect any EE grad to be able to do this with 'unregulated' components from the local radio shack."

Not to worry, we'll close that loophole by regulating the export of EE grads.

seanNovember 22, 2010 9:52 AM

And now that it's in the wild, how soon before we see a little of that asymmetric warfare stuff going on with Stuxnet being repurposed for lower frequency, more common converters. I can think of all sorts of neat applications where you could generally have humorous pranks being pulled for fun and profit.

Clive RobinsonNovember 22, 2010 10:04 AM

Although I'm fairly convinced (and have been for some time) that the target of Stuxnet is the controlers for the centrifuge cascade for enrichment, I'm less certain who the real target is.

That is there are other candidates other than Iran...

One of whom is North Korea.

Unlike the design of the yet to be commisioned Nuke power plant in Iran the design of the centrifuge cascade is well known to the US and many other countries due to a series of technology sales of Pakistani nuclear technology.

Suposadly something like 24 countries purchased the designs of AQ Khan on of which was Libya. Unfortunatly for them the shipment of technology got empounded and thus became very known further since Pakistan cosied up to the US further information became available.

Now the US has an issue with North Korea and it's nuclear weapons and rocket technology. North Korea is a very very closed country and has little or no contact with the outside world so attacking it with malware is going to be difficult.

That is unlees you can get at it through a trusted third party.

It has been said that there is a two way flow of info between Iran and North Korea over nuclear technology for some time.

What better way for the US to spike North Korea's guns than through Iran or one of the other countries that it does technology swaps with?

But who ever get's hit by stuxnet is actually not important, as it sends out a loud and clear message to anybody using AQ Khan technology that they are vulnerable and there is little they can do about it.

Which raises the question of if stuxnet is a political or business weapon. That is it may have been designed and built not by a state player like the US but by an organisation with replacment technology for sale.

Now there are a lot of ex-Russian nuclear physics bods down on their uppers as it where with skills that although highly valuable are not very marketable.

So don't automaticaly assume it is a state actor such as the US hand behind stuxnet (although the probability is high).

bobNovember 22, 2010 10:30 AM

@Winter: Do you write for Fox? You seem to have completely missed the words "Vacon of Finland" and "they can be used for" from your 'analysis'.

RHNovember 22, 2010 10:36 AM

@Clive: while I agree, we aren't 100% sure of the location, I believe I remember reading that Bahshir was the likely target due to the geographic proximity of most of the Stuxnet breakouts, almost isolated in that region.

Shame we won't get to hear the real story. I have a feeling a FOIA application won't be accepted... even after 25 years.

I'm enjoying seeing just how exacting this virus actually was.

LouisNovember 22, 2010 10:36 AM

Thanks Clive for the reality check.

My own risk analysis points to Stuxnet being a Mossad creation, as the threat of [new] nuclear capabilities strikes the most in Middle East, not Asia.

For all interested in building high frequency motors, if it was that simple, it would not be so hard. (Pun intended)

anonymouserNovember 22, 2010 11:32 AM

I wonder how good the QC on the finished product is/was at Natanz. Would the Iranians have ended up with uranium that's poisoned just enough to make a specific type of bomb fizzle?

Them maybe it was just about making the centrifuges break - slowly. I understand they're awfully expensive bits of machinery and hard to set up, too.

The parameters chosen seem awfully specific.

billyjoejimbobNovember 22, 2010 1:32 PM

I must say that I sincerely enjoy the degree of hyperbole in reporting on Stuxnet. The articles on CSM, in particular, are quite nice. The "Son of Stuxnet?" and related article are full of juicy fear-mongering statements without any qualification. Lovely.

JoeNovember 22, 2010 2:05 PM

For all you registered voters in the US. It would make sense to send a message to your representatives to let them know that the "experts" quoted in the CSM article are no more experts on cyberwarfare than Chicken Little was an expert on meterology. And that you disapprove of any funding increases going to deal with such bogus threats.

And yes sending a message has a near zero influence, but not zero.

Rob RosenbergerNovember 22, 2010 5:16 PM

How & when did Symantec obtain a TS/SCI fully operational nuclear SCADA vulnerability testing laboratory where they could dissect Stuxnet?

When did Symantec virus expert Eric Chien become a renowned "Q-cleared" nuclear SCADA authority?

Tony H.November 22, 2010 5:26 PM

So Fararo Paya of Iran makes these frequency converter drives, presumably because the US-made ones are embargoed. There's a nice side effect - maybe the Iranian ones are quite good - who knows? Necessity is the mother... So can I buy one of them for my garage enrichment lab or does the US claim to control technology exports *from* Iran as well as US exports to it?

Doug CoulterNovember 22, 2010 6:31 PM

One of my turbo-molecular pump vacuum systems runs at 933 hz, one in the 600's and those have very nice motor driver/controllers and are available without limit worldwide as far as I know.
And/or, trivial to buy in one country and ship to another if required. When I bought mine I didn't have to sign any papers or agree not to export them.

A turbo is the same thing as a centrifuge as far as the drive goes, it's just that a centrifuge is spinning a different type of rotor with the motor.
In either case, it's not the drive speed that is the limit, it's balance and the fly-apart speed of the driven member with available materials -- both devices push that hard. And bearings...not trivial at all at those speeds.

I think that law is just another case of the govt trying to shut the door way too late, as they did with crypto. For a long time it was illegal to ship overseas things they had already.

As many have pointed out, it's not that hard to make a drive like this, there are two designs on my physics website, because a lot of times surplus (ebay) pumps are bought that have no controller, and the companies want silly amounts of money for them. The only reason there are two designs is that some turbos have a hall sensor and two phase motors, some don't and have 3 phase. It's way not rocket science.

Low harmonics aren't a problem, and they are not needed in general for a motor drive, and were they, a simple passive filter gets the job done off a square wave anyway. The motor is an inductive load as is.

The trick with something that spins real fast is how to ramp up and down in speed and not lose lock with the rotor, but even that's not especially tough as design problems go. Fooling with the rotor speed a little bit wouldn't affect a centrifuge or a turbo all that much, but making one fly apart -- now that's serious. The manuals warn that some of these store about as much energy as a car crash in the rotor, and if you have a "crash" in there, the thing can wring all the bolts off, fly around the room and kill people. My 6" turbo uses 24 3/8" SS fine thread bolts -- we are talking that kind of force for the lightest rotor they can make.

Could be centrifuges used in this work aren't as sophisticated/developed as turbo pumps (they sure aren't made in as large numbers), but turbo pumps do have some speeds they'd rather not run at -- where the speed matches some acoustic resonance and they are noisy. In normal use, you stay off that speed, but it doesn't cause any real harm other than being annoying and noisy.

Actually, anything like that can do "some" separation, as even a turbo pump has an effective "compression ratio" that changes with the weight of the gas involved, and you can see this in a mass spectrum of the residual gas in a vacuum system here easily. Light gases don't pump as well as the heavier ones do...
A common trick to flush things like hydrogen out of a system is to let in argon, then pump the mix out at fairly high pressure (at first) and letting the argon help carry the lighter stuff out.
Works a charm.

JayNovember 22, 2010 7:54 PM

@Doug

Ah, but argon is about 1.8g/L and hydrogen is less than 0.1g/L (according to Wikipedia, anyway.) I think you'd see a leeetle bit less difference between U(235)F6 and U(238)F6...

RobertTNovember 22, 2010 10:49 PM

It looks like Clive was right about this all along, including the PLC for frequency control.

I have to admit I still don't see the technical difficulty because any PWM controller with a frame period of above 20 times the desired frequency (about 20khz) can easily synthesis the required drive frequency. Adding a frequency counter for hall effect feedback would be something even a second year EE could do.

Now doing this at 2KW involves switching some higher voltages , but it's still just a H-bridge rating problem. Most motor drive H-bridges have predrivers available that do all the difficult bits associated with generating High Side voltages and non overlap/ shoot through control. A simple open loop design will easily result in harmonics at -40dB and a first order integrator in the loop will improve that to between -60 to -70 db, so "low harmonic" is no problem at all. This is achievable with out any DSP , PLC's or other processor in the feedback loop.

But apparently the AQ design does it using a PLC controller and that's what they attacked.


gregNovember 23, 2010 12:53 AM

There is one very important difference with isotope separation centrifuges. They run *above* their critical frequency. That is the rotor must spin through several speed ranges where, if they stay there too long, they fly apart.

It was a big deal working how to do this at the time (the next being the thermal gradient). The trick wasn't to make the centrifuge stronger, but more flexible, they have "bellows" that allow the long narrow tube (works better for separation than wide) to flex in its various critical modes. Thus a PLC feedback controller is used.

Also note that each centrifuge is quite small, I would be surprised if each needs a drive motor much bigger than 1kW. Each also has its own controller.

However they still suck at separation. So you need thousands of them, and spin up and spin down needs to be match to gas flow from subsequent stages, otherwise you can get your very small amount of enriched material mix with only a little of the huge amount of depleted material.

However a PLC type thing to do this is still not hard. In a standard industry problem these days.

@Doug Coulter
"I think that law is just another case of the govt trying to shut the door way too late, as they did with crypto. "

Its not just that we "already have it". Its that something that was really hard to do in the 1950's to 1970's is now standard industry practice.

Nukes were hard to do once upon a time. We are no longer in that time. Its twenty ten, and classifying 14MeV neutron data and controlling "high frequency controllers" (anything below 10MHz is not High frequency in my book anymore) is as effective at preventing proliferation as the TSA is at stopping Terrorists.

Nick PNovember 23, 2010 1:25 AM

@ RobertT

Hey, don't give Clive all the credit. I was one of the earliest major promoters of enrichment plants being the target, with Knowledge Brings Fear blog beating me to it as the first by gathering the evidence. Nobody in the comments has mentioned that Natanz was originally thought to be the target. When discussing that in the past, Clive mentioned many sites share similar hardware. This claim's potential hasn't been explored by researchers yet. If he's right, then the Stuxnet authors might be killing half a dozen birds with one stone.

RobertTNovember 23, 2010 5:04 AM

@Greg,
So if I understand you the PLC function is more about matching the spin speed to the flow rate of the previous stage. So messing with the desired spin for a given stage will mess-up the whole centrifuge cascade extraction efficiency. Smart thing to do, especially with a virus.

I wonder if it generated reports showing that everything was running exactly as desired. That would make debug difficult.

BTW: Varying the motor frequency for a PWM H-Bridge is very simple, you would just use an Audio Dac to create the reference frequency for the H-Bridge motor control. No problem.

@Nick P
sorry didn't mean to offend, but I can remember Clive mentioning exactly the targeted PLC as being for centrifuge motor control. I agree if "Evil" people use this PLC arrangement then it must have made this a very attractive target. So much for the theories that this was written by a smart script kiddies...


HarryNovember 23, 2010 8:41 AM

Washington Post
November 23, 2010
Pg. 8

Nuclear Program Said To Be Troubled

Iran's nuclear program has experienced serious problems, including unexplained fluctuations in the performance of the thousands of centrifuges enriching uranium, leading to a rare but temporary shutdown, international inspectors are expected to reveal Tuesday.

The International Atomic Energy Agency, the U.N. unit that monitors nuclear programs, will provide no explanation of the problems. But speculation immediately centered on the Stuxnet worm, a computer virus that some researchers say appears to have been designed specifically to target Iran's centrifuge machines so that they spin out of control.

Iran denies the worm caused any problems.

No country has claimed responsibility for developing the virus, although suspicion has focused primarily on Israel and the United States. James L. Jones, who until recently was President Obama's national security adviser, declined to comment on the worm when asked about it Monday at the Aspen Institute.

The Associated Press first reported on the centrifuge shutdown, which was confirmed by a person familiar with the report. The official said the shutdown is mentioned in a much-anticipated IAEA report expected to be released Tuesday.

U.S. officials did not respond to a request for comment.

-- Glenn Kessler

gregNovember 23, 2010 9:01 AM

@RobertT
Yes that's right. And to move through the critical speeds quickly, usually via a feedback loop (ie a PID controller more or less). If for example there was a critical speed within the variable range of the worm... well it would be very interesting. You would probably trip shutdowns in local controllers all the time or something.

It should also be noted that enriching U235 to 3-5% need for reactors is many many times easier than enriching to the +75% (IIRC) needed for a bomb. But perhaps boosting reduces this requirement to well below 50% U235, but i doubt it.

Carl BussjaegerNovember 23, 2010 9:25 AM

The technical discussions and speculations about the worm's origin is interesting, but I can't help thinking about something more basic:

Do you really want your mission-critical PLCs in a nuclear processing plant run by a Windows machine connected to the Internet in the first place?

Doug CoulterNovember 23, 2010 11:35 AM

@pat,
Yes, absolutely you are correct. I was just pointing out that there's more than one way to skin a cat (in this case, an accidental one), and the difference is one of scale only.

Even a centrifuge doesn't have much "gain" per stage, remember, which is why most setups contain so doggone many of them, and do so many passes to get decent enrichment levels.

Easier by far to separate D from H, or 6Li from 7Li, or to chemically separate Pu. If you have D, you don't need enriched U to make a reactor that will make Pu for you (I don't think I'm giving away any big secret here -- see some Canadian designs).

I work with the light elements here all the time.
http://www.coultersmithing.com/forums
We do it for hobby nuclear fusion, so I'm at least a little up on the topics involved.

Turbos go through a "whirling speed" (or several) as the standard term is used, as well. They're designed to live through that -- but they (and the user) don't like it -- vibration is a bad thing in many systems (think electron microscope etc). Not to mention hard on some expensive, exotic bearings. I agree, centrifuges probably aren't as well developed, and might have more troubles if you sat on a "bad speed" for long, I think I kind of said that above. My systems go through two or three such speeds on the way up (or down), making a lot of noise and vibration doing it. It's normal and occurs in all the types and brands of turbo pumps (which are essentially just driven jet turbines, perhaps with a "drag" stage on the back end).

It's one reason the controllers let you set the run and standby speeds very precisely. You look for a "sweet spot" that's a little different unit to unit due to inevitable tiny variations in the manufacture of the rotors, and try and run on top of that. Mainly because they live longer that way, and run smoother.

The control loop on many makes a PID loop look pretty simple, as things other than just rotor speed are taken into account, such as power drain limiting, and rate of change of speed limiting. But yes, in general it's a closed loop with various logic on top. In the case of my systems (all Pfeiffer) in the event of a power failure, they even use energy stored in the rotor to create a generator that keeps the control electronics alive so nothing bad happens if the power comes back while it is still spinning -- fairly sophisticated stuff compared to a variable speed 3 phase lathe drive. In general, it's more like a controllable phase locked loop than a PID loop, though, with all sorts of provisions for controlled shutdown if things get too far off. These aren't like a 3 phase 60 hz induction motor where you can just slam on power at full frequency and expect the motor to get into sync on its own, you have to guide them along the path, it's simply too big a jump to make from "go" all at once.

The point is, to an EE, this is all fairly simple. Unless a country doing this lacks qualified EE's (or has scientists that drive them off by looking down their noses at people who can actually do things they can only call "conceptually trivial" but can't do themselves, it's pointless to try to ban the tech.
It's pretty basic stuff for any engineer who has had some practical experience. My forum has several members who have designed these from scratch and gotten them to work better than the manufactures designs, actually.

NASA comes to mind as an outfit that has had "engineer drain" as a result of the management and PhD-worship as well. The good ones tend to get tired of putting up with that; which means we get to hire them instead!

So maybe the best thing to do with Iran or N Korea is simply to help increase the friction between academics and practical engineers!

Clive RobinsonNovember 23, 2010 1:23 PM

@ Robert T, Nick P,

Both Nick P and myself thought the unfinished nuclear reactor was an unlikley target.

My reasoning was that the reactor plant had to many unknowns to be an effective target for PLC malware.

Nick beat me on the draw to the enrichment plant because although I had thought of it I actually knew very little about the specific plant and was doing some research to see what was known.

Meanwhile another person on another blog preceaded us both, however it is unclear if they came up with the idea or as has been suggested it was thought up by somebody else.

For those more famliar wth the UN 5 permanent security council members cascade designs most of them don't use that sort of arangment to control their centrifuges (the issues being historicaly reliability and flexibility whic Greg can probably tell you more than I can).

And if I was responsable for designing a cascade myself I would certainly consider doing it the way you suggested and not the PLC route (it would be more responsive, flexible and reliable that way and above a thousand units certainly cheaper).

So why have the Iranians gone down the PLC route...

Well it is a mater of pragmatism and timing.

When AQ Khan designed his cascade custom electronics usually ment involving a company in a country such as the US, so not a good idea from the security asspect.

Also off the shelf parts are usually better for peripheral activities in prototype designs even though the parts may be ten or twenty times more expensive than they need be. They are readily available debuged and have most of the hard work taken out thus alowing you to concentrate on the actual core design.

So once AQ Khan had his design up and running he had no incentive to refine the electronics it was not his field of endevor and would have been carried out by others if required.

But if you think further about it when "illicitly" selling on the design to third parties custom electronics would effectivly have his name written all over it, which would not be a bright idea. Also he would have to support the product which would further put a noose around his neck.

However since the original design custom parts using field programable arrays are now quite cheap to design and manufacture, and with a little thought in the design can be made considerably more general purpose and thus worthwile developing as a sellable product in their own right.

Which might account for why Iran had it's own outlet for the high frequency drivers.

Unfortunatly it did not occur to them to use different firmware on the motor drivers in their cascade, an oversight that is easy to make and may have proved costly for them.

As for Stuxnet what convinced me it was aimed at the enrichment cascade was that it was the AQ Khan design (obvious from photos on the net).

It is this design that was sold to various "Axis of evil" countries and many others. It is just to big a target not to have a go at the effective ROI imense. Where as the incompleate Nuclear reactor was not.

It was the follow on think with regards to "who else" that made me think "What if the target is actually North Korea" and Iran just a usefull stepping stone in the path or even the first step of the journy...

In the case of Stuxnet I suspect that many of us are a bit right and a bit wrong in our reasoning and thus our tentative conclusions. However that is not to say we are incorect just that those paying for the development might not have thought it up (well I guess they don't have to think for Sux II as we've do it for them ;)

Full Cholesterol JacketNovember 23, 2010 7:38 PM

@Clive Robinson:
Interesting thoughts re. N. Korea as the ultimate target. Makes one wonder about the recent revelation of their centrifuge farm. Maybe bragging that they *weren't* affected by Stuxnet by displaying 1000 happy centrifuges, as an in-your-face?

RobertTNovember 24, 2010 3:06 AM

@Greg, Just out of interest, how do you do real time detection of U235F6 from U238F6? assuming this result is used to drive the centrifuge PLC speed control algorithm.

I guess vibrational spectrosopcy could be used, or maybe just molecular density, but these seem more like batch measurement methods than real time feedback control methods.

gregNovember 24, 2010 4:41 AM

@RobertT

You don't (well not in the unclassified literature that i know of). I talk to a few guys at the atomic energy commission (I live in Vienna). But no one breaks the NDA rules!

Its typically just controlled to be low in the enriched channel till at the proper speeds. The problem is that if there is no flow, the rotor dynamics is different. So you need roughly the right flow when spinning up.

Of course this is all for "high ratios per stage" designs.

We could build these things back in the 40s without PLCs, or even transistors. You just have to live with a lower separation ratio per stage and more maintenance (DC brush motors for example).

However you could measure the ratio with a residue gas analyzer. Adds a bit of cost per stage however, so you probably would only do it every 10th or something.

As Clive hinted, the hardest thing to do is Q&A in fact. You want all 10K of the things to work, or fail in a way that they don't Bork the rest.

kangarooNovember 24, 2010 8:09 AM

So, with the usual genius of the bright but terribly narrow-minded we see at this level, the template for a general self-propagating infrastructure eater has been developed to interrupt the development of a primitive nuke?

I guess the same kind of Brzenski thinkers who saw the Commies as a terrible, terrible threat to all that is holy & good, while missing out on the differences between a different tactical approach to the enlightenment program and those who reject that strategy in toto.

Nice. Like a bad sci-fi novel.

RobertTNovember 25, 2010 9:43 PM

I was curious about the AC low harmonic distortion requirement so I ran a few simulations and included a model of a flexible sided centrifuge, as described by Greg.

The problem with an AC drive is that even order distortion of the AC signal causes asymmetric forces on the centrifuge, which results in huge bearing stresses and the possibility of a positive feedback loop where the flexing of the wall adds positively to the drive signal force asymmetry and the whole system tears itself apart.

It was an interesting simulation exercise, but what was more interesting is that ALL the signal drive distortion problems disappear, if you use a PWM controlled Hbridge with a BTL (Bridge Tied Load) motor. (system has no even order distortion)

Sorry for the completely off topic question, but... In the simulations I had a lot of problems with exciting dynamic sub-harmonics / inharmonics on the centrifuge flexible wall surface. This sub-harmonic / in-harmonic rotated around the surface of the centrifuge. Obviously this sort of behavior would be disastrous for a gas separation centrifuge. To get rid of this I had to add lots of material damping, yet suitable damping materials could not withstand the centrifugal, forces, so in reality they would rip off the surface.

Does the centrifuge wall have some special material damping property that I'm not including?


gregNovember 26, 2010 4:49 AM

@RobertT
You are well into classified details there (I presume). Never mind that you could build one or work it out for yourself. But its still classified.

Also i don't know. Since i don't in fact have clearance. What i have regurgitated here, is all unclassified in the literature.

If anyone else knows here (I can think of one), they are probably not permitted to be to clear on the details.

Clive RoinsonNovember 26, 2010 3:51 PM

@ Robert T,

I'm not sure what physical modeling system you are using but it sounds interesting.

I know from other fields of endevor that often classified information is only classified when pertinent to a very specific problem domain.

Thus if you had a hunt around and loocked at say "hollow shaft drives" and "light tube stack chimneys" I suspect you will get the answers you seak. Another area to look at would be high flexibility narrow blade areofoils as used in vertical high efficiency wind turbines. These can have many critical speeds and modes of resonance and antiresonance giving rise to considerable differences in rotor energy storage at verious points on the blade all within their "normal operating" range.

One thing you should consider is that the "bellows" might be at odd non harmonicaly related points along the centrifuge shaft and also act as part of a "turbo bearing" systedm providing sufficient critical damping without direct mechanical contact.

Which brings me on to pipe drilling. If you consider the issues involved with drill pipes going several kilometres into the ground. The drill mud goes down the pipe in one direction and back up the outside of the pipe in another direction. The risk of critical frequences being induced simply from a change of density in drill mud etc is very high. there have been one or two papers on these subjects and some quite interesting books.

RobertTNovember 26, 2010 9:30 PM

@ Clive,
Motors were modeled in Matlab / Simulink and Spice, the mechanical systems were modeled with spice equivalent electrical elements for the mechanical loads (rotor / centrifuge).

The drive force results of the electrical simulation were used as stimulus for the pseudo 3D Finite Element Analysis (FEA) simulation using Comsol. I included some material non-uniformity and aerodynamic effects of some air outside and UF6 inside the centrifuge. Most of the problems seem to originate from vortex shedding (in the semi-evacuated space outside the centrifuge), rather than directly from the mechanical properties of the centrifuge wall itself. Simulation works great if operated completely within a vacuum (inside and outside) unfortunately a useless optimization corner for a gas centrifuge....

I think I need to include an increased plastic / elastic deformation region, that will result in non-linear material losses and an associated damping. On the weekend I'll try to read some Tasneem Shah papers on rotating fluid dynamics and see if I can't damp the whole thing using the UF6 internal gas flow.

A quick literature search seems to suggest that mechanical damping is also provided by the magnetic bearing, so I'll need to add this to the simulation.

I'm basing my whole simulation on a simple model of a Zippe centrifuge, with 10cm diameter, and assuming it is a AQ Khan P2 type.


Clive RobinsonNovember 27, 2010 5:47 AM

@ Robert T,

Hmm sounds like the model should get you quite close to an engineering prototype...

I just hope you don't get any knocks on the door ;)

@CliveNovember 27, 2010 7:45 PM

"just hope you don't get any knocks on the door"

Yea I never though I'd describe running CAD simulations as a thrill seeking / dangerous activity, but .....

Leave a comment

Allowed HTML: <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre>

Photo of Bruce Schneier by Per Ervland.

Schneier on Security is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc..