Pokhran II not fully successful: Scientist - Part-2

[enqyoob]

This is very interesting. So now we see in bold face letters, no less, so it must be true:

Yes so without data there is no basis to claim that higher yield of S1 would definitely mean more damage.

All real world systems are drastically non linear. Even in the trivial model we did we already got a power (exponential) function. And we don't know the correct coefficient either.


No those are not the options that we have to be restricted by -- my option is as follows

S2 worked as expected and caused minor damage to village -- as expected

S1 if it was to work would also not do any further damage -- may be 1" crack instead of 1/2"

S1 may or may not have worked but we can absolutely not use Khetolai to make that call.

The thing that is being ignored here is that unlike all the stuff about sensor raw data, calibrations, locations, radiological data, the precise weapon design features, etc. etc., the Khetolai argument is the only one that does NOT rely on techno-babble about which none of us can have any good infor, and if we have it we cannot say it in public. At the same time, the Khetolai argument is completely sufficient to prove my point.

Why?

Because BARC, DRDO and the Armed Forces had at least 24 years to thoroughly, completely measure, model and understand the nature of shock propagation through the medium between S1-S2 and Khetolai (and their Logistics Base, slightly further away). This region is mostly inside the guarded perimeter, inside which if you are caught, you get at minimum a good thrashing (as the Khetolai brats certified) or far worse if you are found to be a technical person.

They had data from the nuclear blast of 1974 (which actually was located further away, but at a shallow depth), and from thousands of weapon tests because POK is also an IAF test range, and from specific tests conducted to get seismic data.

There is NO reason for uncertainty about the response at Khetolai. IOW, the seismic data at Khetolai vs. "yield" would be based on MEASURED transfer functions for the medium in between, IN ADDITION TO modeling and simulation. Probably BARC/DRDO also have seismic data going far beyond the range, but again, they are the only ones who know what the input was, to measure those transfer functions, so those must be Classified.

But.. given those transfer functions, NO responsible engineer would have designed S1 and S2 to cause "only minor damage" unless the yield actually exceeded the mean expectation by a substantial amount.

As for "minor damage", I did point out that brick-separation cracks as shown in the pictures from Khetolai are very serious, and VERY expensive to fix. Again, this is being repeated for those who are willing to consider facts with an open mind, I know they will be ignored by those who desperately need to "prove" that S1 and S2 were duds. I know about such cracks from personal experience, and I am not exactly without resources as I try to find engineering solutions when it's MY money at stake. Check on the internet for "brick separation cracks" and see if those are described as "minor" by anyone who had to pay to get them fixed.

Again, the Khetolai experience confirms this - the residents say that the Govt came by and promised to fix the damage, but then the engg. estimators came by and since they presented the repair estimate the Govt. has been in hiding. This figures exactly. Those poor people will not be able to sell those houses except at the land value and the price of some re-usable bricks, because the foundation repair is very expensive.

The real clincher is that the kids were asked only to be outside the school building (a sane precaution based on the Italy record, and our knowledge that Indian village schools are not exactly built to the latest earthquake standards (BTW, read the reports and you will find that SINCE 1998, Khetolai has indeed got a NEW school building). So no venting was expected either.

IOW, there is no technical uncertainty here, although the argument uses no Classified technical data, and does no faux-sophisticated WAG techniques.

This is why I am so certain that this is an excellent proof. And because I know that the techno-babble is just people showing off their extreme knawlidj of nyookulear pissicks.

The classic "proof" was from the "Arms Control Donk" who claimed with no sarcasm about himself:

When I published the yield estimates in 1998, the BARC sent me a stack of documents. I gave them to my beer buddies to read those and explain them to me, and they said it was "sh1tty science".

IOW, the donk PUBLISHED HIS ESTIMATES, but he had to get his foul-mouthed beer BUDDIES to EXPLAIN TO HIM real slowly, using all ten fingers and toes to count, and simple matchstick pictures and crayons, what the data sent by BARC meant.

And BARC or whatever desis were stupid enough to imagine that the Arms Control Donk, being a gora in the Bhesht who PUBLISHES stuff, must be authoritative. Like Witzel of Harvard!

Of course, the experts here jumped on this fine demonstration of competence and honesty by the Arms Control Donk as more PROOF.

[shiv]

And one is supposed to believe that your diatribe is just a "honest presentation of debate"? You have a view that India doesn't need to test not because it has TN but because you don't need one. And you have tried every trick in the book and done everything you are pretending you will not do. You were even attacking Arun for some weird name this and that. Come on, get real, who are you kidding?.

ho hum

flamebait

Reach your own conclusions birader. If you must have a rant be my guest. I love your style of "debate" Still - facts are facts and by this rant nothing changes. Not even my views which you have not understood - but then again your vitriol is probably all that I am going to see in terms useful information.

[Bade]

Saar, Please tell us what is the minimum bum tests population sample size required before the above statistics becomes valid?

Well, the numbers are all for 'one' test only. The individual errors used come from a sampling (multiple) of the after effects of the bum going boom. Or else how can you estimate the errors ?

Now if you did multiple bum tests, then each time the yield could be slightly different and then you can give an overall uncertainty of the yield of a given device from the data from multiple tests. This is all high school physics onlee.

But you do not need multiple tests to estimate the error estimate and give confidence level limits of the expected value. This is routinely done in nuclear and particle physics experiments with low samples of gold plated events.

[Bade]

Few observations:

• 1.) Who cares about what NPA think or do.
  2.) Successful bomb of 150kT and above will speak for itself. In fact it will be useful to do many 150kt+ test and not disclose which one of those were fission or TN. That obfuscation is a better for deterrence.
  3) For Indian users I.e. Military, SFC & strategic policy makers, India just needs to break the singular {designer, maker, validator} nexus into two; that will be enough. The Validation function should be ideally done by:
  • A) IISc, or
    B.) TIFR. For that it must be split from DAE and report directly to PMO.[/list]
    
    4) As I mentioned before in earlier avatar of the thread, there should two parallel and autonomous teams of designer (including simulation/modeling physicist teams) to keep the system honest & true.

There is no such split in the US. DOE sponsors all high energy physics and nuclear physics programs, both national labs as well as University funded efforts. Even LLNL or LANL is on DOE money and administered by the universities.

If you split TIFR out of DAE, where will the funding source be appropriated from for TIFR ? We do not have the equivalent of the NSF in India.

[Bade]

Our H bum is like my buying Mega Million lottery ticket, statistically the odds of winning is one in {substitute the grand prize winning amount}, but still I hear that some bum in the great Nation of Texas or Georgia wins the prize. So, If we drop the H bum on somebody and it does not explode then its their paper weight for free on the other hand it just leaks then its their bad karma. Just make sure to mention on the bum "Some components of this device might have origins in China but most parts are made in Bark"

The bums did explode right, unlike you buy a lottery tickets with the odds you mention, but some bloke in Texas wins instead of you. So the analogy is completely off the mark.

The only question that remains is the accuracy of the estimated yields. A different question altogether, no ?

[shiv]

Actually - the pro-testers are now beginning to score self goals. I foresee the Indian government beginning to do a massive stonewalling of the issue and pointedly ignoring all calls for review of the results of the tests. Sikka's cancellation may be one sign that the Goi is tellings its functionaries to stop talking.

The GoI holds all the cards and will not be seen bending to pressure from such interest groups. In fact - beyond a particular stage GoI can start claiming brownie points for "resisting pressure" from "warlike vested interests" while GoI represents peace and disarmament.

In fact the one group that could help people in such an issue is an opposition party. But the main opposition - the BJP is not only in disarray it is also responsible for saying "No more tests". In fact it is the Congress who should say "BJP was wrong". But that is not happening.On this issue Congress and BJP are as thick as thieves. The Congress of course found "No more testing" convenient for the nuke deal.

For the GoI to test now it would be seen as

1) giving in to sundry pressure groups
2) not being confident of a working arsenal
3) going back on its words

It becomes a matter of izzat and so they will stonewall. I suspect India is not going to see any overt nuclear testing for a long long time.

There is the other question of what sort of signals are being sent out by testing part from the 3 I have listed above. It could be seen as Korea style muscle flexing - a show put on by 60 pound weakling unsure of himself.

I personally find India testing a nuke to "send tough signals" laughable. India behaves like a docile nation that bears no ill will towards anyone - and testing is so out of character I would be surprised out of my pants if we ever tested again. Lets see..

Anything more and my views will have to go in the psyche thread.

[Arun_S]

Few observations:

• 1.) Who cares about what NPA think or do.
  2.) Successful bomb of 150kT and above will speak for itself. In fact it will be useful to do many 150kt+ test and not disclose which one of those were fission or TN. That obfuscation is a better for deterrence.
  3) For Indian users I.e. Military, SFC & strategic policy makers, India just needs to break the singular {designer, maker, validator} nexus into two; that will be enough. The Validation function should be ideally done by:
  • A) IISc, or
    B.) TIFR. For that it must be split from DAE and report directly to PMO.[/list]
    
    4) As I mentioned before in earlier avatar of the thread, there should two parallel and autonomous teams of designer (including simulation/modeling physicist teams) to keep the system honest & true.

There is no such split in the US. DOE sponsors all high energy physics and nuclear physics programs, both national labs as well as University funded efforts. Even LLNL or LANL is on DOE money and administered by the universities.

If you split TIFR out of DAE, where will the funding source be appropriated from for TIFR ? We do not have the equivalent of the NSF in India.

SRDE doing on cheap.
You forget that US works the way it does is because it has two separate vertical integrated fully funded (independent) teams in the for of the two labs, that check each other's work.

In terms of funding, I meant separating their reporting structure (to eliminate any perception of incestuous influence under a roof), both IISc and TIFR have people who understand the subject matter well (technical leadership exists) and they can build their department by attracting right skilled people.

[NRao]

Shiv,

ALL that is true IF the GOI is going to test what was tested in S1.

The GoI can always claim to be testing something new/advanced and test?

I am not so sure that the GoI is concerned about internal pressure. It is more than likely that the scare is from other governments and associated threats. The need to be in the good books of others is a lot more important.

[Arun_S]

In fact the one group that could help people in such an issue is an opposition party. But the main opposition - the BJP is not only in disarray it is also responsible for saying "No more tests". In fact it is the Congress who should say "BJP was wrong". But that is not happening.On this issue Congress and BJP are as thick as thieves. The Congress of course found "No more testing" convenient for the nuke deal.

Off topic, but the POK-II saga has shown the compulsions and true colours of BJP (No credit to PVNR/INC for POK-II, and ABV deciding to back off from brink of POK-III test ~2003 and lick the P6/N6 lollipop named NSSP). BMishra as NSA should have seen the missing political need that the test team overlooked, in that PoK-II has to have backup bum that unambiguously demonstrate high yield weapon; I.e. a shaft with high yield FBF as a insurance backup.) That would have given India upper hand negotiating w/Unkill, and not have the Foreign Minister of India negotiate with Talbott two ranks below diplomatic protocol, and plead Indian capability with not so impressive Pok-II field (sic. failed) trial. If BM did his job right all the yield cover-up would have been un-necessary.

[Arun_S]

The classic "proof" was from the "Arms Control Donk" who claimed with no sarcasm about himself:

When I published the yield estimates in 1998, the BARC sent me a stack of documents. I gave them to my beer buddies to read those and explain them to me, and they said it was "sh1tty science".

IOW, the donk PUBLISHED HIS ESTIMATES, but he had to get his foul-mouthed beer BUDDIES to EXPLAIN TO HIM real slowly, using all ten fingers and toes to count, and simple matchstick pictures and crayons, what the data sent by BARC meant.

And BARC or whatever desis were stupid enough to imagine that the Arms Control Donk, being a gora in the Bhesht who PUBLISHES stuff, must be authoritative. Like Witzel of Harvard!

[enqyoob]

I sense that you see the hilarious irony in that demonstration of NPA intellectual standards, Arunji. Isn't that what the donk said? That HE published his estimates, that BARC sent HIM their data - but they he had to show it someone else, to have them explain to him what it meant.

The facts are 400% jay-new-whine. Is it just me that sees the proper interpretation of that? Usually someone who publishes some estimate, does not need any help interpreting countering data. Raises the issue of what his basis for publication was, ne c'est pas?

[geeth]

>>>>The facts are 400% jay-new-whine. Is it just me that sees the proper interpretation of that? Usually someone who publishes some estimate, does not need any help interpreting countering data. Raises the issue of what his basis for publication was, ne c'est pas?

Yeah! counting the holes would be a good start

[enqyoob]

shiv, if I were to do what I do best ("think" like a Paki), then, 2 weeks after the SDRE banias shoot themselves in the mijjile by conducting more "tests" of 150KT, 200KT, 1MT etc.,

I would just send another bunch of soosai BakPakis in to say, Mettupalayam, Bihar, and set off a few bums in the biggest 3.5-star hotel there, housing a dozen western tourists.

A hostage situation, with say 35 dead when it's all over.

And the new 1MT-armed, maccho India will cower again like India did after Kaluchak Attack. Srinagar Legislature Attack. Parliament Attack. Akshardham Attack. Red Fort Attack. Varanasi Attack. Ahmedabad Attack. Bangalore Attack. And Mumbai Attack.

And my Tarrel than Oceans, Deepel than Mountains Fliends will intrude another 5 km and air-dump more expired ratsnake-meat cans inside Indian territory.

There goes the New Improved Deterrent. What stays is the unraveling of the IAEA pact, and the destruction of all hopes of improved electric power in India.

So much for this "Test NOW!" cacophony.

[ramana]

This already happened. It was Kargil. Only it was not just the TSP.

[samuel]

There is NO reason for uncertainty about the response at Khetolai. IOW, the seismic data at Khetolai vs. "yield" would be based on MEASURED transfer functions for the medium in between, IN ADDITION TO modeling and simulation. Probably BARC/DRDO also have seismic data going far beyond the range, but again, they are the only ones who know what the input was, to measure those transfer functions, so those must be Classified.

How is this done?

[NRao]

There seems to be a problem with the link to the latest DTI, so cannot provide a link right now. (We have our 1040, they have "No backend server available for connection: timed out after 10 seconds or idempotent set to OFF.")

However, this from the Sept 2009 issue:

Titled "Instant Sunshine" by Bill Sweetman:

The world's first known operational nuclear weapon fielded without a full-scale test will be the warhead for the ASMP-A missile from MBDA, shown here on a Mirage 2000N

The Obama administration's Quadrennial Defense Review and a parallel review of U.S. nuclear posture could give the go-ahead to two long-debated programs: a next generation missile-launching submarine (SSBN) and a new nuclear warhead. If so, it will be a relief to nuclear insiders who worry that the topic of deterrence has been ignored for too long in the U.S, while nations like France, the U.K., Russia and China outpace U.S. modernization plans.

Does not seem to me that CTBT has ANY value. Why are we concerned about wasting time about such a worthless treaty?

Build the economy and test when needed.

They do it, we talk, waste time and energy and may do it. Then if we do, we test one single sample.

[ramana]

There is NO reason for uncertainty about the response at Khetolai. IOW, the seismic data at Khetolai vs. "yield" would be based on MEASURED transfer functions for the medium in between, IN ADDITION TO modeling and simulation. Probably BARC/DRDO also have seismic data going far beyond the range, but again, they are the only ones who know what the input was, to measure those transfer functions, so those must be Classified.

How is this done?

Even that must be ossified.

[ramana]

To do that NRao, one must not sign the treaty. Right?

[NRao]

Does not signing make a difference?

I doubt it.

What will make a difference IF AT ALL is India committing to "modernize". JUST like them. Nothing different.

During the "modernization" phase - test the 43Kt. Assuming dialing-a-nuke now is in place do the needful to put 15 on an Agni (or tend towards that).

All India needs is a deterrence towards China.

Pakistan modernization will have to impact the rest of the world, let the world take care of that. Why is India bothered about it?

JMTs.

[SaiK]

only hillary made some big radio mouth off about making India commit to CTBT and other strategic treaties.

the worst case, we can make pakis (I mean chinese) explode one at chagai again.. ..and we can measure our yields to a thermo-200kt, on the same event timestamp ., depending on our containment technology.

OR/ all we need is an emphatic statement to this effect from all respected heads, and our mil chiefs says we have got it!.. just like ABV said .. "hamara ...hai"

[NRao]

They ALL are openly modernizing.

At some time the wake up call has to wake you up?

Hit that snooze button.

Do we really need crutches to walk?

[Sanku]

I was under the impression that the Nukes were not deterrence against terrorist attacks in any case, I thought building up military muscle and will all across the spectrum was needed.

All, not either or but and.

The draft doctrine also probably talked only of returning nuclear fire with nuclear fire and not against a terrorist attack.

I am probably an idiot.

-----------

Oh but what if the terrorists explode a dirty bomb? Ah then but I forget Shiv already posted the story of Jingopura.

I am an idiot.

----------

I think my future posts are going to be like John Snow's recent posts and now Raja Ramji's -- its all maya anyway -- we dont need any of this we are never going to use it anyway

Om shanti shanti shanti

[ramana]

Does not signing make a difference?

I doubt it.

What will make a difference IF AT ALL is India committing to "modernize". JUST like them. Nothing different.

During the "modernization" phase - test the 43Kt. Assuming dialing-a-nuke now is in place do the needful to put 15 on an Agni (or tend towards that).

All India needs is a deterrence towards China.

Pakistan modernization will have to impact the rest of the world, let the world take care of that. Why is India bothered about it?

JMTs.

Unfortunately it does. India is not like the others. India follows what it signs. Thats the way it has been. You may not like that but dont say sign and violate that. There are no benefits of signing unless one wants to self constrain.

The modernization drive also will negate the 1998 vintage statements. So what to do?

Caught in a web of difficult choices.

TSP modernization is being done in NOKO. We see how the world is responding to it.

[enqyoob]

How is this done?

Even that must be ossified.

Actually, no. You set off blasts of very well-characterized bums - like 1 stick dynamite, 2 sticks, dynamite, 1000lb bum, 2000lb bum ... and you measure the signals at a number of sensors. For the analysis, pls check "System Identification"

At the very basic level, the technique is the same as how they see if your knee is OK: they hit it with a hammer on which there is an accelerometer and watch the reaction of your knee. The accelerometer tracks the impact level, the reaction of your knee is the "time-domain response" and your is the "acoustic response".

Going beyond, the Hewlett-Packard Fourier Analyzer, and impact System Identification minicomputers have been around since the 1960s. The theory is that an impact can be decomposed into an infinite Fourier series of sine and cosine functions of multiples of a (low) fundamental frequency. The response signal can also be similarly decomposed. Then you essentially find the ratio of the response signal to the input signal at each increment in frequency, and you get your "transfer function", a function of frequency. This is like a signature of the medium.

In the linear regime (until amplitude gets pretty large) this signature is constant. You store it away, "ossified" as ramanaji says. Next time, you take the response, divide by the transfer function, and you should get the "yield" of the input.

That part is very saada level. If the behavior of the medium is LINEAR, things are very simple. Beyond a certain level, response will not be linear, so you have to go way beyond my kindergarten level. Basically, the coefficients of the different terms in the Fourier Series will no longer be constant, but will depend on the amplitude and frequency content of the input.

See Advanced Kindergarten

Here is something on using neural networks to learn the response characteristics of nonlinear vibration systems. This is because nonlinear system behavior has to be modeled.

Here's a book chapter on "Frequency Response Function Measurements in the Presence of Nonlinear Distortions"

Now that we see that there are many peechidees interested in this fine field, let us see about Seismic System Identification. Here is a paper on "SEISMIC SYSTEM IDENTIFICATION OF THE VIENNA BASIN USING ARTIFICIAL MICRO EARTHQUAKES"

Shows pictures of the concept as well.

The excitation shown is a big weight that is thumped down on the ground from a crane. In fact there are books on how someone was causing monster quakes by hitting a fault region with some massive hydraulic hammers.

Maybe in the Pokhran test range they use Pakis like the one that occupied the seat next to me at the Olympics Hockey Match when Supreme HQ went off to the restroom. About 75kT, I am sure.

I would prefer to use well-characterized munitions like aircraft bombs or heavy conventional Prithvi payloads that could be tested in deep shafts. The sensors must be accelerometers, maybe multidimensional etc.

So the point is that very clear "maps" of the response characteristics will be available, AT LEAST between sensor locations and the source locations. Unless Pakis drill many new tunnels under the ground, these should remain pretty constant.

Maybe this explains why there wasn't too much leeway in picking sites for explosions - there are only a finite number of locations for which they have transfer functions.

Whether a given building will collapse, is a much tougher question. It depends on the "system ID" of the building, and what are its resonant frequencies, and what is the actual damping at these frequencies. If the acceleration level in the ground response at those frequencies exceeds some threshold, the building may resonate (large-amplitude response) and collapse. So there are clear threshold levels to avoid.

Brick separation cracks on the other hand are associated with foundation settling, which means that the ground suffers some permanent deformations. This definitely comes under "nonlinear response" which means very substantial amplitudes of excitation at least at certain frequencies. So again, there must be thresholds for this, on which any good test designer would have got good info from measuring the properties of the medium, separately.

Given 24 years and the data from one Big One, they surely had the time and the data to build up all this knowledge. The techniques are not new - OK, the "neural network" came out even in MATLAB TOOLBOXes for college students in the 1990s, but the rest of it is really old hat. Hewlett-Packard Corporation (HP) started in the computer business with a device to test the structural modes of buildings using the hammer/accelerometer technique, and the legend is that the first building they tested (which was just nearing completion) showed an "unconstrained" mode near 0 Hz. Of course the machine was immediately dissed, as this meant that the building's foundation was completely loose - they they checked and found that this was indeed the case. Saved hajaar lives and more $$M, and made Believers out of the construction industry.

[SaiK]

the problem is all about testing.. since, even the p5 nations think our power does not qualify for us being the 6th power. we have to either say it we have it or show it.

now we have only once choice.. sign up., but test if you want before that. i am not sure, they would agree to put that in writing, saying allow us to test atleast n number of times, before we get drafted.

that way, the path to 6th element is predrawn, and many would start retesting.. on that premise.


it is better in writing first, then testing rather test, and get banned.

[enqyoob]

The smart way is to get the DDM into a locked room, slap a Security Clearance classification on them, offer them the Inside Scoop on something in exchange for signature on the Offishial Secretz Act.

Tell them to whip up panic about the PRC-NoKo-TSP-Myanmar CTBT violations, and the resulting security threat. Keep this up for a year, and even SeeEnnEnn etc. may pick up the message and the duniya will start accepting it.

THEN declare national security imperative (OK, DECLARE IT in public, don't mumble-mumble that "China-India relashuns are excellent" etc as the mantris are saying even as Indian farmers are being forced off their land by Chinese invaders in Ladakh and Arunachal) and proceed with 6, or 10, or 20 tests, then make a big show of getting the PRC also to the negotiating table and have a MuTT (Mutual Testban Treaty).

[NRao]

since, even the p5 nations think our power does not qualify for us being the 6th power.

Does India WANT to be the 6th "power"?

I have to doubt it. There is more intra-India quibbling than with the P-5!!!

IF that is want the country wants - that is fine. Nothing wrong with that. BUT, what is wrong is to state one thing and do something else.

Unfortunately it does. India is not like the others. India follows what it signs.

I am not suggesting that India sign and then break out. But, after signing, if others break the agreement where is an agreement for India to follow what it signed?

The DTI article by Sweetman (none the less) clearly states that ALL are upgrading. So, they have to be breaking some aspects of the treaty. So, where is a treaty there to sign (perhaps) or if signed to follow?

We are very good at statistics, proofs, etc - and nothing wrong with that. But, there is a very dire NEED to be very good at strategy and policy that addresses current and future environments.

Yes, it is not about my liking or not. But, how about if it matters to the country?

[ramana]

My take is dont sign the treaty. Others are all NPT powers and so them upgrading all is legitimate per their treaties. India is not an NPT power so such acts are not allowed.


By not signing the option to proof is retained. Nor does India violate any treaty that it hasnt signed if it exercises the option. This is per international law- Vienna Convention of Treaties that says no one can be forced a treaty that they didnt sign. The NPA came up with a BS thing called 'international norms' which they make up as they go along and want all to toe to them. The US NPAs are hand in glove with their national security apparatus. They flit in and fleet out of govt.

[abhiti]

There goes the New Improved Deterrent. What stays is the unraveling of the IAEA pact, and the destruction of all hopes of improved electric power in India. So much for this "Test NOW!" cacophony.

This is where you are totally wrong. The problem is you are thinking like an engineer not as an economist.

The reason Indian is in power crisis is not due to some nuke technologic denial regime but because state power companies have no money to invest and private sector is unwilling or unable to invest on large scale due to either profitability or govt controls. Your argument may have made some sense had nuclear power been cheaper than coal fired plant or gas fired plant. In reality nuclear power is at least 2 times as expensive as power generated from coal fired plants or gas fired plants. It is so far cheaper than solar power but that may also change in less than a decade.

Therefore what makes sense is expansion of gas fired or coal fired plants, not nuke plants. No point buying a BMW when a Maruti 800 will do just fine which btw is also made in India which means when you buy Maruti it helps grow Indian economy while a BMW just helps German economy. Economically imported nuke power for India is the worst solution.

MMS understands the above much better than I do because he is leading economist, not a 2 cent economist like me. Now that you know that think what was 123 all about?

[ramana]

Note Praful Bidwai writes in Frontline:

Fizzle?

And follow his main argument being made here also.

[enqyoob]

ramana, it is sad when one has to cite Bidwai for support. And yes, his main line of argument reflects that taken by several of the "TEST NOW!" crowd and conveniently ignores the facts staring him in that handsome face in front of that equally brilliant Marxist brain.

Example:

Another important example is the success of the fusion (hydrogen or thermonuclear) bomb, codenamed Shakti-I, that India claimed to have tested on May 11, 1998, whose yield was reported to be 43 kt. This was one of three devices exploded that day, with a combined yield of 50 kt (revised to 60 kt in February 1999).

Thermonuclear or fusion weapons usually have yields in the megaton (1,000 kt) range. {there goes any need for such nuances as "boosted fission", "dial-a-yield" etc. Bidwai is obviously the sort of expert we need to quote here }

But DAE Secretary R. Chidambaram said the Shakti-I yield was kept deliberately low to avert seismic damage to villages in the vicinity of the test site. He claimed the tests were “perfect” and “weaponisation is now complete”.

So his claim is that S-1, if it were a REAL thermonuke, should have produced 1MT if the "heat produced by the primary" had indeed been "enough to ignite the secondary". And I come back to the same question: What would have happened to Khetolai in that case - i.e., if the secondary had indeed "ignited" per Bidwai and his "Classified" western sources?


Of course, this would not bother Praful Bidwai any more than it bothers several of the experts here, because the murder of dozens of villagers at Nandigram by Bidwai's Commie masters didn't either.

But ramanaji, this is the problem: The Khetolai argument, which was made by Santanam and all other GOI soures right then, still rings absolutely true. I see no way around it.

[samuel]

OK, but how do you do system identification for a volume -- i.e. many degrees of freedom -- with few measurements?

These are examples of inverse problems. To give you a simple example, suppose you have a problem Y = A X, where Y is the measurement vector A is a mixing matrix, the medium, the convolution function, the density function, the mass function or whatever, and X is the state or "input" and then you say I don't know A and so if you had sufficient measurements you can invert for A. That is the simplest case, other examples are construction of seismic stacks from shots, tomography and multi-baseline stereo. The problem is
a) When you don't know X (how much was the blast) and you don't know A (what is the SPATIAL distribution of density) then this simple equation becomes at best bilinear at worst non-linear; no need to resort to neural nets, they are not much more powerful than what you will find in bryson and ho in 1962.

On the other hand, if all you were interested in is "transfer functions" (the old fashioned way), a calibration curve if you will, then that makes a certain assumption about the SPATIAL distribution of density. What is it? This is a high-school question.

So what happened to the bricks was a result of spatial variability in whatever mass or density function through which the wave was propagating, or was it X? In my view, meanfield approaches are not good enough to explain such small scale variability of the effect of whatever it is that was propagating.

Let's just start with a simple problem, a simple tomographic problem with one source at the center and K-sensors at the periphery. I will tell you exactly what the source intensity is and i will tell you exactly that the measurements are noise free. If you assume the medium is constant, uniform, duh, don't need too many measurements. Oh but now it has some mud here, salt domes there, rock here, shale over there, oil sands somewhere else and suddenly the problem becomes COMPLEX.

Worse, when you are NOT sure what the source intensity is, and rho is a spatial function then the problem is REALLY complex.

But we can start with some simple examples of sparse reconstruction, a square wave on a grid with few measurements. Come up with the best neural net you can and tell me it reconstructs anything like the square wave. Only very recently have we been able to use random bases and other results in compressed sensing and sparsity constraints under L1 etc. to beat the old (regularization/tikhonov) type reconstructions.
Added: In an Aero Astro context, you may want to look at Charbel Farhat's work using the grasmannian manifold. That is a good example of how to beat sparse measurement problems. But this is all in 2000s!
But the geoscientists are really stuck back in the iron age when it comes to solving these problems.

S

[samuel]

I'll give another example and here it really is a basic idea of something affecting a spatially extended medium. Frequently it is said that the damage due to winds such as in cyclones goes as the cubed of the wind speed. This is an example of a calibration curve. But the next time a cyclone hits over a town, use this to estimate the damage in the city or why the neighboring house's bricks cracked and yours did not. Would you need to know what the spatial distribution of construction is or not? If not what will you assume, an "average" house? If that is what you are going to do then good luck, check your error bars.

I am not an expert at nuclear shock waves etc but I think I understand enough of inverse problems in various contexts to see what the issue may be. At least in simulation it will be possible to show there are alternate explanations for the khetolai effect. And if it really was a "more powerful than advertised" effect, then that is a bias, a DC term that pulls up all measurements. Did this happen?

[enqyoob]

Samuel, the whole point of System ID is that you don't need to know the detailed density distribution etc. etc. to get the relationship by measurement.

The people who do the modeling use all those to build the relationship. If those start matching the test data, the confidence level goes up. The number of validation points is huge, so the uncertainty can be brought way down.

The problem is much simpler than the full 3D tomographic reconstruction of the entire density/ impedance field, because finally, for a margin of safety at the village and the logistics base, you need only response at the surface (several points at and near the surface) at those locations.

This is why the shaft locations had to be where they were. They could not afford much variation in the source location.

Then the transmission paths of the signal to the village and base are pretty much independent of time, over many decades, unless there is some huge earthquake there in the meantime that puts big fractures into the medium. There could be multiple paths, but always the same every time (as they could confirm). So it is actually a simple transfer function - except for the dependence of the coefficients on the magnitude of the excitation.

So here's how I would have done it - use data from a number of tests, with sensors at and around the village, and at a few different depths. Also, sensors at intermediate locations all over the test range.

Think how they validated the CONVENTIONAL payloads of the Prithvi, Agni etc. What better way than to set them off in a well-instrumented test range, where the sensors are built to take far worse blasts? Every such blast improves the transfer functions.

But now that you mention it, a full tomographic reconstruction is not that hard either for this problem. Its a static medium. A few sensors can be moved around and put in wells at various different depths. I imagine that this is all part of standard oil / water prospecting anyway, so no problem with its being detected.

The trouble in matching the modeling to the measurements is, as you say, that the uncertainty in measured data (from one blast, say) does not allow confirmation of a unique solution. But as the number of measured paths and measurements increases, the number of possible solutions that fits all the measurements, comes down drastically, and very soon you focus in on the one distribution that matches all.

It is really that simple - just takes a fair number of people digging wells, locating sensors, and taking and analyzing the data.

[Rahul M]

.......
So his claim is that S-1, if it were a REAL thermonuke, should have produced 1MT if the "heat produced by the primary" had indeed been "enough to ignite the secondary". And I come back to the same question: What would have happened to Khetolai in that case - i.e., if the secondary had indeed "ignited" per Kidwai and his "Classified" western sources?

just a nitpick, who is kidwai in this case ?
(are you insulting that fine gentleman by twisting his name ?)

I might add, in which case I'm fully with you !

[enqyoob]

Just saw your cyclone analogy. There is a huge difference. In the case of the cyclone, there is no fixed spatial relationship between houses, etc. that stays constant over the years. And the cyclone wind pattern is completely different from one cyclone to the next. If you know exactly from which point the wind is going to come next time, sure, you can predict quite accurately what will happen, vs. wind speed.

Also, the forces and stresses are proportional to v-squared in the wind, not v-cubed. And so what you can predict is the threshold wind speed that a given roof can withstand, below which there is no danger of damage from the wind. So roofs in hurricane-prone areas have to be designed to that wind speed. If the roof fails at levels below that, the insurance company can go after the roof-builders (they do). These things are relevant here, to show that the test designers would have been asked to keep the shock at Khetolai below the threshold for damage, but of course they would have also been under pressure from the weapon designer to go for the absolute maximum that they thought they could get away with.
***************************

But from what you wrote, I sense (apologies if this is not the case) that you have not understood how System ID is different from a 1-D calibration. You have to think in the frequency domain, not time domain, to understand what is the signature that remains constant for the particular source - target combination. In the time domain, the proper relationship is a Convolution Integral, as you correctly state.

The time-domain signal can be converted to frequency domain using the Fourier Transform (a Laplace transform with s = i*2*pi*frequency ) . IN THE FREQUENCY domain, there is a linear relationship between the source signal and the response at each frequency. The proportionality constant is a complex number (both magnitude and phase). If the magnitude is reasonably small (below threshold at which the material yields or the displacements cause permanent shifts), what happens at one frequency does not affect what happens at other frequencies, which is the beauty of this approach. So you can develop a "signature" which is a fixed function of frequency.

When the magnitude gets to be too big, there are distortions, and then you use things like neural networks to understand the nonlinear relationship from many test datasets, not to find different paths through the medium.

[samuel]

OK, let's do simple examples then.

X - source 0 0 0 0 0 0 0 0 K 0 0 0 0 0 0

K is the location where the shot was fired.
and length of X is the whole domain's volume. How big is this? Let's say 1km x 1km x 10m in resolution and I just pulled that out of my hat. Let's consider 300kmx300kmx300m domain and that's a grid of 300x300x30 = 27,00,000 grid nodes. This would be a "small problem"
your matrix A

in Y = H AX is 2,700,000 x 2,700,000 (there may be some reduced rank approximations)
Your measurement vector is
0 0 0 0 a 0 0 0 0 0 b 0 0 0

A few sparse locations. H is the corresponding measurement operator.

No problem with stationarity in time and glad we can keep shooting and measuring over 20 years.

How many Xs and Ys do you need to well reconstruct A? A lot, unless there are strong assumptions about the structure of A. right? Suppose you take 20% measurements (which in
many problems is grossly insufficient) that's 540,000 source-measurement pairs. If you measure at 10 places, non repeating, that 54,000 times you blow something up. Over 20 years that's 2700 times a year or 225 times a month. Ok, doable. Then consider that the medium will respond nonlinearly. What it does to a little blast is different than what it does to a big blast and suddenly you can't have a singular SI problem, i.e. all X are nearly the same set point. So, I don't know and devil in details.

May be some additional constraints from geology are used. But even people such as shell who I have talked to face immense problems walking around their shots year after year over the same sight in search for a reconstruction of the density. I mean when the reconstruct your brain using tomography, you know it spins around a bazillion times right?

[enqyoob]

kidwai

Aaiiyyooooo! Rahulji! A mere typo onlee, saar!! Will correct it immediately, saar! Besides, kidwai is actually a legit naam onlee, though I would not insult any kidwai by association with the person to whom I was referring.

Since poor Arun_S got beaten up mercilessly here for misspelling Dr. PC's name, with all sorts of motives ascribed, I have been terrified of doing anything that would be twisted into and from the usual quarters!!!!

[samuel]

I think whether you resort to time or frequency the underlying issue is really of sparsity right? The advantage you realize in frequency domain is based on the fact that in the frequency the convolution function (matrix) diagonalizes, but this is true if and only if the convolution funciton is block circulant i.e. medium doing the same thing and is therefore isotropic. At this point, the calibration curve and transfer functions become trivial. But it won't be isotropic right, which it because of spatially varying density, right? -- Then, how many measurements do you need to reconstruct the volume and what are the inherent degrees of freedom?
****
You can check from standard papers in meteorology that the kinetic energy generated by a storm is proportion to the wind speed cubed.

[enqyoob]

Samuel: Pls read some of those papers - or the Wikipedia page. Note that System ID is used routinely to get the complete description of skyscrapers, which have much more complexity. Try your example on that problem, and you can quickly prove to yourself that it is impossible to characterize even a 3-storey building. So that is obviously not the way to think about or apply the technique.

At most, you need one sensor at every house, if you wanted to REALLY push the limits. Otherwise, I would have used, say, 5 permanent sensors at Khetolai and 5 at the logistics base, from which I record signals every time there is an explosive test conducted. As long as the test stays below the threshold, I don't need to worry about individual houses, just the signals at one sample location, or, for overkill, at all ten locations.

A number of other sensor locations would be used all over the range, over a period of years, just to get a much more detailed simulation model, but this is not needed for the Khetolai / Si yield relationship.

In this business, instead of being buried under gazillions of node points and time-domain data points, you deal with, say, at most 1024 frequency-domain points at each of the ten sensors, and even there, the computer does the work, not you. There is NO matrix inversion to deal with.

hint in the interests of Truth In Arguments: Maybe I have used these things long enough for the green radiation from the oscilloscopes to have fried my brain...

P.S. Samuel, it is aerodynamic lift or drag that makes roofs, tree-trunks and cows fly in storms, not "kinetic energy". These are forces, and are proportional to wind speed-squared. I don't really need to check papers to know that...

Of course, once a tree starts flying, then what hits the house in its path does indeed have a lot to do with the kinetic energy of the tree. That still depends on the forces that accelerate the tree, and in turn, those forces depend on square of wind speed.