Here it is!
Beiting, E., Investigation of the nickelhydrogen anomalous heat effect 2017, The Aerospace Corporation: Report No. ATR201701760
http://lenrcanr.org/acrobat/BeitingEinvestigat.pdf
See also the two abstracts on pages 7 and 8:
Here it is!
Beiting, E., Investigation of the nickelhydrogen anomalous heat effect 2017, The Aerospace Corporation: Report No. ATR201701760
http://lenrcanr.org/acrobat/BeitingEinvestigat.pdf
See also the two abstracts on pages 7 and 8:
QuoteThe results presented here indicate a source of heat energy was measured that is not chemical in origin.
This measurement and conclusion are in general agreement with those of Arata and Zhang [4] and Ahern
[5]. Given the importance a source of high specific energy has for space technology, this work indicates
that additional research should be pursued. If future work is better funded, calorimetry should be used
instead of thermometry to measure the heat power.
Beiting is quoted:
"If future work is better funded, calorimetry should be used instead of thermometry to measure the heat power."
That has already begun. See the second abstract listed above.
Some comments on the Breiting / Aerospace corp report…
Executive summary (for those who want to quit now): Looks like a good chance it’s noise.
Comments:
Breiting (‘B’ hereafter) does the same thing the rest of the CF community does and fails to consider the calibration constants in his cal. equations as experimental variables. He talks about doing a Propagation of Error calc but the only variable he looks at is T explicitly. So, I did my usual, and plugged in a 16% variation in cal. constants. I did it simply by increasing and decreasing the constants in his cal equations by a fixed percentage using the same number on all the constants. That doesn’t mean they couldn't vary independently, I just didn’t want to bother with all that.
But to start we need to know what the supposed error is that we are trying to explain. B also tries to do the energy per unit mass trick, which is bogus since he doesn’t know what ‘mass’ is the right mass to use. So backing it up, his 173 MJ/g number works out to ~0.94W excess power signal, which he also refers to as ‘~1W’ in his exec summary.
So using the PwrC2T1Vac equation, I calculated the power for a span of T’s, but I will only discuss the one for 300C. His numbers give 12.94W at 300C, increase all the constants by 4% and you get 13.45W, decrease by 4% and you get 12.42W. That’s a span of 1.03W. B only reports one calibration, so we have no idea what the reproducibility of his calibrations is, but these results are in the same range as the span on the Storms’ data, and on the various reports by Miles.
But further, I had to ask why he used a cubic equation. To get a rough idea what might happen with a different equation, I used his equation to make a small set of data similar in size to what he used, and then fit it with a quadratic. At 350C. the cubic gave 17.34W and the quadratic gave 16.41W, for a difference of 0.93W. At 300C the diff was 0.88W. (Note that using the cubic gave a positive excess heat.)
In other words, simple piddling around with the calibration equation covered the signal detected. That means to believe the calibrations, we need a lot more info on how he chose his equations, and we need to see some replication to assess the reproducibility of those equations.
After that we can consider the rest of the waffling he does.
I am not going to defend this. You do the calcs if you don’t like what I say.
So you’re saying that if you shift the entire calibration curve up the page by a minimum of 8%, the excess heat disappears  Or if you use a less wellfitted* regression curve, the same thing happens?
Meanwhile, in other news, the pope is still catholic, and some bear excrement was found in a forest.
Care to suggest a reallife, feasible, nonzoological reason for such a potential dramatic swing in calibration constant? If indeed that doesn’t count as you ‘defending’ your take on reality?
* I notice that, unlike Beiting, you don’t bother to quote your new quadratics R^2 value...
The R^2 of a quadratic fit to a known cubic’s data is not particularly useful. Instead I hand digitized the Cell 2 Vacuum (red) data from Figure 4.4 as an approximation to the real data. I then fitted it with a quartic, cubic, and a quadratic. Results are:
Quartic eqn coeff.: 5.36754e10 2.46591e7 1.26090e4 1.09850e2 .280572 R^2 = .999990
Cubic eqn coefficients: 1.15621e7 4.741444e5 1.69908e2 .384213 R^2 = .999970
Quadradtic eqn. coeff.: 1.06138e4 9.05552e3 .171434 R^2 = .999743
The R^2 values are not adequate to distinguish which model is best. The t values of the coefficients are needed (or the significance levels) so the full analytical results are needed.
The interesting thing however is the differences in the preducted power for a given temp in the upper temperature operating range. 3 sets are shown below for 300, 325, and 350C.
T 
Power 
modelcubic 

quart 
350 
16.49228 
0.197372 
cubic 
350 
16.29491 

quad 
350 
15.9999 
0.295 
T 
Power 
Modelcubic 

quart 
325 
14.13117 
0.044798 
cubic 
325 
14.08638 

quad 
325 
13.98244 
0.10394 
T 
Power 
modelcubic 

quart 
300 
12.05278 
0.02491 
cubic 
300 
12.07769 

quad 
300 
12.09764 
0.019952 
Notice that a) the difference (model error) increases as the T increases and b) the magnitude is significantly larger that the 0.050 W claimed by Breiting for some cases. Again, Breiting needs to show more data and analysis results. Kudos to him for giving the cal. equations, because this is extremely rare in the CF field. That is what allows the above analysis.
As well, the 'sensitivity analysis' I did with the variation of 16% in the cal constants of the given cubic cal equation still stands as well. Proper error propagation is required...
My general conclusion still stands, the observed excess could well be noise (or a combination of noise and modelling error).
If there were an 8% shift, as Shanahan claims, much of the test shown on p. 20, the calibration runs, and the control runs would be endothermic. They would be swallowing up megajoules of heat. It would be a fantastic coincidence that the calibrations fell exactly on the zero line. This is impossible. Shanahan has a rare talent for inventing impossible physics.
It's recurringly amazing how some people keep believing.
As expected...
If there were an 8% shift, as Shanahan claims, much of the test shown on p. 20, the calibration runs, and the control runs would be endothermic. They would be swallowing up megajoules of heat. It would be a fantastic coincidence that the calibrations fell exactly on the zero line. This is impossible. Shanahan has a rare talent for inventing impossible physics.
Readers may first notice that p.20 of the report deals with the first experiment, which is denoted a learning experience and is not used to report apparent excess heats in the report's main body. Thus it is not relevant to my discussion to date. There may well be other things going on with it. I didn't look at it since it was not used to make excess heat claims.
Further, I clearly showed that slight changes in the calibration constants have significant impacts on the predicted power computed from measured temperature. At 350C, the model changes I looked at can cause an ~200 mW error, much larger than the claimed 50 mW. Likewise at 350C, a 5% (not 8%) change induces an 820 mW error (and as noted, the global numbers imply something on the order of 944 mW). Table E1 of the report shows that between the vac and N2 cal curves there is a 574 mW difference at 300C, so the small changes in cal constants I am discussing give changes on the same order of magnitude as changing the gas space contents. This implies knowing the variation in cal constants is as important as knowing what is in the gas phase of the cell which Breiting spends considerable time discussing and compensating for.
Why Jed thinks such changes would drive the curves endothermic I don't follow. Unless he is talking about much lower temps. As one can see, at 0C the cal equations give negative powers, but that is just an indication of the fact that the real cell behavior doesn't fit the polynomial form well outside the data span, which is typical of statistical fits. And in any case, the cell is going to do what the cell does. The question is how well the equations used to interpret the data correspond to the real behavior.
Also note that a 50mW error at 12 W input is a 0.4% error. That's a better calorimeter than McKubre ever made. Do you all really believe Breiting made that good of a calorimeter on his first shot? Better error analysis is required.
I will also say at this point I looked a little bit at the chemistry used by Breiting and I find it inadequate. It is incorrect to assume the observed heat (if assumed to be real) is not at least partially explained by chemistry.
I think that using such a high degree polynomial risks hide the measurement error in the fit. It would have been nice to have two independent measurements at each
point in order to clearly demonstrate the size of the measurement error. Using low degree polynomial will lead to model errors that are absurdly big in
case the points is almost exact e.g. small measurement variation but follows a high degree polynomial. When repeating values one could vary some parameters not included
in the model but which is known to have a small effect this way we increase the noise a bit but get to be more on the safe side in the analysis. The fit indicates that the measurement
error is very small and the callibration essentially follows a deterministic law, my question is if that is the case.
Also note that a 50mW error at 12 W input is a 0.4% error. That's a better calorimeter than McKubre ever made. Do you all really believe Breiting made that good of a calorimeter on his first shot?
Yes, I am sure he did. I expect McKubre agrees. The state of the art has improved and the people at The Aerospace Corp. are world class. (See: http://www.aerospace.org/) McKubre described his calorimeter as easy to understand but "stupid" (simple minded, rather than ultraprecise). The error in it was larger than several other calorimeters, such as Fleischmann's and Miles'.
Beiting has made a second, flow calorimeter that confirms the first one. Shanahan cannot explain this, either, except with his impossible handwaving.
Beiting has made a second, flow calorimeter that confirms the first one.
Link? I can't be responsible for what I can't get.
I gave the link already, but here it is again. See the second abstract, p. 8:
An abstract is simply a claim until substantiated with the standard types of information that would come out in the supposed subsequent paper. Thus, at this point, your claim is probably just wishful thinking.
It do look like the calormetry is very exact, as taken for granted by those who work with it, and the method used e.g. advanced interpolation using higher order polynomial which is appropriate for deterministic systems. This means that statistics is out of the window and is not a good method to apply here because the estimated variance is way to high compared to the real one due to the low degrees of freedom. In any case a figure of e.g. the relative error is xxx% would be helpful just to checkbox that non are fooled in any way. There are method that does statistics based on known variances and they could be used here in case one want to dwell on it and I think that would be the appropriate way to do it in this case. We don't have a figure
yet, good researchers know the variance of their systems and my bet is that Breiting knows quite well that all is ok. But no proof on the table for the dining party.
It do look like the calormetry is very exact, as taken for granted by those who work with it, and the method used e.g. advanced interpolation using higher order polynomial which is appropriate for deterministic systems. This means that statistics is out of the window and is not a good method to apply here because the estimated variance is way to high compared to the real one due to the low degrees of freedom. In any case a figure of e.g. the relative error is xxx% would be helpful just to checkbox that non are fooled in any way. There are method that does statistics based on known variances and they could be used here in case one want to dwell on it and I think that would be the appropriate way to do it in this case. We don't have a figure
yet, good researchers know the variance of their systems and my bet is that Breiting knows quite well that all is ok. But no proof on the table for the dining party.
Stefan, several comments...
Very exact calorimetry means very small changes can have very big impacts.
I checked the problem with low degrees of freedom. When the number of parameters (p) used in the fit approaches the number of data points (N), the use of R^2 is inappropriate and one must use what is known as the 'adjusted' R^2 = Ra^2.
Ra^2 = 1  ( ( 1  R^2) ( N1 ) / ( Np1) )
Didn't really change the values enough to matter in the cases I looked at, but one always needs to check of course.
It is a wellknown problem that linear regression can 'overfit' data sets, especially smaller ones. The chemometric procedure that is preferred is to used the PRESS value (Predictive Residual Error Sum of Squares) instead of R^2, which is determined through a process that uses jackknifing or bootstrapping the data set to check the predictive accuracy of the statistical model. (That's where the 'P' in PRESS comes from.)
However, the requirement for determining the error in the computed output power is still the same and requires that all experimentally determined variables be included in a propagation of error calculation, which does, in the end, only give an estimate of the error. My calculations in the prior posts simply show that the 50 mW claimed error seems to be overly optimistic (which is typical in CF calorimetry) based on the fact that even in the best calorimeters the cal constants seem to have 15% relative standard deviations. It is always possible that Breiting has done better, but he needs to show this, not just assert it, as you indicate as well with your 'dining party' comment.
and the people at The Aerospace Corp. are world class. (See: http://www.aerospace.org/)
Out of curiosity, how do you measure that?
McKubre described his calorimeter as easy to understand but "stupid" (simple minded, rather than ultraprecise).
Which one? The one he used for his 19921993 studies? Or one of the ones from his 19931994 studies (the L, M, HH, T, D, F, G, OHF type) reported on in the 1998 EPRI report? Or some other one?
The error in it was larger than several other calorimeters, such as Fleischmann's and Miles'.
Given that no one in the CF community sees fit to define the error in their calibration curves properly and carry that through to the final results, this doesn't mean much.
Kirk: I suggest you explain why this 8% shift only occurs in cells with the material produced at Ames in specified conditions, and not during calibrations or with control cells.
Out of curiosity, how do you measure that?
It is an observation, not a measurement. People familiar with the Corporation say this, and I think the web site backs it up. If you do not think they are worldclass experts, perhaps you should tell us why. Have they made serious errors? Have their other studies been discredited?
Kirk: I suggest you explain why this 8% shift only occurs in cells with the material produced at Ames in specified conditions, and not during calibrations or with control cells.
Why would I try to explain your proposition?
It is an observation, not a measurement. People familiar with the Corporation say this, and I think the web site backs it up. If you do not think they are worldclass experts, perhaps you should tell us why.
Well, if it is 'just' an observation with no numerical quality to it at all, then you can't claim it establishes any type of priority to the organization vs. others. I believe that in fact you are asserting that the work coming from the Aerospace Corporation is better than that from many other organizations. And my intent is to point out that that kind of 'observation' is pointless. If that method was valid, I could claim that MIT was a flybynight university since Hagelstein doesn't understand the difference between random and systematic. That of course would be completely incorrect about MIT. What Hagelstein does reflect primarily on Hagelstein, and only marginally on MIT. Likewise, what any Aerospace Corp. employee does reflects only marginally in a technical sense on Aerospace. Of course political and social impacts can be more prominent. So in particular, I have no idea whether the employees of Aerospace in aggregate are good, bad, or ugly, nor do I really care, because all I am interested in at the moment is Beiting. Where Beiting works is of little importance. What Beiting does is much more relevant.
Furthermore, anyone can make mistakes, so the rep of the employer has no relevance at all to the question of the quality of the specific work.
Have they made serious errors? Have their other studies been discredited?
Beiting, yes, he has. He failed to compute the error of his calibration curve properly and he failed to take into account the proper chemistry in his sample prep and subsequent experimentation. There might be more if I study the paper more, but what I've seen so far is enough to class his efforts as 'typical soso CF community work'. And that isn't 'worldclass'. With regards to other Aerospace people, no idea, don't care.
It is disappointing to read all this quibbling and fussing over claims for small amounts of excess power. If LENR reactions make energy, why hasn't someone, in 30 years since P&F, found a way to make say, ten times the current amounts, so that statistical methods are not necessary for evaluation and errors of the type Shanahan is calculating simply don't matter? Of course, JedRothwell will say people have. But then it turns out from the papers that it was for a short time, with questionable methods of measurements, or only one time and not currently reproducible. Or worse yet, the claim is based only on an anecdote. Frustrating!
so that statistical methods are not necessary for evaluation and errors of the type Shanahan is calculating simply don't matter?
The methods and errors that Shanahan is calculating are impossible. They do not show up in calibrations. In other words, there is no test he can propose to show them, or to falsify his claims. If the problems existed he could propose a method of calibration that would reveal them. It is not possible that the choice of metals would change the arithmetic.
Well, if it is 'just' an observation with no numerical quality to it at all, then you can't claim it establishes any type of priority to the organization vs. others.
Yes, you can, using realworld observations and common sense. This is the world's leading aerospace engineering organization. Aerospace engineering is the most difficult, advanced and demanding engineering there is, and it is extraordinarily successful. We know that because airplanes seldom crash. A team of experts there who worked for years on this experiment. It was reviewed and approved by the management. It seems unlikely that such people overlooked a problem that Shanahan found in an hour or so.
Shanahan's hypothesis is even more unlikely because it is a magic problem that cannot be detected by calibration or any other test, and thus cannot be falsified. It only happens when there is a particular choice of metal, which cannot affect the calorimetry. There can be no physical explanation for such a thing. It resembles his claims that people cannot feel an object is hot by sense of touch, or a 1liter object will remain hot for three days with no input power, or that a bucket of water left in a room will magically evaporate overnight. In other words, once again he makes claims that anyone should instantly recognize are preposterous. I doubt he believes these claims. I suppose he is trolling us, or hoping to fool people such as seven_of_twenty who apparently cannot tell that Shanahan is spouting impossible nonsense.
It is disappointing to read all this quibbling and fussing over claims for small amounts of excess power. If LENR reactions make energy, why hasn't someone, in 30 years since P&F, found a way to make say, ten times the current amounts,
My stock answer is that if you had any idea how difficult it is, you would be amazed that they can do it at all.
But then it turns out from the papers that it was for a short time, with questionable methods of measurements, or only one time and not currently reproducible.
Either you did not read the paper, you are deliberately misrepresenting it, or you have a strange notion of what is a short time. Every statement you made here is false, and contradicted by the paper. I will not bother to point out where because you will probably repeat the same nonsense even if I do. Depending on whether you are ignorant or lying I suggest you should:
1. Read the paper before commenting on it. You make yourself look foolish saying that 42 days is a short time. What would be a long time?
2. Stop trying to fool people. Anyone can read the paper and see that it does not say what you claim.
OK then. All of mainline science is fooled. LENR is a raving success.
I'm not trying to fool people. I couldn't begin to compete with the likes of Rossi.
BTW, I was not referring to any particular paper. I was writing about your usual response to the "low power" or low Pout/Pin or "short time" critiques.
OK then. All of mainline science is fooled. LENR is a raving success.
It is a success, when you judge it based on mainstream, peerreviewed journals of physics and chemistry. It is not a success if you get your information from Wikipedia or people who say it is impossible to tell an object is hot by sense of touch. I suggest you stick to conventional, legitimate sources of information.
BTW, I was not referring to any particular paper. I was writing about your usual response to the "low power" or low Pout/Pin or "short time" critiques.
You are not referring to any paper at all. Your description is nonsense. But, in any case, this thread is a discussion of the Beiting paper. Shanahan's "statistical methods . . . for evaluation and errors" apply the Beiting papers. So obviously, you were referring to that paper. It seems you have not read it, so you don't know what you are talking about.
You should not assume that Shanahan's "evaluation and errors" are real, and he actually did find something that experts failed to see over several years. Oh, and that magically does not show up in calibrations. When you believe such things, you are as naive and misguided as a Rossi supporter.
It seems unlikely that such people overlooked a problem that Shanahan found in an hour or so.
That is the nature of systematic errors. Or lack of training.
That is the nature of systematic errors.
Invisible systematic errors that cannot be detected with a calibration, or by any other physical test. Unfalsifiable errors. Metaphysical errors that you alone, in all the world, believe. Perhaps you are delusional. Surely you are an egomaniac who thinks he knows better than a team of experts who spent years on this experiment.
Either that or you are trolling us.
Either that or you are trolling us.
Or you are incapable of understanding the issues, an interpretation your comments seem to favor.
Why would I try to explain your proposition?
Because it is actually less ridiculous than your own proposition?
Beiting, yes, he has. He failed to compute the error of his calibration curve properly
Well really, he failed to compute your whimsical ideas about the possible error inherent in his calibration curve.
You must be crazy to think that some strange and unexplained (and apparently also inexplicable by yourself) failure/CCSH causes a temporary shift of that size.
Or you are incapable of understanding the issues, an interpretation your comments seem to favor.
This is not about me. You are saying the experts at The Aerospace Corporation are incapable of understanding the issues. You are also saying that even though their calibrations do not show these errors, the errors exist. You should tell us how the researchers might reveal the errors and confirm your hypothesis, if not by calibration. What other tests should they do?