[Stoves] Going back to 3-Stone Fire [Was Re: Chinaandcookstoves]

Crispin Pemberton-Pigott crispinpigott at outlook.com
Tue Dec 26 03:52:43 CST 2017


Dear Kirk

I hesitated to address your use of the word 'projected' several times in the previous message and again in this one.  Please look up the definition. Then read where you wrote that how I 'am not interested enough to make it happen'.‎ There are always others who will bully, sneer and jeer in the vain hope it will apply pressure or diminish the effect of a message.

Your 'goodwill offer' to test properly is of benefit to you, not me. I am sharing what I consider beneficial when testing stoves. I am not selling a test method. None are complete and fully informed.  I do not have or control a laboratory.

Claims advanced rooted in a discredited test method do not help you or your customers. More, there is a definite possibility of repercussion or harm. The Gold Standard, which is aware of the mis-rating the WBT delivers, has a fascinating method of covering up the financial implications of it.

What you describe as 'my agenda' is not adequately captured so let me write it clearly for you. I have said this many times to many people over the years, so there are plenty of witnesses:

"I intend to overthrow the entire world of stoves and re-establish it on a new ‎foundation of modern science and engineering."

"Overthrow" refers to a paradigm. There are very few stove models that have been developed with a community of users and a community of scientists, both social and material, chemical and mathematical. ‎Notable achievements are the clay Darfur stove by Practical Action, the Improved Kenyan Jiko, the Thai Bucket Stove, the traditional Indonesian Keren Stove to which there is more than meets the eye, the Mongolian metal wood stove for yurts (gers) based on the original Russian design, and the high mass heating and cooking 'tile stoves' as perfected by the 16th century Dutch and Scandinavians. The Franklin Stove (1742) is a technical wonder that still has more potential than many stoves now being produced.

Sadly, the cooking stove ‎has few heroic devices. This sector, if it exists as an identifiable arena of endeavour, is suffused with agendas, none clearer than the 'how', and what is contained in, the IWA.

If we want broad acceptance, adequate funding,‎ access to equipment, staff, production resources, media, we have to up the game. We must cooperate to create a new paradigm.

There are many tests better than the WBT, and many excellent elements of tests and field assessments. A wave of your hand is not enough to dismiss their contribution‎ prior to your having made an independent investigation of how they work and what they provide. If you can't do that, read analyses by others. If you can't understand their analysis, read their conclusions. They are probably supported by the text.

Fabio Riva's work is difficult, but accessible, sobering in its implications, filled with guidance for test analysis, and conceptually brilliant in its approach. I guess that's why they will give him a doctorate degree. Yixiang Zhang's pedantic analysis of the formulas, constants and concepts of several test methods has provided clear guidance for analysts and authors.

The road ahead is very bright. ‎Walk towards the light.

Happy New Year everyone
Crispin





Crispin,

Thank you for taking the time to write a response.  You went on a lot about how the WBT works and the weaknesses you see in it.  I actually haven’t defended the WBT in this exchange.  I just wanted to know what those who have issue with the test include when they say WBT.  I made a good will offer to try one of the “good” tests you support.  It doesn’t appear that you are interested enough to make it happen, but I am.   Once again I think you have read a lot of your own agenda into what I said and asked.  I made a choice to stay with the WBT because it is available to me.  The “better” tests you mention are not.  I feel sad that it appears this will continue.  I hesitate to write again because I am not certain that you would respond without projecting your agenda.

Kirk H.

Sent from Mail<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgo.microsoft.com%2Ffwlink%2F%3FLinkId%3D550986&data=02%7C01%7C%7C234fa3b606174b46146b08d54c0d8454%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636498543301844696&sdata=4zagS9cXmwCASBjHw7DFr%2BD2G4FopkcrOFV0FJ0g2OM%3D&reserved=0> for Windows 10

From: Crispin Pemberton-Pigott<mailto:crispinpigott at outlook.com>
Sent: Monday, December 25, 2017 4:23 PM
To: Discussion of biomass cooking stoves<mailto:stoves at lists.bioenergylists.org>
Subject: Re: [Stoves] Going back to 3-Stone Fire [Was Re: Chinaandcookstoves]

Friends, this is a long post.

If you have every wondered why the WBT gets so criticized, and why so many people have tried to develop an alternative, this message is for you. The current WBT spreadsheet v4.2.3 is attached with a few sample values entered for demonstration only. If someone requires an explanation why you would like some other method used to determine the fuel efficiency of a stove, show them this message.

Dear Kirk

Regarding the information on your stove tests by others, thanks for trying and sending. It is not the WBT spreadsheet, it is a set of outputs from the spreadsheet. I understand (from you) that you do not have access to the sheets used to process the data. The sheet we need filled in is attached.

>My concern at the beginning of this exchange was to find out if the sensors, filters and computers were a part of the disagreement against the WBT.  It has been established that they are not.

I am not sure why this was a concern to you. The WBT does not prescribe sensors in a standard manner.

Here is how equipment for making measurements is normally provided in a Standard or a Test Method. I am using a Chinese one for gas stoves because nearly no one can read the bits that don’t matter, it is just to show how to create such a list:

[cid:image003.jpg at 01D37DB5.7BE627F0]

For each thing to be measured, it specifies a type, a range for the instrument to make that measurement, and a precision. This sets a limit on the uncertainty for each measurement. All measurements are accompanied by an uncertainty. When designing a test method, one should have in mind a final uncertainty so the result is acceptable. In order to determine the final uncertainty, all the measurement uncertainties have to be propagated through all the formulas used so as to produce a metric following strict rules<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fen.wikipedia.org%2Fwiki%2FPropagation_of_uncertainty&data=02%7C01%7C%7C234fa3b606174b46146b08d54c0d8454%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636498543301844696&sdata=V1Xqahxt8Hu2kVkz23lUdVtc1LRUyiPzo0A6QAlkozY%3D&reserved=0>.

For the WBT, there are a number of measurements to be made, and any instrument used to make that measurement will have an inherent uncertainty. The propagation of uncertainties is not shown in the WBT document. An attempt to use test data to determine the uncertainty of results of WBT’s was conducted by Fabio Riva in Milan as part of his PhD. We have discussed the paper on this list. He demonstrated that the results are highly uncertain, so much so that no wood stove could be shown to be better than the baseline stoves.  It was a devastating conclusion.  There is no need for me to add to that analysis which included a look at how the statistics for multiple tests is approached: WBT results are not ‘normally’ distributed.

An example of how the WBT increases the uncertainty to be greater than the “error propagation” gives is how it creates the ‘per litre’ values. (There is so much to learn from analyzing this protocol.) It is very unusual for a protocol to have calculations that increase the uncertainty of computed values not present in the raw data, beyond the increase from propagation of measurement uncertainties. In this case the WBT is an object lesson in how not to write a test protocol.

Boiling Water (Nikhil’s Nemesis)
The energy released from fuel to bring 5 litres of water to a boil is calculated using the ‘char deducted’ efficiency formula unique to the WBT. This involves starting with the mass of fuel fed into the stove, determining the energy in that fuel, deducting from it the energy in an unspecified fraction of the total char produced, resulting in an energy value in kiloJoules. This is divided by the ‘unit’ dry fuel energy (unless the version of the sheet you are using has the “E21 error”) per kg and this calculates a mass of dry fuel equivalent to the number of kiloJoules from the first calculation.

Because the mass of char is determined imprecisely, problems arises. There are two methods: taking the total mass of what is in the char container (including the ash) or ‘recovering’ most of it.  In the latter case, popular these days, this is not the energy used during the test, it is most of it. If the former method is used, it is more than the total char.  (See Penn Taylor 2009 for a correction of the ash=char problem.)  The total energy in the remaining fuel and char is not correctly determined – just the theoretical energy in the removed fuel with char knocked off (accepted as ‘unburned’), and the energy in the recovered char, whatever fraction that is of the total. The recoverable fraction of char is based on the opinion of the tester. If it is ash-in, the error depends on the fuel.

Next, the dry mass of fuel equivalent to the number of kiloJoules is divided by the mass of water remaining in the pot at the time the boiling portion of the test stopped. Suppose this was 4.8 litres. If the energy liberated was 5000 kJ, then the result is 5000/4.8 = 1014.66 kJ/litre boiled.  This is expressed in g/litre.  (Cell W37). The answer is incorrect because the number of litres boiled was 5, not 4.8. The error in this example is 4.2%.

Because the boiling point reached during the test may not be the same as the local ‘official’ boiling point (established using a procedure in the WBT protocol) this 4.8 litres is ‘corrected’ for the temperature actually achieved, and again corrected for the ∆T measured. This idea is to normalize the result to a standard 75 degree ∆T . From an engineering point of view, this is not necessary, other metrics are to be preferred, but the idea is understandable.  The first correction for the boiling point is not legitimate and introduces a random error that appears in the ‘effective mass of water boiled’ (Cell W32). In this case (attached) the error is 2.44%. The error in the char mass v.s. char+ash mass, alternatively char mass recovered v.s. char mass is unknown. We could suppose 5% of the char is unrecoverable or that ash is 5% of the char mass. In the attached example that has a 2.5% error impact for the energy number

The total error (before considering the instrumental uncertainties and propagation through formulae) is 9.1% of value on this metric, well over the total uncertainty permitted by the IWA (see the text for how that is stated).  These are, by definition, systematic errors because they are contained in the spreadsheet, however they are also random because it depends on fuel composition and what happens during the test. There are numerous other errors of this nature in the spreadsheet.  Thus, even having ‘perfect’ instruments, there is no possibility of meeting the uncertainty requirement of the IWA even with 100 tests or 1000.

>The disagreement seems to be only against the portion of the test that measures how much heat enters the pot.

That would be to misunderstand the nature and extent of the conceptual errors underlying the WBT. This set of calculations is indefensible. It cannot be argued with some clever words, that the WBT calculations deliver a reasonable assessment of fuel efficiency, emissions per task performed, even the heat transfer efficiency (which is not reported) would be defective.

The efficiency metric requires that the two high power tests be averaged. The intent is to report the effect of a high mass stove on fuel efficiency. No conceptual argument is presented showing that averaging a cold and hot start of a high mass stove provides an assessment of the impact of thermal mass, or that the result tells us something about performance. Let’s look at how this is done:

Cold start, fuel mass minus recovered char, however defined by the tester = energy = fuel mass burned.
Char energy is deducted from wood energy and the result turned into a dry fuel mass equivalent.

Hot start, fuel burned, minus not the char produced, but the same mass of char that was produced during the cold start. Why? Because it is difficult to follow the protocol if the test is stopped and the char mass determined directly. It would be easy to determine the char mass produced if the test were conducted on a scale as per HTP, and water heating instead of boiled, however this is not the case with the WBT.

So we have for the Hot start: a fuel mass burned, minus an indeterminate fractions of the char mass produced during the Cold Start test. The error this introduces can be estimated by taking the two fuel consumption numbers and subtracting to obtain the difference. In this attached example, it is 17.4% of energy (cold v.s. hot) which should have produced a difference of 14 g of char, ≈28 g of fuel = 7.2% error for the fuel consumption calculation. This error is additive to the others.

To get the average efficiency, we are required to sum the total heat gained by the pot for the two runs and divide by the energy in the total fuel fed during both runs. In the WBT, the total fuel consumed is not recorded, and the calculation is instead done by simple-averaging the two efficiency numbers. This is not a valid procedure as anyone in Grade 10 knows. It is permissible to average fractions, but not efficiencies.  It is permissible only if the result of the two runs is expected to be the same, while the entire purpose of performing the hot start is because the result is expected to be different. The difference in the example between the correctly calculated efficiency and the WBT formula is 0.5%. This brings the total error to 9.6% before starting to add uncertainties from the measurement apparatus. The error is not an uncertainty, it is just wrong. Under some conditions, it might be right, but you have no way to know when that is the case.

>I see this as not the most important part of the test since as soon as the stove goes to a new home, a different cooking vessel will be used and these numbers no longer apply.

That is an important point you make. Of what value is a number that does not represent performance in use? For a national standard, the performance assessment has to inform the regulator what the product will do. That is how you choose a baby seat for a car – a device that is supposed to assure the good health of your baby. Why should it be different for a cooking stove? A decontextualized testing method informs neither the user nor the regulator nor the funder. And that is quite independent of providing the wrong answer even if the contexts happen to coincide.

>The  measurements taken by the sensors and filters will continue to apply, though effected by different fuels and ambient humidities and temperatures, again with whatever test is used.

And their uncertainties will continue to be propagated through the defective calculating method to produce answers that are highly unreliable for any practical purpose.

>I have found the water boiling portion of the test to be useful for designing the stove to make the maximum amount of heat produced available at the top of the stove for the cooking vessel.

I agree. That is the best portion of the test in that it provides the heat transfer efficiency (or something close to it). If you maximise the heat transfer efficiency, you are probably going to minimise the fuel consumption and have the highest heating power, assuming that the burn rate is constant. The problems start when you assert that the fuel consumption is somehow provided by converting the heat transfer efficiency input numbers into a mass of fuel equivalent. It is not a valid step and the result highly misleading where the stove produces a significant amount of char with the error easily reaching 100% of value.

>I don’t see what all the disagreement is about.

Long ago I warned you that there were serious errors in the calculations in the WBT. You chose to take a certain path rather than investigating my assertion. If you can understand the delineation above, you have progressed. All these errors have been available in documented form for a decade. That they have not been corrected to date is a shocking abdication of responsibility by those responsible for creating and managing the Protocol.

I realise it is difficult for the average Joe to analyse a spreadsheet like the one attached, but it has to be done. One of the scariest things to happen is that this test was introduced in the IWA as ‘valid’ without expert review or correction, and that the EPA and others tried to get it accepted as the ISO test method without its ever having been submitted for expert review.  We can leave it to historians to analyse what they were up to.

Now, having gone through the high power metrics and covering only a few of the errors contained in them, I assure you that the low power metrics are far worse that the high power ones, to the point of ruling out the final numbers having any value at all. The ‘per litre’ numbers are literally of no scientific value because even after replicating all the other errors, they divide the result by a random number of litres, that number having no relationship with any of the fuel consumption numbers. (The mass of water in a simmering pot has no effect on the energy needed to keep it hot.)
You were assuring us on this list that your stove performed very well on the low power metrics, not realizing that these numbers have no meaning at all. It is not your fault. You were misled.

>The numbers are only useful for the test setup, whatever test protocol is used.  The useful field information is produced by the sensors and filters, this being CO, CO2, and particulates.

Field numbers run through the same defective labyrinth will still produce garbage results.

>I have been working on a new stove that improves on the Wonderwerk 316 TLUD-ND stove, looking for lower turn-down (hopefully to 1/5 power), cleaner very-high power levels, and simpler construction.

That is a good use of your time. Defending the WBT is not. Next time, or now, start using a properly drafted test method with valid metrics and calculations that are mathematically ‘legal’.

>My proposal is that in one years time (December 2018) we test this new stove with one of the “better” tests you suggest.  We will also test it with the WBT.

That will be helpful if you still harbor doubts about my assessment, and the assessments of numerous others in published reviews.

>I suggest this stove because its turn-down capability allows the low power measurements.

The IWA and WBT low power metrics are not valid (Zhang, Y et al, 2014). If you do not know why, it is worth your time to find out.

>We can then compare the two, showing the results of both tests with the same stove.  We will then have all of the scientific information possible for the stove and both test protocols, hopefully without to much contention.

Comparisons will be useful. The Chinese Agricultural University held a conference three years ago at which four different protocols were run side by side to see what the differences are for their common metrics.

They were pretty close when normalized for the different conditions, however the WBT fuel efficiency metrics was wildly wrong for char making stoves, with an error of more than 200% of value (19% v.s. 44% for example). The Indian Test method was mostly good conceptually and in the calculations. Changing the pot, as you mentioned above, would not have changed the result because the thermal mass of the pot is considered in the calculation of efficiency. That test dates from 1991 (written after reviewing the WBT) but the lesson learned in India never reached the West Coast.

>We may find that both tests have something to contribute.

In my view the WBT has nothing to contribute. It only adds confusion and error. Nothing ‘learned’ from a WBT could not be better and more correctly learned from another test. There is no published test that is as bad and error-riddled as the WBT with the possible exception of the EPTP from Colorado State U. That is why all results from it should be disposed of and everything re-tested.

>Since you are the major supporter of these “better” tests, you make sure that a “better” test is available at a neutral lab that has the sensors and filters that are not in contention.

I never waited for such a request. Better tests have been available for three decades. IS-13152 is a better test, but not nearly good enough, It has at least 11 errors in it, but it is a darn sight better than a WBT.

>Also the lab should be located where I can get myself and the stove to it without excessive effort or expense.  Perhaps CSU Fort Collins might be willing to host it.

CSU is not a fan of the WBT – they use their own test which introduces two new errors. The results are less variable but more in error.

>For my part, I will complete development of the new stove and make sure that it is tuned to its best performance.

I suggest that the new ISO test, which, if you do not use the char-deduction, provides the fuel efficiency at three different (unspecified) power levels. It will be useful to you as you are trying to increase the turn-down ratio. Dr Nurhuda has achieved a turn down of >4:1 with a stove very similar to yours. If you use a contextual test sequence, the result will be helpful for making claims about ‘those people’ using it.

>It would also be a way to introduce and prove one of the test protocols that you have been supporting.

I can retire from the ‘list of approvers’ – it is not me who has to approve of it. There are many technically competent regulators and scientists who understand how to design and conduct experiments. When I began I had no intention of writing test protocols. It was only out of necessity because I was being paid to conduct tests and design better products. If the test method doesn’t report the actual performance, it is difficult to know when small improvements have been achieved. Like anyone, I was surprised at how poorly the WBT was written.

>If you would rather use a different TLUD which will produce more emissions so as to make it easier to measure (LBNL had some difficulty at first and had to re-calibrate their equipment and get better filters to be able to make any measurements), that is ok too, as long as the stove remains unchanged between tests.  However, if we do that the turn-down limitation will degrade the low power results.

The LBNL test bench uses the EPA dilution tunnel system. This is not suited to small clean stoves, and is having difficult even with the modern masonry heaters which are very large.

>I have placed the data given to me by Lawrence Berkley National Laboratory on this list twice.

True, but we do not know how they produced it and what formulas were used. Did they put the raw data into a WBT spreadsheet and let the numbers fall out? When I wrote to them they said they did not use it. They have their own methods which are unknown and no sheet has been provided. The did report using the IWA metrics but as I have pointed out, the low power ones are not valid and we can’t un-divide by the number of litres because we don’t know what that number was for each run.

>I have included both pages again for your convenience so you don’t have to look through the archives.  You didn’t accept the data on these two pages before so I don’t know why you would accept it now, but it is all I have.  See the two pages of results in the one attachment.

That is a summary result, not the WBT spreadsheet. If we have the actual input data, say from Aprovecho (which does not usually provide it) I can re-calculated results based on the measurements they started with. This does not give a completely useful result, but it is how we can take a WBT and remove as much of the error and uncertainty as possible.

If you know what instruments were used for each measurement, it is possible to produce the final uncertainty in the outputs.

>Proving the performance of the stove was not my first intent in this exchange.  Answering my original question about what exactly is included in the disagreements with the WBT was my first intent, and I have the answer I needed.

I think you did not, but now may have it. The deconstruction of a test method takes time and I do not expect many here to do it.

>If we can settle the disagreements about the WBT so that further discussion will be productive, it will be worth the effort.

I am not yet convinced that you have a good understanding how now defective the WBT is. In other words, if you examined it, we would not disagree at all. There is no gain in ‘picking a side’ on a matter like this. There are hundreds of thousands of mechanical engineers all over the world making measurements daily to check the performance of solid fuel-burning devices like power stations and boilers. They do not make such fundamental errors as dividing by the wrong number of litres of water input to the system or mis-counting the mass of fuel fed into the combustion chamber.

Stove designers have no special dispensation to violate the rules of arithmetic because they want to bury char in the ground or match some other erroneous result.

Nikhil is in error to consider that these ‘engineering’ numbers are of no use or consequence. Many aspects of consumer acceptance and prediction of future performance can be determined in a lab setting. When the test protocol is informed by high quality social study of the intended users, the rating has far more meaning. You may have noticed that most widely accepted and ‘best fit’ stoves were designed in their final form by industrial designers. They are the specialists who bring the engineers and marketers and manufacturers and social scientists, ergonomists and invertors together. There is something to learn from that fact.

Regards
Crispin


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20171226/a8945e57/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 451638F3FA894EC7904CB2099DEC83A4.jpg
Type: image/jpeg
Size: 81265 bytes
Desc: 451638F3FA894EC7904CB2099DEC83A4.jpg
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20171226/a8945e57/attachment.jpg>


More information about the Stoves mailing list