[Stoves] Testing Cookstoves: Autocorrelation and White Swans

[Julien Winter]

I haven't got as far as water-boiling tests yet, because I have been
working on fundamental issues of burner design, and that takes a while.

However I can see a some problems with the way stoves are being tested.
1)  Measurements made over the course of a run are autocorrelated.
2)  Proper testing of a stove involves a range of fuels.
3)  Test to find the boundaries of failure, not success.
In summary, proper testing of a stove, prior to manufacturing thousands, or
making it an exemplar for millions can't be done with one test.  It takes
many tests; tests that try to find out not only where the stove succeeds,
but most importantly, where it fails.

1)  AUTOCORRELATION and Correlation vs. Independent Observations.

Autocorrelation, or serial correlation, is a statistical term use to say
that two observations, for example, of temperature, are not independent,
because they are related in space or time.  Autocorrelation can be a good
thing to study if you are looking at spatial patterns in soils, but it can
be a problem if you are trying to measure properties of a stove.

If we are trying to measure energy transfer during boiling, followed by
energy transfer during simmering, all in the same run, then these two
measurements of energy transfer will be autocorrelated.  They are
autocorrelated, because there history to the combustion reaction,
especially in a TLUD.  In a TLUD, the depth of char increases over time,
and changes in temperature.  This change can alter the chemical composition
of the pyrogas.  In TLUDs and other stoves burning thick pieces of
fuel, char combustion can increase over time.

Now it may be that boiling followed by simmering is so common that energy
transfer from fire to pot over this the sequence should be
measured.  However, if we a primarily interested in how efficient energy
is transferred at different power levels, then having a separate run for
each power level would make the observations at different power
levels independent of each other.

There are actually two different turndowns to measure:
a) the turndown of fuel consumption rate
b) the turndown of energy transfer rate to a pot

2) A RANGE OF FUELS

It is important to test stoves over a range of fuels, because they behave
quite differently depending on moisture content, volatile content, particle
thickness and shape.  In a ND-TLUD the fire in wood chips invariably,
channels; with thick fuel (e.g., sticks) there is char combustion on the
surface while the interior pyrolysis; and, if air spaces are vertical then
a very strong draft develops in the fuel bed.   Channeling of the ignition
can increase as primary air is cut back, especially in wood chips.

Across all these fuels there is >5 fold change in ND-TLUD gasification
rate.  In other words, turndown is not properly represented by a single
fuel.

3) TEST FOR FAILURE NOT SUCCESS

Critical testing of stove should try to find where it fails.  Although it
is useful to see where a stove succeeds, repeated observations of success
is not critical testing.

Scientist are encouraged to design critical experiments that reject
hypotheses, not confirm them.  If, under critical test, our hypothesis is
not rejected then it is probably true.  We owe this line of reasoning to
the philosopher, Karl Popper.  Queen Elizabeth (of England) gave him a
knighthood, therefore, he must the right!!

In the case of cookstoves, we should cover a range of conditions (fuels and
turndowns) to see where they fail.  Using a single fuel is not subjecting a
stove to critical testing.

To use a more simplistic example, let us say that a European looks in the
sky and sees only white swans, and comes up with the hypothesis that "all
swans are white".   To confirm the hypothesis is not the way to go.  We
can count thousands of white swans in the skies of Europe, and we still
haven't put our hypothesis to a critical test.  We have to devise a
circumstance or an experiment, where our hypothesis could fail.  So lets
search all corners of the Planet to see if we can find a non-white swan.
Lo, in Australia, the swans are black.  The crucial point here is that
in our critical test, we only had to see one black swan for us to
reject our hypothesis that "all swans are white"; all the thousands of
white swans, previously seen, now count for naught.


To test cookstoves (i.e., ND-TLUDs) on wood pellets is to count white swans.


To conclude:  before a cookstove is promoted as an exemplar, or sold by the
thousands, there is a lot of testing to be done to characterize stove
performance over a range of conditions.  One water-boiling test just
doesn't make the grade.

Cheers,
Julien.

P.S.:  Scientists must really get inventors and engineers pissed-off.  Not
only do the want to try to break the stove, they want to do it four times
so that they can say, "Yup, I am 95% sure that your stove is broken."


-- 
Julien Winter
Cobourg, ON, CANADA
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20150218/5438386e/attachment.html>