[Stoves] Western Electric - rules for analysis of divergence

Crispin Pemberton-Pigott crispinpigott at gmail.com
Fri Mar 11 19:50:55 CST 2011


Dear Frank

I was thinking of applying it in a very different context: not looking at
the performance of a stove with different ingredients, so to speak, but if a
test is useful for predicting performance.

Suppose you test a stove using 4 different methods in a lab. The might in
fact be 4 different methods of analysing data from a single test. In other
words the data collected is well agreed by there are alternative ways of
looking at the data and determining the 'efficiency' or emissions or
savings. So you get 4 different outputs, or more accurately, 4 sets of
outputs. Which analysis tool delivers better information?

Then you put the stoves into the field and measure in-field performance. The
method used in the field may matter a great deal. One thing I have noticed
is that the people working outside the WBT-CCT orbit use uncontrolled
cooking tests and get what appear to be realistic measures of what is
actually happening in terms of fuel savings. The measurement of 'real'
performance can be debated separately. The point is to compare lab tests to
what is an agreeable measurement of actual field performance.

As you know, most stove selections are made on the basis of lab tests
because that is what is available. Many stove projects promulgate new
technologies, or newly adapted technologies. Testing and 

Field proofs' are few.

If you plot the actual performance as a centre line and then plot the
claimed performance from lab tests as above or below that line, and the
claimed performance is consistently above or below (as defined by the W-E
method) then a red light should go on. If lab results are widely distributed
both above and below the line, it would indicate that the method is
inaccurate. If it is consistently on one side, it would indicate a
calculation error which can be corrected. If the correlation is random then
is has no predictive ability (i.e. is not making a useful assessment of the
performance.

The point of the W-E Rules as I understand it is to look for systemic
problems: if there are lab results that are inconsistent with real
performance, one should look into the method for causes of the difference.
The result would be a better test.

Similarly, if the field measurements are inconsistent with lab tests of how
a product should perform, one can also look at the field tests for
inaccuracies or different calculation methods. I see the field side as
consisting of two different types of evaluation: one which is pretty much
the duplication of the lab test but in the field (CCT) and one which
measures what people actually do when they cook what they want, measured in
a similar manner (UCT).  It is the latter which I am convinced provides an
assessment of real performance.

There is a paper coming from the SeTAR Centre looking at the number of UCT's
that would have to be conducted to get a statistically meaningful assessment
of performance (DUE Conference, Cape Town in April). The development of this
method (also used by others in 'stoves') provides an opportunity to check
the efficacy of lab test methods. As I see it now, the various test methods
(protocols) give results/accuracies which vary a lot from the real numbers.
The W-E Rules can show us how they deviate and under what conditions.

Regards

Crispin

++++++++++

Dear Crispin, and all,

It seems the Western Electric Rule (WER) is used when 1) we think we are
ready to put all the systems together and want to determine if they work
combined or 2) We want to convince (prove to) some NGO that we need money
funded into an international stove QC program so to move forward.

>From what I could understand (way over my head in many parts) is this looks
at nothing more than 'noise levels' of a system. We have enough (IMO) in
each segment of the systems to work on first before there is a reason to put
all systems together and look at the whole.  

 The small part that I look at (biomass fuel) is not even quantified for
ranges of parameters for any of the stoves developed. Moisture, particle
size and description, particle density, proximate analysis and their ranges
need to be researched per stove to market that stove for a fuel area and
compare that stove against others marketed for the same fuel criteria. That
noise level in itself is enough to make the WER pointless to look at. 

Regards

Frank

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20110312/269673be/attachment.html>


More information about the Stoves mailing list