Writer: TKT Janne Kovanen
In many older facilities, the determination of the fuel moisture content is still based on sampling and laboratory analyses. By far the greatest error in moisture content determination comes from sampling, so the aim is always to take samples as representative as possible.
In Finland, the quality guidelines for wood fuels (VTT-M-07608-13) are generally applied. They have been composed utilizing i.a. European solid biofuel standards. Corresponding national guidelines exist also in other countries that use a lot of biomass. The guidelines define how many fuel samples are to be taken from a fuel lot of a given size in order to reach the desired accuracy in the moisture content result. In principle, this is easy: follow the guidelines and reach the desired measurement accuracy.
Sampling, however, involves challenges. The computation formulas for sampling standards applied in the quality guidelines are applications of formulas used for coal (coal standards ISO 1390-7, ISO 13909-8, ISO 18283). They are therefore generally based on much larger delivery lots and sublots than in the case of biofuels.
The greatest challenge are moisture variations within a load, which are often ignored. The quality guidelines for wood fuels provide examples of the internal variations for various fuel types. These have been applied, even though the actual moisture variations, especially for forest fuels, are significantly greater.
Even perfect sampling won’t solve all problems
In Finland and Sweden, several studies have been carried out over the years on the internal variations in loads of forest fuels, for example the extensive studies by Metla (Finnish Forest Research Institute) in 1987, which demonstrated great internal variations (Verkasalo Erkki. Measurement of moisture content and dry weight of forest chips by load sampling methods. Folia Forestalia 694 Metla. 1987). This study is not mentioned in the references of the quality guidelines for wood fuels.
The recently published VTT study report ”Kiinteiden biopolttoaineiden CEN näytteenottostandardin soveltaminen Suomen oloihin” (Application of the CEN sampling standard for solid biofuels to conditions in Finland) also states that the internal variations in loads are greater in practice than has been assumed. If one wanted to reach the accuracy so far assumed in the quality guidelines, the numbers of samples should be increased.
Figure. The required number of individual samples from a load of forest residue when aiming for a +/- 3 weight-% accuracy.
The quality guidelines determine the numbers of samples, and after this the individual samples are either mixed or divided by the squaring method into a single sample, or alternatively the individual samples are added into a sample bin containing the same fuel from the same supplier. And after that, some part is taken from the big bin as a bulk sample for oven drying. This would all work if the fuel was of uniform consistency.
What is uniform, then? Pulpwood chip is considered uniform, but if you ask about it at a pulp mill, you are told that it isn’t always so and that there can be great variation between loads.
Based on our own experiments, it is possible to take samples of very different moisture levels from screened pulpwood chip of a single wood, when sufficiently many samples are taken.
Why is the unreliable standard method still used, then?
There is a custom of “one bulk sample per day per fuel per supplier” in general use. The motivation for this is keeping the amount of manual labour in laboratories, i.e. handling samples, weighing, recording results, and oven drying, at a reasonable level. And not so much getting a statistically correct result. Consciously or unconsciously, one trusts one’s own sample and its evidential value (”Belief in the law of small numbers”, Tversky-Kahneman, 1971).
Some facilities, instead of bulk samples, use per-load sampling and analysis. There samples are collected from each truck, and typically one of them is analysed using oven drying. A single 300-500 gramme sample then represents, for instance, a 35 ton load.
Sticking to the old method is justified by “this is what we’ve always done and will do” thinking which is combined with religious-scientific philosophising about the shape, size, trajectory, and rotation direction of the sampling instrument, theory of probability, and historic results. It might also be the so-called Ikea effect, when one is evaluating the functioning of one’s own sampling machine. Especially if the choice made back in the day cost a million.
At its worst or best, many graduation theses ignore bulk sample results that differ from the other results because they are “errors”! That is to say, the end result desired according to the pre-adopted theory is not reached if all the results are accepted.
What can go wrong in sampling?
The commonly used theory of “the driver is taking too dry samples” might just as well be “the facility is taking too wet samples”. Often the reason can be the duration the samples are let to dry in the sampling bins or that the division of the sample is done carelessly.
An automatic sampling instrument or a robot can be imagined to take impartial samples. And so it does, but a machine does what it is told, or what it in terms of mechanics, automation, and conditions is able to do.
As an example of this, in most of the sampling instruments we have seen, even sampling is implemented so that the instrument grabs samples every minute and is satisfied when six have been taken. This leads to the question, how long does it take to unload a 3-part container truck? And how long does it take the supplier to notice that he should put the best fuel in the first container?
Another example could be about the piece size distribution of the sample. Either the mechanics of the sampling instrument favours small or big pieces, or the division happens in the sample handling during crushing, mixing, or squaring.
A third example comes from the sampling instrument providing an unrepresentative sample, at worst, wrong fuel from the previous truck. If you ask the workers in the facility what is done in such a situation, they often answer: “analyse, since the stuff has ended up in the sample bin”. There is no time to figure out who brought what, or was supposed to.
A fourth example is related to the determination of the calorific value and the impurities or the degree of rotten of the sample. Calorific value determined in a laboratory from a single small crumb can very well introduce variation, from which the conclusion is drawn that the calorific value of the same fuel varies greatly and therefore has to be constantly analysed. In this, the basic principles and quantity requirements of sampling in order to reach any kind of accuracy are forgotten.