Measurement standards, source of problems?

Material transformations involve strong couplings between material and processing conditions, such that the same material or combination of materials gives rise to different products depending on the conditions. Such a particularity has multiple implications that I have been trying to highlight and analyze for several articles.

The paradox does not end there. The measurement standards, on which most industrial methods are based, are likely to obscure the identification of the causes of difficulties and their control. To the point of paradoxically constituting indirect sources of problems. Let's try to see how.

The measure, this evidence

The value of measurement to quantify properties and behaviors is generally self-evident in technical fields. Joseph Juran sums up the consensus surrounding the principle of measurement well: "Measurement is the most effective remedy for vagueness and multiple dialects — 'Say it in numbers.'. To measure is to assign numbers to properties or phenomena by processes.

In the language of metrology, measurement (measurement) is a "process consisting in obtaining experimentally one or more values ​​that can reasonably be attributed to a quantity" ; a magnitude being a “property of a phenomenon, of a body or of a substance, which can be expressed quantitatively in the form of a number and a reference”. Note that these are values ​​assigned reasonably, after precisely ou scientifically. The link between the measured value and the magnitude is not as strict and precise as is generally imagined. I will come back to it.

There is therefore consensus on the importance of the measurement and generally on its methods as well: an actor in a given sector knows which measurements must be carried out to meet normative constraints or its R&D needs and how to do it. In principle at least.

Measurement standards, the heart of the measurement policy

Measurement standards most often constitute the backbone of companies' measurement policy, obviously implemented in quality control, but also often reproduced or declined in R&D.

Ishikawa's sharp observation

Ishikawa's view is unequivocal: “There are no standards -whether they be national, international or company-wide- that are perfect. […] Standards that were adequate when they were first established, quickly become obsolete […] If standards and regulations are not revised in six months, it is proof that no one is seriously using them.”

In many sectors of material transformation, it can be noted that measurement standards date back several decades. The evolution of the properties of raw materials, the complexity of the functionalities of finished products, customer expectations are all elements to abound in the direction of a progressive obsolescence of standards for technical reasons. - the six-month period mentioned by Ishikawa deserves a justification that he unfortunately does not give.

The imperfection of the standards can be better understood through the analysis of the usual use of the data sheet.

The bias of the technical sheet

How in this case to identify the origin of the variability between two products considered equivalent according to standard measurements? What measurement will make it possible to qualify the parameter(s) associated with the real difference in behavior between these two products? This problem is that of the representativeness of the measurement.

The issue of the representativeness of the measurement

The question of the representativeness of a measure is crucial, but most of the time considered obvious. Measurement is therefore reduced to questions of precision, uncertainty or repeatability.

If scientific teaching seems to me in general not to really address this question, such a gap seems to me to find its origin at the origin of measurement theories. Indeed, the preoccupation of the pioneers at the end of the XNUMXth century. (Helmoltz, Campbell) concerned macroscopic properties (weight, dimensions) directly associated with a body and their objective was to build a theory of the validity of such a measurement, carried out using a device directly measuring said property.

The representativeness of the measurement (more precisely of the measurand to speak the language of metrology) is in these cases trivial: the property to be measured is intuitively identifiable.

The case of complex phenomena

When is it on the other hand when one encounters a difficulty in the conditioning of a emulsion by dosage in jars? Which property is representative of this phenomenon? Is there only one? By what miracle the viscosity measured under arbitrary conditions - shear in particular - can it hope to be representative, that is to say lead to the discrimination of behavioral differences? And when such a measurement is effectively representative, what is its limit of validity?

Similarly, the particle size of a powder is an important factor influencing its behavior; however, it is not sufficient in most cases either to discriminate or to predict problematic behaviors.

In the case of concrete phenomena, one can “reasonably” expect that certain properties are related to the phenomenon. Reasonably means that the variations of certain values ​​related to the measured property can be correlated with the variabilities observed in the occurrence of the phenomenon. It is a question of investigating to show it. The lack of a clear correlation between the parameter measured and the occurrences of the phenomenon forces most actors to proceed according to the traditional method of empiricism.

Conversely, precise answers to these questions open the way to predictive instrumental methods allowing R&D to anticipate concrete behaviors in industrialization and production, as well as to guide the settings.

Measurement-Sensor, Measurement-Experiment and Parametrability

In contexts where the coupling between material and conditions of implementation is strong - from the storage of powders with its effects of hygrometry or vaulting in silos to applications by spraying or coating roll of a liquid product or pasty non-Newtonian-, most phenomena can be considered complex.

The data produced by the use of measurement standards are therefore not systematically, far from it, representative of the phenomena targeted. In many cases, they indicate at most trends between very different subjects. But this is often insufficient to compare subtle formulation or processing effects.

In order to highlight this limitation linked above all to the methods of implementing a measure -and not only to the instrumentation-, I have proposed in the article “another look at measurement” to distinguish between two types of implementation:

• The "sensor-measurement", aiming to produce a value for a parameter associated with the product by arbitrary and fixed modalities - either natively in a non-configurable device, or parameterized;
• The "measurement-experiment", aimed at mobilizing the responses of matter under controlled conditions according to the parameters available on the instrumentation.

This simple distinction aims to propose a transversal reading grid of instrumental practices capable of recognizing the respective merits of these approaches according to the contexts. The other advantage of this dual reading is to clarify the constraints weighing on each type of implementation modality for effective use.

"Measurement-Sensor"

The "Measurement-Sensor" mode is particularly representative of the implementation of measurement standards, delivering a value following the mobilization of the material or product according to a fixed protocol.

As far as the physical behavior of matter is concerned, when the conditions for mobilizing matter in the instrumentation are not representative of real conditions, the representativeness is generally low, or even zero. The example of the limitations of the use of the viscosity cut developed in article #3 illustrates this case. We could follow the same approach to show the limitations of Hall funnel or tapped density powder flowability tests, such as measurements of viscometry in general and many other techniques.

Conversely, when the instrumentation is configurable and the measurement protocol has been developed and validated in such a way as to be representative of a phenomenon, the "Measurement-Sensor" mode makes it possible to capture the essential characteristics of the phenomenon by a value unique.

To summarize, at the risk of sounding trivial: when we want to have to use only one value to characterize the behavior of a material, this value must be “reasonably” representative of the real behavior.

Develop internal measurement standards

In conclusion, many sectoral standards are used, generally for convenience, both in the framework of normative control and for R&D needs.

Some confusion arises from the fact that measurement values ​​are seen as objectively - and absolutely - representative of differences between products. However, depending on the phenomenon in question, the values ​​measured according to a certain standard may be of no use. On the contrary, in a situation of resolution of problem of variability, the measurement carried out can become a source of problem since it makes pass for identical what is different, thus preventing to qualify what in the matter can be with the origin of the observed variability.

Measurement standards are by definition considered capable of standardizing views, that is, of homogenizing them. In practice, however, Many frictions both internally and between customers and suppliers are clearly associated with the difficulty of building a common representation of the phenomena and the factors that influence them.

At the present time, where Data processing tools are becoming increasingly important in companies, the issue of measurement, without which no data exists, seems to me all the more crucial. “Garbage in, Garbage out” say the Datascientists. In practice, it is not only a question of hoping to detect “problem” data, it is above all a question of avoiding generating them.