Technical problems and their resolution

What do the manufacture of a lip balm have in common, theextrusion of elastomer parts or the emptying of a silo of powders ? The coupling between matter and process is crucial and its control often poses difficulty. In the event of inappropriate adjustment of the process, the finished product risks being non-compliant; in the worst case, production is shut down.

These difficulties, critical in production, are encountered right from the development and industrialization phases of a new product.. The problems of transposition of new products resulting from R&D, the technical difficulties in achieving a satisfactory adjustment of the product/process, the problems appearing subsequently when everything is "under control", customer feedback or misunderstandings and tensions likely to Emerging between services are all symptoms of the problems associated with a limited mastery of couplings.

Why are these difficulties so common? Are they inevitable? I would like to show here that the Quality methods are not completely unrelated to the fact that such difficulties can persist. I take this opportunity to outline ways of progress.

La universal sequence quality

The more the difficulties are common and experienced by a wide variety of actors, the more it seems useful to seek their origins in the depths of practices and cultures. In this respect, the Quality literature is a living source for tracing the common principles of industrial practices.

Product Development Organizational Sequence

The sequence of technical problems

Behind the presumed universality of this sequence may arise the question of its effectiveness, which is based on the assumption that the development of product functionalities can be done relatively independently of manufacturing conditions on an industrial scale. In the processing industries of “complex” material, the independence between product and process is lacking: the final properties of the product depend on the materials as much as on the conditions of the process. It is through its intimate coupling with the material that a process acts - unlike a process for assembling mechanical parts for example.

These couplings are at the origin of a cascade of operational difficulties throughout the technical chain, from R&D to end use, including industrialization or production.. We can cite production stoppages, customer feedback or implementation difficulties on the customer site, recurring non-compliance, difficulties in scale-up or reformulation, etc.

These difficulties have the particularity of often mobilizing interactions with R&D, thus disrupting the universal sequence (see following diagram). For example, the industrialization stages often involve development cycles in which it is not only a question of testing implementation configurations, but also of adjusting the formulation conditions - which in the logic of the universal sequence are fixed once and for all.

Inter-departmental interactions related to difficulties in the product development sequence

However, such couplings between material and process are not new, nor are the associated problems obviously. Would they have gone unnoticed or simply anecdotal?

Trial-and-error methods at the heart of the problems

It is hard to imagine that such an observation has not already been made by many other observers, a fortiori within organizations, if only because the direct and indirect costs associated with these difficulties, both in terms of consumption of resources (material, human, production tool) as well as deadlines, performance or quality, are considerable.

On the other hand, it seems to me quite plausible that these problems seem inevitable to many, as intrinsic to the industrial approach. One of the methodological reasons for this is that the vast majority of industrial methods, and among them Quality methods, are based on field approaches. Whether solving a problem or improving the efficiency of the organization, everything is conducted by trial and error: this is the heart of the philosophy of continuous improvement.. Deming's famous PDCA illustrates this well: we plan (Plan), we do (Do), we verify (Check), we validate (Act).

We find this type of approach, ranging from intuitive trial and error to more methodical approaches such as experimental plans, both in R&D, in industrialization and in production. At the heart of these practices, technical know-how in particular is central.

The invisible know-how

We essentially associate know-how with the technical act, often forgetting that it also underlies each methodological step, whether it involves analyzing a problem, identifying avenues of investigation, interpreting results or to implement preventive or corrective actions. In the vast majority of Quality approaches, theknow-how most often plays a technical background role, indefinite but essential, specific but always silent.

In practice, know-how is never self-evident, especially insofar as it necessarily evolves - by increasing in the process of continuous improvement or by stagnation or even reduction in environments with high turnover. Similarly, the dissemination of knowledge and the sharing of know-how is not systematic along the organizational chain. R&D, industrialization and production teams are thus likely to encounter various problems all associated with a single phenomenon and to resolve their occurrences independently and by pure trial and error.

Thus, the prospect of "integrating quality into the design" of products and processes or of favoring "right the first time" seems difficult in the absence of consideration of the role and place of the know-how(s) in the organization.

Statistical abstraction

The statistical dimension of Quality approaches also reflects this ambiguous relationship to know-how. Indeed, ofn the mathematical approach, neither the concrete nature of the problem dealt with, nor that of the processes or products concerned, nor the entities to which the mathematical operations are applied come into play.. It is absolutely neutral for the mathematical tool whether the data relate to the manufacture of yoghurt pots or the coating of airplane wings, whether they are valid and representative or whether we add cabbages and carrots or how we interpret the results of the treatment to operate in practice. This is precisely what makes its power. However, the calculator is unable to tell us if we are doing the wrong calculation.

Here too, common sense, know-how, business expertise, knowledge, are essential to oversee both upstream and downstream the validity of data, measurements, results, practical consequences, etc. The slightest gap in the validity of the data can make the use of the mathematical arsenal completely ineffective.

At present, the introduction of machine learning or IoT techniques to statistical approaches does not in any way modify this problem of representativeness of the measurement, which is crucial for the industry and unfortunately considerably underestimated - I will have to come back at length in a future article.

Current issues

At the present time - for about fifty years all the same... - the intensification of international competition and the acceleration of economic dynamics are forcing organizations to bring innovations to market more and more quickly. the time-to-market has become crucial in many fields to seize a market or to position oneself advantageously there.

The increasing complexity of products, through the development of ever finer functionalities, the use of materials that are themselves more complex or the need to respond to regulatory constraints or environmental issues, requires manufacturers to better control developments. The substitution of an ingredient often induces a complete reformulation of the product and a complete development cycle.

Faced with these challenges, trial-and-error approaches are far from effective.. They are a symptom as much as a cause of the operational difficulties mentioned. Behind the strange treatment accorded to know-how in Quality methods, another dimension seems completely forgotten: the understanding of phenomena.

The challenge of understanding the phenomena

One finds in the statistical foundation of Quality methods – for example at Shewhart, pioneer of statistics for Quality Control, or Fischer, that of experimental plans – the origin of a deeply rooted idea: the causes of variability should be considered by default as unknowns. Statistical methods would then be preferred to collect information on phenomena and events whose reality would be essentially statistical.

Concretely, in industrial processes, many phenomena are completely deterministic. By deterministic, it is a question of understanding that a certain configuration of initial conditions (linked to the material as much as to the implementation) induces certain final properties. Statistical variations related to random factors play a negligible role. Worse: the judgment that a factor is random translates in many cases above all a misunderstanding of the concrete factors. Most industrial phenomena can thus be understood scientifically -even if in most cases no equation makes it possible to simply predict the final state from the initial state.

Let's take the example of the manufacture of a gelled product (a gel shower or a dessert cream) and a problem of loss of texture at the line outlet. The texture is based (1) on the presence of ingredients having the capacity to form a physical network and (2) on implementation conditions allowing the structuring of the network. The nature of the minerals, the proportions, the mechanical and thermal conditions of preparation, the thermal kinetics, the other species present will influence the stability of the texture. These influencing factors are known, even if for each particular case, their relative impacts are always to be determined in order to claim to control the phenomenon and its variability.

Understanding an industrial phenomenon means giving yourself a representation of the various parameters influencing the phenomenon as well as their relative impact, but also defining the mechanisms of occurrence of the phenomenon. Such an approach to understanding has largely proven itself in the sciences, so it may seem surprising that Quality approaches rely so little on scientific knowledge and approaches to avoid trial-and-error approaches.

Towards a “behavioural” industrial science?

On the contrary, it is the Quality approaches that have innervated the R&D approaches. We have seen that R&D departments often carry out their work with Quality methods and tools, in which it is essentially the properties of the finished product that are quantified and the questions associated with the process are left unanswered.

However, for several decades, the scientific concepts allowing to build a dynamic understanding of matter have been firmly established. The concepts and techniques of a certain region of science that could be described as "behavioural" - mobilizing in particular physics, mechanics of complex fluids, thermomechanical engineering, rheology, tribology, instrumentation, etc.- make it possible to identify the complex behavior of matter under thermomechanical stress.

The implementation of these approaches for industrial issues is neither direct nor simple, and most often it involves developing specific approaches to guarantee their effectiveness. These approaches make it possible in particular to anticipate the stages of industrialization, to solve problems, to make innovation and R&D more reliable more quickly than any empirical or statistical approach.

There are thus methodological, technical and scientific paths to help industrial organizations progress from approaches that are still essentially based on control to approaches that favor mastery through the understanding of phenomena.

It is another perspective of the Industry of the Future which then takes shape, taking head on the real complexity of the phenomena of the transformation of matter, the importance of know-how and taking advantage of scientific methods adapted to industry.