unternehmen & trends DIGITAL 1/2021

containing process parameters which serve as the settings of the individual equipment and components so as to achieve the desired results. This means that there is no specific set of machine parameters you have to use to be able to produce the desired products. Some- times it depends on the conditions that prevai l in the environment whether a process by which a specific product is manufactured works or does not work. Let me illustrate this by an example, for instance, a large manufacturer of plas- tic film. This film is made by blown film extrusion machines. In this process, the film is blown upwards like a big hose so it can cool down and be wound up af- terwards. To this end, certain parame- ters have to be set so the film has the strength required by the customer, but also a specific elasticity. Sometimes, special weather conditions do not allow this film to be produced on a certain day, because it simply does not work. But nobody knows why this is actually the case. There are experienced machine operators saying: ‘Today we do not even have to start working on that product at all. It won’t work.’ But why this is the case remains within the minds or the intuition of the operators. Now our idea is to try to make use of this existing knowledge that experienced machine operators have and to model it. In addition, we also start to collect data from the processes right from the beginning. We do this by means of over-instrumentation by adding comprehensive sensor technology to the plant. In doing so, we collect data and use it to make models and learn about the process parameters. On this basis, we identify the parameters that need to be changed in order to maintain the same high quality at the end of the day. We can only manage to do so by bun- dling the aforementioned competences and this is what we mean when we say that we industrialize immature processes. Is it wrong or absurd to compare this phenomenon of immature processes with the profane activity of making a cake? Sometimes it turns out nicely and sometimes it doesn’t, even though we think that we did it the same way in both cases? Sauer: Indeed, sometimes the dough ris- es and sometimes it doesn’t although you did it in exactly the same way in both cas- es. And still you do not get the end prod- uct you intended to have. Our working assumption is: By means of over-instru- mentation, i. e. the excessive installation of all kinds of sensor technology at vari- ous points within the process, we want to get evidence, on the basis of data-driven modelling, what parameters are actually responsible for certain product proper- ties. As soon as the process is running in a stable, reliable way and we have learned what process parameters are responsible for what measuring results, we can re- move the surplus instrumentation again. After all, the process has to be economic, too, as sensor technology, data genera- tion and processing are rather costly. For this reason, we only want to have exactly the sensor technology needed to control or regulate the process in a targeted way at the end of the day. Over-instrumentation is thus only a merely functional method to identify the parame- ters responsible for particular quality fea- tures within the overall process. At the Karlsruhe Research Factory, you also experiment with the issue of “em- bedded scientists”. What exactly is this? Sauer: We define embedded scientists as employees from other enterprises