Industry 5.0

Industry 4.0

is on everyone’s lips; many have an idea, but rather few have a precise definition. I’ll let selected vocabulary affect you:

Digitalization; complete connectivity; linking of all data; revolution; “processes and decisions will run at unprecedented speed”; systems “control themselves”; “analyze everything”; self-organized production; highly individualized products; comprehensive flexibilization; digital transformation; new level of organization and control of the value chain over the complete product life cycle; a concept, not just technology; unstoppable.

In particular, digitalization, networking, and intelligence are central elements of the descriptions.

We imagine Industry 4.0 like that: extensive data is collected, everything is analyzed, and everyone talks to everyone else. The hopes for “unimagined possibilities” are great and, in my opinion, justified. But I also read out something that I would like to call “displacement”. What should it look like when systems “manage themselves”? Where do we get suggestions for decisions that are being made at unprecedented rates? What do we do with the analyses that are becoming available on an enormous scale?

I believe that in the euphoria, too little emphasis is placed on the fact that neither data nor connectivity represent added value in themselves. Data are not yet insights. Even insights are not yet decisions. Nota bene: even self-deciding machines have to make decisions! And if the number of possible decision alternatives grows rapidly, who can keep track of choosing reasonably from these alternatives? Who can judge whether the machine has planned well?

This is where mathematics comes into play, or more precisely mathematical optimization. It can and must bridge the gap between technological possibilities and concrete recommendations for action. In Production 4.0, too, the core questions are always the same: in what order, at what times, in what quantities, for what customers, in what quality, at what price, on what resources, with what capacities, etc. should we produce when, where, how much, of what? For me, Industry 4.0 means first and foremost: creating opportunities to move away from decisions based (in the best case) on the evaluation of a few scenarios. If, for example, we create the technological scope for customized production through data collection and connectivity, then we should also apply the mathematical possibilities to make the best possible use of this scope. Otherwise, the potential will be wasted.

Descriptive, Predictive, Prescriptive

Today, data is already collected in dashboards and spreadsheets to even obtain an overview of the status quo. People then make decisions based on their vast experience. This is called descriptive analytics, and it has arrived in all companies.

Perhaps you employ data scientists who prepare and evaluate your manifold data, use it to make forecasts and develop a basis for decision-making. This is called predictive analytics. This also includes the large growth area of machine learning. Some companies are already involved in this field.

Ultimately, however, only a few decision alternatives are evaluated in a technically high-quality manner. Does this lead to good decisions? This question can only be answered by prescriptive analytics. Here, based on mathematical models and methods, a decision proposal is calculated that considers the confusingly large number of all alternatives and selects one of the best among them. The mathematical subfield that deals with this is called mathematical optimization. Some people know the somewhat dusty term of operations research. (Prof. Lübbecke discusses possible applications of operations research, such as the division of electoral districts, in the article “The Mathematics of Decisions” in “RWTH Themen”).

Let's say yes to mathematics

Whatever terms we use, Industry 4.0 will confront us (or machines) with unknown complex planning challenges. These should be met with means that are appropriate to the complexity. This also means: Let’s say yes to mathematics and computer science not only as a “service” and “additional work”, but as a central element, as part of the manager 4.0’s toolbox.

When I call for this, I obviously think that this is not happening enough today. I think that this “denial” is also due to a fear that the mathematics and computer science that awaits us will be “too hard” for us. That not only processes and systems are insanely complex, but so are the models and methods for mastering them. That we no longer understand all this and are even regarded as stupid by others. So, is it better not to deal with it at all?

It is true that modern mathematical algorithms can suggest good or even best decisions to humans or machines. Think quite banally of the algorithms in a navigation device that guides you along the shortest route. Such examples are already numerous today in production, logistics, transport, energy, healthcare, etc. It is true that these processes are not easy to understand, even for experts. However, it is also true that the operation and use of these algorithms have become much easier today, because there is outstanding standard software for mapping even complex decision-making and planning tasks. If you really want intelligence, you need mathematics.

So, if I may give some advice today: "Before you “blindly” collect data and connect (or make connectable) “everything”, get advice on the mathematical possibilities of what can be achieved with what data. Collect smartly and properly, with a goal in mind. Don't buy software that doesn't have the “math inside” label on it. Don't settle for anything less than mathematical optimization. Your competitors won't either."

Maybe we’ll soon be talking about the mathematized factory, who knows? Maybe the mathematical revolution will only come with Industry 5.0. But it will come. I’m looking forward to it.