Maximum flexibility thanks to software-defined manufacturing

Question: What are the key basic principles of SDM?

Riedel: One key basic principle is that thought is given to generalizing manufacturing processes. Take for example a more complex process like the manufacturing of car doors. Why do carmakers need a different door production system for every vehicle instead of a general-purpose system that the door itself provides with the geometric dimensions and the position of the clamping claws for welding? You could produce any door on this system without having to make any changes to the jig as long as you stick to a certain product class. You wouldn't be able to manufacture a refrigerator door, but you would be making your manufacturing process much more flexible by ensuring that the parametric information, workflows, controllers, etc. are no longer hard-coded and instead have properties derived from the variance of the product.
A second SDM paradigm is quality assurance, an area in which you have very narrowly defined quality control loops. Instead of making modifications, you try to optimize your processes in situ by thinking about whether to use a direct or indirect measuring method to monitor quality in the process. You can check the quality of the weld seams during arc welding with the help of structure-borne sound tests, for example. This of course assumes that you have already defined the entire process in the software. You need a digital shadow to know what your patterns need to look like and a digital twin so that you can monitor the actual process or simulate it proactively.

Question: Is it possible to implement SDM without a complete digital twin of the production systems?

Riedel: Yes. Full stop. As far as I'm concerned, the term digital twin is a buzzword. What is a digital twin? It's a way of mapping processes and resources in digital form with a degree of accuracy that approximates reality. The exact nature of the digital twin depends on the type of processes involved. Conventional processes don't pose a problem, but have you ever seen a digital twin model that takes account of temperature drift? I haven't.

Question: What do you mean by temperature drift?

Riedel: It's simple: your process works one way at 20 degrees, another way at 25 degrees and yet another way at 35 degrees. There are already numerous parameters in manufacturing that are well suited for use in virtual commissioning – to use another buzzword. But if the digital twin worked 100 percent correctly, there would be no need to do the last bit of the actual commissioning.

Question: Doesn't the fact that the digital twin is never 100 percent accurate limit the possibilities of SDM?

Riedel: No, but you don't necessarily need it either. Although we don't have the completely accurate models, we do have observations made in the past and can enrich a sufficiently accurate model with historical data. This is referred to as a "gray box", and this is where SDM is heading and also where the journey into simulation technology is taking us.

Question: Does SDM signal the end of programmable logic controllers?

Riedel: I wouldn't put it quite that drastically, but the PLC is being reduced to the elementary processes. Although it is still characterized by special hardware, it is slowly moving in the direction of microcontrollers and is in the process of disappearing into the software completely. The next step is virtualizing the PLC in its entirety and running multiple controllers on a single server. SDM can use this technology to make the entire manufacturing process flexible by specifying which controller is to be used and under which process condition. It is even possible to replace different versions during live operation. This already works even for time-critical operations in the millisecond range.

Question: What challenges need to be overcome when implementing SDM?

Riedel: There are number of challenges involved. One is the long depreciation period for production equipment. Companies have made a lot of investments that run using old technology. Another is the skill set required. That's why we here at the university and in the research community see it as our job to awaken an understanding of digital processes in traditional mechanical engineers and to make IT nerds understand that their solutions also have to work in manufacturing. Achieving the requisite level of training and awareness at every level of a company isn't easy. You need decision-makers who understand that information management is a factor in manufacturing.

Question: Is SDM really compatible with the brownfield approach in manufacturing?

Riedel: You can proceed in a way similar to Industry 4.0 and package the brownfield in a shell model. A machine tool that has no mains supply is given a shell and a gateway at field bus communication level, i.e. it is not connected natively but via an interface instead. You can't use it to control all the functionality of the machine but at least you receive data from the system. With SDM, you can also go and encapsulate the brownfield. But of course, a brownfield limits your options. The previously described solution with the clamping claws won't work if the claws are screwed on and don't have any drives. Which means that you have to invest.  

Question: Does SDM result in additional requirements in terms of cyber security?

Riedel: No more than we're already dealing with in the context of Industry 4.0. However, the further you move in the direction of software, the greater the potential danger if someone hacks the system. If in Industry 4.0 you've specified that you're to receive my measurement data from the machine and someone hijacks the machine, they will receive the measurement data. If you can use the same line to parameterize your machine or even replace all the software, then a hacker could cause a lot more damage. That's why you must go to great pains to ensure security.

Question: How should companies proceed when switching to SDM?

Riedel: Based on our triangle comprising manufacturing, digitalization and business processes, you would start with the processes to see where they are too expensive or not running smoothly rather than starting with production technology or digitalization. You can look at the classic KPIs, like overall equipment efficiency, process efficiency, process costs, number of acceptable parts, etc., and then think about where there is potential for improvement that you can leverage with digitalization.

Question: Are there already companies that have implemented SDM in productive use?

Riedel: There are, but I don't know if I'm allowed to name names. In the automotive industry, it tends to be the suppliers who press ahead because they are subject to entirely different pressures than the OEMs.
And there are a few brave companies in the capital goods industry, i.e. larger plant manufacturers, who see advantages for themselves and their customers.

Question: What benefits do companies expect from SDM, aside from greater flexibility?

Riedel: SDM goes beyond mere flexibilization. You can use it to map completely new and unexpected manufacturing processes. You don't need to know exactly what and how you will be manufacturing in the future. As long as it is physically feasible with the equipment you have, you only need to adapt the software or replace parts of it. Which means that you are completely adaptable to change. This offers enormous potential for optimization, for new business models, and also for AI if its performance capabilities continue to grow.

Question: What SDM-specific research projects are being carried out at the University of Stuttgart?

Riedel: Together with the University of Karlsruhe, we carried out a project involving car manufacturing called SMD4FZI, which received a huge amount of funding from the federal government. In the project, we placed great focus on manufacturing processes in the automotive industry, primarily in the areas of metalworking and cycled variant production, i.e. areas in which carmakers today have incredibly long lead times and, any time a change is made, downtimes of a few weeks. The demonstrator that we developed is in my office at the Institute for Control Engineering of Machine Tools and Production Equipment (ISW) at the University of Stuttgart.