Sie erreichen uns persönlich montags bis freitags von 8:00 bis 17:00 Uhr.
Production Data Being the Oil of the Future
If sentences could collapse from exhaustion due to excessive quoting, the one about data being the oil of the future would probably be one of the high-risk candidates. It cannot possibly be left out of any text on big data applications. Some IT providers have indeed used their power of disposal over gigantic amounts of data to make good money. A number of these providers came to the conclusion that vast amounts of data were valuable per se and that all you had to do was find out how to get to the treasures hidden within—for this is where the real potential supposedly lies.
Always Keep Process and Business Perspectives in Mind
Now let’s talk about production systems. Nowadays, any reasonably modern system is equipped with sensors for just about any relevant indicator down to the last screw and is capable of exchanging the resulting data with other systems. Put a slightly different way, there is a very large supply of data, and you are the one whose task it is to select the part that will generate a benefit.
Naturally, the question as to what constitutes a sensible selection of data cannot be answered without taking a look at the process and business perspectives. Data that generates a quantitative or at least qualitative benefit in some place or other but does so in a way that is incomprehensible is simply not of interest, regardless of how “big” it is.
Looking at the purely technical side of things and what is possible from a communications technological point of view rarely gets you anywhere. Collecting and linking data for its own sake may please IT providers but will lead nowhere in the end. Of course, it is quite easy to create one or two data lakes, quickly flood them with data from a great many tributaries and hope that a gifted math wizard will manage to catch the big fish with his algorithmic nets – provided there are any big fish. Because THAT is by no means a given.
Two Worlds – Two Speeds
As regards technical communication on the shop floor side, a world of real-time communication with very short clock cycles encounters a business world where considerably longer cycles are the rule. Connecting these worlds is a challenging task, but it is becoming increasingly simple to solve these days, as many solution providers focus on this interface and transition layer.
Given the variety of solutions offered, it is no easy feat to find something like a best practice approach. But it is possible to name a number of technical elements that are highly likely to appear in a good solution. A good solution will try to avoid proprietary communication protocols, which means that the protocols commonly used on the Internet will play a significant part in one way or another. From the perspective of the ISO/OSI layer model, the business side involves, from bottom to top, IP, UDP, TCP socket, and HTTP/HTTPS. Applications are fed via REST, and possible data brokers include MQTT, OPC/UA, Apache Kafka and the like. Then we have the traditional middleware functions for data and message transformation in our toolbox. At the top end of the food chain, we have SAP PI/PO, while Node-RED might even be sufficient for less complex tasks.
On the shop floor side, we are faced with a vast number of industry protocols and industry-specific infrastructure. But there is good news here. There are easily scalable gateways that can handle all protocols and procedures commonly used in business systems for almost every combination used at the transition between the system environments. You just have to know what you want or need.
We Used to Call It Production Data Acquisition (PDA)
Most companies already have some amount of experience with data utilization and acquisition from the shop floor in the context of PDA/MDA processes. The automated acquisition of data for the purpose of maintenance is bread-and-butter business for the IT departments of many companies. Recipe management for the process industry is a slightly more complex and time-consuming task. It involves exchanging structured recipe data between SAP systems and process control systems. Generally speaking, linking machines is old hat.
Don’t Believe Anything When It Comes to Digitization and Cloud
The range of new technical possibilities is, of course, a result of the IoT/Industry 4.0 hype. The meta topic of digitalization hovers over all these technologies, and the corresponding invocations by marketing departments cultivate said topic relentlessly. I am absolutely NOT joining in the chorus of digitalization/IoT/Industry 4.0 disruptor doorstep brigades who want to persuade you that your company will drop dead tomorrow if you don’t invest in total networking of everything with everything, machine learning, and smart something or other right this minute.
If, on top of that, someone tries to lure you into the cloud, look very closely, and not just at the prospects of cost cutting, and don’t believe anything they say about the safety of your data and the availability of services right away. Before you take the step into the cloud, it is absolutely necessary to have the risks assessed thoroughly by a neutral expert. Otherwise, you may have lowered your IT costs, but you might have also gambled with your company’s ability to act and survive. This is one of the main reasons why German SMEs are being cautious and skeptical — and that’s a good thing.
More and More Production Data Thanks to Increasing Integration
Of course, machines appear not only as sources of data but are just as often the receivers of data. It would be quite easy to supply a system with all the data from the SAP PPS, for example. Given the increasing integration of systems and processes, it can be assumed that there will be more and more data that a production system can use in one way or another. Even today, modern systems are capable of using planning data on the basis of a medium time frame to optimize their own behavior.
A system will simply “know” that it does not need to operate at the limit that day because the planning data for the next three shifts does not require this kind of operation. This opens up possibilities for operating a system with little wear and great energy efficiency. For example, the system could perform some of the scheduling itself. Generally speaking, the development of modern production systems is characterized by a larger amount of leeway when it comes to delegating and performing tasks relating to production control. Machines plan independently, control the flow of material and can “see” beyond their own system boundaries. What used to be done in the MES, the control station or the planning board in the foreman’s office, systems can now do themselves.
Of course, this provides data-based possibilities for optimization, but nobody should succumb to the illusion that these possibilities are among the low-hanging fruit. They are not. The development of optimization procedures, their mathematical modeling and, ultimately, their implementation are a highly complex undertaking, and various CIM initiatives have failed because they underestimated this task. Many people on the side of industry suppliers have learned something new from their experiences. But it is up to you to decide whether you consider it useful.
Networking and Complexity
Whenever the degree of networking in a system environment increases, so does the complexity of this environment. New possibilities are accompanied by an increase in variety and, with that — remember Ashby’s Law? — the requirement to increase the complexity of system control. One of the things that killed CIM were the immense costs that arose in this context. The cost advantages made possible through networked and automated systems regularly fell victim to the high investments in new control systems.
This symptom is one of the weak points of using machine learning procedures. If it is impossible or far too expensive to program a control system, there may be other AI-assisted procedures that can do it.
Everything Is Nothing without a Clear Vision
Either way, any such project is certainly doomed to fail if there is no clear vision with regard to the objectives of investing in machine integration. The question of a constructive vision of what is to be is the first and most important question you should ask yourself.
The new toolboxes of industry suppliers and IT specialists are filled to the brim and look promising, but that doesn’t invalidate the good old rule that a task should already be present and the tools must fit the task. You wouldn’t just hammer a nail into your living room wall for the sole purpose of trying out a new hammer — or at least not without consequences. Some will say that it doesn’t hurt to practice with the hammer — but then, unfortunately, the next problem comes along in the form of a screw.
Have you already heard about SAP Connected Manufacturing? Watch the video for more information.
Feel free to contact me if you have any questions or suggestions.
– by Mario Lütkebohle, Consultant, NTT DATA Business Solutions AG –