Mr. Munz, if you look back 20 years: Where were we then in the evolution of automation technology?

Heinrich Munz: 20 years ago, we were discussing three key questions: Will the industrial PC replace the PLC? Will Ethernet-based fieldbuses replace conventional fieldbuses? And thirdly: Will there only be one Ethernet-based fieldbus standard that everyone agrees on?

Has the industry found answers?

Heinrich Munz: We know the answer to question 1: PLCs and IPCs continue to enjoy a harmonious symbiosis on the plant floor, with each device contributing its specific strengths for the respective use case. In some cases, the boundaries between IPC and PLC are also blurred in pure software solutions, so-called soft PLCs or PLCs with integrated visualization.

Regarding point 2: Even 20 years after the introduction of Profinet, we have only just reached break-even in 2018 - in other words, we now have roughly the same market penetration for Ethernet-based Profinet as for Profibus. - Once a technology has been introduced in our industry, it does not disappear so quickly. This also leads us to the answer to the third question: at that time, the time was right to agree on a common fieldbus standard due to the transition to the new Ethernet technology, but it could not succeed because there was no general and neutral standard for real-time capability for Ethernet on the one hand and no company-neutral information modeling technology on the other.

What happened instead was that some automation providers saw the opportunity for competition and their own USPs in the development and distribution of self-developed real-time extensions to the standard Ethernet, including device profiles. What we then experienced was the continued madness of the fieldbus war for the industry in terms of business and economics. The ultimate irony was the IEC standard 61158, in which no fewer than 19 incompatible and non-interoperable fieldbuses were 'standardized'. Thankfully, this has changed recently: With OPC UA and TSN, we now seem to have a single information and communication standard for the industry.

Did you expect automation to develop as we see it today?

Heinrich Munz: On the whole, yes - but I was very wrong about the speed of implementation.

Why was it that the speed of evolution was completely different from what you imagined at the time?

Heinrich Munz: One reason was and is certainly that the major automation providers and their customers - quite rightly - want to secure their return on investment for as long as possible. The innovation cycles in our industry are simply much slower than in the consumer sector, for example.

Another reason for the slow acceptance and implementation of IT in OT was and still is the considerable cultural difference in the way these areas think and act. While topics such as agility, fail fast, fail often and a start-up mentality are commonplace in IT, waterfall models - such as zero error tolerance, continuous improvement processes (CIP) and lean management - continue to dominate in OT.

Ultimately, it is also simply a question of qualifications. OT personnel who are trained to program PLCs and commission fieldbuses are not yet qualified to properly install and operate IT networks internally and possibly even externally to the Internet with all the security measures that need to be observed. I see this as one of the biggest hurdles to the rapid introduction of Industry 4.0 in manufacturing companies.

What are the future key technologies for mechanical engineering?

Heinrich Munz: Essentially, I see five key technologies: Mechanics, electronics, firmware / software, information / communication - including wireless - and services. Up to now, mechatronic component thinking has dominated in this sector. The system thinking that the customer's component or machine is only ever part of the customer solution is often missing. And service is only understood in the context of maintenance and warranty cases.

To remain competitive in the future, machine and plant manufacturers need to rethink. Of course, mechanics must continue to play a central role, as it covers the 'physical' part of 'cyber physical systems', which is what Industry 4.0 ultimately represents as a further development of Mechatronics 3.0. However, mechatronics should no longer be seen as an end in itself, but as a means to an end - for new business models.

Thanks to Moore's Law, the performance of electronics will continue to virtually explode, doubling every 18 to 24 months. Software, on the other hand, must play the new leading role. 'Software is eating the world', as Marc Andreesen, founder of Netscape, once put it, or 'software-defined machines' are the new buzzwords. Anyone who has ever installed a software update, like a smartphone app, in their Tesla and thus received a car with completely new capabilities, such as automatic lane keeping or autonomous parking, will be familiar with this.

However, in order for these new 'super brains with mechanical attachments' to be able to bring their horsepower to the Industry 4.0 road, they need to be capable of information and communication. And not just to fulfill their control task as before, but also with higher-level IT systems in the so-called edge on the plant floor, as well as with IT systems such as MES, ERP and the cloud. Information modeling technology plays a crucial role here: the machines must be able to provide semantic self-disclosure and thus describe themselves so that the communication partners know who they are dealing with and what data and functions the respective machine offers.

And this brings us to the fifth key discipline: services! And not just the familiar maintenance and warranty services, but completely new services and business models for machine manufacturers in connection with their machines. All the way to 'Machine as a Service', whereby the means to an end - the machine - remains in the possession of the manufacturer, it is merely provided to the customer and the customer benefit is billed directly. This is because the customer is not actually interested in a robot, but only in its movement performance. So why sell him a robot and not the movement itself? The latter can be transmitted to the manufacturer in a tamper-proof manner using Industry 4.0 machine communication, which automatically generates an invoice at the end of the month for the number, distance, weight, speed, etc. of the robot movements performed. The responsibility for financing, commissioning, trouble-free operation, servicing, replacement and return remains with the machine manufacturer, who is paid accordingly for this new type of service. In this way, the machine manufacturer escapes the unstoppable fundamental trend that every good that is in high demand for a long time will sooner or later degenerate into a commodity with high competition and low margins.

Where do you see the challenges and opportunities for automation technology in the next 10 to 20 years?

Heinrich Munz: For me, the biggest challenge is and remains the transition in automation technology from the previously dominant mechatronics and therefore component-centered thinking - mechanics, electronics and a bit of software - to cyber-physical systems thinking - i.e. mechanics, electronics, a lot of software, machine information and communication as well as services. We need to move away from mapping the data and functions of a machine in bits and bytes on narrow fieldbuses and instead adopt powerful tools from IT, such as information modeling and broadband real-time communication.

The next step is to use these IT tools accordingly. For example, all automation devices must be given a semantic self-description using information modeling. Fortunately, with OPC UA, both the necessary modeling and communication technology has already been found and selected with great consensus for Industry 4.0. The VDMA has taken on the task of organizing the creation of manufacturer-independent semantic self-descriptions of machines. Who else? Only the machine manufacturers themselves are in a position to fulfill this task properly. The previous approach, in which the fieldbus associations took on this task, was not the right one. After all, it is the machine and not the transport vehicle for communication that is important.

Of course, all this has a lot to do with the acceptance and qualification of the employees concerned. It remains to be seen whether and how quickly the 'old hands' will be able to adapt, or how quickly new generations of employees will follow. Unlike banks, insurance companies or the consumer world, for example, OT has neglected to make use of the blessings of IT, including its security, for around 30 years and now has to make up for this lost time.

"The edge level, distributed ledger technologies and artificial intelligence will be the exciting topics in automation over the next few decades," says Heinrich Munz.

© Computer&AUTOMATION / Heiko Stahl

What topics do you expect to be important for automation in the future?

Heinrich Munz: In addition to the semantic self-description of devices and machines, there is the edge, i.e. the interface on the plant floor between the automation devices and the higher-level IT systems. The edge will play a very large and important role, as it is fueled from two sides.

Firstly, the cloud is decentralizing downwards into the edge, because the classic cloud, which was made for pure software applications without physical devices, has various disadvantages in connection with the physical things of the IoT, such as 'too far away', too high data transport costs, too little Internet bandwidth, too slow response times, too little reliability even at 99.5 %, operators' fear of giving the data out of their hands, fear of possible new attack vectors regarding sabotage and espionage.

Secondly, tasks coming from the devices that are currently solved decentrally in the devices will be centralized in the edge. For example, data filtering, data pre-processing, data concentration or the programming and configuration of devices such as robots, which are not component-related but system-related. One example: In terms of system-related solution thinking, it is not optimal if several robots, which all have to perform a common task together with corresponding process controllers, all have to be programmed or configured individually in order to then synchronize with each other via a few bits and bytes on fieldbuses - and everything is then also developed with different engineering tools in each case. It would be much easier if all components, including robots, process controllers and I/Os, could be programmed, configured and then controlled for this joint task from a single central engineering tool. And this will be one of the tasks for the Edge.

We also need to see which of the following new technologies will really take hold in industrial automation: New human machine interfaces such as smart phones, tablets, data glasses with augmented reality. Or even more visionary: implants in the retina, in the ear or under the skin for ID purposes, for example.

Another important topic could be distributed ledgers. This family of technologies is better known under the brand name blockchain, which has gained dubious fame as a cyber currency, but is only one of many - such as IOTA and C-Chain - possible types of implementation and application. The basic principle is a distributed, virtually tamper-proof database in which entries cannot be changed once they have been made. A community determines online who is allowed to enter something, creating a kind of consensus-based administration. This can be used in our industry for all use cases, whereby reliable evidence is important, such as smart contracts, i.e. forgery-proof instructions, forgery-proof logbooks and machine 'flight recorders', which can store data in order to be able to process old and new business models on this basis. Possible applications: Proof of warranty, liability claims, pay per use.

And last but not least, artificial intelligence - AI for short - and related topics such as machine learning and deep learning will play a major role. It is not yet possible to estimate what impact all of this will have on industrial automation. The data lakes that these technologies require are still too small. At Kuka, we are seeing the first promising results from the machine learning family, although 'machine' is taken literally: Robots are no longer programmed down to the last detail, but after programming the rough framework, it is left to the robot to acquire the final, optimizing details itself through repeated trial and error. For example, flipping a half-full bottle so that it comes to rest on the bottom of the bottle. This is called 'reinforcement learning'. It becomes particularly exciting when several robots learn tasks together, such as 'reaching into a box'. As soon as a robot has learned something, it does not keep it to itself, but passes on its knowledge to other robots, for example by sharing it in the cloud. As a result, the learning process takes on an exponential speed. Google has tried this out with some Kuka robots, which were supposed to learn to randomly pick up everyday household objects from boxes. But beware, today the following still applies: "If it's in Python, it's machine learning. If it's in PowerPoint, it's AI."