The footprint of machines and systems is getting smaller and smaller. Integrated components and systems must adapt to this trend. What requirements do drive systems have to meet as a result?

Sebastian Markert: In principle, the use of materials and resources in a drive system must continue to decrease while maintaining the same performance. This is achieved through a higher degree of integration: the more individual components are brought together and coordinated, the better. Our aim is also to offer additional functions that the user does not expect, but which will make up the USP in the future. We have various functions in the pipeline that enable data evaluation and also allow certain compensation algorithms to be implemented or condition monitoring to be introduced. This allows the drive as a whole to be made smaller and more compact, while at the same time increasing overall sustainability.

Can individual components be made so much smaller without compromising their robustness and load-bearing capacity?

Mathias Blaskovic is responsible for new product strategies, among other things, in Product Management Precision Gearboxes at Sumitomo Drive Technologies.

© Sumitomo

Markert: Of course, materials also play a role when it comes to weight. That's why this is on our agenda.

Mathias Blaskovic: We are already working with different materials and are investigating these materials in order to make our gearboxes fit for different requirements and areas of application. For example, in some applications in the food sector, components such as our gearboxes must not rust, so manufacturers have to consider components made from alternative materials such as stainless steel or titanium.

Precision drive technology in robotics - The role of software

What role does software play in terms of higher performance and energy efficiency?

Markert: Software helps us to fully exploit the potential of our drive unit. With intelligent control technology, i.e. an intelligent driver, we can regulate and compensate for many special features of our gearboxes and find a perfect match. This should be done ex works so that the user receives a plug & play solution. For example: Imagine a drive axle consisting of a gearbox, motor, a control unit and optionally a brake, which we combine as an actuator. At the factory, we adapt the control model directly to the individual components so that inertia and friction, for example, are compensated for. This means that the driver is tuned to every detail of its axis and negative effects can be eliminated.

Can you explain this using the example of a robot?

Markert : It's difficult to explain on an entire robot, but easier on a single axis: In the teaching process, I specify a target position. The motor moves to this position and the controller constantly compares: where am I, where do I want to go, where am I, where do I want to go. At some point it gets there, but it still has too much torque, too high a speed, so that it overshoots the position. It readjusts this overshoot and settles down to the target position; the robot settles down, so to speak. An intelligent control system knows the friction and other effects and parameters of the axis. As soon as it approaches the target position, it removes torque and moves smoothly to the position. It is effectively in position x milliseconds earlier and the robot can begin its work. Of course, we are talking about milliseconds here, but in terms of the cycle, the time adds up, especially as up to six axes are used in a robot.

"Gearboxes have almost reached their technological limits."

Robots are increasingly being equipped with vision systems to independently detect and pick up objects and their position. How is real-time control of the axes realized in such a case?

Markert: We offer an open interface via which we can integrate additional components. Let's stick with the example of lightweight robots: at the end of the last axis, vision systems can be integrated directly via the interface mentioned. If the control master is programmed accordingly, information from the vision system can be taken into account and the axis can be continuously readjusted.

Software is essential in these applications. Does Sumitomo develop this itself?

Markert: In this case, we are dependent on the manufacturer of the control technology.

To what extent is the development of new drive technology influenced by digitalization?

Blaskovic : Gearboxes have almost reached their technological limits. Only with a great deal of effort can performance be improved. In addition, gearboxes in and of themselves are very simple components that are now rather subordinate to the USP and are usually bought in. This makes fully integrated systems all the more important. Users can simply plug and play them into a machine and concentrate on the important things, such as digitalization. In principle, the user takes development to a new level, because it is no longer just about the individual components, but about the overall system and how its parts communicate and coordinate with each other. Of course, the individual components have to be smart in some way, and as a drive supplier we have to deliver that.

This is where sensor technology plays a role. How do you integrate sensors into the drivetrain?

Markert: In the cobot sector in particular, the current state of the art is to integrate one or more encoders in the drive. This provides me with a lot of information, for example if my drive behaves differently than expected. If, for example, the relative angle between the input and output shaft is outside my expected range during operation, this indicates an overload, which causes tension in the gearbox and may even cause damage. I can prevent this by removing the current from the motor or issuing an error message, for example.

The drive as a sensor

What information - classic operating parameters and other data - is particularly interesting for developers and users?

Markert: Actual torque information at the output is interesting because I know exactly how many Newton meters my output axle sees as counter torque. In a current project, we are working on an integrated torque sensor for our gearbox that dynamically records real-time information about the torque in the output as well.
One application example here would be a surgical robot that performs a biopsy on the spine. Without dynamic torque detection, the distance from the skin surface to the bone is measured using an X-ray image or another technique, the needle moves to this preset position and removes the sample. If I have measured myself in advance, this can be uncomfortable for the patient. With the additional torque sensor, the immersion depth is irrelevant. This means that I switch to torque control: As soon as the guided needle reaches the spine, the torque on the output will increase. The system switches off at a defined threshold value. This means that the robot is in exactly the position it should be in. Of course, torque sensors are already available on the market today, but in terms of price and size class, they are still a long way from what we have in mind. The torque sensor directly in the gearbox makes it space-neutral and, in a second step, significantly cheaper.

"Both the power density and the integration are at the highest level for us."

Real-time information about the torque is also interesting for comparison with the control models; for example, the friction of a gearbox changes over its lifetime. With the torque information, I can continuously adjust these models to the actual state and make optimum use of the real drive throughout its entire service life.

How do you develop a modular principle with standard components that then fits the variety of customer-specific applications?

Markert: Our gearboxes form the basis. The motors and brakes are matched to the gearbox. Additional components, such as encoders or even the complete driver, can be added on a modular basis. The housing follows at the end. It is also possible to configure various accessories as an option. The user can now assemble an axis from the elements on offer to meet their exact requirements and get exactly what they need for their application, no more and no less.

Blaskovic: However, we don't just develop with performance in mind, we also make sure that integration is as comprehensive as possible. This means that there are no superfluous components such as bearings in the system and that both installation space and weight are optimally utilized. Both the power density and the integration are at the highest level with us; you can't achieve this by buying individual components.

How do you approach new markets in development?

Blaskovic: Every new market has its own requirements. Our aim is to check whether we can place technologies from our existing portfolio there with adaptations. Take the food industry, for example: we already supply numerous units, for example in the transportation sector. The closer our products get to the end product "food", the higher the requirements become. Here, for example, no industrial oils may escape from the gearbox, as otherwise the goods would be contaminated. Our products meet the technical requirements with special sealing elements and food-grade lubricants. We meet industry requirements such as resistance to acids and alkalis as well as non-corrosive surfaces by adapting the material of our gearboxes to these requirements.

What role does the digital twin play throughout the entire life cycle?

Blaskovic: The digital twin is often used to increase efficiency in production: A production process is simulated in order to be able to produce a component faster and more efficiently or to increase machine utilization. This target/actual comparison will become increasingly important in the future. As a component manufacturer, we are required to find suitable interfaces for this.

AI in drive technology - hype or real help?

How does Sumitomo use artificial intelligence and how useful is it?

Markert: You always have to ask yourself the question "How do I use artificial intelligence? What am I doing and what are others doing with artificial intelligence? AI is a topic that moves the whole world. It's clear that the more I can integrate into a computer, the more I can take a technology to the extreme.
In development, we use artificial intelligence to optimize components. A classic example is the topology optimizer of FEM software (finite element method), which optimizes components according to the prevailing loads. Of course, artificial intelligence is much more far-reaching. As a rule, we only talk about it when machines can learn for themselves using neural networks. We do not yet use this form of intelligence in development itself.

Are you talking more about data analysis or artificial intelligence?

Blaskovic: Big data is a little bit the key to everything. In fact, it is a challenge that was not even recognized a few years ago. It's no longer about simply collecting data at random, but about collecting good quality data in a targeted manner and making it available for specific applications. Only then can artificial intelligence be developed in a meaningful way. If you can't see, hear, smell or feel anything, you can't draw any conclusions about your environment. The situation is similar with artificial intelligence: sensors and data are its sensory organs.

Is the topic still being 'hyped' too much?

Blaskovic: That's difficult to judge. I think the development is so rapid that society needs time to come to terms with it. There are certainly some areas of application in which AI works, but rapid implementation is also a question of company size: a start-up will be able to do this more quickly than a group with established structures and processes. After all, it's not just about the technical requirements; the employees also need to be brought on board. However, I strongly suspect that younger generations will adopt new technologies more quickly than someone who has been doing something in a certain way for a very long time.

Markert: Artificial intelligence needs to be thought about from the front. Let me put it this way: there is a bottom-up and a top-down approach. The hype surrounding AI is more top-down, which can lead to AI being discussed as an extreme. In other words, robots build themselves and replace the worker. Top-down also means that we offer users the deluxe package with extensive functions at a very high price. Whether they need the functions is another question. We tend to think of artificial intelligence from a bottom-up approach: what do I want to know, what do I want to achieve? And then you start to make a selective choice of information. The advantage of this approach is that we can solve specific problems for ourselves and our customers. This is where the modular principle comes in again: we offer various functions from which the user selects the right one.