is Head of the Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW) at Stuttgart University. He studied engineering cybernetics and for over 20 years has been involved with virtual validation methods in product development and production.
is Sales Manager at Bosch Connected Devices and Solutions. The specialist field of the certified technician in communications
engineering, who also holds a degree in business administration, is MEMS – Micro Electrical Mechanical Sensors, which process
both physical and electronic impulses. – “Machine and process monitoring using MEM sensors” will also be one of the topics
at the Grinding Symposium 2019.
studied electrical and industrial engineering and joined STUDER in 1979, where for many years he was responsible for business
in North America and was later CEO of the Cylindrical Grinding Technology Group. He is now CTO and Chairman of the Board of the UNITED GRINDING Group.
It can be assumed that the importanceof process data will also increase in the machine tool industry. How well are companies prepared for this?
It’s not as if we’ve only just begun handling data. But of course it’s very trendy to talk about data right now. I have been concerned with digitalization throughout my professional career. So I’m a bit of a digital native. Today people always assume that digitalization is something completely new. But what’s new is actually that the awareness of digitalization has increased. What we also mustn’t forget is the consideration: What does it do for the customer? Which benefits can we provide with it?
Exactly, much has already been achieved in terms of automation and digitalization. But in recent years and undoubtedly also with Industry 4.0, the development has been given fresh impetus and this has also resulted in the attempt to network entire processes and make them more transparent. Bosch is in a luxurious position: We have a large number of manufacturing plants and the ability to support our processes with sensors produced in-house.
We are now in the second wave of digitalization, which is based on networking. In my experience however, in machine manufacturing
many companies are still close to zero. They could learn something from the automotive and aviation industry, which are a lot further on. The aerospace industry is forced to work with data, because by law it must ensure that everything it produces and develops is documented.
What type of data is usually collected nowadays in industrial production processes?
Basically, data is collected with regard to whether a product is fault-free or not. If we refine this, there is also the topic
of measuring data and measuring records. And in the next step we can add the data required to perform condition monitoring, predictive or prescriptive maintenance. But the crucial point is: The more intelligent my data processing system, the more my data will make sense afterwards. Pure data acquisition is no use to anyone.
In our field we have the machine data on the one hand, which documents the condition of the machine. This enables condition monitoring. On the other hand, the process data is important for the customer because it is based on production know-how.
The triangular relationship between end user, machine manufacturer and sensor manufacturer must be shaped so that the end user can see the benefit of sharing his data.
And in terms of data acquisition, I recommend collecting more rather than less data. Storage capacity and computing power cost virtually nothing nowadays.
So is too much or too little data collected today? Is the volume of data collected already sufficient to enable reliable predictive maintenance?
I would say that it’s currently not enough. At present data is only collected locally from an isolated machine, or a section of
a production system. This is insufficient. The present scenario of predictive maintenance is generally very engineered. Naturally it would be fantastic – but is by no means a matter of course –, if I as the manufacturer could see the machine after six months, or one or two years, and check whether the calculation which I used when designing the machine corresponds to reality.
We have a unit which has been measuring data for ten years. It is very difficult to access the data, however, and only possible when maintenance is required on the machine. Yet from this data we can identify deviations from a machine’s standard behavior, for instance.
There is a critical data volume from which reliable statements can be made. If the machine manufacturer has access to
the data of many customers, and if he has large volumes of data, this becomes significant. And this also brings the advantage
for the customer: If he puts his data into the manufacturer’s pool, he can also benefit from other customers’ data. And if the manufacturer sees a pattern the next time, we already know what is behind this from the volume of data which we have collected. In order for us to evaluate this, the value of data needs to be quantified. The benefit is then immediately apparent and there is an incentive to participate.
How can a data-based quality management system be implemented under the current conditions?
If I want to implement a data-based quality management system, the underlying data must naturally be of high quality, and I
must also validate that this data is correct. To guarantee this I must perform a preprocessing, i.e. get closer to the sensor and qualify
the data there first of all, in order to then make quality statements. The question here is how quickly you can gain an insight from the collected data and how quickly you can then initiate measures based on this insight. I think that we face rather bigger hurdles here.
The company group offers digital services under the label of UNITED GRINDING Digital Solutions™. Which of these are already based on data-based production processes?
One of the services we offer is the Production Monitor, an important component of UNITED GRINDING Digital
Solutions™. It provides a real-time analysis of the machine and, with its 24/7 monitoring, enables optimization of the machine’s
availability and capacity utilization as well as early identification of imminent production backlogs. This belongs in the broadest sense
to the field of OEE, Overall Equipment Effectiveness, i.e. the representation of how effectively a system is working. The Production
Monitor also provides a point of entry to further applications. We equip our machines with sensors wherever it makes sense, in order to identify weak points. This data is then fed into the Production Monitor.
If the importance of data in the production process increases however, one question becomes more and more crucial for the company: How can I combine data specialists with process experts among my employees?
All machine manufacturers should receive a short basic training course in IT. When a machine manufacturer enters the
labor market today, he has little knowledge of IT. And an IT’er knows nothing about machine manufacturing. At the moment companies have to think about how to get the employee up to speed in the other discipline. But we have a fantastic dual training system, which should make this possible. Stuttgart University was the first to create a professorship offering exactly this combination. Each person receives a basic understanding of the other person’s work, this is important so that both can then program a production line together, for instance.
To what extent are data-based production processes necessary to produce small lot sizes? The buzzword is customizing.
Lot size is a huge topic. The most famous example, which has received a great deal of publicity in the press, is sportswear manufacturer Adidas, which produces individual shoes. If one takes a closer look, these are actually not so individual. They are simply
shoes which consist of several modules, where you modify two or three parameters and the design. I still wouldn’t call that an individual product. Genreally with small lot sizes the challenge is to achieve 100 percent quality straight away. And naturally you can
only get that if you operate a data-based quality management system.
A lot size of one in the production process is an issue for us, if our customer’s customer only requires a workpiece as a one-off. The biggest challenge then is how you achieve the desired quality straight away. It’s a question of data, but also experience. Another question is how quickly I can retool my machine. We have been working very intensively on minimizing changeover times for many years.
Changeover and machine set-up are also issues in large scale production. With car doors, for instance. They may only be
doors, but they all differ depending on manufacturer and model. The challenge lies in handling the data: How can I store the setting
parameters of the individual variants, so that the production system can quickly change between the models when needed.
Sensors collect data in industrial production processes. What requirements are placed on these sensors?
From the manufacturer’s point of view it is important that sensors are robust, reliable and standardized. So that I don’t have to constantly reinvent the wheel. And availability is also important. It needs to be ensured that I can still work with a sensor or an interface even after five or more years.
We have been using an acceleration sensor on every machine for over 40 years. We developed and manufactured this sensor ourselves and called it Sensitron. It enables us to detect when the grinding wheel contacts the workpiece. The period before this is known as air grinding, which is naturally not efficient, so you must approach the workpiece quickly.
What makes MEMS sensors, i.e. Micro Electrical Mechanical Sensors, different? What advantages do they offer in comparison to conventional sensors?
You can fit a great deal of functionality into a small space on MEMS, i.e. around five or six sensor principles, which saves a lot of space. As well as being active from the sensor perspective I also have the option of preprocessing the collected data, in other words: Bringing intelligence to the sensor.
MEMS sensors don’t just process electronic impulses but also mechanical ones. What are these in industrial production processes?
If we define industrial production processes rather more broadly and integrate logistical processes, then there are numerous possibilities for using such sensors on machine systems. This relates to the machine’s position. It is never good for machine tools to be at an angle or installed in a twisted position. This is information which I can record with a MEMS sensor. With electrical components such as pumps or drives, for instance, I can record acceleration values. Other values are rotation rate or temperature. What is important to me is this: As sensor manufacturers we are just the enablers. We aren’t the ones who generate the information and shape the benefit. Without the feedback and integration of the machine operators we are toothless paper tigers.
When transporting our machines we can use acceleration sensors to determine exactly which vibrations the machine is exposed to. During transport by truck and ship there is a significant risk of vibrations. During machine operation the sensor-based monitoring of roller bearings, for example, is of great importance. These are vital wear parts, which are exposed to high dynamic loads. And if we record parameters such as temperature, vibrations and speed on these bearings, we can draw conclusions about the technical condition of the bearings and estimate their life span.
The complete acquisition and representation of a production system’s data would be a digital twin. If you connect this twin and the production system with a feedback loop, they could control themselves as an autonomous system. How realistic do you consider this?
Extremely realistic in the distant future, I would say. Just give us a bit of time. As things stand today this would be expensive and completely impractical but the notion is a delightful one.
When we redesign a machine we already have a digital twin, as it were. From the CAD drawings we can generate simulations and assess the dynamic behavior of the machine, for example. We can thus optimize a machine which has just been developed. And there are also opportunities for autonomous production control: Grinding is always the last link in a process chain. This makes it possible to assess the previous process steps. The grinding machine could say: Hello, turning machine, why don’t you make your allowance a bit smaller, then I wouldn’t have to grind as much. For this, however, we need the ability to communicate; one machine must understand what the other one is saying. We are therefore working with the Association of German Machine Manufacturers (VDW) on a so-called VDW connector and on a data formatting definition for the OPC UA machine protocol. Our aim as a company group is to support our customers in this area too, for instance if they deal with frequent production changeovers.
How can data security be guaranteed when generating data in the production process?
We have implemented a Security Engineering Process in our product creation process. This means that at each milestone of the development process we look at which data is collected and which conclusions can be drawn.
We use standardized systems, which transmit the data in encrypted form and are also TÜV-IT-certified. The security of customer data is our top priority.
Who owns the data? And how must the assignment of user rights and roles be regulated in relation to this data?
The data initially belongs to the operator of the machine. However, if the machine manufacturer – as we have already seen –, can obtain benefit from the data for both sides, both should consider how they can work together. I think everyone now realizes how valuable data is.
The system which is installed on the machines and which collects data is naturally our property. This software contains the know-how of the UNITED GRINDING Group. The data which is collected is the customer’s machine and process data. This naturally belongs to him, and we do not access it ourselves. If the customer wishes to share this data with us within a regulated framework, obviously we can think of many things we can do with it for his benefit. Enhanced with artificial intelligence and machine learning, there would certainly be enormous potential here for process optimization in future. Offering such comprehensive solutions might also mean a different business model for us.
Offering digital solutions often results in changes to the business model. For instance, we had a case where we provided a GPS tracker. A customer wanted to track his supply chain, but he didn’t want to buy the hardware and software, just have the service. You then go from being a producer to a service provider and have to consider what the business model is here.