Intralogistics Specialist TGW Anticipates Problems with Predictive Maintenance

• TGW relies on condition monitoring with the multiple award-winning picking robot Rovolution

• Digital twin as a central innovation in the Industry 4.0 era

(Marchtrenk) As part of the Industry 4.0 development, downtime of intralogistics systems is supposed to belong to the past, as the condition of critical components is monitored. The challenge is to continuously optimize system availability, emphasizes Dr. Maximilian Beinhofer, Head of Cognitive Systems Development at intralogistics specialist TGW, in an interview:

What are the disadvantages of performing maintenance only when a component fails?

Dr. Maximilian Beinhofer: In such a case, we are talking about corrective maintenance. This can lead to a machine or system being partially or completely out of operation. This means: Availability is reduced, which can result in economic disadvantages for the user. Another problem: The fault may need to be diagnosed first, which costs valuable time.

One can alternatively simply replace wear parts regularly…

Yes, this is called preventive maintenance. It involves establishing certain maintenance or replacement cycles based on experience. This guarantees very high system availability. The downside: Costs are higher because parts are replaced that still have wear reserves. The great art lies in finding the ideal timing – both for the maintenance provider and the maintenance recipient. A good solution is therefore condition monitoring and predictive maintenance. Based on so-called digital twins, it is considered one of the central innovations in the field of Industry 4.0.

How does predictive maintenance work?

Using condition monitoring of components via sensors, one can represent in the software whether a problem is looming. Ideally, this happens in real-time or with minimal delay. The core of our approach is: With smart algorithms, that is, methods from the fields of machine learning and data science, we at TGW intelligently network or fuse existing data from the sensors so that we can make very accurate statements about the condition or wear of components. This saves costs because we do not need to install additional sensors.

Do you have a practical example?

 In our award-winning picking robot Rovolution, we measure the vacuum condition in the gripping device. If, for example, there is a pressure loss due to dust in the environment, we see it immediately and can react.

How do you deal with older systems that do not have the necessary sensors?

Additional sensors, for example, for measuring vibrations, can be installed. Depending on the size of the system, a few to more than a hundred sensors may be needed, which is why a cost-benefit analysis should be conducted beforehand. However, it is generally true that existing intralogistics systems can be retrofitted.

What is the difference between predictive and prescriptive maintenance?

The two approaches build on each other. Predictive maintenance requires condition monitoring as a basis. It is not enough to know whether a sensor is occupied or not. It is about how far the wear has progressed. Once this data is available, predictive maintenance software can create a forecast that a component will last approximately three more months from a certain value. Prescriptive maintenance then advises what to do in three months.

What are the main advantages of predictive maintenance?

In principle, it is about optimizing system availability at the lowest costs. Additionally, the feedback loop is continuously improved. Algorithms ensure that the self-learning system is permanently optimized.

In which areas is predictive maintenance useful?

In principle, everywhere in a system. However, the connecting elements are the most interesting. For example, if one of ten picking workstations fails, 90 percent of the picking capacity is still available. However, if a sorting system fails through which all goods pass, then the failure means immediate downtime.

What are the challenges of predictive maintenance?

On the one hand, it is about generating the greatest leverage with the least effort. On the other hand, the technical challenge is to utilize the networks of the system so that the necessary data can be transmitted to the predictive maintenance software. The third challenge is the feedback loops. If problems occur in the conveyor technology, technicians on-site must report them. As a manufacturer, one must develop intelligent methods to ensure that feedback is provided immediately and is also machine-readable.

How is this ensured?

To train the algorithms, one must know exactly when maintenance occurred and what was done. Otherwise, the system believes that an improvement has occurred on its own. This report must not be a freely formulated text from the technician. It must consist of standardized answers from a drop-down menu because machine-readable data is needed to train the machine learning system. At the same time, the feedback loop must be quick and easy to use so that the maintenance technician can provide feedback promptly.

For which modules is TGW developing predictive maintenance?

Condition monitoring is already available for the picking robot Rovolution. In parallel, we are developing a special cloud solution for data collection and processing. In principle, the goal is to capture all data from mechatronics to IT in the future – of course, adhering to GDPR regulations and data security. We collect data from multiple customers. This has the advantage that a new customer benefits from the data of existing customers and receives advice from the software on what to do to optimize their system. At the end of the process is the digital twin. One can not only analyze what happened in replay mode but also see in real-time what is currently happening. In a further step, one can also look into the future and make predictions.

How will the demand for solutions in the field of predictive maintenance develop?

The topic is currently popular. I expect that in five to ten years, only systems that offer this service will be sold. For large individual machines, it is already common to use a vibration sensor. For the extensively networked intralogistics systems, there are still various strategies at the moment.

 Do customers see the benefit and are they willing to pay for such services?

I believe that the business models behind maintenance contracts will change in the long term. The new tools and services offer advantages for customers – and these advantages will ultimately be visible in the total cost of ownership (TCO) consideration. Accordingly, we will also adjust our business models.

Title photo: © TGW / Caption: PickCenter Rovolution by TGW: Condition monitoring ensures high availability of the multiple award-winning picking robot.

Dr. Maximilian Beinhofer leads the Cognitive Systems Development department at the TGW Logistics Group headquarters in Marchtrenk, Austria. He studied mathematics at the universities of Aachen and Freiburg and earned his doctorate in computer science in probabilistic robotics there. In 2014, the scientist began his career at TGW as a logistics and IT consultant. Since 2016, he has been leading the development team in the field of Cognitive Systems Development.