Prof. Dr. Florian Kerber heads the Technology Transfer Center 'Flexible Automation Nördlingen' (TTZ) at Augsburg University of Applied Sciences. The TTZ supports companies in the digital transformation of their production technology.
The 'ModProFT' research project, which is funded by the Free State of Bavaria, aims to improve human-robot cooperation. Where is the digital twin used?
In this case, the product is the printed circuit board, and our partner is an EMS service provider for small series. The information that belongs to such a printed circuit board is crucial. But customers often don't provide detailed information, only a kind of parts list. That's why we use the digital twin as a first step: we enrich the product information from the company's IT with the information collected during the manufacturing and testing process (camera-based object registration, AOIs, etc.). The second step: Automatic placement machines are often used, which are fed with the necessary information and report back completed production steps. However, there is usually no complete data consistency in a heterogeneous machine park. Which test program should be loaded? How does the feedback take place? A standardized architecture must be created with the help of the digital twin in order to be able to track the complete flow of PCBs through production.
Step three is the robot-assisted testing process. There is a cobot and several test stations at a workstation. What individual movement steps does the robot have to perform in order to recognize the PCB, pick it up, open the tester, insert it and trigger the test? The entire process is to be mapped digitally in order to prepare path planning with sensor information and identify sources of error. Up to now, an employee has been required in the event of a fault. The environment model that we record with various sensors increases the tolerance for ambiguities and uncertainties in detection.
Sometimes the term digital twin seems like a mere buzzword ...
I have a similar view. The term is used in an inflationary way and often without a clear definition. In principle, simulations are nothing new and have been in use for decades. What is fundamentally new, however, is the coupling of real and virtual models. The vision of a true digital twin would be if the virtual model ran parallel to the real process and a permanent comparison took place. This would make it possible to detect errors and optimize processes. But there is still a long way to go before the digital twin is complete.
What is the problem?
Robot manufacturers very often already supply an exact image of their robots. However, in the case of special machines in production, there are usually no exact models of the conveyor belts, drives, valves, etc. that can be used for a simulation. If the descriptions were standardized, the widespread use of digital twins would be possible in the first place. But we still have to enrich our model with our own sensor technology to get around this problem.
