June 2020

26 n ROBOTICS, AUTOMATION AND MACHINE-BUILDING June 2020 www.drivesncontrols.com Audi mines its data for predictive maintenance M any users fail when trying to implement predictive maintenance because they don’t have access to the right data. This was the situation faced by Mathias Mayer, who plans automation technology at Audi’s Neckarsulmplant in Germany. From experience he knew that that“90% of the data in body construction isn’t used or accessed.”This usually means that extra sensors are needed, but that was not the path Mayer wanted to go down. On the contrary, he thought,“let’s process the unused data first. If an additional sensor really ends up being necessary, I’d certainly be willing to talk about it.” To Mayer, better utilisation of available data is the most important requirement for reducing downtime and working more efficiently. It is going to become even more decisive as the complexity of production processes and the degree of automation continue to increase. Why is data collection so difficult, though? A glimpse into the body construction processes at Neckarsulm reveals the challenge. Audi assembles its A4, A6, A7, A8, R8 and A5 Cabrio models at the site using around 2,500 PLC- controlled industrial robots. “We see the PLC as a puppet master making up to ten robots dance,”says Mayer. The value creation takes place at the robot, which is why access to robot data is so immensely important. As well as Audi’s large number of plants, its variety of production methods also complicate data access and evaluation. For example, cutting weight while maximising durability can only be achieved by combining different materials. This entails the use of a variety of connection technologies. For the A8 alone, these range fromwelding processes, through to gluing and riveting – in all, 15 processes need to be coordinated. If production falters, experts in each of these processes are needed. In three-shift production, this ends up being expensive and time-consuming, because of the number of employees that need to be trained and qualified to cover all of the technologies. Breaking new ground To Mayer, deviating from tried-and-tested processes is out of the question.“Our qualification process is definitely expensive and time-consuming, but our customers expect top quality.”If different employees examine the same thing, the results can differ – unlike data, which is always the same. “It’s this data that we have to use to optimise both production and processes,”says Mayer. The data has to be processed so that, for example, even a non-expert can restart a friction-welding process. In this way, unplanned production downtimes can be reduced and availability, process efficiency and quality can be increased – for example, through live systemmonitoring and adapting process parameters automatically. Previous process monitoring and optimisation methods based on expert knowledge need to be kept. Ultimately, this will reduce maintenance costs andminimise testing efforts. How does this work in practice, though? In future body construction processes, data will be collected from devices, and integrated and visualised directly – without needing additional gateways – because the robots will have enough capacity. If necessary, there’s an employee who understands the process and can intervene. In Mayer’s view, this division of labour is the key to success. It’s only on this basis that data mining and machine learning can be implemented successfully. Audi is using OPC UA and MQTT to transport data, which is routed to an edge layer, above which there is a“big data” platform. Applications such as diagnostic Although there various ways of implementing preventive maintenance, they often fail. A key challenge is accessing and using the data that could identify flaws. This article from Profibus & Profinet International looks at how Audi is achieving this using a combination of Profinet and OPC UA. Installing a panoramic glass roof on an Audi A6 at the Neckarsulm plant