How can automotive production remain competitive in times of ever greater variety and ever faster product cycles? The new study "At the end of the line - How automakers can embrace flexible production" [1] sees the solution, at least for the premium segment, in intelligently networked, self-organizing production.
This means a rethink compared to today's established bead chain principle, as the authors of the study from Strategy& Germany, the strategy consultancy of PwC, and the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB write. However, the costs could pay for themselves within a very short time.
Autonomous on the road, even before completion: instead of on the assembly line, the car body navigates through the factory on a driverless transport system, on an individually optimized course between the modular, versatile and fully networked machines and systems. Instead of human planning and control, this is based on digitalized, AI-driven self-organization that is not limited to the factory premises, but extends across the entire supply chain. This is the vision of the study now published for the future production of cars that vie for the favor of individual buyers - in contrast to the production of volume models for car sharing, for example.
Pearl chain principle reaches its limits
"The idea of self-organizing production can already be found in the first documents on Industry 4.0 - and even then it was nothing new," explains Dr.-Ing. In its 2013 "Implementation recommendations for the future project Industry 4.0", acatech described the vision that "intelligent products (...) are capable of autonomous production control thanks to their ad-hoc networking capability and the inclusion of a digital product description". The state of the art in body construction, painting and assembly today is the control principle of the pearl chain, consistently implemented right through to "just-in-sequence" delivery: the components arrive in exactly the right order for the vehicles or pearls to be produced. "However, the increasing diversity of vehicle types, variants and derivatives is pushing the bead chain principle to its limits," explains Sauer, who coordinates the Automation and Digitalization business unit at Fraunhofer IOSB. "For example, the actual work content and effort involved in a particular processing step can vary greatly from vehicle to vehicle - yet a uniform average cycle time must apply to all of them."
Promising examples of the implementation of self-organizing production
The required adaptability is also difficult to implement within this framework. The standard today is automated equipment that is designed for specific series, engine variants or assembly scopes. "Such systems are very efficient at high capacity utilization - but if demand fluctuates, as is the case now during the coronavirus crisis, fixed costs become a problem," says Sauer. More flexible systems, more universally applicable operating resources and modularized production that can be converted for new tasks without high engineering costs are necessary for cost-effective production under these circumstances.
"Of course, this is challenging to implement - but it is possible with the current state of the art and economically worthwhile, as demonstrated by a few examples that we examine in more detail in the study," says Sauer. The challenges include the permanent localization and online tracking of car bodies, components and means of transport as well as the simulation and operational control of the entire system. Methods such as reinforcement learning and scalable edge cloud data centers (GAIA-X nodes directly in the production hall) provide the appropriate tools.
Fraunhofer IOSB conducts case studies on the introduction of self-organized production and logistics for customers from the manufacturing industry. Furthermore, algorithms for distributed planning are developed and evaluated in simulation environments.
Flexible and resilient supply chains
However, self-organization does not only take place in the factory, but can also help at the level of global supply chains. "Suppliers are reluctant to hand over planning sovereignty and therefore confidential information to a central supply chain orchestrator," explains Olaf Sauer. Self-organization and agent-based decentralized planning offer a way out of this dilemma.
With the Smart Factory Web [2], which has been declared an official testbed by the Industrial Internet Consortium, Fraunhofer IOSB has therefore developed a marketplace for the flexible utilization of production capacities available worldwide. This marketplace is linked to a detailed description model for existing production capabilities and allows specific production processes to be monitored online. At the same time, the Smart Factory Web meets the highest standards of data security and sovereignty.
"However, self-organization in a network of production facilities and suppliers will require the systematic use of artificial intelligence and machine learning methods," says Dr. Julius Pfrommer, research group leader at Fraunhofer IOSB and another author of the study. This also includes learning models generated across companies ("federated learning"). "However, in order to make AI methods an easy-to-use tool for planning and development engineers, a certain gap still needs to be bridged in practice." Fraunhofer IOSB is addressing this challenge together with the Karlsruhe Institute of Technology (KIT) and the FZI Research Center for Information Technology in the new Karlsruhe Competence Center for AI Engineering (CC-KING) [3]. Funded by the Baden-Württemberg Ministry of Economic Affairs, methodological foundations, tools and demonstration scenarios for AI engineering are being advanced here, and consulting services and training courses are being offered for industry and SMEs.
[1] The study "At the end of the line - How automakers can embrace flexible production" by Strategy& Germany and Fraunhofer IOSB is now available and can be downloaded free of charge. Further information and download link at www.iosb.fraunhofer.de/end-of-the-line.
[2] www.smartfactoryweb.de
[3] www.ki-engineering.eu
