Addressing the Scalability Bottleneck of Semantic Technologies at Bosch

At the heart of smart manufacturing is real-time semi-automatic decision-making. Such decisions are vital for optimizing production lines, e.g., reducing resource consumption, improving the quality of discrete manufacturing operations, and optimizing the actual products, e.g., optimizing the sampling rate for measuring product dimensions during production. Such decision-making relies on massive industrial data thus posing a real-time processing bottleneck. Indeed, consider an example of automated welding that is present in multiple Bosch production sites where real-time decisions include welding machine adjustment when welding quality (welding spots) degrades. Such processes are data-intensive, including sensor measurements, e.g., temperature, pressure, and electrical conductivity, settings of welding parameters, and replacement of accessories (welding caps), etc. When the welding is performed during car body manufacturing, each such body has up to 6.000 welding spots, generating a large amount of data instances. The data is distributed across several analytical pipelines in charge of feed statistics, training traditional ML models, quality control measures, and many more [8]. This decision-making for welding requires both integration of heterogeneous data and real-time computation on top of it, thus leading to scalability bottleneck. Delivering a consistent and accurate industry-grade solution as long as the data grows in time, increases the complexity in scalability and performance terms. These challenges combined with the need of maintaining daily operations, increase further scalability requirements, creating the need to implement additional machine learning, knowledge engineering, and data management solutions transversely into a unified framework. Bosch is a multinational company with a strong emphasis on manufacturing and engineering in automotive, energy, consumer goods, and other industries. Smart and AI-powered manufacturing is one of the central pillars of the Bosch strategy, thus real-time effective and efficient processing of extreme data chains of heterogeneous, distributed, fast-growing, and often disconnected or hardly compatible information is critical for the company’s success.