Siemens AG (headquartered in Berlin and Munich) is a global powerhouse in electronics and electrical engineering. Operating in the fields of automation, electrification and mainly also digitalization, Siemens holds leading market positions in all its business areas. The company has roughly 379,000 employees – of which 41,000 were newly hired in FY18 and of which 117,000 or 31% (newly 4,700) are based in Germany – working to develop and manufacture products, design and install complex systems and projects, and tailor a wide range of solutions for individual requirements. In FY18, Siemens had revenue with business in more than 200 countries of €83.04 billion. In FY17, arising from Siemens software and digital services alone, €5.2 billion could be achieved (e.g. via MindSphere), making it a growth rate in this area of 20%. For monitoring and control purpose Siemens offers the customer the whole value chain starting from sensors over data acquisition and data communication to cloud services for data visualization and so on. Since many years Siemens is a trendsetter in converting advanced embedded electronic devices into interconnected industrial IoT-devices for control, measurement, data acquisition and communication. So, Siemens could benefit from efficient deep learning methods which enable robust, self-adaptive, safe and secure industrial IoT-devices.

For the significantly rising number of dynamically configured infrastructure solutions in transportation, energy and manufacturing, it is necessary to provide a fast and reliable, but also a self-adaptive, distributed sensor and communication network, to detect the current state of motors, production lines, power converters or electrical distribution systems and thus provide a high degree of flexibility, safety, availability, efficiency at reasonable cost at the same time.

On the one hand side, due to the recent advances in artificial intelligence and especially in deep learning methods, deep learning methods seems to become a powerful tool to identify the increasingly complex states of dynamically configured infrastructure. On the other hand, the time-consuming communication to get the data from a sensor network to a cloud service which offers the whole range of existing deep learning-Tools is still limiting the use of deep learning methods for real time, control and safety applications in distributed dynamically configured infrastructure. Thus, Siemens is interested to overcome the limitations by using deep learning methods on all communication layers for control and safety applications in distribute dynamically configured systems.

Key persons involved:

  • Dr. Ing. Roland Weiss
  • Andreas Tobola.