Machine Learning in Industrial Control Engineering


Today’s control and computer technology enable industrial application of model-based controllers. The required models must be real-time capable - i.e. they must not be too complex - in order to be able to calculate them at runtime. In addition, system behavior can rarely be described exactly by mathematical models due to physical or chemical principles are not known or they are difficult to describe mathematically. For this reason, machine learning approaches are increasingly used in industrial control engineering. For this purpose, data is collected either at runtime or offline, reduced with respect to the control task and the controller models are created using machine learning. Controllers can also be designed and optimized using machine learning. This offers the potential for controllers to adjust themselves in the future.

September 15, 2021


Information on RWTHmoodle Course Rooms

Registration for RWTHmoodle course rooms is usually performed by the students themselves, starting from the lecture site in the RWTHonline system. If you have trouble registering or you need a late registration, please contact your respective advisor.



Sebastian Stemmler

Head of Industrial Systems and Production Systems


+49 241 80 27479



Target Audience

The target audience of this course are students who want to further deepen their knowledge of control engineering and who want to develop the future of industrial control engineering. In this context, we will discuss how machine learning methods can be integrated and effectively used within control loops. Students will be enabled to create and evaluate data-driven models using various machine learning methods on the basis of industrial application examples. Furthermore, you will learn to design sensors, observers and model-based controllers using data-driven models.



Lecture „Regelungstechnik“ or similar



The focus of this lecture is primarily the development of model-based controllers whose models are created by machine learning. This enables a high transferability of this controller design method to almost any industrial application. In addition, it shifts the focus from setting abstract controller parameters to generating suitable control-oriented system models. This also offers the potential for controllers to adjust themselves independently in the future.

In detail the lecture addresses the application of machine learning methods in the context of industrial control engineering. This is supported by the consideration of various application examples, so that the lecture answers the following questions:

  • How can we learn from data and what methods are suitable?
  • What data is required and how can it be reduced using suitable algorithms?
  • How can control engineering models be generated using machine learning?
  • How can sensors be implemented using machine learning to calculate important values?
  • How can observers (virtual sensors) be realized using machine learning to estimate non-measurable values?
  • How can machine-learned models be used for model-based control?

Information About the Exam

Information will follow.