The Production Systems Group is researching various topics of automation in the context of different production and manufacturing processes. Future production plants will be characterized by a high degree of reproducibility, flexibility and interconnectivity. With the goal of intelligent manufacturing systems both automatic process control as well as self-sufficient monitoring is required. In the light of increasing production volumes, the desire for short turnaround times and ever decreasing tolerances a shift from the classical control of machine-related variables is necessary. Instead, the tracking of higher-level process and quality variables is of interest, which not only reduces the adjustment effort but also allows for an efficient and high-performance control of the process over various product cycles. In this context, the research group develops control concepts for production systems from a variety of industries. Together with our interdisciplinary project partners, these solutions are then implemented on systems at shop floor level.
Due to the wide range of application areas, a multitude of challenges linked to control engineering arises:
- Most processes are complex, non-linear or time-variant and can only be modeled using multidisciplinary approaches. The resulting models must be either reduced in order or substituted by a simplified description to enable real-time model based control.
- Process steps not accessible to physical modelling must be described using databased approaches.
- Often relevant process variables cannot be measured directly. In the majority of cases, suitable sensors are expensive or not available and must therefore be replaced by a proper processing of the measured data.
- In general, quality-related variables cannot be measured during the process, but can only be determined retrospectively. Key factors influencing the process must therefore be identified at runtime and mapped model-wise.
- Changing process conditions or varying product designs require a high flexibility of the automation solution. This also raises the question about the robustness of the designed control algorithm.
- In order to prevent damage to both plant and product, predefined limit values, such as for machine and process variables, as well as secondary conditions, for example energy consumption, have to be met.
- Many production processes are cyclic. The influence of previous machine cycles has to be considered for the control of the current process.
- Depending on the application, production systems differ fundamentally in their dynamics and thus present different challenges regarding response time and real-time capability of the control concept.
In order to meet these challenges, the Production Systems Group works together with interdisciplinary partners from industry and research to develop control engineering solutions that extend the current state of the art and transform the considered manufacturing processes in view of a digital factory. These include among others:
- Model-based optimization for process control of complex systems.
- Use of model-based predictive control, iterative learning control and robust control in real-time critical systems.
- Adaptive control concepts for tracking cyclical production processes.
- Data driven identification methods to describe hard to model subsystems using
- for examlpe: Neuronal Networks, Support Vector Machines or Gaussian-Process-Regression
- Synthesis of reduced white box and grey box models.
- Development of soft sensors to monitor unmeasurable variables.
- Increase of process stability for the purpose of an improved reproducibility.
- Rapid control prototyping and transfer of control concepts to industry-graded controllers, for example SPS
Current topics include questions from the following areas:
- Digitalization of production systems
- High dynamic process control in textile systems
- Data-driven optimization in rough milling
- Feature centered control in forming processes, for example cold rolling
- Cross-phase concepts for quality control in injection molding