Intelligente Regelungsstrategien als Schlüsseltechnologie selbstoptimierender Fertigungssysteme

  • Intelligent control strategies as an enabler for self-optimizing manufacturing systems

Stemmler, Sebastian; Abel, Dirk (Thesis advisor); Brecher, Christian (Thesis advisor)

Aachen (2020)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020


The settings of manufacturing systems are often based on the experiences of the machine operator, previously identified technology tables or various simulation tools. Machine oriented control loops ensure the reproducibility of the machine behavior. However, changing material properties and environmental conditions lead to a varying process behavior, which generally results in deviations in the quality of the manufactured parts. To establish process and quality control loops, the quality measurement of the workpiece is necessary, though it can usually be determined only after the manufacturing process. Therefore, this thesis targets the concept of model-based self-optimization of manufacturing systems. An essential element of this concept is the quality model which describes the relationship between the considered quality variables and the relevant process variables. On this basis, an optimal trajectory for process variables can be determined, which are controlled by process control loops. This thesis aims at investigating intelligent controllers for process and quality control loops in manufacturing technology. Suitable modeling approaches are first presented, which can explicitly be used in the control scheme. Furthermore, the concept of state estimation is introduced to predict non-measurable state variables. Apart from this, a model-based predictive controller as well as an iterative learning controller are introduced, which allow the explicit consideration of model knowledge, constraints and arbitrary control objectives. The mentioned concept of model-based self-optimization is applied to roughing in milling and plastic injection molding. For plastic injection molding, a quality control approach is developed which aims to control the weight of the manufactured workpieces. Considering the desired weight of the workpiece, the pvT-optimization determines an optimal trajectory for the cavity pressure. A model-based, predictive control as well as an iterative learning control scheme are examined regarding their ability to control the nonlinear process behavior. Additionally, it is shown that the accuracy of the weight can significantly be increased by the combination of the mentioned control approaches and the pvT-optimization. In milling, a model-based predictive force controller is developed which aims at decreasing the manufacturing time. The controller predicts the future cutting force which occurs on the milling cutter by a force model and adjusts the feed velocity online. The explicit consideration of a machine model in the controller enables the definition of constraints for a maximum desired cutting force and thus the feed velocity. Exceeding cutting forces lead to increased tool wear or in worst case to tool damage. In order to increase the accuracy of the cutting force model, the parameters of the force model are identified at runtime. The feed rate as well as the cutting force can be high-dynamically controlled by the mentioned controllers. Furthermore, variable process conditions such as tool wear ormaterial fluctuations are compensated by the online identification of the force model. The presented results show that novel process control loops can be realized by the establishment of intelligent controllers. These are in turn prerequisites for the implementation of the model-based self-optimization in manufacturing. The controller is parameterized using the process models. Thus, the mentioned approaches can be applied to other manufacturing process by the adaption of the process model.