Cognition-enhanced, Self-optimizing Manufacturing Processes


Enhance the level of controllability of the manufacturing processes in order to ensure high product quality and process productivity by means of model-based self-optimization
phase 2

(1) Realization of multidimensional, model-based control systems (up to self-optimization)
(2) Extension of the manufacturing system’s transparency up to a level that the systems’ behavior can be described at any instant of time
(3) Development of interfaces for the embedding of the manufacturing system into the production environment


Self-optimizing concepts are implemented as prototypes for different production technologies: metal cutting, welding, laser cutting, injection molding, weaving and braiding. The project covers all active research areas from simulation, sensor technology and monitoring to control and quality inspection. This allows for a complete view at the different technologies to identify overlapping research questions and synergies. Each of the different sub-projects has a different focus concerning their particular technology and application. The overall objective is a self-optimizing production. Each demonstrator is a model to gain inside knowledge of the underlying process. The model based control forms a joint link between the demonstrators. It has been implemented in milling, injection molding, weaving and braiding. In this regard, the cluster’s interdisciplinarity was a great advantage.

  Copyright: © Image: Thilo Vogel

To determine the optimum of the real process, virtual process models are used for numerical optimization. In the past two years, all demonstrators made significant process towards a self-optimizing production.

A new axis control has been implemented for the laser cutting process. Now it is possible to control the cutting speed on the basis of the thermal emission. This reduces the thermal input in the material. To generate empirical process models for gas-shielded metal-arc welding, a software tool has been developed. The model quality is estimated with statistical parameters. For the gun drilling process, the self-developed sensor has been investigated to monitor the chip removal rate. The system and the results of this work were published in a dissertation in 2016.

  Copyright: © Image: Thilo Vogel

The simulation model for milling forces has been extended to micro simulation and variable feed velocities. It can be used for model predictive force control in milling. There, the transfer function of the machining center can now be identified online. It has been used to continuously update a force model. This in turn has been used to control the predicted milling force. A significant productivity increase of the complete system could be demonstrated compared to established solutions. The approach and the results have been documented in a dissertation. The same control principle but with different parameters has been used in plastic injection molding. Variation of the viscosity due to material quality can be compensated. To improve the controller, a new tool has been constructed. It integrates pressure sensors into the hot channel. The controller has been transferred to a second machine, to demonstrate that the approach is generic.

A commercial camera system has been improved to measure the fiber orientation in the radial braiding process. The system is capable of real-time monitoring and, therefore, can be used for controlling. The algorithm for offline optimization of the weaving process has been validated. The user interface has been integrated into an Android-app. In addition to that, the exception handling for unexpected process dead time has been improved.

The flexible inspection system has been extended to a kinematic model and a path planning system. Now, simulation can be used to determine time and energy optimized trajectories.


Further improvements are planned for all of the demonstrators.

The intense interdisciplinary collaboration with the IRT is central due to improve the controllers. In particular, the controller in laser cutting should obtain real-time capability. The results in gun drilling will be included to a model which can be used in controlling. The model predictive control in milling will be transferred to a modern machining center. To further increase the practical relevance, machine internal sensor signals should be used for automatic control. It is planned to extend the simulation model to abrasive tool wear. Furthermore, experiments will be conducted with the prototypes of the injection molding process and the weaving process for validation. To increase the autonomy of the flexible inspection system, an automatic object recognition and inspection characteristic extraction should be developed.

Find out more:

Technical Demonstrator "Self-optimizing Injection Molding"

Technical Demonstrator "Self-optimizing Manufacturing Systems for Laser Cutting"

Technical Demonstrator "Model-based Modules for the Self-optimization of Gas Metal Arc Welding Processes"

Technical Demonstrator "Self-optimization of the Weaving Process"

Technical Demonstrator "Self-optimized Metal Cutting Processes"

Technical Demonstrator "Self-optimization of the Radial Braiding Process"