Refine
Year of publication
Institute
- Fachbereich Medizintechnik und Technomathematik (1575)
- Fachbereich Elektrotechnik und Informationstechnik (715)
- IfB - Institut für Bioengineering (567)
- Fachbereich Energietechnik (563)
- Fachbereich Chemie und Biotechnologie (541)
- INB - Institut für Nano- und Biotechnologien (533)
- Fachbereich Luft- und Raumfahrttechnik (484)
- Fachbereich Maschinenbau und Mechatronik (272)
- Fachbereich Wirtschaftswissenschaften (209)
- Solar-Institut Jülich (161)
Has Fulltext
- no (4735) (remove)
Language
- English (4735) (remove)
Document Type
- Article (3194)
- Conference Proceeding (1065)
- Part of a Book (197)
- Book (146)
- Conference: Meeting Abstract (34)
- Doctoral Thesis (32)
- Patent (25)
- Other (10)
- Report (10)
- Conference Poster (5)
Keywords
- Gamification (6)
- avalanche (6)
- Additive manufacturing (5)
- Earthquake (5)
- Enterprise Architecture (5)
- Industry 4.0 (5)
- MINLP (5)
- Natural language processing (5)
- solar sail (5)
- Additive Manufacturing (4)
Non-intrusive measuring techniques have attained a lot of interest in relation to both hydraulic modeling and prototype applications. Complimenting acoustic techniques, significant progress has been made for the development of new optical methods. Computer vision techniques can help to extract new information, e. g. high-resolution velocity and depth data, from videos captured with relatively inexpensive, consumer-grade cameras. Depth cameras are sensors providing information on the distance between the camera and observed features. Currently, sensors with different working principles are available. Stereoscopic systems reference physical image features (passive system) from two perspectives; in order to enhance the number of features and improve the results, a sensor may also estimate the disparity from a detected light to its original projection (active stereo system). In the current study, the RGB-D camera Intel RealSense D435, working on such stereo vision principle, is used in different, typical hydraulic modeling applications. All tests have been conducted at the Utah Water Research Laboratory. This paper will demonstrate the performance and limitations of the RGB-D sensor, installed as a single camera and as camera arrays, applied to 1) detect the free surface for highly turbulent, aerated hydraulic jumps, for free-falling jets and for an energy dissipation basin downstream of a labyrinth weir and 2) to monitor local scours upstream and downstream of a Piano Key Weir. It is intended to share the authors’ experiences with respect to camera settings, calibration, lightning conditions and other requirements in order to promote this useful, easily accessible device. Results will be compared to data from classical instrumentation and the literature. It will be shown that even in difficult application, e. g. the detection of a highly turbulent, fluctuating free-surface, the RGB-D sensor may yield similar accuracy as classical, intrusive probes.
In this paper, the use of reinforcement learning (RL) in control systems is investigated using a rotatory inverted pendulum as an example. The control behavior of an RL controller is compared to that of traditional LQR and MPC controllers. This is done by evaluating their behavior under optimal conditions, their disturbance behavior, their robustness and their development process. All the investigated controllers are developed using MATLAB and the Simulink simulation environment and later deployed to a real pendulum model powered by a Raspberry Pi. The RL algorithm used is Proximal Policy Optimization (PPO). The LQR controller exhibits an easy development process, an average to good control behavior and average to good robustness. A linear MPC controller could show excellent results under optimal operating conditions. However, when subjected to disturbances or deviations from the equilibrium point, it showed poor performance and sometimes instable behavior. Employing a nonlinear MPC Controller in real time was not possible due to the high computational effort involved. The RL controller exhibits by far the most versatile and robust control behavior. When operated in the simulation environment, it achieved a high control accuracy. When employed in the real system, however, it only shows average accuracy and a significantly greater performance loss compared to the simulation than the traditional controllers. With MATLAB, it is not yet possible to directly post-train the RL controller on the Raspberry Pi, which is an obstacle to the practical application of RL in a prototyping or teaching setting. Nevertheless, RL in general proves to be a flexible and powerful control method, which is well suited for complex or nonlinear systems where traditional controllers struggle.
Often, detailed simulations of heat conduction in complicated, porous media have large runtimes. Then homogenization is a powerful tool to speed up the calculations by preserving accurate solutions at the same time. Unfortunately real structures are generally non-periodic, which requires unpractical, complicated homogenization techniques. We demonstrate in this paper, that the application of simple, periodic techniques to realistic media, that are just close to periodic, gives accurate, approximative solutions. In order to obtain effective parameters for the homogenized heat equation, we have to solve a so called “cell problem”. In contrast to periodic structures it is not trivial to determine a suitable unit cell, which represents a non-periodic media. To overcome this problem, we give a rule of thumb on how to choose a good cell. Finally we demonstrate the efficiency of our method for virtually generated foams as well as real foams and compare these results to periodic structures.