Conference Proceeding
Refine
Year of publication
Institute
- Fachbereich Elektrotechnik und Informationstechnik (286)
- Fachbereich Energietechnik (226)
- Fachbereich Luft- und Raumfahrttechnik (198)
- Fachbereich Maschinenbau und Mechatronik (188)
- Solar-Institut Jülich (164)
- Fachbereich Medizintechnik und Technomathematik (150)
- IfB - Institut für Bioengineering (114)
- Fachbereich Bauingenieurwesen (103)
- ECSM European Center for Sustainable Mobility (59)
- Fachbereich Wirtschaftswissenschaften (54)
Has Fulltext
- no (1413) (remove)
Document Type
- Conference Proceeding (1413) (remove)
Keywords
- Enterprise Architecture (5)
- Gamification (5)
- Energy storage (4)
- Natural language processing (4)
- Power plants (4)
- hydrogen (4)
- solar sail (4)
- Associated liquids (3)
- Concentrated solar power (3)
- Hybrid energy system (3)
Magnetic nanoparticles (MNP) are investigated with great interest for biomedical applications in diagnostics (e.g. imaging: magnetic particle imaging (MPI)), therapeutics (e.g. hyperthermia: magnetic fluid hyperthermia (MFH)) and multi-purpose biosensing (e.g. magnetic immunoassays (MIA)). What all of these applications have in common is that they are based on the unique magnetic relaxation mechanisms of MNP in an alternating magnetic field (AMF). While MFH and MPI are currently the most prominent examples of biomedical applications, here we present results on the relatively new biosensing application of frequency mixing magnetic detection (FMMD) from a simulation perspective. In general, we ask how the key parameters of MNP (core size and magnetic anisotropy) affect the FMMD signal: by varying the core size, we investigate the effect of the magnetic volume per MNP; and by changing the effective magnetic anisotropy, we study the MNPs’ flexibility to leave its preferred magnetization direction. From this, we predict the most effective combination of MNP core size and magnetic anisotropy for maximum signal generation.
In this paper we investigate the use of deep neural networks for 3D object detection in uncommon, unstructured environments such as in an open-pit mine. While neural nets are frequently used for object detection in regular autonomous driving applications, more unusual driving scenarios aside street traffic pose additional challenges. For one, the collection of appropriate data sets to train the networks is an issue. For another, testing the performance of trained networks often requires tailored integration with the particular domain as well. While there exist different solutions for these problems in regular autonomous driving, there are only very few approaches that work for special domains just as well. We address both the challenges above in this work. First, we discuss two possible ways of acquiring data for training and evaluation. That is, we evaluate a semi-automated annotation of recorded LIDAR data and we examine synthetic data generation. Using these datasets we train and test different deep neural network for the task of object detection. Second, we propose a possible integration of a ROS2 detector module for an autonomous driving platform. Finally, we present the performance of three state-of-the-art deep neural networks in the domain of 3D object detection on a synthetic dataset and a smaller one containing a characteristic object from an open-pit mine.
After a brief introduction of conventional laboratory structures, this work focuses on an innovative and universal approach for a setup of a training laboratory for electric machines and drive systems. The novel approach employs a central 48 V DC bus, which forms the backbone of the structure. Several sets of DC machine, asynchronous machine and synchronous machine are connected to this bus. The advantages of the novel system structure are manifold, both from a didactic and a technical point of view: Student groups can work on their own performance level in a highly parallelized and at the same time individualized way. Additional training setups (similar or different) can easily be added. Only the total power dissipation has to be provided, i.e. the DC bus balances the power flow between the student groups. Comparative results of course evaluations of several cohorts of students are shown.
Interplanetary trajectories for low-thrust spacecraft are often characterized by multiple revolutions around the sun. Unfortunately, the convergence of traditional trajectory optimizers that are based on numerical optimal control methods depends strongly on an adequate initial guess for the control function (if a direct method is used) or for the starting values of the adjoint vector (if an indirect method is used). Especially when many revolutions around the sun are re-
quired, trajectory optimization becomes a very difficult and time-consuming task that involves a lot of experience and expert knowledge in astrodynamics and optimal control theory, because an adequate initial guess is extremely hard to find. Evolutionary neurocontrol (ENC) was proposed as a smart method for low-thrust trajectory optimization that fuses artificial neural networks and evolutionary algorithms to so-called evolutionary neurocontrollers (ENCs) [1]. Inspired by natural archetypes, ENC attacks the trajectoryoptimization problem from the perspective of artificial intelligence and machine learning, a perspective that is quite different from that of optimal control theory. Within the context of ENC, a trajectory is regarded as the result of a spacecraft steering strategy that maps permanently the actual spacecraft state and the actual target state onto the actual spacecraft control vector. This way, the problem of searching the optimal spacecraft trajectory is equivalent to the problem of searching (or "learning") the optimal spacecraft steering strategy. An artificial neural network is used to implement such a spacecraft steering strategy. It can be regarded as a parameterized function (the network function) that is defined by the internal network parameters. Therefore, each distinct set of network parameters defines a different network function and thus a different steering strategy. The problem of searching the optimal steering strategy is now equivalent to the problem of searching the optimal set of network parameters. Evolutionary algorithms that work on a population of (artificial) chromosomes are used to find the optimal network parameters, because the parameters can be easily mapped onto a chromosome. The trajectory optimization problem is solved when the optimal chromosome is found. A comparison of solar sail trajectories that have been published by others [2, 3, 4, 5] with ENC-trajectories has shown that ENCs can be successfully applied for near-globally optimal spacecraft control [1, 6] and that they are able to find trajectories that are closer to the (unknown) global optimum, because they explore the trajectory search space more exhaustively than a human expert can do. The obtained trajectories are fairly accurate with respect to the terminal constraint. If a more accurate trajectory is required, the ENC-solution can be used as an initial guess for a local trajectory optimization method. Using ENC, low-thrust trajectories can be optimized without an initial guess and without expert attendance.
Here, new results for nuclear electric spacecraft and for solar sail spacecraft are presented and it will be shown that ENCs find very good trajectories even for very difficult problems. Trajectory optimization results are presented for 1. NASA's Solar Polar Imager Mission, a mission to attain a highly inclined close solar orbit with a solar sail [7] 2. a mission to de ect asteroid Apophis with a solar sail from a retrograde orbit with a very-high velocity impact [8, 9] 3. JPL's \2nd Global Trajectory Optimization Competition", a grand tour to visit four asteroids from different classes with a NEP spacecraft