Publications

Analyzing time series data like EEG or MEG is challenging due to noisy, high-dimensional, and patient-specific signals. Deep learning methods have been demonstrated to be superior in analyzing time series data compared to shallow learning methods which utilize handcrafted and often subjective features. Especially, recurrent deep neural networks (RNN) are considered suitable to analyze such continuous data. However, previous studies show that they are computationally expensive and difficult to train. In contrast, feed-forward networks (FFN) have previously mostly been considered in combination with hand-crafted and problem-specific feature extractions, such as short time Fourier and discrete wavelet transform. A sought-after are easily applicable methods that efficiently analyze raw data to remove the need for problem-specific adaptations. In this work, we systematically compare RNN and FFN topologies as well as advanced architectural concepts on multiple datasets with the same data preprocessing pipeline. We examine the behavior of those approaches to provide an update and guideline for researchers who deal with automated analysis of EEG time series data. To ensure that the results are meaningful, it is important to compare the presented approaches while keeping the same experimental setup, which to our knowledge was never done before. This paper is a first step toward a fairer comparison of different methodologies with EEG time series data. Our results indicate that a recurrent LSTM architecture with attention performs best on less complex tasks, while the temporal convolutional network (TCN) outperforms all the recurrent architectures on the most complex dataset yielding a 8.61% accuracy improvement. In general, we found the attention mechanism to substantially improve classification results of RNNs. Toward a light-weight and online learning-ready approach, we found extreme learning machines (ELM) to yield comparable results for the less complex tasks.

Die automatische optische Inspektion ist das wichtigste Werkzeug der Qualitätskontrolle in der modernen Elektronikfertigung. Durch die automatisierte Bildaufnahme und das Ausführen vordefinierter Bildverarbeitungsschritte haben diese Systeme die manuelle optische Inspektion weitestgehend verdrängt. Trotz des großen Maßes an Automatisierung sind menschliche Experten an vielen Schritten der Prüfung unverzichtbar und damit potenzielle Fehlerquellen. In den letzten Jahren wurden zahlreiche Ansätze untersucht, welche einzelne Aspekte der optischen Qualitätssicherung durch die Anwendung von Methoden der künstlichen Intelligenz deutlich verbessern. Für den Wandel der optischen Inspektion hin zu einer verlässlichen und voll autonomen Prüfung wird in dieser Arbeit ein Modell mit fünf Phasen vorgestellt, welches die Entwicklungsschritte auf diesem Weg abbildet. Das neue Modell unterscheidet sich von bisherigen Ansätzen durch einen ganzheitlichen Blick auf die Qualitätskontrolle und die Berücksichtigung aller Prozessschritte. Um die Umsetzung dieses Modells zu tragen, zeigt diese Arbeit ein neues Architekturmuster auf, welches Lösungen auf Basis von künstlicher Intelligenz trainieren und ausführen kann. Durch seine hohe Flexibilität kann die neue Architektur über unterschiedliche Auslieferungen auf einer heterogenen Menge an Systemen angewendet werden und viele unterschiedliche Anwendungen von künstlicher Intelligenz über das Feld der optischen Inspektion hinaus umsetzen. Für die allgemeingültige Beschreibung von KI-Lösungen basiert diese Architektur auf einer Menge an Objekten, welche in dieser Arbeit definiert werden. Eine Umsetzung dieser Architektur wird diskutiert und ihre Anwendbarkeit anhand von drei Experimenten bewiesen. Die Implementierung der beschriebenen Architektur ist unter einer OpenSource-Lizenz veröffentlicht.

The systematic manipulation of components of multimodal particle solutions is a key for the design of modern industrial products and pharmaceuticals with highly customized properties. In order to optimize innovative particle separation devices on microfluidic scales, a particle size recognition with simultaneous volumetric position determination is essential. In the present study, the astigmatism particle tracking velocimetry is extended by a deterministic algorithm and a deep neural network (DNN) to include size classification of particles of multimodal size distribution. Without any adaptation of the existing measurement setup, a reliable classification of bimodal particle solutions in the size range of 1.14 μm–5.03 μm is demonstrated with a precision of up to 99.9 %. Concurrently, the high detection rate of the particles, suspended in a laminar fluid flow, is quantified by a recall of 99.0 %. By extracting particle images from the experimentally acquired images and placing them on a synthetic background, semi-synthetic images with consistent ground truth are generated. These contain labeled overlapping particle images that are correctly detected and classified by the DNN. The study is complemented by employing the presented algorithms for simultaneous size recognition of up to four particle species with a particle diameter in between 1.14 μm and 5.03 μm. With the very high precision of up to 99.3 % at a recall of 94.8 %, the applicability to classify multimodal particle mixtures even in dense solutions is confirmed. The present contribution thus paves the way for quantitative evaluation of microfluidic separation and mixing processes.

The paper presents a method for extending the diversity of virtual experiments in remotely controlled laboratories. Using the example of experiments controlling the model of a manufacturing cell, it is shown how a variety of different experiments can be achieved by synthesizing new variants of the structure of this manufacturing cell. The manufacturing cell is used to process workpieces and consists of elements such as conveyors, turntables, processing devices, and others. Each of these devices is represented by a variety of discrete sensors and actuators, where the sensors are inputs and the actuators are outputs to the control system, which must be designed by the student as a finite state machine (FSM). The devices with their sensors and actuators, as well as visual and virtual device models, are grouped together as a structure constructor. Here, the visual model of a device is a set of graphic objects whose visual properties depend on the values of the tags of the virtual model. The virtual model of the device specifies the graphical (tag values) and functional (values of the inputs of the FSM) response of the device model to the control signals of the FSM. The requirements for the structure of the manufacturing cell are determined by the sequence of operations for processing the workpiece. As a result of the synthesis of the manufacturing cell structure, the following are determined: the composition and location of the devices on the working field of the system; their visual reactions to temporal changes in the values of the outputs of the FSM; the values of the inputs in response to the control signals (=outputs of the FSM); and the values of the variables of the relationship with neighboring devices. The proposed method can be useful in organizing a laboratory workshop on the design of parallel and hierarchical control automata.

There is substantial intersubject variability of behavioral and neurophysiological responses to transcranial electrical stimulation (tES), which represents one of the most important limitations of tES. Many tES protocols utilize a fixed experimental parameter set disregarding individual anatomical and physiological properties. This one-size-fits-all approach might be one reason for the observed interindividual response variability. Simulation of current flow applying head models based on available anatomical data can help to individualize stimulation parameters and contribute to the understanding of the causes of this response variability. Current flow modeling can be used to retrospectively investigate the characteristics of tES effectivity. Previous studies examined, for example, the impact of skull defects and lesions on the modulation of current flow and demonstrated effective stimulation intensities in different age groups. Furthermore, uncertainty analysis of electrical conductivities in current flow modeling indicated the most influential tissue compartments. Current flow modeling, when used in prospective study planning, can potentially guide stimulation configurations resulting in individually effective tES. Specifically, current flow modeling using individual or matched head models can be employed by clinicians and scientists to, for example, plan dosage in tES protocols for individuals or groups of participants. We review studies that show a relationship between the presence of behavioral/neurophysiological responses and features derived from individualized current flow models. We highlight the potential benefits of individualized current flow modeling.