Band 89, Heft 5

The fluctuations of the renewable energies request storage systems as a significant part of the future energy system. Here we introduce a measurement approach regarding the regional coverage of the power demand and the influence of a storage system on it. The approach is based on measured power data with given uncertainties. The resolution is 15 minutes in the time domain and 2-digit ZIP code regions in the spatial domain. An idealized storage model is introduced, and one storage is implemented for each region. As a result, the regional coverage is measured for different storage parameters and the respective measurement uncertainty is assessed to be below the set goal of 2 %. Furthermore, the inverse problem, i. e. the storage dimensioning, is investigated by estimating the necessary storage parameters to achieve a desired regional coverage. Findings: electrical self-sufficiency can be reached in 23 out of 95 regions, in contrast in 56 regions the regional coverage is increased insignificantly regardless of the parameters of the storage system. This illustrates the necessity of power transport to enable the increase of regional coverage. Concluding, the proposed measurement approach enables the estimation of storage parameters and can be used for storage dimensioning.

Electrical impedance tomography (EIT) is an imaging method for characterizing the inner conductivity distribution of an object based on the measured boundary voltages resulting from the injection of an AC signal, followed by an image reconstruction procedure. An algorithm tries to solve an ill-posed inverse problem making it challenging to reconstruct an accurate image. To overcome this, we propose a genetic algorithm (GA) for the image reconstruction with a non-blind search method considering prior knowledge about the possible conductivity distribution in the initial search space. To validate the algorithm, experiments have been conducted in a water tank. The algorithm’s performance was evaluated regarding image quality and processing time, being able to minimize the corresponding quality function to 0.0505 with 100 generations using the non-blind search and the uniform crossover/random mutation. Compared to traditional methods, the GA achieves significantly better image quality. It has been implemented as an image reconstruction algorithm for gesture recognition. EIT measurements have been conducted with six persons performing American sign numbers (0–9) resulting in 1800 reconstructed images. They were classified by a previously developed convolutional neural network (CNN), reaching a 92 % accuracy, which is a very good achievement in the case of multiple subjects.

A central issue in Wireless Body Sensor Networks (WBSNs) is the large amount of measurement data for monitoring vital parameters, which need to be continuously measured, immediately processed and timely transmitted. This requires a big storage space and computing effort leading to a high-power consumption. Reducing the amount of transmitted data contributes significantly to an extension of the sensor operation time. In this contribution, we focus exactly at this aspect. We propose a data aggregation method based on Artificial Neural Networks (ANN) combining multiple physiological signals, which are the ElectroCardioGram (ECG), ElectroMyoGram (EMG) and Blood Pressure (BP), in one signal before transmission. The simulation and implementation results reveal a reduction of energy consumption to 87.32 %, ensuring a high accuracy level (80.53 %) and a relatively execution time (48.47 ms).

Thermophysical properties of materials and the optimization of the heat transfer are becoming more and more important for industrial applications of micro- and nanoelectronic devices. Thin layers in the micrometer to nanometer range are used to give specific functions to the devices. Since the thermophysical properties of thin layers differ from bulk material, this data is required for precise predictions of thermal management. One way to obtain the thermal properties of thin layers is the optical-based Time Domain Thermoreflectance (TDTR) method. To carry out TDTR measurements with a low level of uncertainty, the samples under study must meet requirements related to the surface roughness and a low level of optical scattering. The range of samples analysable by TDTR can be extended by applying the so-called bidirectional heat flow approach. This approach opens the possibility to assess thermal properties of materials with rough surfaces as well. The validity of the implemented model was shown by the characterisation of a test sample with well-known thermal properties fabricated for this purpose out of poly(methyl methacrylate) (PMMA) roughened with acetone:ethanol. The results obtained by TDTR measurements were compared to literature values, demonstrating the applicability of the bidirectional heat flow approach for this setup.

We propose a model to classify axial, radial loading profiles and rotation speed of wire-race bearings using Convolutional Neural Networks with spectrograms. Software-based models are able to measure quantities of machine elements. In many cases a physical sensor takes up a lot of space and is expensive. Wire-race bearings are lightweight bearings that can be installed in a space-saving manner. For predictive maintenance of the load-dependent lifetime estimation there is the need to record the stress conditions on the bearing. While force sensors have the ability to record these necessary stresses, they cannot be installed space-savingly. With the proposed setup using an accelerometer attached to the housing of a wire-race bearing, it is possible to both save space and reduce the number of physical sensors. Furthermore, we can show that the model is able to predict unseen data from 168 different load profiles with an overall accuracy of 86.3 %. In several experiments we generated vibration data which is processed using the fast Fourier transform algorithm and spectrograms. We show that image classification of spectrograms with Convolutional Neural Networks significantly outperforms other models with fast Fourier transforms fed into Artificial Neural Networks or logistic regression.

Zusammenfassung Vermehrt werden Daten in der Produktion erfasst und mit Hilfe maschineller Lernverfahren ausgewertet. Ziel dieser Datenanalysen ist es, Informationen über die Produktion zu erhalten. Typische Anwendungsbeispiele sind die Vorhersage der Produktqualität sowie die vorausschauende Wartung und Instandhaltung. Allerdings führen Datenanalysen aufgrund der domänenspezifischen Herausforderungen häufig zu keinen ausreichenden Analyseergebnissen. Die Kombination von Datenanalysen mit Expertenwissen stellt hierbei einen vielversprechenden Ansatz zur Lösung der domänenspezifischen Problemstellungen dar. Dieser Beitrag stellt hierzu eine neue Vorgehensweise für die Durchführung von Projekten des maschinellen Lernens mit wissensbasierten Ansätzen in der Produktion vor. In einem Ablaufplan werden notwendige Schritte und die Beteiligung der entsprechenden Experten vorgestellt. Für die Nutzung von vorhandenem Domänenwissen in der Produktion werden in diesem Artikel verschiedene Methoden zur Wissensmodellierung aufgezeigt und diskutiert. Der Ablaufplan und die Methoden zur Wissensmodellierung werden anschließend am Beispiel einer servopneumatischen Schweißzange validiert.

Zusammenfassung Sensorlose Messverfahren ermöglichen eine platzsparende und gegen Umgebungseinflüsse unempfindliche Bestimmung der Läuferposition in elektrischen Maschinen. Die konventionellen Verfahren weisen jedoch einen verhältnismäßig hohen Rauschpegel auf und sind empfindlich gegenüber magnetischer Sättigung sowie Veränderungen der elektrischen Betriebsparameter. Beim Fluss-Bypass-Verfahren wird über einen im Stator integrierten Fluss-Bypass die zeitliche Änderung eines Streuflusses im Stator der Maschine gemessen. Die Sättigung des Stators kann damit effektiv kompensiert werden. Die sonst übliche Differenziation des Stroms oder ein aufwändiges Modell zur Ermittlung von Induktivität und Wicklungswiderstand sind nicht erforderlich. Mittels theoretischer und experimenteller Untersuchungen werden der Rauschpegel, die Linearität sowie der Einfluss von Strom, Schaltfrequenz und Tastverhältnis auf das Lagemesssignal untersucht. Das Fluss-Bypass-Verfahren weist eine höhere Linearität, erheblich geringere Rauschpegel sowie eine größere Sättigungstoleranz auf, als eine konventionelle sensorlose Messung mittels Induktivitätsbestimmung. Ein wesentlicher Störeffekt sind Wirbelströme im Material. Bei entsprechender Auslegung des Fluss-Bypasses können innerhalb eines gewissen Messwertbereich parasitäre Efekte verschiedener Art gleichzeitig gemindert werden.