Refine
Year of publication
- 2020 (136) (remove)
Document Type
- Article (44)
- conference proceeding (article) (42)
- conference proceeding (presentation) (8)
- conference proceeding (draft) (7)
- Part of a Book (6)
- conference proceeding (summary) (6)
- Book (4)
- Other (4)
- Bachelor-/Diplom Thesis (3)
- Part of Periodical (3)
Reviewed
- Begutachtet/Reviewed (55)
Keywords
- regenerative Energien (14)
- Europäische Metropolregion Nürnberg (12)
- Energieversorgung (9)
- Infektionskrankheit (6)
- Laufwasserkraftwerk (6)
- Metropolregion Nürnberg (6)
- Energiespeicher (5)
- COVID 19 (4)
- Energiewetter (4)
- Erneuerbare Energien (4)
Institute
- Fakultät Informatik (16)
- Fakultät für Elektrotechnik Feinwerktechnik Informationstechnik (16)
- Fakultät Maschinenbau und Versorgungstechnik (13)
- Fakultät Verfahrenstechnik (13)
- Fakultät Betriebswirtschaft (10)
- Fakultät Sozialwissenschaften (9)
- Institut für leistungselektronische Systeme ELSYS (9)
- Institut für Angewandte Wasserstoffforschung, Elektro- und Thermochemische Energiesysteme (5)
- Fakultät Angewandte Mathematik, Physik und Allgemeinwissenschaften (3)
- Institut für Chemie, Material- und Produktentwicklung (3)
In einer Machbarkeitsstudie erforschte das Projektteam, inwiefern der Begriff der Figuration (Elias) geeignet ist, um bislang nicht berücksichtigte oder unentdeckte Aspekte des sozialen Phänomens Wohnungslosigkeit sichtbar zu machen und in für die Praxis anwendbare Erkenntnisse zu transformieren. Mit diesen Ergebnissen sollten weitere Mittel für ein mindestens zweijähriges Forschungsprojekt eingeworben werden.
Digitalisierung für alle? Zur Auswirkung digitaler Angebote auf Teilhabechancen von Wohnungslosen
(2020)
Das an der Technischen Hochschule Nürnberg Georg Simon Ohm angesiedelte Forschungsprojekt „SIWo – Smart Inklusion für Wohnungslose“ widmet sich der Fragestellung, welche Informations‑, Unterstützungs- und Beratungsbedarfe bei wohnungslosen Menschen vorliegen und in welcher Form sich die Teilhabechancen für die heterogene Zielgruppe wohnungsloser oder von Wohnungsnot betroffener Menschen mittels digitaler Hilfe und Beratungsangebote verbessern lassen. Hierfür werden zum einen mit Gruppendiskussionen und qualitativen Interviews Bedarfe identifiziert. Zum anderen sollen im Rahmen des Projekts prototypische Lösungsansätze in Form einer plattformübergreifenden Hilfe-App entwickelt, getestet und mit Evaluationsverfahren begleitet werden. Charakteristisch ist dabei die enge interdisziplinäre Zusammenarbeit zwischen Sozialwissenschaft, Informatik und Energietechnik, die partizipative Begleitung des Projekts durch Wohnungslose und andere Stadtakteur*innen, sowie der Fokus auf interaktive Funktionen, die digitalisierte Begegnungsräume ermöglichen und Interaktionen auch zwischen Bevölkerungs-gruppen fördern, die i. d. R. eher wenig miteinander zu tun haben. Dieses Vorgehen begreift somit Wohnungslosigkeit nicht ausschließlich als professionelles Arbeitsfeld der Sozialen Arbeit, sondern als gesamtgesellschaftliche Aufgabe.
In view of the advent of autonomous driving, the automotive industry is longing for adequately trained employees in a wide range of fields; these include not so much the classic automotive engineering and embedded systems topics, but rather software development, robotics, machine learning, statistics and design. As a particularly specific, industry-oriented discipline, automotive software engineering is hardly represented in academic teaching, except for a few AUTOSAR courses. But as a beacon project of digitization, autonomous driving can arouse broad interest among students who are interested in how diverse disciplines can be profitably combined to realize an impressive project. This interest is an excellent chance for both universities and industry to engage in constructive dialogue, to carry out exciting multidisciplinary projects at universities and to meet the needs of industry with graduates who have a relevant and versatile education. This workshop provides an opportunity for both sides to meet, share requirements and experiences and develop a best practice course according to industry requirements and academic feasibility.
Machine learning classifiers for image recognition are prevalent in many applications. We study the problem of finding adversarial examples for such classifiers, i.e., to manipulate the images in such a way that they still look like the original images to a human but are misinterpreted by the classifier. Finding adversarial examples corresponds to a search problem in the image space. We focus on black-box attacks that can only use the original classifier to guide the search. The challenge is not to find adversarial examples, but rather to find them efficiently, ideally in real time. We show two novel methods that increase the efficiency of black-box search algorithms for adversarial examples: The first uses a relevance mask, i.e., a bitmask on the original image that restricts the search to those pixels that appear to be more relevant to the attacked classifier than others. The second exploits the discovery of merge drift, a phenomenon that negatively affects search algorithms that are based on the merging of image candidates. We evaluate both concepts on existing and new algorithms.
A novel modeling strategy is proposed which allows high-accuracy predictions of aerodynamic and aeroacoustic target values for a low-pressure axial fan, equipped with serrated leading edges. Inspired by machine learning processes, the sampling of the experimental space is realized by use of a Latin hypercube design plus a factorial design, providing highly diverse information on the analyzed system. The effects of four influencing parameters (IP) are tested, characterizing the inflow conditions as well as the serration geometry. A total of 65 target values in the time and frequency domains are defined and can be approximated with high accuracy by individual artificial neural networks. Furthermore, the validation of the model against fully independent test points within the experimental space yields a remarkable fit, even for the spectral distribution in 1/3-octave bands, proving the ability of the model to generalize. A metaheuristic multi-objective optimization approach provides two-dimensional Pareto optimal solutions for selected pairs of target values. This is particularly important for reconciling opposing trends, such as the noise reduction capability and aerodynamic performance. The chosen optimization strategy also allows for a customized design of serrated leading edges, tailored to the specific operating conditions of the axial fan.
In this article we show how fashion brands communicate with their follower on Instagram. We use a continuously update dataset of 68 brands, more than 300,000 posts and more than 40,000,000 comments. Starting with descriptive statistics, we uncover different behavior and success of the various brands. It turns out that there are patterns specific to luxury, mass-market and sportswear brands. Posting volume is extremely brand dependent as is the number of comments and the engagement of the community. Having understood the statistics, we turn to machine learning techniques to measure the response of the community via comments. Topic models help us understand the structure of their respective community and uncover insights regarding the response to campaigns. Having up-to-date content is essential for this kind of analysis, as the market is highly volatile. Furthermore, automatic data analysis is crucial to measure the success of campaigns and adjust them accordingly for maximum effect.
Using Mask R-CNN for Image-Based Wear Classification of Solid Carbide Milling and Drilling Tools
(2020)
In order to ensure high productivity and quality in industrial production, early identification of tool wear is needed. Within the context of Industry 4.0, we integrate wear monitoring of solid carbide milling and drilling cutters automatically into the production process. Therefore, we propose to analyze wear types with image instance segmentation using Mask R-CNN with feature pyramid and bounding box regression. Our approach is able to recognize the five most important wear types: flank wear, crater wear, fracture, built-up edge and plastic deformation. While other methods use image classification and classify only one wear type for each image, our model is able to detect multiple wear types. Over 35 models with different hyperparameter settings were trained on 5,000 labeled images to establish a reliable classifier. The results show up to 82.03% accuracy and benefit for overlapping wear types, which is crucial for using the model in production.
Heterogeneous catalysis, a process in which the reaction of gaseous or liquid chemical reagents is facilitated at the surface of a solid material, is responsible for the majority of industrial-scale chemical and fuel production reactions. The energy required to drive these reactions has historically been derived from the combustion of non-renewable fossil fuels and carries an unavoidably large carbon footprint. More recently, the development of environmentally responsible and sustainable chemical industries is increasingly motivated by greenhouse gas-induced climate change, thus creating demand for eco-friendly heterogeneous catalytic processes. This includes innovative approaches enabled by renewable forms of energy, such as the electrification of chemical and petrochemical processes, utilization of CO2 as a feedstock and the incorporation of light into catalytic reactions. Herein we review the conversion of solar energy to chemical energy using CO2, and describe how the photophysical and photochemical properties of nanostructured metal oxide photocatalysts have been engineered to efficiently incorporate light into heterogeneous gas–solid CO2 hydrogenation reactions. Realizing high photonic and energy efficiencies in these systems has demanded innovation in not only photocatalyst engineering, but also photoreactor and process engineering. Rather than exclusively providing an in-depth discussion of the chemistry and science within each individual study, this Tutorial Review highlights the multidisciplinary character of photocatalysis studies by covering the four essential components of a typical research work in this field (materials engineering, theoretical modelling, reactor engineering and process development) via case studies of the archetypal indium oxide catalyst materials. Through advances in these four components, progress has been made towards the ultimate goal of industrializing the production of CO2-derived chemicals and fuels.
Heterogeneous thermal catalytic processes are vital for industrial production of fuels, fertilizers, and other chemicals necessary for sustaining human life. However, these processes are highly energy-intensive, requiring a vast consumption of fossil fuels. An emerging class of heterogeneous catalysts that are thermally driven but also exhibit a photochemically enhanced rate can potentially reduce process energy intensity by partially substituting conventional heat (where fossil fuels are needed) with solar energy. Such catalyst systems have yet to be practically utilized. Here, we demonstrate a compact electrically heated photo- and thermal annular reactor module to reduce CO2 to CO, via the reverse water gas shift reaction. A first-principles-based design approach was taken in developing a SiO2 on an Al photo- and thermal catalyst system for the model photo- and thermal indium oxide hydroxide (In2O3–x(OH)y) catalysts. A 5-fold light enhancement in the CO production rate and over 70 h of stable CO production were achieved. This represents the highest light enhancement effect reported for this model photocatalyst to date. The reactor presented herein allows continuous operation and a significant reduction of 31% in heater power consumption when provided with an additional 2 suns of irradiation, demonstrating the strong photo- and thermal-harvesting performances of the catalyst system developed in this work.
The ineffectiveness of many gamification projects can be attributed to wrong decisions made during the conceptual design phase, especially in the selection of game design elements. This paper introduces a data driven method of creating a gamification pattern language similar to software design patterns to help gamification designers select such elements. Thanks to modern machine learning technologies such a pattern language can be based on a comprehensive empirical analysis to assess the actual use of game design elements in games. This paper is the first report on an ongoing research project that has been carried out since the beginning of 2017 in cooperation with the German Games Archive to extract game design elements from more than 30,000 board games using machine learning techniques. Initial tests based on support vector classification and 4,000 games show that game design elements can be reliably identified with accuracy rates between 80 and 90%.
High-Performance, Scalable, and Low-Cost Copper Hydroxyapatite for Photothermal CO2 Reduction
(2020)
An essential element for sustainable use of renewable resources is an efficient comminution process for which the relation between energy input and size reduction is of great importance. Fine comminution of beech, oak, and spruce wood chips in cutting and hammer mills at different moisture content levels is investigated. The influence of the different process parameters as well as the size reduction performed by the hammer and cutting mill on the specific comminution energy is reported and the particulate properties of the comminution products are reported. Considering the energy requirements, functional relations were derived from the experimental results, which describe the relation between comminution energy and size reduction.
A novel jet dispersion technique was developed and investigated which enabled excellent carbon nanotube (CNT) dispersion by high exfoliation at even very low pressure drops. Suitable procedures were developed for the characterization of agglomerate size and fraction of individual CNTs. The appropriate characterization enabled the definition of a dimensionless dispersing parameter and the development of a kinetic model that describes CNT dispersion in dependence of the volumetric energy input. The rheological behavior of CNT suspensions in steady‐shear flows was investigated and demonstrated how the agglomerate fracture and CNT individualization influence the suspension viscosity.
Influence of the Temperature on Coccolith‐Containing Systems from Emiliania huxleyi Cultivations
(2020)
Thermogravimetric analysis of a coccolith‐containing biogenic broth showed a three‐step degradation process. According to this system behavior, the biogenic broth was heated to specific temperatures and characterized in terms of its morphology, surface chemistry, and crystallinity. The elemental and organic composition of the treated samples was also evaluated and compared to the reference material. The presented results were acquired in an effort to exploit pretreatment scenarios for such a biogenic system that would improve and support a separation process.
Herausforderungen bei der Etablierung einer Circular Economy von carbonfaservertärkten Kunststoffen (CFK) ergeben sich insbesondere durch die nötige, jedoch problematische Trennung des Faser‐Matrix‐Verbunds. Das Pyrolyseverfahren stellt eine mögliche, bereits industriell umgesetzte, Technologie zur Verbundtrennung dar. Eine Weiterentwicklung ist die mikrowellenunterstütze Pyrolyse, bei der das CFK‐Material durch die Mikrowellen aufgeheizt und so die Faser freigelegt wird. Prozessgrundlagen und Einflüsse auf die Faser‐Matrix‐Trennung werden im vorliegenden Beitrag betrachtet.
Aufgrund steigender energetischer Anforderungen an Gebäude, werden Wärmedämmverbundsysteme (WDVS) seit den 1970 Jahren zur Dämmung von Gebäudefassaden verwendet. Angesichts der Langlebigkeit dieser Kompositmaterialien fallen seit den letzten Jahren vermehrt WDVS‐Abfälle an. Eine werkstoffliche Verwertung dieser Systeme existiert aktuell noch nicht, wodurch ein enormes Ressourceneffizienzpotenzial verloren geht. Aufgrund der Komplexität dieser Kompositmaterialien wurden im Rahmen dieser Arbeit Wärmedämmverbundsysteme durch verfahrenstechnische Prozessschritte aufbereitet.
Die elektrischen Eigenschaften von Partikeln insbesondere der Durchgangswiderstand von Partikelschüttungen haben in vielen verfahrenstechnischen Prozessen, bei der Produktqualifizierung oder dem Brand‐ und Explosionsschutz eine entscheidende Bedeutung. Da sich der Ladungstransport in einem dispersen System von dem in einem homogenen Festkörper unterscheidet, ist der Durchgangswiderstand entscheidend von den Produkteigenschaften der Partikel und des Fluids abhängig. Im Rahmen dieser Arbeit wurde die Abhängigkeit des Durchgangswiderstandes von der inneren Struktur der Schüttung, der relativen Feuchtigkeit, der Konditionierungszeit bei unterschiedlichem Klima und der mittleren Partikelgröße in einer konzentrischen Messzelle untersucht.
Austernschalen stellen als Abfallprodukt eine große Umweltbelastung dar und die Wiederverwendung des Materials gewinnt zunehmend an Bedeutung. Da der Zerkleinerungsprozess beim Recycling von Sekundärrohstoffen die Eigenschaften des Materials definiert, wurden Austernschalen mit einer Planetenkugelmühle und einer Schlagradmühle zerkleinert und die gewonnenen Fraktionen in Bezug auf Größe und Form analysiert und miteinander verglichen. Ein Zerkleinerungsmaterial der Schlagradmühle wurde für eine Reihe von Benetzungsversuchen verwendet, die dazu beitrugen, die freie Oberflächenenergie des Materials nachzuweisen.
Calculation of external climate costs for food highlights inadequate pricing of animal products
(2020)
Although the agricultural sector is globally a main emitter of greenhouse gases, thorough economic analysis of environmental and social externalities has not yet been conducted. Available research assessing agricultural external costs lacks a differentiation between farming systems and food categories. A method addressing this scientific gap is established in this paper and applied in the context of Germany. Using life-cycle assessment and meta-analytical approaches, we calculate the external climate costs of foodstuff. Results show that external greenhouse gas costs are highest for conventional and organic animal-based products (2.41€/kg product; 146% and 71% surcharge on producer price level), followed by conventional dairy products (0.24€/kg product; 91% surcharge) and lowest for organic plant-based products (0.02€/kg product; 6% surcharge). The large difference of relative external climate costs between food categories as well as the absolute external climate costs of the agricultural sector imply the urgency for policy measures that close the gap between current market prices and the true costs of food.
Good management requires proper measurement, yet little is known about anthropogenic climate effects of agriculture. To remedy this, a precise measurement of negative externalities is urgently needed. Therefore, the authors of this article describe the heterogeneity of results from previous studies on climate effects and – focusing on the agricultural sector – identify reasons for this phenomenon. The authors conduct a meta-regression analysis, based on 53 primary studies that cover the period between 1951 and 2015. All countries or country groups are included in the 1345 reported results on emitted amounts of CO2e and SO2e. Our findings confirm the well-known result that an increase in livestock quantities corresponds with a significant increase in emission levels. By integrating culture-related country data, the authors conclude that the level of “humane orientation” and the amounts of emissions follow opposite courses. Furthermore, studies conducted while the first author of this study was working for an NGO, report significantly higher emissions. Based on an adaptation of a meta-regression analysis to the field of environmental performance measurement, we are able to provide new insights about the influence of the change of individual drivers on the emission level. Examples of these insights include a one-third increase in cattle stock is associated with an increase in emissions of 29.45 t/km2. We also find that publications where the first author works for an NGO reported an emission level 87.04 t/km2 higher than other publications. These findings enable the identification of the main drivers of emissions, while helping to explain the heterogeneity of existing studies. Based on the findings of this study, companies can take reliable measures to reduce the external climate effects of their products.
Good management requires proper measurement, yet little is known about anthropogenic climate effects of agriculture. To remedy this, a precise measurement of negative externalities is urgently needed. Therefore, the authors of this article describe the heterogeneity of results from previous studies on climate effects and – focusing on the agricultural sector – identify reasons for this phenomenon. The authors conduct a meta-regression analysis, based on 53 primary studies that cover the period between 1951 and 2015. All countries or country groups are included in the 1345 reported results on emitted amounts of CO2e and SO2e. Our findings confirm the well-known result that an increase in livestock quantities corresponds with a significant increase in emission levels. By integrating culture-related country data, the authors conclude that the level of “humane orientation” and the amounts of emissions follow opposite courses. Furthermore, studies conducted while the first author of this study was working for an NGO, report significantly higher emissions. Based on an adaptation of a meta-regression analysis to the field of environmental performance measurement, we are able to provide new insights about the influence of the change of individual drivers on the emission level. Examples of these insights include a one-third increase in cattle stock is associated with an increase in emissions of 29.45 t/km2. We also find that publications where the first author works for an NGO reported an emission level 87.04 t/km2 higher than other publications. These findings enable the identification of the main drivers of emissions, while helping to explain the heterogeneity of existing studies. Based on the findings of this study, companies can take reliable measures to reduce the external climate effects of their products.
SafeML: Safety Monitoring of Machine Learning Classifiers Through Statistical Difference Measures
(2020)
ZusammenfassungDigital angebotene psychologische Interventionen gegen Schlafstörungen sind aktuell ein sehr intensiv bearbeitetes Forschungsthema. In dieser Übersichtsarbeit werden Originalarbeiten und Metaanalysen zu diesem Thema zusammengefasst. Hierbei zeigt sich, dass die internetbasierte kognitive Verhaltenstherapie für Insomnie (KVT-I) bei Erwachsenen durchweg sehr effektiv ist mit allenfalls leicht geringeren Effektstärken als die gleiche Behandlung mit physischer Präsenz von Therapeuten und Patienten. Behandlungseffekte zeigen sich dabei auch für sekundäre Outcome-Parameter wie Depressivität, Angst, Fatigue und Lebensqualität. Hingegen lassen die Forschung zur Wirksamkeit der internetbasierten KVT‑I bei Erwachsenen mit komorbiden psychischen Störungen oder körperlichen Erkrankungen sowie die Forschung zur Frage, wieviel Therapeutenkontakt in die Behandlung integriert werden sollte, anscheinend noch keine abschließenden Antworten zu. In diesen Bereichen scheint weitere Forschung notwendig zu sein scheint.
Shiftwork can be a risk factor for a number of different somatic and psychological health conditions, especially sleep disorders. Shiftworkers sleep less than dayworkers, and 20–40% of them suffer from difficulties initiating and maintaining sleep, which result in reduced capacity for work and social life. A common coping strategy might be the use of alcohol, which presents a health and safety hazard as it further impairs sleep quality and exacerbates sleepiness in the workplace. This review aimed to assess the extent of such possible connections.
The vapor pressures of two binary mixtures containing 2-trifluoroethanol (TFE) + 1,3-dimethyl-2-imidazolidinone (DMI) and TFE + 2-pyrrolidone (PYR), were investigated at temperatures T = (274.15 to423.15) K using two different static method installations. Both combinations were modelled using an extended
Clausius-Clapeyron equation with concentration dependent parameters and the NRTL equation with temperature dependent parameters. The best fit was obtained using the NRTL equation.
This publication describes an application of a
Truncated Signed Distance Mapping approach for disaster
intervention in underground mine shafts through geometrical
change detection of the shaft walls. The paper describes two
main problems of such an approach (aligning two potentially
huge point clouds and automatic change detection by comparing
the reconstructed volumes) and explains in detail the proposed
solution.
The influence of corporate diversification on firm value is an important field in strategy research. Studies in strategic management and finance research have analyzed value creation through product and geographic diversification from a shareholder’s perspective. This study completes this picture by analyzing the bondholders’ perspective. It is suggested that product diversification creates value for bondholders, while geographic diversification destroys bondholder value. The hypotheses are tested on a sample of S&P 1,200 firms in 2001–2011 using a fixed-effects panel model. Drawing on prior research, bondholder value creation is measured using the Merton model. The empirical results support the hypothesis that bondholders gain value through product diversification but lose value through geographic diversification. Considering prior research results, these results show that product diversification is preferable for bondholders, while geographic diversification is preferable from a shareholder’s perspective. The opposite effects of both diversification strategies on shareholders, respectively, bondholders offer an important new perspective on corporate diversification. The results show that firms with a high level of corporate debt should struggle to justify a strategy involving geographic dispersion of activities and support a more diversified product portfolio strategy. This study also offers several avenues for investigating the bondholder’s perspective on corporate diversification in more detail.
In this paper, a novel state machine for mobile robots is described that enables a direct use for exploration and inspection tasks. It offers a graphical user interface (GUI) to supervise the process and to issue commands if necessary. The state machine was developed for the open-source framework Robot Operating System (ROS) and can interface arbitrary algorithms for navigation and exploration. Interfaces to the commonly used ROS navigation stack and the explore_lite package are already included and can be utilized. In addition, routines for mapping and inspection can be added freely to adapt to the area of application. The state machine features a teleoperation mode to which it changes as soon as a respective command was issued. It also implements a software emergency stop and multiplexes all movement commands to the motor controller. To show the state machine's capabilities several simulations and real-world experiments are described in which it was used.
Qualitative content analysis is a systematic method commonly used in the social sciences to analyze textual data from interviews or online discussions. However, this method usually requires high expertise and manual effort because human coders need to read, interpret, and manually annotate text passages. This is especially true if the system of categories used for annotation is complex and semantically rich. Therefore, qualitative content analysis could benefit greatly from automated coding. In this work, we investigate the usage of machine learning-based text classification models for automatic coding in the area of psycho-social online counseling. We developed a system of over 50 categories to analyze counseling conversations, labeled over 10.000 text passages manually, and evaluated the performance of different machine learning-based classifiers against human coders.
For the production of brittle materials like technical ceramics it is important to prohibit big defects in the material to achieve high mechanical strength. Due to that already the granule properties are important because they can cause defects in the final product. Characterization and quantification of the granule properties are important steps for improving the production process and properties of the granules for further processing. In this work a novel method for the characterization and quantification of granule morphology, especially sphericity, and internal structure like the ratio of macro pores is provided via digital microscopy. The new quantitative characterization method can be done in an academic and industrial field with appropriate effort. Differences between granules can be characterized, quantified and be used for further development.
As part of a collaborative project, we took the window shutter — a widely used traditional facade element — and reinterpreted it for “adaptive” solutions using new, sustainable technologies. The aim was to design a modular construction kit that varies radiant energy, light and air flow to weather conditions and occupant requirements in new and existing residential and office buildings.
Die Insomnie ist eine der häufigsten Gesundheitsstörungen im höheren Lebensalter. Sie erzeugt Leidensdruck und verursacht zahlreiche gesundheitliche Probleme. Daher ist ihre Behandlung oft indiziert. Die Verhaltenstherapie ist auch bei älteren Menschen die Behandlung der Wahl. Zudem hat die Lichttherapie ihren Stellenwert. Pharmakologische Maßnahmen sind weniger gut untersucht, sind unklar in ihrem Nutzen bei einer Langzeitanwendung und sollten nur kurzfristig zur Reduktion von Leidensdruck sowie eingebunden in ein therapeutisches Gesamtkonzept zur Anwendung kommen.
Performing machine learning tasks in mobile applications yields a challenging conflict of interest: highly sensitive client information (e.g., speech data) should remain private while also the intellectual property of service providers (e.g., model parameters) must be protected. Cryptographic techniques offer secure solutions for this, but have an unacceptable overhead and moreover require frequent network interaction.In this work, we design a practically efficient hardware-based solution. Specifically, we build OFFLINE MODEL GUARD (OMG) to enable privacy-preserving machine learning on the predominant mobile computing platform ARM—even in offline scenarios. By leveraging a trusted execution environment for strict hardware-enforced isolation from other system components, OMG guarantees privacy of client data, secrecy of provided models, and integrity of processing algorithms. Our prototype implementation on an ARM HiKey 960 development board performs privacy-preserving keyword recognition using TensorFlow Lite for Microcontrollers in real time.
In the industrial application of additive manufacturing processes, a significant amount of time and resources are dedicated to the orientation and pre-print setup of the geometry. Steps such as the generation of support structures and the process simulation are among the most time-consuming. For the thorough assessment of an orientation of a given geometry, even more criteria, like print time or surface quality, need to be considered. POEAM proposes a method for an efficient assessment of a set of orientations, by means of well formulated criteria and an early elimination of insufficient orientations. The goal is to narrow the search field, so costly preparation steps will only be performed on orientations that promise a superior end result. Furthermore, POEAM is an automated process, which means it can be performed with minimal human interaction, resulting in an optimum regarding cost-efficiency and evaluation time. The method was applied to a representative geometry and has shown results that confirm the above-mentioned advantages.
Basierend auf einer experimentellen Messkampagne
werden zwei Simulationsmodelle mit unterschiedlichen Modellierungstiefen, welche die Anlagenkomponenten einer Sole-Wasser-Wärmepumpe und eines
thermischen Energiespeichers abbilden, validiert. Auf
Grundlage von Referenzlastprofilen nach VDI 4655
werden Messungen an einem Hardware-in-the-Loop
Wärmepumpenprüfstand unter realitätsnahen Betriebssituationen und Randbedingungen durchgeführt.
Die Validierungsergebnisse zeigen generell eine hohe
Genauigkeit des detaillierten Modells gegenüber den
gemessenen Systemgrößen, aber auch dem vereinfachten Modell, auf. So können beispielsweise absolute Fehler im Ladezustand des Speichers von
höchstens 7,2 Prozentpunkten und von der mittleren
Speichertemperatur von 1,4 °C ermittelt werden. Der
Einsatz des detaillierten Modells ist für eine zukünftige Entwicklung von Regelungsstrategien zu bevorzugen
Der steigende Einsatz erneuerbarer Energien im Wohnungsbau erfordert intelligente Lösungen, um die lokal erzeugte Energie vor Ort nutzen zu können. Das Ziel dieser Forschungsarbeit ist es, optimierte Betriebsführungsstrategien für modulierende Wärmepumpen (MWP) mit Speichern zu entwickeln und diese in einem Plusenergie-Reihenhauskomplex anzuwenden, um den Eigenverbrauch der Photovoltaikerzeugung (PV) und die Effizienz zu erhöhen. Der Betrieb und der Vergleich zweier optimierter Betriebsführung-sstrategien zeigen, dass beide einen ähnlich hohen PV-Eigenverbrauch erzielen, aber zu unterschiedlichen Arbeitszahlen führen.
Stuttering is a complex speech disorder that can be identified by repetitions, prolongations of sounds, syllables or words and blocks while speaking. Severity assessment is usually done by a speech therapist. While attempts at automated assessment were made, it is rarely used in therapy. Common methods for the assessment of stuttering severity include percent stuttered syllables (% SS), the average of the three longest stuttering symptoms during a speech task or the recently introduced Speech Efficiency Score (SES). This paper introduces the Speech Control Index (SCI), a new method to evaluate the severity of stuttering. Unlike SES, it can also be used to assess therapy success for fluency shaping. We evaluate both SES and SCI on a new comprehensively labeled dataset containing stuttered German speech of clients prior to, during and after undergoing stuttering therapy. Phone alignments of an automatic speech recognition system are statistically evaluated in relation to their relative position to labeled stuttering events. The results indicate that phone length distributions differ in respect to their position in and around labeled stuttering events.
The magnitude of current harmonics depends on the design of an electrical machine. By suppressing these harmonics noise can be reduced and efficiency improved. Iterative Learning Control (ILC) has proven effective in reducing harmonics. One of the challenges of working with ILC is operation at varying speeds. Variable speeds are particularly important for applications like automotive drives. The ILC period length changes during the learning process at varying speeds. Due to fixed sample rates, the number of values processed by the ILC varies with motor speed.This paper proposes a method to solve this problem and uses ILC at varying speeds. The ILC used to eliminate the harmonics is based on the inverse system. The usage of a two-dimensional memory array is proposed. This data structure holds rows for specific speeds between which interpolation is performed, enabling the elimination of errors which are periodically cyclic to one electrical rotation. This includes the reduction of the motor current harmonics. To verify the presented method a permanent magnet synchronous motor with distinctive 5th and 7th harmonics is used. In real-time implementations, limitations of memory and computational capacity occur.
Parkinson's disease is a neurodegenerative disorder characterized by the presence of different motor impairments. Information from speech, handwriting, and gait signals have been considered to evaluate the neurological state of the patients. On the other hand, user models based on Gaussian mixture models - universal background models (GMMUBM) and i-vectors are considered the state-of-the-art in biometric applications like speaker verification because they are able to model specific speaker traits. This study introduces the use of GMM-UBM and i-vectors to evaluate the neurological state of Parkinson's patients using information from speech, handwriting, and gait. The results show the importance of different feature sets from each type of signal in the assessment of the neurological state of the patients.
Offline Model Guard
(2020)
Performing machine learning tasks in mobile applications yields a challenging conflict of interest: highly sensitive client information (e.g., speech data) should remain private while also the intellectual property of service providers (e.g., model parameters) must be protected. Cryptographic techniques offer secure solutions for this, but have an unacceptable overhead and moreover require frequent network interaction. In this work, we design a practically efficient hardware-based solution. Specifically, we build Offline Model Guard (OMG) to enable privacy-preserving machine learning on the predominant mobile computing platform ARM - even in offline scenarios. By leveraging a trusted execution environment for strict hardware-enforced isolation from other system components, OMG guarantees privacy of client data, secrecy of provided models, and integrity of processing algorithms. Our prototype implementation on an ARM HiKey 960 development board performs privacy-preserving keyword recognition using TensorFlow Lite for Microcontrollers in real time.
Deep neural networks (DNN) have recently been widely used in speaker recognition systems, achieving state-of-the-art performance on various benchmarks. The x-vector architecture is especially popular in this research community, due to its excellent performance and manageable computational complexity. In this paper, we present the lrx-vector system, which is the low-rank factorized version of the x-vector embedding network. The primary objective of this topology is to further reduce the memory requirement of the speaker recognition system. We discuss the deployment of knowledge distillation for training the lrx-vector system and compare against low-rank factorization with SVD. On the VOiCES 2019 far-field corpus we were able to reduce the weights by 28% compared to the full-rank x-vector system while keeping the recognition rate constant (1.83% EER).
Speaker recognition performance has been greatly improved with the emergence of deep learning. Deep neural networks show the capacity to effectively deal with impacts of noise and reverberation, making them attractive to far-field speaker recognition systems. The x-vector framework is a popular choice for generating speaker embeddings in recent literature due to its robust training mechanism and excellent performance in various test sets. In this paper, we start with early work on including invariant representation learning (IRL) to the loss function and modify the approach with centroid alignment (CA) and length variability cost (LVC) techniques to further improve robustness in noisy, far-field applications. This work mainly focuses on improvements for short-duration test utterances (1-8s). We also present improved results on long-duration tasks. In addition, this work discusses a novel self-attention mechanism. On the VOiCES far-field corpus, the combination of the proposed techniques achieves relative improvements of 7.0% for extremely short and 8.2% for full-duration test utterances on equal error rate (EER) over our baseline system.