Conference Proceeding
Refine
Year of publication
Document Type
- Conference Proceeding (1016) (remove)
Language
- English (1016) (remove)
Has Fulltext
- no (1016) (remove)
Keywords
- Enterprise Architecture (5)
- Energy storage (4)
- Gamification (4)
- Natural language processing (4)
- Power plants (4)
- hydrogen (4)
- solar sail (4)
- Associated liquids (3)
- Concentrated solar power (3)
- Hybrid energy system (3)
- MASCOT (3)
- Out-of-plane load (3)
- earthquakes (3)
- Additive manufacturing (2)
- Adjacent buildings (2)
- Case Study (2)
- Clustering (2)
- Deep learning (2)
- Digital Twin (2)
- Diversity (2)
Institute
- Fachbereich Elektrotechnik und Informationstechnik (224)
- Fachbereich Luft- und Raumfahrttechnik (171)
- Fachbereich Energietechnik (158)
- Fachbereich Medizintechnik und Technomathematik (131)
- IfB - Institut für Bioengineering (109)
- Solar-Institut Jülich (108)
- Fachbereich Maschinenbau und Mechatronik (98)
- Fachbereich Bauingenieurwesen (70)
- ECSM European Center for Sustainable Mobility (50)
- Fachbereich Wirtschaftswissenschaften (42)
- MASKOR Institut für Mobile Autonome Systeme und Kognitive Robotik (42)
- INB - Institut für Nano- und Biotechnologien (33)
- Fachbereich Chemie und Biotechnologie (23)
- Kommission für Forschung und Entwicklung (16)
- Nowum-Energy (11)
- Fachbereich Architektur (7)
- Fachbereich Gestaltung (3)
- Institut fuer Angewandte Polymerchemie (2)
- ZHQ - Bereich Hochschuldidaktik und Evaluation (2)
- Arbeitsstelle fuer Hochschuldidaktik und Studienberatung (1)
Heavy metal detection with semiconductor devices based on PLD-prepared chalcogenide glass thin films
(2007)
In recent years, the development of large pretrained language models, such as BERT and GPT, significantly improved information extraction systems on various tasks, including relation classification. State-of-the-art systems are highly accurate on scientific benchmarks. A lack of explainability is currently a complicating factor in many real-world applications. Comprehensible systems are necessary to prevent biased, counterintuitive, or harmful decisions.
We introduce semantic extents, a concept to analyze decision patterns for the relation classification task. Semantic extents are the most influential parts of texts concerning classification decisions. Our definition allows similar procedures to determine semantic extents for humans and models. We provide an annotation tool and a software framework to determine semantic extents for humans and models conveniently and reproducibly. Comparing both reveals that models tend to learn shortcut patterns from data. These patterns are hard to detect with current interpretability methods, such as input reductions. Our approach can help detect and eliminate spurious decision patterns during model development. Semantic extents can increase the reliability and security of natural language processing systems. Semantic extents are an essential step in enabling applications in critical areas like healthcare or finance. Moreover, our work opens new research directions for developing methods to explain deep learning models.
Multi-attribute relation extraction (MARE): simplifying the application of relation extraction
(2021)
Natural language understanding’s relation extraction makes innovative and encouraging novel business concepts possible and facilitates new digitilized decision-making processes. Current approaches allow the extraction of relations with a fixed number of entities as attributes. Extracting relations with an arbitrary amount of attributes requires complex systems and costly relation-trigger annotations to assist these systems. We introduce multi-attribute relation extraction (MARE) as an assumption-less problem formulation with two approaches, facilitating an explicit mapping from business use cases to the data annotations. Avoiding elaborated annotation constraints simplifies the application of relation extraction approaches. The evaluation compares our models to current state-of-the-art event extraction and binary relation extraction methods. Our approaches show improvement compared to these on the extraction of general multi-attribute relations.
Supervised machine learning and deep learning require a large amount of labeled data, which data scientists obtain in a manual, and time-consuming annotation process. To mitigate this challenge, Active Learning (AL) proposes promising data points to annotators they annotate next instead of a subsequent or random sample. This method is supposed to save annotation effort while maintaining model performance.
However, practitioners face many AL strategies for different tasks and need an empirical basis to choose between them. Surveys categorize AL strategies into taxonomies without performance indications. Presentations of novel AL strategies compare the performance to a small subset of strategies. Our contribution addresses the empirical basis by introducing a reproducible active learning evaluation (ALE) framework for the comparative evaluation of AL strategies in NLP.
The framework allows the implementation of AL strategies with low effort and a fair data-driven comparison through defining and tracking experiment parameters (e.g., initial dataset size, number of data points per query step, and the budget). ALE helps practitioners to make more informed decisions, and researchers can focus on developing new, effective AL strategies and deriving best practices for specific use cases. With best practices, practitioners can lower their annotation costs. We present a case study to illustrate how to use the framework.
The progress in natural language processing (NLP) research over the last years, offers novel business opportunities for companies, as automated user interaction or improved data analysis. Building sophisticated NLP applications requires dealing with modern machine learning (ML) technologies, which impedes enterprises from establishing successful NLP projects. Our experience in applied NLP research projects shows that the continuous integration of research prototypes in production-like environments with quality assurance builds trust in the software and shows convenience and usefulness regarding the business goal. We introduce STAMP 4 NLP as an iterative and incremental process model for developing NLP applications. With STAMP 4 NLP, we merge software engineering principles with best practices from data science. Instantiating our process model allows efficiently creating prototypes by utilizing templates, conventions, and implementations, enabling developers and data scientists to focus on the business goals. Due to our iterative-incremental approach, businesses can deploy an enhanced version of the prototype to their software environment after every iteration, maximizing potential business value and trust early and avoiding the cost of successful yet never deployed experiments.
Scientific questions
- How can a non-stationary heat offering in the commercial vehicle be used to reduce fuel consumption?
- Which potentials offer route and environmental information among with predicted speed and load trajectories to increase the efficiency of a ORC-System?
Methods
- Desktop bound holistic simulation model for a heavy duty truck incl. an ORC System
- Prediction of massflows, temperatures and mixture quality (AFR) of exhaust gas
In energy economy forecasts of different time series are rudimentary. In this study, a prediction for the German day-ahead spot market is created with Apache Spark and R. It is just an example for many different applications in virtual power plant environments. Other examples of use as intraday price processes, load processes of machines or electric vehicles, real time energy loads of photovoltaic systems and many more time series need to be analysed and predicted.
This work gives a short introduction into the project where this study is settled. It describes the time series methods that are used in energy industry for forecasts shortly. As programming technique Apache Spark, which is a strong cluster computing technology, is utilised. Today, single time series can be predicted. The focus of this work is on developing a method to parallel forecasting, to process multiple time series simultaneously with R and Apache Spark.
This study investigates the influence of pressure on the temperature distribution of the micromix (MMX) hydrogen flame and the NOx emissions. A steady computational fluid dynamic (CFD) analysis is performed by simulating a reactive flow with a detailed chemical reaction model. The numerical analysis is validated based on experimental investigations. A quantitative correlation is parametrized based on the numerical results. We find, that the flame initiation point shifts with increasing pressure from anchoring behind a downstream located bluff body towards anchoring upstream at the hydrogen jet. The numerical NOx emissions trend regarding to a variation of pressure is in good agreement with the experimental results. The pressure has an impact on both, the residence time within the maximum temperature region and on the peak temperature itself. In conclusion, the numerical model proved to be adequate for future prototype design exploration studies targeting on improving the operating range.
Experimental and numerical investigation on the effect of pressure on micromix hydrogen combustion
(2021)
The micromix (MMX) combustion concept is a DLN gas turbine combustion technology designed for high hydrogen content fuels. Multiple non-premixed miniaturized flames based on jet in cross-flow (JICF) are inherently safe against flashback and ensure a stable operation in various operative conditions.
The objective of this paper is to investigate the influence of pressure on the micromix flame with focus on the flame initiation point and the NOx emissions. A numerical model based on a steady RANS approach and the Complex Chemistry model with relevant reactions of the GRI 3.0 mechanism is used to predict the reactive flow and NOx emissions at various pressure conditions. Regarding the turbulence-chemical interaction, the Laminar Flame Concept (LFC) and the Eddy Dissipation Concept (EDC) are compared. The numerical results are validated against experimental results that have been acquired at a high pressure test facility for industrial can-type gas turbine combustors with regard to flame initiation and NOx emissions.
The numerical approach is adequate to predict the flame initiation point and NOx emission trends. Interestingly, the flame shifts its initiation point during the pressure increase in upstream direction, whereby the flame attachment shifts from anchoring behind a downstream located bluff body towards anchoring directly at the hydrogen jet. The LFC predicts this change and the NOx emissions more accurately than the EDC. The resulting NOx correlation regarding the pressure is similar to a non-premixed type combustion configuration.
Investigation Of The Seismic Behaviour Of Infill Masonry Using Numerical Modelling Approaches
(2017)
Masonry is a widely spread construction type which is used all over the world for different types of structures. Due to its simple and cheap construction, it is used as non-structural as well as structural element. In frame structures, such as
reinforced concrete frames, masonry may be used as infill. While the bare frame itself is able to carry the loads when it comes to seismic events, the infilled frame is not able to warp freely due to the constrained movement. This restraint results in a complex interaction between the infill and the surrounding frame, which may lead to severe damage to the infill as well as the surrounding frame. The interaction is studied in different projects and effective approaches for the description of the behavior are still lacking. Experimental programs are usually quite expensive, while numerical models, once validated, do offer an efficient approach for the investigation of the interaction when horizontally loaded. In order to study the numerous parameters influencing the seismic load bearing behavior, numerical models may be used. Therefore, this contribution presents a numerical approach for the simulation of infill masonry in reinforced concrete frames. Both parts, the surrounding frame as well as the infill are represented by micro modelling approaches to correctly take into account the different types of failure. The adopted numerical model describes the inelastic behavior of the system, as indicated by the obtained results of the overall structural response as well as the formation of damage in the infilled wall. Comparison of the numerical and experimental results highlights the valuable contribution of numerical simulations in the study and design of infilled frames. As damage of the infill masonry may occur in-plane due to the interaction as well as out-of-plane due to the low vertical load, both directions of loading are investigated.
During the development of a Competence Developing Game’s (CDG) story it is indispensable to understand the target audience. Thereby, CDGs stories represent more than just the plot. The Story is about the
Setting, the Characters and the Plot. As a toolkit to support the
development of such a story, this paper introduces the UserFocused Storybuilding (short UFoS) Framework for CDGs. The Framework and its utilization will be explained, followed by a description of its development and derivation, including an empirical study. In addition, to simplify the Framework use regarding the CDG’s target audience, a new concept of Nine Psychographic Player Types will be explained. This concept of Player Types provides an approach to handle the differences in between players during the UFoS Framework use. Thereby,
this article presents a unique approach to the development of
target group-differentiated CDGs stories.