- Textile materials and evaluations
- Structural Analysis and Optimization
- Mechanical Behavior of Composites
- Multimodal Machine Learning Applications
- Additive Manufacturing and 3D Printing Technologies
- Domain Adaptation and Few-Shot Learning
- Structural Analysis of Composite Materials
- Software System Performance and Reliability
- Data Stream Mining Techniques
- Neural Networks and Applications
- Cellular and Composite Structures
- Time Series Analysis and Forecasting
- Air Quality Monitoring and Forecasting
- Ectopic Pregnancy Diagnosis and Management
- Advanced Bandit Algorithms Research
- biodegradable polymer synthesis and properties
- Assisted Reproductive Technology and Twin Pregnancy
- Maternal and Perinatal Health Interventions
- Reproductive Health and Technologies
- Prenatal Screening and Diagnostics
- Advanced Neural Network Applications
- Advanced Database Systems and Queries
- Crime Patterns and Interventions
- Software Reliability and Analysis Research
- Engineering and Materials Science Studies
Aker BP (Norway)
2025
Technische Universität Dresden
2018-2023
Institute of Textile Technology and Process Engineering
2023
Existing continual learning methods use Batch Normalization (BN) to facilitate training and improve generalization across tasks. However, the non-i.i.d non-stationary nature of data, especially in online setting, amplify discrepancy between testing BN hinder performance older In this work, we study cross-task normalization effect where normalizes data using moments biased towards current task, resulting higher catastrophic forgetting. This limitation motivates us propose a simple yet...
Objective: To evaluate the short-term outcomes of percutaneous closure ventricular septal defects in patients all age. Summary: Ventricular defect (VSD) is most common congenital heart and can be detected during prenatal postnatal period. In 1987, Lock et al first applied technique interventional occlusion VSD using devices through a catheter this method quickly spread worldwide. Compared with surgery to repair VSD, intervention occlude septum helps minimize complications after procedure,...
Objectives: To review the indications and results of cesarean sections for nulliparous pregnancies at theNational Hospital Tropical Diseases in 2023.Subjects methods: A cross-sectional descriptive study combined with a retrospective medical records who delivered ≥ 28 weeks gestational age National from May 2021 to November 2023.Results: Research on 200 shows that rate first-born babies accounted 67% total sections. Indications section were due genital tract issues (13.5%), fetal position...
Abstract Over the past decade, adoption of drilling process automation solutions has significantly enhanced operation efficiency. This improvement been achieved because Drilling Control System (DCS) providers have enabled external service companies to offer dynamic advice that can be used parameterize Automatic (ADCS) functions. However, developing externally managed ADCS functions poses challenges, particularly given risk potential failures due spurious parameters. Therefore, it is...
Weft-knitted fabrics offer an excellent formability into complex shapes for composite application. In biaxial weft-knitted fabric, additional yarns are inserted in the warp (wale-wise) and weft (course-wise) directions as a reinforcement. Due to these straight yarns, mechanical properties of such better than those unreinforced fabrics. The forming process flat 3D preforms is challenging requires numerical simulation. this paper, behavior simulated by means macro- meso-scale finite element...
The fast adaptation capability of deep neural networks in non-stationary environments is critical for online time series forecasting. Successful solutions require handling changes to new and recurring patterns. However, training forecaster on the fly notoriously challenging because their limited ability adapt catastrophic forgetting old knowledge. In this work, inspired by Complementary Learning Systems (CLS) theory, we propose Fast Slow learning Networks (FSNet), a holistic framework...
Numerical modelling of textile materials is important for developing new forming processes reinforcements composite parts, namely to decide on tool geometry and process parameters such as blank holder forces. A numerical model reduces the application trial-and-error, which significantly costs material resources. The magnitude distribution forces play a decisive role in quality formed textile. In general, no or even low will lead wrinkles useful part preform. This mechanical parts reinforced...
Shell-rib structures made of textile-reinforced composites are used in a wide range applications to increase bending, buckling and torsional stiffness. Such usually manufactured differential construction at the preform level by assembling several textile or component subsequent joining separately shells stiffening structures. Integral production is one way overcome disadvantages forenamed methods, such as high manual effort, failure during fiber distortion. Weft-knitting technology excellent...
Organizations are awash in data. In many cases, they do not know what data exists within the organization and much information is available when needed, or worse, gets recreated from other sources. this paper, we present an automatic approach to spatio-temporal indexing of datasets organization. The process automatically identifies spatial temporal fields, normalizes cleans those then loads them into a big store where can be efficiently searched, queried, analyzed. We evaluated our on 600...
Abstract The models for textile structures based on the Finite Element Method (FEM) are diverse. In general, they can be divided into three groups their length scale: macro‐, meso‐, and micro‐scale models. macro‐scale simplest approach require least computational cost. On contrary, most realistic complex, thus involving extremely high costs. meso‐scale stand in middle with sufficient representation degree at a reasonable Every group of has its own advantageous application field. For forming...
Deep Neural Networks (DNNs) are typically trained by backpropagation in a batch learning setting, which requires the entire training data to be made available prior task. This is not scalable for many real-world scenarios where new arrives sequentially stream form. We aim address an open challenge of "Online Learning" (ODL) DNNs on fly online setting. Unlike traditional that often optimizes some convex objective function with respect shallow model (e.g., linear/kernel-based hypothesis), ODL...
Orthogonal parameterization is a compelling solution to the vanishing gradient problem (VGP) in recurrent neural networks (RNNs). With orthogonal parameters and non-saturated activation functions, gradients such models are constrained unit norms. On other hand, although traditional vanilla RNNs seen have higher memory capacity, they suffer from VGP perform badly many applications. This work proposes Adaptive-Saturated (asRNN), variant that dynamically adjusts its saturation level between two...
Recent advancements in Code Large Language Models (CodeLLMs) have predominantly focused on open-ended code generation tasks, often neglecting the critical aspect of understanding and comprehension. To bridge this gap, we present CodeMMLU, a comprehensive multiple-choice question-answer benchmark designed to evaluate depth software LLMs. CodeMMLU includes over 10,000 questions sourced from diverse domains, encompassing tasks such as analysis, defect detection, engineering principles across...
Fiber-reinforced plastic (FRP) structures are established in numerous lightweight solutions. Among textile products available for technical applications, flat woven fabrics commonly used to produce 3D components. In order convert the into geometries, draping and cutting processes applied primarily. This leads structural distortions, yarn interruption overlaps of layers. Beside resulting reduction mechanical properties, manual work steps necessary. offer outstanding possibilities realizing...
Lightweight panels are of high relevance for various applications, such as automotive, aerospace, civil engineering, and achieve stiffnesses strengths at low self-weight. The sandwich principle is commonly used to manufacture the panels, although conventional materials reinforcement structures often limit design application in a wide range possible applications. reason this that lightweight fail either under combined compressive/shear loading or result delamination individual layers. In...
Die numerische Simulation hilft bei der Analyse textiler Strukturen. Zudem unterstützt sie die technische Entwicklung verschnittfreier und endkonturgerecht gewebter 3D-Preformen, was Voraussetzung für einen einstufigen Fertigungsprozess ist (direktes Preforming). Der simulationsgestützte Entwicklungsprozess reduziert nicht nur Trial-and-Error-Versuchsreihen, sondern sagt auch Grenzen aktuellen Technik voraus. Flexibilität Maschinentechnik ermöglicht Herstellung mehrlagiger stark...
The bending and shearing properties have a major influence on the drapability of 2D fabrics while forming 3D preforms for composites. behaviour are dependent yarn stiffness, interaction between systems, density systems in textile structure. Weft-knitted known their excellent due to internal Meso-scale models two biaxial reinforced weft-knitted with different configurations used predict fabrics. Results show an agreement simulations compared experimental trials, thus, making model valid tool...
The goal of this paper is to summarize methodologies used in extracting entities and topics from a database criminal records newspapers. Statistical models had successfully been studying the roughly 300,000 New York Times articles. In addition, these also analyze related people, organizations, places (D Newman, 2006). Additionally, analytical approaches, especially hotspot mapping, were some researches with an aim predict crime locations circumstances future, those approaches tested quite (S...
Continual learning aims to learn continuously from a stream of tasks and data in an online-learning fashion, being capable exploiting what was learned previously improve current future while still able perform well on the previous tasks. One common limitation many existing continual methods is that they often train model directly all available training without validation due nature learning, thus suffering poor generalization at test time. In this work, we present novel framework named...
According to the Complementary Learning Systems (CLS) theory~\cite{mcclelland1995there} in neuroscience, humans do effective \emph{continual learning} through two complementary systems: a fast learning system centered on hippocampus for rapid of specifics, individual experiences; and slow located neocortex gradual acquisition structured knowledge about environment. Motivated by this theory, we propose \emph{DualNets} (for Dual Networks), general continual framework comprising supervised...
Preformen für komplex gekrümmte Schalenstrukturen den Einsatz in faserverstärkten Kunstoffen werden aktuell meist mittels sequentiellem Preforming gefertigt, wodurch zeit- und materialaufwändige Prozessschritte notwendig sind. Um diese Arbeitsschritte zu minimieren hohen Bedarf seitens der Industrie an dreidimensionalen, schalenförmigen decken, wurde im Rahmen des IGF-Projekts 19805 BR die neuartige Technologie abzugsfreien Webens entwickelt. Diese ermöglicht das direkte Weben sphärisch...
Das Open-Reed-Weaving (ORW) ermöglicht die Einbringung einer formschlussbildenden Profilierung in Bewehrungsstrukturen für Textilbeton. Die Herausforderung bei diesem Verfahren liegt hauptsächlich der Realisierung Betonarmierung benötigten geringen Strukturdehnung. Im Rahmen des IGF-Forschungsprojekts wurde untersucht, inwieweit mit simulationsgestützt entwickelten ORW-Bewehrungsstrukturen eine formschlussbildende gleichzeitig geringer Strukturdehnung umgesetzt werden kann. Darüber hinaus...