1,165 research outputs found
Deriving translational acoustic sub-word embeddings
No full textThere is a growing interest in understanding the representational geometry of acoustic word embeddings (AWEs), which are fixed-dimensional representations of spoken words. However, not much research has been conducted on acoustic sub-word embeddings (ASWEs), which can provide a better understanding of the AWE space. This work focuses on decomposing AWEs to obtain ASWEs while retaining the ability to reconstruct AWEs by translating ASWEs in the embedding space, under constrained settings. Initially, high-quality AWEs are obtained with an Average Precision (AP) score of 0.97 on the word discrimination task. Subsequently, ASWEs are derived through the decomposition of AWEs. Three adapted versions of the AP metric, utilized for evaluating the quality of the derived ASWEs and their translational properties, are proposed. The results demonstrate that the derived ASWEs exhibit high quality, and the reconstruction of AWEs from the ASWEs is achievable by translating them in the embedding space
Single Skin Kite Airfoil Optimization for AWES
No full textAirborne Wind Energy is a technology where wind energy is harvested with tethered flying devices. Kitepower uses flexible leading edge inflatable kites, but these have a scaling disadvantage in that they become heavier with size. A single skin kite has the potential of negating this disadvantage while at the same time being more aerodynamically efficient. An airfoil of this type is therefore investigated using Computational Fluid Dynamics and optimized using Surrogate Modelling techniques. A hybrid mesh was generated with hyperbolic extrusion and triangulation. The RANS solver that was used produced good results.The results of the optimization were unsatisfactory. The parametrization did not provide enough local control and unique airfoil shapes. The surrogate modelling approach is promising due to the computationally expensive CFD analyses.Aerospace Engineerin
Sistema de monitoreo multivariable para electrolizadores alcalinos de baja escala
No full textEn este estudio se enfatiza la relevancia de la transición hacia energías renovables mediante la implementación de electrolizadores alcalinos (AWEs) para la producción de hidrógeno verde. Los AWEs, desde su introducción en el siglo XVIII, han evolucionado desde capacidades iniciales de 0.08 Nm³/h hasta alcanzar 1.200 Nm³/h en configuraciones industriales modernas. Este trabajo introduce un sistema de monitoreo multivariable y modular (EMM) diseñado para AWEs de baja escala (hasta 2.5 kW), que incorpora una interfaz gráfica de usuario para facilitar una operación eficiente y segura. El EMM monitoriza en tiempo real las variables críticas incluyendo temperatura (T), presión (P), flujo de hidrógeno (Fh), voltaje (Vy), y corriente (C), lo que permite adaptaciones dinámicas a las condiciones operativas. Durante las pruebas, el sistema demostró un incremento del 250 % en la producción de hidrógeno al ajustar la temperatura en un 60 % y la presión en un 10 %. El diseño compacto del EMM, con dimensiones de 14 cm x 5 cm x 15 cm y un peso de 2 kg, junto con su capacidad de operar autónomamente por 18 horas, subrayan su aplicabilidad en diversos entornos operativos. Estos resultados destacan la importancia de sistemas de monitoreo integrados y adaptativos en la optimización de la producción de hidrógeno y la integración de AWEs con las redes de energía renovable.This study emphasizes the importance of transitioning to renewable energies through the deployment of alkaline water electrolyzers (AWEs) for green hydrogen production. AWEs, since their introduction in the 18th century, have evolved from initial capacities of 0.08 Nm³/h to modern industrial configurations reaching up to 1,200 Nm³/h. This work introduces a multivariable and modular monitoring system (EMM) designed for low-scale AWEs (up to 2.5 kW), incorporating a graphical user interface to facilitate efficient and safe operation. The EMM monitors critical variables in real-time including temperature (T), pressure (P), hydrogen flow (Fh), voltage (Vy), and current (C), allowing dynamic adaptations to operational conditions. During testing, the system demonstrated a 250 % increase in hydrogen production by adjusting the temperature by 60 % and the pressure by 10 %. The compact design of the EMM, with dimensions of 14 cm x 5 cm x 15 cm and a weight of 2 kg, along with its capability to operate autonomously for 18 hours, underscore its applicability in various operational environments. These results highlight the importance of integrated and adaptive monitoring systems in optimizing hydrogen production and integrating AWEs with renewable energy networks.Pregrad
Airborne Wind Energy Resource Analysis: From Wind Potential to Power Output
No full textAirborne Wind Energy Systems (AWES) have different power generation characteristics than conventional wind turbines, which can not be accurately captured in the traditional power curve. One important aspect is that it can harvest wind energy in a much wider range of altitudes than conventional wind turbines. Theoretically also High Altitude Winds (HAW) can be harnessed and the systems can be placed at a larger variety of sites.Wind Energ
Improving acoustic word embeddings through correspondence training of self-supervised speech representations
No full textAcoustic word embeddings (AWEs) are vector representations of spoken words. An effective method for obtaining AWEs is the Correspondence Auto-Encoder (CAE). In the past, the CAE method has been associated with traditional MFCC features. Representations obtained from self-supervised learning (SSL)-based speech models such as HuBERT, Wav2vec2, etc., are outperforming MFCC in many downstream tasks. However, they have not been well studied in the context of learning AWEs. This work explores the effectiveness of CAE with SSL-based speech representations to obtain improved AWEs. Additionally, the capabilities of SSL-based speech models are explored in cross-lingual scenarios for obtaining AWEs. Experiments are conducted on five languages: Polish, Portuguese, Spanish, French, and English. HuBERT-based CAE model achieves the best results for word discrimination in all languages, despite HuBERT being pre-trained on English only. Also, the HuBERT-based CAE model works well in cross-lingual settings. It outperforms MFCC-based CAE models trained on the target languages when trained on one source language and tested on target languages
- …
