Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

RUC. Repositorio da Universidade da Coruña
Publicación Artículo científico (JournalArticle)

Thermal and Rheological Properties of Fischer–Tropsch Wax/High-Flow LLDPE Blends

  • Mhlabeni, Thobile
  • Ramjee, Shatish
  • López-Beceiro, Jorge
  • Díaz-Díaz, Ana-María
  • Artiaga, Ramón
  • Focke, Walter W.
[Abstract]: Waxes find use as processing aids in filled compounds and polyethylene-based masterbatches. In such applications, the thermal and physical property changes they impart to the polymer matrix are important. Therefore, this study details results obtained for blends prepared by mixing a Fischer–Tropsch (F–T) wax with a high-flow linear low-density polyethylene (LLDPE). The melting and crystallization behavior are studied using hot-stage polarized optical microscopy (POM) and differential scanning calorimetry (DSC). The calorimetry results are consistent with partial cocrystallization of the two components. The melting and crystallization exo- and endotherms for the wax- and LLDPE-rich phases remained separate. However, they change in shape and shift toward higher- and lower temperature ranges, respectively. It is found that increasing the wax content delays the crystallization, decreases the overall crystallinity, and reduces the size of the crystallites of the polyethylene-rich phase. Rotational viscosity is measured at 170 °C in the Newtonian shear-rate range. The variation of the zero-shear viscosity with blend composition is consistent with the assumption of a homogeneous melt in which the chains are in an entangled state. Therefore, it is concluded that the wax and LLDPE are, in effect, miscible in the melt and partially compatible in the solid state., Generous financial support from Sasol is gratefully acknowledged. Sasol research grant agreement SAP No. 126/20 GT
RUC. Repositorio da Universidade da Coruña
Publicación Artículo científico (JournalArticle)

Hard and soft clustering of categorical time series based on two novel distances with an application to biological sequences

  • López-Oriona, Ángel
  • Vilar, José
  • D'Urso, Pierpaolo
Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG., [Abstract]: Two novel distances between categorical time series are introduced. Both of them measure discrepancies between extracted features describing the underlying serial dependence patterns. One distance is based on well-known association measures, namely Cramer's v and Cohen's κ. The other one relies on the so-called binarization of a categorical process, which indicates the presence of each category by means of a canonical vector. Binarization is used to construct a set of innovative association measures which allow to identify different types of serial dependence. The metrics are used to perform crisp and fuzzy clustering of nominal series. The proposed approaches are able to group together series generated from similar stochastic processes, achieve accurate results with series coming from a broad range of models and are computationally efficient. Extensive simulation studies show that both hard and soft clustering algorithms outperform several alternative procedures proposed in the literature. Two applications involving biological sequences from different species highlight the usefulness of the introduced techniques., Xunta de Galicia; ED431G 2019/01, Xunta de Galicia; ED431C-2020-14, The research of Ángel López-Oriona and José A. Vilar has been supported by the Ministerio de Economía y Competitividad (MINECO) grants MTM2017-82724-R and PID2020-113578RB-100, the Xunta de Galicia (Grupos de Referencia Competitiva ED431C-2020-14), and the Centro de Investigación del Sistema Universitario de Galicia “CITIC” grant ED431G 2019/01; all of them through the European Regional Development Fund (ERDF). This work has received funding for open access charge by Universidade da Coruña/CISUG. The author Ángel López-Oriona is very grateful to researcher Maite Freire for her lessons about DNA theory.
RUC. Repositorio da Universidade da Coruña
Publicación Artículo científico (JournalArticle)

Comparison by thermal analysis of Joule‑cured versus oven‑cured composites

  • Vázquez, Laura
  • López-Beceiro, Jorge
  • Díaz-Díaz, Ana-María
  • Álvarez García, Ana
  • Pereira Rodríguez, Mercedes
  • Artiaga, Ramón
Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG, [Abstract]: The current technology for curing high-performance composites, such as those used in industries like such as aeronautics and the automotive industry, is based on the use of autoclaves, where the material is cured by external heating, in large ovens. This type of curing requires enormous amounts of energy, of which only a small part is invested in the actual curing of the material, and the rest is mainly used for heating and maintaining the temperature of the autoclave. An alternative method that entails a lower energy cost compared to the traditional methodology is curing through the Joule effect, in which an electric current is passed through the material, so that it acquires temperature from the inside due to the passage of current through the carbon fibres, triggering and accelerating the curing process of the composite. While Joule curing may provide a much more efficient and faster curing, a control technology is needed to ensure that temperatures all throughout the composite match the temperature programme. In this work, a procedure has been developed to control the Joule effect curing of carbon fibre/epoxy composites in order to compare, by means of differential scanning calorimetry (DSC) and dynamic mechanical analysis (DMA), the curing obtained by this method with that obtained by the traditional oven curing method., We acknowledge the financial support provided by the Ministerio de Ciencia e Innovación (Spain), under grant PID2020-113578RB-100, and the Programa de Doutoramento Industrial 2022, funded by Xunta de Galicia, through the grant number 07_IN606D_2022_2695330., Xunta de Galicia: 07_IN606D_2022_2695330
RUC. Repositorio da Universidade da Coruña
Publicación Tesis doctoral (DoctoralThesis)

New methodological contributions in statistical learning for time series

  • López-Oriona, Ángel
[Abstract] Time series databases are becoming omnipresent in several fields (e.g., computer sciences, finance, biology...), which makes the development of data mining algorithms for these objects a top priority among machine learning and statistics practitioners. This thesis deals with the construction of dif- ferent algorithms for time series data mining addressing a wide variety of problems. A novel dissimilarity measure between real-valued multivariate series defined in the frequency domain is proposed and used to construct several clustering algorithms whose main goal is to group the time series according to their underlying dependence structures. Both the crisp and the fuzzy clustering approaches are considered. The proposed spectral di- vergence measure is also combined with different statistical tools (including techniques of dimensionality reduction, wavelet decomposition, and func- tional data analysis) to introduce new and powerful algorithms to perform classification and outlier detection of multivariate time series. Clustering of times series with categorical response is also addressed by considering distances taking advantage of the specific properties of these data objects. The dissimilarity measures used for clustering purposes are also employed for testing the equality of the generating processes of two time series, which results in several hypothesis tests. A whole new different paradigm to per- form time series clustering is introduced by establishing the clustering pur- pose in terms of prediction accuracy. This realistic and challenging problem is addressed by means of a specific algorithm based on global forecasting methods. All the proposed techniques are properly formalized, their ad- vantages with respect to alternative methods are shown through extensive simulation experiments and their usefulness is illustrated by using real time series databases. Three software packages written in the R programming language are created to implement many of the constructed algorithms and their functionalities are explained by means of practical examples., [Resumen] Las bases de datos de series temporales se están volviendo omnipresentes en varios campos (por ejemplo, informática, finanzas, biología...), lo que hace que el desarrollo de algoritmos de minería de datos para estos objetos sea una de las principales prioridades para los profesionales del aprendizaje automático y de la estadística. Esta tesis trata sobre la construcción de diferentes algoritmos para la minería de datos de series temporales que abordan una amplia variedad de problemas. Se propone una nueva medida de disimilaridad entre series numéricas multivariantes definida en el dominio de la frecuencia que se utiliza para construir varios algoritmos de clustering cuyo principal objetivo es agrupar las series temporales en base a sus estruc- turas de dependencia subyacentes. Se consideran los enfoques de clustering crisp y clustering fuzzy. Adicionalmente, la medida espectral de divergencia propuesta se combina con diferentes herramientas estadísticas (por ejemplo, técnicas de reducción de la dimensión, descomposición wavelet, y análisis de datos funcionales) para introducir poderosos algoritmos de clasificación y de detección de anomalías en un contexto de series multivariantes. El clustering de series temporales con respuesta categórica también se aborda mediante el uso de distancias que explotan las propiedades específicas de este tipo de series. Las medidas de disimilaridad utilizadas para clustering se emplean asimismo para contrastar la igualdad de los procesos generadores de dos series de tiempo, lo que da lugar a varios contrastes de hipótesis. Adicionalmente, se introduce un paradigma completamente novedoso para abordar el problema de clustering de series temporales, que se basa en con- siderar la precisión de las predicciones como criterio para el agrupamiento. Se trata de un problema realista y complejo, que se aborda mediante un algoritmo específico basado en modelos globales de predicción. Todos los métodos propuestos se formalizan adecuadamente, sus ventajas con respecto a enfoques alternativos se corroboran mediante completos estudios de simulaci on y su utilidad se ilustra mediante el uso de bases de datos reales. Muchos de los algoritmos construidos se implementan en alguno de los tres paquetes que se desarrollan en el lenguaje de programación R. Las funcionalidades de los mismos se describen mediante varios ejemplos., [Resumo] As bases de datos de series temporais están a volverse omnipresentes en varios campos (por exemplo, informática, finanzas, bioloxía...), o que fai que o desenvolvemento de algoritmos de minería de datos para estes obxectos sexa unha das principais prioridades para os profesionais da aprendizaxe automática e da estatística. Esta tese trata sobre a construción de diferentes algoritmos para a minería de datos de series temporais que abordan unha ampla variedade de problemas. Proponse unha nova medida de discrepancia entre series numéricas multivariantes definida no dominio da frecuencia que se utiliza para construír varios algoritmos de clustering cuxo principal obxectivo é agrupar as series temporais en base as súas estruturas de dependencia subxacentes. Considéranse os enfoques de clustering crisp e clustering fuzzy. Adicionalmente, a medida espectral de diverxencia proposta combínase con diferentes ferramentas estatísticas (por exemplo, técnicas de redución da dimensión, descomposición wavelet, e análise de datos funcionais) para introducir poderosos algoritmos de clasicación e de detección de anomalías nun contexto de series multivariantes. O clustering de series temporais con resposta categórica tamén se aborda mediante o uso de distancias que explotan as propiedades específicas deste tipo de series. As medidas de discrepancia utilizadas para clustering empréganse asímesmo para contrastar a igualdade dos procesos xeradores de dúas series de tempo, o que dá lugar a varios contrastes de hipóteses. Adicionalmente, introdúcese un paradigma completamente novo para abordar o problema de clustering de series temporais, que se basea en considerar a precisión das predicións como criterio para o agrupamento. Trátase dun problema realista e complexo, que se aborda mediante un algoritmo específi co baseado en modelos globais de predición. Todos os métodos propostos formalízanse adecuadamente, as súas vantaxes con respecto a enfoques alternativos corrobóranse mediante completos estudos de simulación e a súa utilidade ilústrase mediante o uso de bases de datos reais. Moitos dos algoritmos construídos impleméntanse nalgún dos tres paquetes que se desenvolven na linguaxe de programación R. As funcionalidades dos mesmos descríbense mediante varios exemplos., Xunta de Galicia; ED431G-2019/01, Xunta de Galicia; ED481A2020/298, Xunta de Galicia; ED431C2020/14, Xunta de Galicia; ED481A-2020/298, Xunta de Galicia; ED431G-2019/01
RUC. Repositorio da Universidade da Coruña
Publicación Artículo científico (JournalArticle)

Machine learning for multivariate time series with the R package mlmts

  • López-Oriona, Ángel
  • Vilar, José
Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG, [Abstract]: Time series data are ubiquitous nowadays. Whereas most of the literature on the topic deals with univariate time series, multivariate time series have typically received much less attention. However, the development of machine learning algorithms for the latter objects has substantially increased in recent years. The R package mlmts attempts to provide a set of widespread data mining techniques for multivariate series. Several functions allowing the execution of clustering, classification, outlier detection and forecasting methods, among others, are included in the package. mlmts also incorporates a collection of multivariate time series datasets often used to test the performance of new classification algorithms. The main characteristics of the package are described and its use is illustrated through various examples. Practitioners from a wide variety of fields could benefit from the general framework provided by mlmts., This research has been supported by the Ministerio de Economía y Competitividad (MINECO) grants MTM2017-82724-R and PID2020-113578RB-100, the Xunta de Galicia (Grupos de Referencia Competitiva ED431C-2020-14), and the Centro de Investigación del Sistema Universitario de Galicia, “CITIC” grant ED431G 2019/01; all of them through the European Regional Development Fund (ERDF). This work has received funding for open access charge by University of A Coruña/CISUG., Xunta de Galicia; ED431C-2020-14, Xunta de Galicia; ED431G 2019/01