Ximena Alvarez

Bio: Ximena Alvarez is an academic researcher from University of Cuenca. The author has contributed to research in topics: Sparse approximation & Fault (power engineering). The author has an hindex of 3, co-authored 7 publications receiving 19 citations.

Feature engineering based on ANOVA, cluster validity assessment and KNN for fault diagnosis in bearings

Abstract: El numero de caracteristicas para el diagnostico de fallas en maquinaria rotativa puede ser grande debido a las diferentes senales disponibles que contienen informacion util. De un amplio conjunto de caracteristicas disponibles, algunas de ellas son mas adecuadas que otras, para clasificar adecuadamente ciertos modos de falla. El enfoque clasico para la seleccion de caracteristicas tiene como objetivo clasificar el conjunto de caracteristicas originales; sin embargo, enseleccion de caracteristicas, se ha reconocido que un conjunto de mejores caracteristicas individuales no necesariamente conduce a una buena clasificacion. Este documento propone un marco para la ingenieria de caracteristicas para identificar el conjunto de caracteristicas que pueden producir grupos de datos adecuados. Primero, el marco utiliza ANOVA combinado con la prueba de Tukey para clasificar las caracteristicas significativas individualmente; a continuacion, se realiza un analisis adicional basado en las distancias entre grupos y dentro del grupo para clasificar subconjuntos de caracteristicas significativas previamente identificadas. Nuestra contribucion tiene como objetivo descubrir el subconjunto de caracteristicas que discrimina mejor los grupos de datos asociados a varias condiciones defectuosas de los dispositivos mecanicos, para construir clasificadores de fallas multiples mas robustos. Clasificacion de gravedad de fallas enlos rodamientos se estudian para verificar el marco propuesto, con datos recopilados de un banco de pruebas en condiciones reales de velocidad y carga en el dispositivo giratorio.

ANOVA and Cluster Distance Based Contributions for Feature Empirical Analysis to Fault Diagnosis in Rotating Machinery

Abstract: The number of extracted features for fault diagnosis in rotating machinery can grow considerably due to the large amount of available data collected from different monitored signals. Usually, feature selection or reduction are conducted through several techniques proposing a unique set of representative features for all available classes; nevertheless, in feature selection, it has been recognized that a set of individually good features do not necessarily lead to good classification. This paper proposes a general framework to analyse the feature selection oriented to identify the features that can produce clusters of data with a proper structure. In the first stage, the framework uses Analysis of Variance (ANOVA) combined with Tukey's test for ranking the significant features individually. In the second stage, a further analysis based on inter-cluster and intra-cluster distances, is accomplished to rank subsets of significant features previously identified. In this sense, our contribution aims at discovering the subset of features that are discriminating better the clusters of data associated to several faulty conditions of the mechanical devices, to build more robust fault multi-classifiers. Fault severity classification in rolling bearings is studied to test the proposed framework, with data collected from an experimental test bed under real conditions of speed and load on the rotating device.

Poincaré plot features from vibration signal for gearbox fault diagnosis

Abstract: This paper describes a method for fault diagnosis in gearboxes using features extracted from the Poincare plot of the vibration signal. Several features describing the geometrical shape of the Poincare plot are calculated and three of these features are selected for performing the classification of 10 types of faults recorded in the gearbox vibration signal dataset. A multi-class Error-Correcting Output Code Support Vector Machine is trained for performing the classification of faults. The cross-validation performed show that the highest accuracy attained is 95.3% when signals recorded using the load L1 are considered.

Neutrosophic analysis of variance: application to university students

Abstract: The existing analysis of variance (ANOVA) test cannot be applied when the sample is selected from the population having come imprecise, fuzzy and uncertain observations. The neutrosophic statistics will be applied to analyze the data having uncertain observations or the parameters. In this paper, we will introduce the neutrosophic analysis of variance (NANONA). The NANONA is an extension of the classical ANOVA. We presented the NANONA table. We performed the NANONA to test teaching methods using data collected from the university students.

Analysis of Machine Learning Classifiers for Early Detection of DDoS Attacks on IoT Devices

Abstract: Distributed denial-of-service attacks are still difficult to handle as per current scenarios. The attack aim is a menace to network security and exhausting the target networks with malicious traffic from multiple sites. Although a plethora of conventional methods have been proposed to detect DDoS attacks, so far the rapid diagnosis of these attacks using feature selection algorithms is a daunting challenge. The proposed system uses a hybrid methodology for selecting features by applying feature selection methods on machine learning classifiers. Feature selections methods, namely chi-square, Extra Tree and ANOVA have been applied on four classifiers Random Forest, Decision Tree, k-Nearest Neighbors and XGBoost for early detection of DDoS attacks on IoT devices. We use the CICDDoS2019 dataset containing comprehensive DDoS attacks to train and assess the proposed methodology in a cloud-based environment (Google Colab). Based on the experimental results, the proposed hybrid methodology provides superior performance with a feature reduction ratio of 82.5% by achieving 98.34% accuracy with ANOVA for XGBoost and helps in early detection of DDoS attacks on IoT devices.

Vibration signal analysis using symbolic dynamics for gearbox fault diagnosis

Abstract: This paper addresses the use of two algorithms based on symbolic dynamics analysis of vibration signal for fault diagnosis in gearboxes. The symbolic dynamics algorithm (SDA) works by subdividing the phase space described by the Poincare plot into several angular regions; then, a symbol is assigned to each region. The probability distributions generated by the set of symbols are considered as features for classification of faults in a gearbox. The peak symbolic dynamics algorithm (PSDA) is a variant that extracts a sequence of peaks from the vibration signals and then performs the phase-space subdivision and symbol coding. A gearbox vibration signal dataset is analyzed for classifying 10 types of faults. Fault classification is attained using a multi-class support vector machine. The highest accuracy attained using k-fold cross-validation is 100.0% for load L3 with SDA and 100% with load L2 with PSDA. The accuracy considering all signals in the gearbox dataset is 99.2% with SDA and 99.8% with PSDA. The algorithms proposed have the advantage of being simple, accurate, and fast, and they could be adapted for online condition monitoring.