I.Estudios

Artículo 3:

“AN ONTOLOGY-BASED ARCHITECTURE FOR EVENT DETECTION AND PREDUCTION ON MONITORING SYSTEMS OF WIRELESS SENSOR NETWORKS” 

AUTOR: Ingenieros: Ricardo González,  María E. Vidal y Claudia Barenco.

Resumen:

En sistemas de monitoreo de gran escala, la integración y el análisis de datos sencillos es un requerimiento importante, ya que permite detectar y predecir la ocurrencia de eventos relevantes en el sistema. En este artículo se propone  la arquitectura HAEDEP, que consta de varios niveles y que emplea una ontología para representar el conocimiento del dominio. Este conocimiento es usado a su vez en la detección y predicción de estos eventos relevantes, específicamente en un sistema que emplea redes inalámbricas de sensores para capturar información del mundo físico. Este trabajo ilustra los beneficios de usar una arquitectura como la de HAEDEP mediante un caso de estudio asociado al campo de la refinación de petróleo. HAEDEP también puede ser usado en otros dominios donde actividades de verificación, almacenamiento y control se requieran.

Palabras Clave: Redes Inalámbricas de Sensores, monitorización de sistemas, sistemas orientados a eventos, conocimiento del dominio, ontologías.

 

 

 

 

AN ONTOLOGY-BASED ARCHITECTURE FOR EVENT DETECTION AND PREDUCTION ON MONITORING SYSTEMS OF WIRELESS SENSOR NETWORKS

ABSTRACT

In large-scale monitoring systems, integration of detailed data is required to detect and predict the occurrence of events that impact the performance of the whole environment. In this paper, we present an ontology-based multi-level architecture called HAEDEP, for event detection and event prediction on large-scale monitoring systems based on Wireless Sensors Networks. In HAEDEP, an ontology is used to represent domain knowledge that can be used at different levels, and enable the system to detect and predict the events that will affect the system behavior. We illustrate the benefits of HAEDEP by using a case study in the oil refinery field. However, HAEDEP could be used in other domains where the checking, recording and controlling activities are required.

Keywords: WSN, monitoring system, event oriented, domain knowledge, ontologies. 

NNlsߩЄt-align:justify'> 

 

ABSTRACT:

The prediction of chaotic systems has been and continues to be a subject with great scientific interest. This is because the benefits that can be obtained knowing the future behaviors of this kind of systems based on previous known data. However, the search for solution strategies with an acceptable error margin presents some levels of complexity that must be treated conveniently, based on the difference between the predicted values and the real values. Enough is to say, for instance, specific problems like climate and seismic predictions and the stock market, to realize of the benefit that would mean to have a good strategy based on prediction. Since these problems are related to managing knowledge, most attempts to obtain a predictive tool with acceptable results are related to technical heuristics. In this regard, Neural Networks is one of the most commonly used, specifically by configuring Multi-layer Perceptron and Radial Based Functions models. We must ask ourselves then: Which one of these models is most appropriate to solve this kind of problems? Can we guarantee a tool of prediction with low error margins using this kind of technique? This article presents a comparative study of patterns of Neural Networks above, responding to the previous two questions as well as providing conclusions that are intended to assist the scientific world to address a specific prediction problem with this technique.

Keywords: Prediction, Neural Networks, Multi-layer Perceptron, Radial Based Function.