Journal bearings play an important role on many rotating machines placed on industrial environments, especially in steam turbines of thermoelectric power plants. Babbitt damage (BD) and excessive clearance (C) are usual faults of steam turbine journal bearings. This paper is focused on achieving an effective identification of these faults through an intelligent recognition approach. The work was carried out through the processing of real data obtained from an industrial environment. In this work, a feature selection procedure was applied in order to choose the features more suitable to identify the faults. This feature selection procedure was performed through the computation of typical testors, which allows working with both quantitative and qualitative features. The classification tasks were carried out by using Nearest Neighbors, Voting Algorithm, Naïve Associative Classifier and Assisted Classification for Imbalance Data techniques. Several performance measures were computed and used in order to assess the classification effectiveness. The achieved results (e.g., six performance measures were above 0.998) showed the convenience of applying pattern recognition techniques to the automatic identification of BD and C.
En el diagnóstico de fallos de chumaceras actualmente no se considera el conocimiento experto expresado en variables no numéricas, a pesar de constituir una fuente de información importante. Este trabajo se desarrolló con el objetivo de identificar los rasgos más relevantes para clasificar un grupo de fallos ocurridos en las chumaceras de una turbina de vapor. Las variables que soportan el trabajo corresponden a los datos almacenados en reportes de diagnóstico y mantenimiento de una termoeléctrica en explotación. Las técnicas aplicadas para procesar los datos cuantitativos y cualitativos son herramientas de reconocimiento lógico combinatorio de patrones (RLCP). Se determinó la confusión de los rasgos del conjunto inicial y posteriormente los testores y testores típicos, y se calculó el peso informacional de los rasgos. Los resultados alcanzados mostraron que la relevancia de los rasgos cualitativos presentes en la descripción de los fallos es superior a la de los rasgos numéricos típicamente empleados.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.