Poincaré Images Extracted from Vibration Signals are Useful Features for Fault Classification in a Reciprocating Compressor

Ruben Medina; Mariela Cerrada Lozada; Diego Roman Cabrera Mendieta; Rene Vinicio Sanchez Loja

Poincaré Images Extracted from Vibration Signals are Useful Features for Fault Classification in a Reciprocating Compressor

Título traducido de la contribución: Las imágenes de Poincaré extraídas de las señales de vibración son características útiles para la clasificación de fallas en un compresor alternativo

Ruben Medina
, Mariela Cerrada Lozada
, Diego Roman Cabrera Mendieta
, Rene Vinicio Sanchez Loja

Grupo de Investigación y Desarrollo en Tecnologías Industriales (GIDTEC)

Producción científica: Contribución a una conferencia › Documento

Resumen

La clasificación de fallas es una herramienta esencial para Pronósticos y Gestión de Salud (PHM), cuyo objetivo es la evaluación del estado de salud de una máquina y sus componentes. En esta investigación, proponemos utilizar imágenes de Poincaré (PI) como características para la clasificación de fallas utilizando redes neuronales convolucionales (CNN). El enfoque se valida para clasificar 17 fallas de válvulas en un compresor alternativo (RC). El enfoque propuesto logró una precisión de clasificación promedio del 94,97 %, lo que representa una mejora de alrededor del 15 % con respecto al uso de características discretas extraídas del gráfico de Poincaré para clasificar fallas con modelos Random Forest. Además, los resultados muestran que el rendimiento de la clasificación con PI es comparable al uso de espectrogramas como imágenes de entrada alternativas a los clasificadores de fallas basados en CNN.

Título traducido de la contribución	Las imágenes de Poincaré extraídas de las señales de vibración son características útiles para la clasificación de fallas en un compresor alternativo
Idioma original	Inglés estadounidense
Estado	Publicada - 28 dic 2022
Evento	XIX The conference on Latin America Control Congress - CU Duración: 28 nov 2022 → 2 dic 2022 https://lacc-2022.github.io/

Conferencia

Conferencia	XIX The conference on Latin America Control Congress
Período	28/11/22 → 2/12/22
Dirección de internet	https://lacc-2022.github.io/

Palabras clave

Poincaré
Reciprocating compressor
Vibration signals

Areas de Conocimiento del CACES

827A Mantenimiento industrial

Otros archivos y enlaces

Other Link

Monitoreo Inteligente de la Condición de Maquinaria Rotativa Mediante la Fusión de Señales de Audio, Emisión Acústica, Vibración y Corriente
Llerena Pizarro, O. R. (Investigador Secundario), Sanchez Loja, R. V. (Investigador principal), Cabrera Mendieta, D. R. (Investigador Secundario), Lucero Otorongo, P. M. (Investigador Externo), Macancela Poveda, J. C. (Investigador Externo), Perez Rivera, I. A. (Investigador Externo), Pacheco Cordova, E. E. (Investigador Externo), Vacacela Costa, A. S. (Estudiante Investigador), Pacheco Montilla, F. K. (Investigador Externo), Villacis Marin, M. L. (Investigador Secundario), Guaman Buestan, A. D. P. (Investigador Secundario), Torres Diaz, C. P. (Investigador Externo), Valente De Oliveira, J. L. (Investigador Externo), Vásquez, R. (Investigador Externo), Lojano Armijos, F. J. (Estudiante Investigador), Chingal Imaicela, D. E. (Investigador Externo), Siguencia Urgiles, J. F. (Investigador Externo), Cajas Muñoz, F. D. (Investigador Externo), Montalvan Pulla, F. I. (Estudiante Investigador), Quinteros Espinoza, M. E. (Investigador Externo), Ortega Lucero, L. R. (Estudiante Investigador), Llivicura Orellana, H. F. (Estudiante Investigador), Calle Lazo, A. K. (Estudiante Investigador) & Li, C. (Investigador Externo)
17/01/19 → …
Proyecto: Investigación y Desarrollo

Citar esto

@conference{f4124ea21bd64d5b927f37e5aa8f8fd6,

title = "Poincar{\'e} Images Extracted from Vibration Signals are Useful Features for Fault Classification in a Reciprocating Compressor",

abstract = "Fault classification is an essential tool for Prognostics and Health Management (PHM), whose goal is the assessment of the health condition of a machine and its components. In this research, we propose using Poincar{\'e} Images (PI) as features for fault classification using Convolutional Neural Networks (CNN). The approach is validated for classifying 17 valve faults in a reciprocating compressor (RC). The proposed approach attained an averaged classification accuracy of 94.97\% which represents an improvement of about 15\% concerning using discrete features extracted from the Poincar{\'e} plot for classifying faults with Random Forest models. Additionally, results show that the classification performance with PI is comparable to using spectrograms as alternative input images to CNN-based fault classifiers.",

keywords = "Poincar{\'e}, Reciprocating compressor, Vibration signals, Poincar{\'e}, Reciprocating compressor, Vibration signals",

author = "Ruben Medina and \{Cerrada Lozada\}, Mariela and \{Cabrera Mendieta\}, \{Diego Roman\} and \{Sanchez Loja\}, \{Rene Vinicio\}",

year = "2022",

month = dec,

day = "28",

language = "English (US)",

note = "XIX The conference on Latin America Control Congress ; Conference date: 28-11-2022 Through 02-12-2022",

url = "https://lacc-2022.github.io/",

}

TY - CONF

T1 - Poincaré Images Extracted from Vibration Signals are Useful Features for Fault Classification in a Reciprocating Compressor

AU - Medina, Ruben

AU - Cerrada Lozada, Mariela

AU - Cabrera Mendieta, Diego Roman

AU - Sanchez Loja, Rene Vinicio

PY - 2022/12/28

Y1 - 2022/12/28

N2 - Fault classification is an essential tool for Prognostics and Health Management (PHM), whose goal is the assessment of the health condition of a machine and its components. In this research, we propose using Poincaré Images (PI) as features for fault classification using Convolutional Neural Networks (CNN). The approach is validated for classifying 17 valve faults in a reciprocating compressor (RC). The proposed approach attained an averaged classification accuracy of 94.97% which represents an improvement of about 15% concerning using discrete features extracted from the Poincaré plot for classifying faults with Random Forest models. Additionally, results show that the classification performance with PI is comparable to using spectrograms as alternative input images to CNN-based fault classifiers.

AB - Fault classification is an essential tool for Prognostics and Health Management (PHM), whose goal is the assessment of the health condition of a machine and its components. In this research, we propose using Poincaré Images (PI) as features for fault classification using Convolutional Neural Networks (CNN). The approach is validated for classifying 17 valve faults in a reciprocating compressor (RC). The proposed approach attained an averaged classification accuracy of 94.97% which represents an improvement of about 15% concerning using discrete features extracted from the Poincaré plot for classifying faults with Random Forest models. Additionally, results show that the classification performance with PI is comparable to using spectrograms as alternative input images to CNN-based fault classifiers.

KW - Poincaré

KW - Reciprocating compressor

KW - Vibration signals

KW - Poincaré

KW - Reciprocating compressor

KW - Vibration signals

UR - https://link.springer.com/chapter/10.1007/978-3-031-26361-3_11

M3 - Paper

T2 - XIX The conference on Latin America Control Congress

Y2 - 28 November 2022 through 2 December 2022

ER -

Poincaré Images Extracted from Vibration Signals are Useful Features for Fault Classification in a Reciprocating Compressor

Resumen

Conferencia

Palabras clave

Areas de Conocimiento del CACES

Otros archivos y enlaces

Huella

Proyectos

Monitoreo Inteligente de la Condición de Maquinaria Rotativa Mediante la Fusión de Señales de Audio, Emisión Acústica, Vibración y Corriente

Citar esto