Publication: Metodología para la selección de los parámetros WPT óptimos para el análisis de la condición de elementos mecánicos rotativos
Loading...
Identifiers
Publication date
2022-02
Defense date
2022-03-04
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
La mejora del mantenimiento es actualmente uno de los principales cometidos de las industrias, ya que es de vital importancia detectar defectos a tiempo y así evitar fallos catastróficos.
La tendencia actual es la de automatizar todo el proceso de mantenimiento, y la de conocer el estado de los componentes en todo momento. Esto anticipa la detección de defectos, aumentando la fiabilidad de los sistemas. Este concepto se denomina monitorización de estado, y se basa en conocer el estado de los sistemas en tiempo real y, junto con la conexión con las diferentes áreas de la industria, aspira a integrarse en el paradigma actual de la Industria 4.0.
Dentro de la industria, la maquinaria más habitual está compuesta por componentes rotativos. Una de las formas más comunes de analizar estos elementos es mediante un análisis de las señales vibratorias obtenidas durante su funcionamiento. Los tipos de defectos que se pueden encontrar en este tipo de máquinas son muy variados, así como la forma de detectarlos.
Parece bastante extendido el uso de las herramientas para llevar a cabo el procesamiento de las señales vibratorias, siendo cada vez más habituales las herramientas que proporcionan información tanto en el dominio del tiempo como de la frecuencia, como la WT (Wavelet Transform), pero no hay muchos estudios que se centren en encontrar los parámetros de dichas herramientas que garanticen unos resultados óptimos, sino que se utilizan en base a la experiencia.
En la presente Tesis Doctoral se ha desarrollado una metodología que permite la selección de la wavelet madre óptima para llevar a cabo un análisis WPT (Wavelet Packets Transform) para diferentes tipos de defectos en maquinaria rotativa. La correcta elección de este parámetro es fundamental para poder realizar un buen análisis de las señales vibratorias y poder detectar así los posibles defectos a partir de los patrones obtenidos. Para llevar a cabo la metodología se ha definido un parámetro denominado DEV (grado de variación de energía) para calcular la diferencia de energía entre estado sano y defectuoso, por lo que el objetivo principal para la selección de la wavelet madre es maximizar dicho valor. La selección de la wavelet madre involucra el estudio de las propiedades matemáticas de cada una de ellas.
El criterio final para la selección de la wavelet madre óptima para cada aplicación estará basada en la tasa de acierto de cada una de ellas cuando se introducen como entrada en un sistema de clasificación inteligente, en este caso una SVM (Support Vector Machines) lineal.
Con el fin de validar la metodología, ésta se aplica a diferentes elementos mecánicos, con diferentes defectos y bajo distintas condiciones de operación: fisura lateral en un eje, desequilibrio de un eje, holgura de un eje y defecto en pista interna y en pista externa de un rodamiento a 20, 40 y 60 Hz.
Una vez seleccionada la wavelet madre óptima, ésta se utiliza para un estudio más avanzado de los patrones obtenidos con la WPT como identificador de la condición del elemento, comprobando así la bondad de la función seleccionada y caracterizando la máquina para poder detectar los defectos de forma temprana. Por tanto, los patrones seleccionados serían útiles para una monitorización de estado óptima de la máquina, en la que se detectarían con la mayor anticipación posible todos los tipos de defectos caracterizados.
The improvement of maintenance is currently one of the main tasks in the industries, since it is of vital importance to detect defects in time and thus avoid catastrophic failures. The current trend is to automate the entire maintenance process, and to know the status of the components at all times. This anticipates the detection of defects, increasing the reliability of the systems. This concept is called condition monitoring, which is based on knowing the status of the systems in real time and, together with the connection with the different areas of the industry, aspires to be integrated into the paradigm of Industry 4.0. Within the industry, the most common machinery is made up of rotating components. One of the most common ways to analyse these elements is by analysing the vibratory signals obtained during their operation. The types of defects that can be found in this type of machine are very varied, as well as the way to detect them. The use of tools to carry out the processing of vibratory signals seems to be quite widespread, with tools that provide information in both time and frequency domain, such as the WT (Wavelet Transform), are becoming more and more common, but there are not many studies that focus on finding the parameters of these tools that guarantee optimal results, but they are used based on experience. In this Doctoral Thesis, a methodology has been developed that allows the selection of the optimal mother wavelet to do a WPT (Wavelet Packets Transform) analysis for different types of defects in rotating machinery. The correct choice of this parameter is essential to be able to do a good analysis of the vibratory signals and thus be able to detect possible defects from the obtained patterns. To carry out the methodology, a parameter called DEV (Energy Variation Degree) has been defined to calculate the energy difference between healthy and defective conditions, so the main objective for selecting the mother wavelet is to maximize said value. The selection of the mother wavelet involves the study of the mathematical properties of each of them. The final criterion for the selection of the optimal mother wavelet for each application will be based on the success rate of each of them when they are entered as input in an intelligent classification system, in this case a linear SVM (Support Vector Machines). In order to validate the methodology, it is applied to different mechanical elements, with different defects and under different operating conditions: lateral crack in a shaft, unbalance in a shaft, clearance in a shaft and defects in the inner race and in the outer race of a bearing at 20, 40 and 60 Hz. Once the optimal mother wavelet has been selected, it is used for a more advanced study of the patterns obtained with the WPT as an identifier of the condition of the element, thus checking the goodness of the selected function and characterizing the machine to be able to detect defects early. Therefore, the selected patterns would be useful for an optimal condition monitoring of the machine, in which all the types of the characterized defects would be detected as early as possible.
The improvement of maintenance is currently one of the main tasks in the industries, since it is of vital importance to detect defects in time and thus avoid catastrophic failures. The current trend is to automate the entire maintenance process, and to know the status of the components at all times. This anticipates the detection of defects, increasing the reliability of the systems. This concept is called condition monitoring, which is based on knowing the status of the systems in real time and, together with the connection with the different areas of the industry, aspires to be integrated into the paradigm of Industry 4.0. Within the industry, the most common machinery is made up of rotating components. One of the most common ways to analyse these elements is by analysing the vibratory signals obtained during their operation. The types of defects that can be found in this type of machine are very varied, as well as the way to detect them. The use of tools to carry out the processing of vibratory signals seems to be quite widespread, with tools that provide information in both time and frequency domain, such as the WT (Wavelet Transform), are becoming more and more common, but there are not many studies that focus on finding the parameters of these tools that guarantee optimal results, but they are used based on experience. In this Doctoral Thesis, a methodology has been developed that allows the selection of the optimal mother wavelet to do a WPT (Wavelet Packets Transform) analysis for different types of defects in rotating machinery. The correct choice of this parameter is essential to be able to do a good analysis of the vibratory signals and thus be able to detect possible defects from the obtained patterns. To carry out the methodology, a parameter called DEV (Energy Variation Degree) has been defined to calculate the energy difference between healthy and defective conditions, so the main objective for selecting the mother wavelet is to maximize said value. The selection of the mother wavelet involves the study of the mathematical properties of each of them. The final criterion for the selection of the optimal mother wavelet for each application will be based on the success rate of each of them when they are entered as input in an intelligent classification system, in this case a linear SVM (Support Vector Machines). In order to validate the methodology, it is applied to different mechanical elements, with different defects and under different operating conditions: lateral crack in a shaft, unbalance in a shaft, clearance in a shaft and defects in the inner race and in the outer race of a bearing at 20, 40 and 60 Hz. Once the optimal mother wavelet has been selected, it is used for a more advanced study of the patterns obtained with the WPT as an identifier of the condition of the element, thus checking the goodness of the selected function and characterizing the machine to be able to detect defects early. Therefore, the selected patterns would be useful for an optimal condition monitoring of the machine, in which all the types of the characterized defects would be detected as early as possible.
Description
Keywords
Maquinaria rotativa, Fallos en máquinas, Detección de fallos, Vibraciones mecánicas, Wavelet Packets Transform (WPT), Mother wavelet, Condition monitoring, Vibration analysis, Crack detection