Publication: Instrumentación multicanal con implementación de técnicas de reducción de ruido y procesado en-línea para la detección de ultrasonidos: aplicación a la localización de emisiones acústicas de descargas parciales
Loading...
Identifiers
Publication date
2015
Defense date
2015-09-24
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
La medida de descargas parciales (Partial Discharges - PD) es un indicador muy fiable de
la degradación del aislamiento eléctrico. Por esta razón, resulta muy interesante la monitorización
de dicha actividad con el objetivo de analizar equipos eléctricos de alta tensión, tales como los
transformadores, y de estudiar su probabilidad de fallo. En este sentido, se han desarrollado
nuevas tecnologías para la detección de PDs, las cuales incorporan técnicas digitales de
procesado de señal (filtrado digital, wavelets, etc.). Dentro de los métodos de detección de PDs en
transformadores, la detección acústica posee grandes ventajas en términos de inmunidad a
interferencias electromagnéticas y por su más fácil instalación en campo. Para la detección
acústica se usan típicamente sensores ultrasónicos piezoeléctricos (PZT), que van ubicados
externamente en las paredes del tanque del transformador y distribuidos para proporcionar
información sobre la localización de la fuente de PDs. El proceso de localización de la fuente de
emisión se puede aplicar sobre la base de los tiempos de llegada (Times Of Arrival – TOAs) de
las emisiones acústicas (Acoustic Emissions – AEs) a los sensores. En la actualidad, se están
desarrollando otros sensores que utilizan fibra óptica (Optical Fiber – OF). Estos sensores van
inmersos en el medio aislante, lo que permite detectar una señal acústica más intensa, libre de
interferencias por múltiple trayectoria y con una mejor relación señal a ruido.
En este trabajo de investigación, se propone un sistema de instrumentación multicanal
para la detección y la localización de AEs de PDs basado en la utilización combinada de sensores
intrínsecos de OF internos y sensores PZT externos. Este sistema posee bloques hardware
específicos de acondicionamiento y adquisición de señales acústicas en plataformas PXI y varios
bloques software para la detección y localización de señales acústicas.
Para la implementación del software, se proponen dos tipos de arquitecturas. La primera
de ellas es una arquitectura simple que utiliza sólo instrumentos virtuales (Virtual Instruments –
VIs) para implementar todas las etapas y operaciones en LabVIEW. Esta arquitectura incluye
bloques software de adquisición, detección y reducción de ruido (denoising). Se han adaptado
diferentes técnicas de reducción de ruido procedentes de las utilizadas para medidas eléctricas de
PDs y se han implementado para medidas acústicas. Las más útiles son el filtrado wavelet junto
con una implementación digital de filtros recomendados en el estándar (WT + IEEE FILTER).
Este sistema de procesamiento obtiene mejoras significativas en ambientes ruidosos con señales
débiles, lo que permite reducir el umbral de detección en un orden de magnitud y mejorar la
precisión en la medida de los TOAs, que son la base del proceso de localización. Este hecho se ha
demostrado experimentalmente con el sistema desarrollado.
Esta primera arquitectura integra como etapa de localización una basada en VIs, la cual
implementa diferentes estrategias de trilateración para averiguar la posición de la fuente de
emisión acústica (Acoustic Emission Source – AES) a partir de la información de los TOAs. El
primer enfoque tiene dos pasos de refinamiento progresivo con el fin de reducir el tiempo de
cómputo, pero obteniendo al mismo tiempo, una buena resolución espacial. En primer lugar, se
realiza una localización gruesa en dm y en segundo lugar, se completa una localización más fina
en cm limitada a la región que conforma la solución del primer paso. Con el fin de mejorar la
relación entre las prestaciones y el coste computacional y la complejidad del sistema se
implementó un segundo enfoque. Éste se basa en patrones de diferencias de tiempos, los cuales se
calculan para cada posición del transformador y se almacenan en tablas de búsqueda. La posición
de la fuente se obtiene al comparar el vector de diferencias de tiempos medido con las tablas de
diferencias de tiempos que mapean espacialmente el transformador.
La segunda arquitectura es híbrida, combina dos herramientas de programación
(LabVIEW - Matlab) con el fin de mejorar el rendimiento del sistema. La comunicación entre
ambas partes se realiza mediante un protocolo de transferencia de paquetes (Transmission
Control Protocol / Internet Protocol – TCP / IP). El sistema está divido en dos partes claramente
diferenciadas. La primera de ellas es un diseño inspirado en VIs y realiza los procesos de
adquisición, reducción de ruido y detección. Como parte de este procesamiento se obtienen los
TOAs y se transfieren a la siguiente parte. La segunda parte está programada en Matlab y se
dedica a la localización de la AES. Resuelve el sistema de ecuaciones de trilateración para un
modelo 3D de localización. En ella se implementan y se compararon varias estrategias: el Método
Solve, el Método de Mínimos Cuadrados, el Método de Mínimos Normalizados, el Método de
Cramer, el Método No-Iterativo Indirecto y el Método de Optimización por Nube de Partículas.
Adicionalmente, se ha instalado el sistema en un esquema de monitorización de PDs en
condiciones reales de alta tensión en AC (HVAC) y en DC (HVDC). Es importante mencionar
que la caracterización del aislamiento en estos sistemas HVAC-DC representa un desafío para la
red actual de transporte de energía eléctrica y este sistema de monitorización de PDs proporciona
una herramienta que está a la vanguardia en este campo de investigación ya que incluye detección
acústica, sensores de fibra óptica y localización de la fuente de PDs.
The measurement of partial discharges (PD) is a very reliable indicator of the degradation of the electrical insulation. For this reason, it is very interesting to monitor this activity with the aim of analyzing high-voltage electrical equipment such as transformers, and to study the probability of failure. In this sense, new technologies have been developed for the detection of PD, which incorporate digital techniques of signal processing (digital filtering, wavelets, etc.). Among the methods of PD detection in transformers, the acoustic detection has great advantages in terms of immunity to electromagnetic interference and easier installation in the field. Ultrasonic piezoelectric sensors (PZT) are typically used for the acoustic detection, which are located externally on the transformer tank walls and distributed to provide information about the location of the PD source. A localization process can be applied based on the times of arrival (TOA) of the acoustic emissions (AE) to the sensors. Currently, other sensors that use optical fiber (OF) are being developed. These sensors are immersed in the insulating medium, thus they detect more intense sound signals, free of interference from the multipath and with better signal to noise ratio. In this research work, a multichannel instrumentation system for the detection and the localization of AE from PD is proposed on the base of using together internal OF intrinsic sensors and external ultrasonic PZT sensors. This system has specific hardware blocks of acoustic signal conditioning and acquisition on PXI platforms and several software blocks for detecting and locating the acoustic signals. For implementing the software, two types of architectures are proposed in this work. The first one is a single architecture that utilizes only virtual instruments (VI) to implement all the stages and operations in LabVIEW. This architecture includes software blocks of acquisition, detection and noise reduction (denoising). Different techniques were adapted from electrical PD denoising and implemented for the acoustic measurements; the most useful are a wavelet filtering plus a digital implementation of the filters recommended in the standard (WT + IEEE FILTER). This denoising system obtains significant improvements in noisy environments with weak signals, allowing to reduce the detection threshold one order of magnitude and to improve the precision in the determination of the TOA, which is the base of the accuracy in the localization process. These were experimentally characterized with the developed system. The architecture integrates a location stage in the VI structure and different trilateration strategies are implemented to find out the position of the acoustic emission source (AES) through the TOA. The first approach is divided into two parts of progressive refinement to reduce the computational time, but obtaining an adequate spatial resolution. First, a coarse location is performed in dm and second, a finer location in cm is obtained within the solution of the first part. In order to improve the ratio of performance over computation + complexity costs a second approach was implemented. This is based on time-differences patterns that are formed for each position in the transformer by mapping the internal structure and dimensions of the transformer as lookup tables. The source is located by comparing the measured time-differences vector with the time-differences lookup tables. The second architecture is hybrid, which combines two programming tools (LabVIEW – Matlab) in order to improve the performance of the system. The communication between both tools is through a packet transfer protocol (Transmission Control Protocol / Internet Protocol - TCP/IP). The system is divided in two parts: the first is a VI inspired design and is devoted to the processes of acquisition, denoising and detection. As part of this processing the TOA are obtained and they are transferred to the next part. The second part is programmed in Matlab and it is dedicated to locating the AES. It solves the equations system of trilateration for a 3D model of localization. Several strategies are implemented and compared: the Solve Method, the Least Squares Method, the Least Norm Method, the Method of Cramer, the Indirect Non-Iterative Method and the Particle Swarm Optimization Method. Additionally, the system was installed for monitoring experiments of PDs in real conditions of High Voltage Alternating Current (HVAC) and High Voltage Direct Current (HVDC). It is worth to mention that the characterization of the insulation in these high-voltage AC-DC systems is a challenge for the transportation network and this PD monitoring system provides a state-of-the-art tool for this purpose by combining acoustic detection, optical fiber sensing and localization of the source of PD.
The measurement of partial discharges (PD) is a very reliable indicator of the degradation of the electrical insulation. For this reason, it is very interesting to monitor this activity with the aim of analyzing high-voltage electrical equipment such as transformers, and to study the probability of failure. In this sense, new technologies have been developed for the detection of PD, which incorporate digital techniques of signal processing (digital filtering, wavelets, etc.). Among the methods of PD detection in transformers, the acoustic detection has great advantages in terms of immunity to electromagnetic interference and easier installation in the field. Ultrasonic piezoelectric sensors (PZT) are typically used for the acoustic detection, which are located externally on the transformer tank walls and distributed to provide information about the location of the PD source. A localization process can be applied based on the times of arrival (TOA) of the acoustic emissions (AE) to the sensors. Currently, other sensors that use optical fiber (OF) are being developed. These sensors are immersed in the insulating medium, thus they detect more intense sound signals, free of interference from the multipath and with better signal to noise ratio. In this research work, a multichannel instrumentation system for the detection and the localization of AE from PD is proposed on the base of using together internal OF intrinsic sensors and external ultrasonic PZT sensors. This system has specific hardware blocks of acoustic signal conditioning and acquisition on PXI platforms and several software blocks for detecting and locating the acoustic signals. For implementing the software, two types of architectures are proposed in this work. The first one is a single architecture that utilizes only virtual instruments (VI) to implement all the stages and operations in LabVIEW. This architecture includes software blocks of acquisition, detection and noise reduction (denoising). Different techniques were adapted from electrical PD denoising and implemented for the acoustic measurements; the most useful are a wavelet filtering plus a digital implementation of the filters recommended in the standard (WT + IEEE FILTER). This denoising system obtains significant improvements in noisy environments with weak signals, allowing to reduce the detection threshold one order of magnitude and to improve the precision in the determination of the TOA, which is the base of the accuracy in the localization process. These were experimentally characterized with the developed system. The architecture integrates a location stage in the VI structure and different trilateration strategies are implemented to find out the position of the acoustic emission source (AES) through the TOA. The first approach is divided into two parts of progressive refinement to reduce the computational time, but obtaining an adequate spatial resolution. First, a coarse location is performed in dm and second, a finer location in cm is obtained within the solution of the first part. In order to improve the ratio of performance over computation + complexity costs a second approach was implemented. This is based on time-differences patterns that are formed for each position in the transformer by mapping the internal structure and dimensions of the transformer as lookup tables. The source is located by comparing the measured time-differences vector with the time-differences lookup tables. The second architecture is hybrid, which combines two programming tools (LabVIEW – Matlab) in order to improve the performance of the system. The communication between both tools is through a packet transfer protocol (Transmission Control Protocol / Internet Protocol - TCP/IP). The system is divided in two parts: the first is a VI inspired design and is devoted to the processes of acquisition, denoising and detection. As part of this processing the TOA are obtained and they are transferred to the next part. The second part is programmed in Matlab and it is dedicated to locating the AES. It solves the equations system of trilateration for a 3D model of localization. Several strategies are implemented and compared: the Solve Method, the Least Squares Method, the Least Norm Method, the Method of Cramer, the Indirect Non-Iterative Method and the Particle Swarm Optimization Method. Additionally, the system was installed for monitoring experiments of PDs in real conditions of High Voltage Alternating Current (HVAC) and High Voltage Direct Current (HVDC). It is worth to mention that the characterization of the insulation in these high-voltage AC-DC systems is a challenge for the transportation network and this PD monitoring system provides a state-of-the-art tool for this purpose by combining acoustic detection, optical fiber sensing and localization of the source of PD.
Description
Mención Internacional en el título de doctor
Esta tesis contiene artículos de investigación en anexo
Esta tesis contiene artículos de investigación en anexo
Keywords
Emisión acústica, Detección de ultrasonidos, Localización de la fuente, Descargas parciales, Sistema de iInstrumentación multicanal, Sensores de fibra óptica, Reducción de ruido, LabVIEW, Sistema de programación híbrido (LabVIEW- Matlab), Acoustic Emission, Ultrasound Detection, Source Location, Partial Discharges, Multichannel Instrumentation System, Optical Fiber Sensors, Denoising, LabVIEW, Hybrid Programming (LabVIEW - Matlab)