Fault Detection

Abstract — Civil infrastructures such as bridges, highways and buildings are important components of any region and their health should be monitored. Sensor networks can be used to assess the infrastructure conditions. These networks can consist of up to thousands of different sensors and continuously generate large amounts of data. Analyzing the data to find anomalies in a timely manner is very critical, as it can prevent disasters from occurring. A reliable integrated system that efficiently incorporates and analyses data from sensors must be designed. This paper proposes an architecture that envisions the deployment of a sensor network, collecting data of various physical parameters. Stochastic models, incorporating data from all sensors, were generated. A meta-heuristic algorithm solved the models for several scenarios and successfully identified anomalies. The proposed methodology aims to identify anomalies, which allows appropriate preventive actions in timely manner.

Keywords—civil infrastructure, health monitoring system, integrated sensor data statistical model, meta-heuristic algorithms, fault identification.

This  paper  describes  a  methodology  that  aims  to find and diagnosing faults in transmission lines exploitation image  process  technique.  The  image  processing  techniques have  been  widely  used  to  solve  problem  in  process  of  all areas.  In  this  paper,  the  methodology  conjointly  uses  a digital  image  process  Wavelet  Shrinkage  function  to  fault identification  and  diagnosis.  In  other  words,  the  purpose  is to  extract  the  faulty  image  from  the  source  with  the separation  and  the  co-ordinates  of  the  transmission  lines. The  segmentation  objective  is  the  image  division  its  set  of parts and objects, which distinguishes it among others in the scene,  are  the  key  to  have  an  improved  result  in identification of faults.The experimental results indicate that the  proposed  method  provides  promising  results  and  is advantageous both in terms of PSNR and in visual quality.

This paper presents a new algorithm for fault detection and classification using discrete wavelet transform (DWT) and back-propagation neural network (BPNN) based on Clarke's transformation on parallel transmission. Alpha and beta (mode) currents generated by Clarke's transformation were used to convert the signal of discrete wavelet transform (DWT) to get the wavelet transform coefficients (WTC) and the wavelet energy coefficient (WEC). Daubechies4 (Db4) was used as a mother wavelet to decompose the high frequency components of the signal error. The simulation was performed using PSCAD / EMTDC for transmission system modeling. Simulation was performed at different locations along the transmission line with different types of fault and fault resistance, fault location and fault initial angle on a given power system model. Four statistic methods utilized are in the present study to determine the accuracy of detection and classification faults. The results show that the best Clarke transformation occurred on the configuration of 12–24-48-4, respectively. For instance, the errors using mean square error method, the errors of BPNN, Pattern Recognition Network and Fit Network are 0.03721, 0.13115 and 0.03728, respectively. This indicates that the BPNN results are the lowest error.

This document describes the DIR net, a distributed environment which is part of the EFTOS fault tolerance framework. The DIR net is a system consisting of two components, called DIR Manager (or, shortly, the manager) and DIR Backup Agent (shortly, the backup). One manager and a set of backups is located in the system to be `guarded', one component per node. At this point the DIR net weaves a web which substantially does two things: 1) makes itself tolerant to a number of possible faults, and 2) gathers information pertaining the run of the user application. As soon as an error occurs within the DIR net, the system executes built-in recovery actions that allow itself to continue processing despite a number of hardware/software faults, possibly doing a graceful degradation of its features; when an error occurs in the user application, the DIR net, by means of custom- and user-defined detection tools, is informed of such events and runs one or more recovery strategies, both built-in and coded by the user using an ancillary compile-time tool, the rl translator. Such tools translates the user-defined strategies into a binary `R-code', i.e., a pseudo-code interpreted by a special component of the DIR net, the Recovery Interpreter, rint (in a sense, rint is a r-code virtual machine.) This document describes the generic component of the DIR net, a function which can behave either as manager or as backup.

The fault diagnosis problem of a mobile robot is approached via a technique which is based on the local model neural network (LMNN)  structure. A number of parameters are modeled on a healthy plant, and then compared on line with the actual values.This leads to the creation of residuals. The
neural network training is performed off line. The technique was applied for the detection and diagnosis of faults on the wheels and motors of a mobile robot, with encouraging results.

The Fault detection which is based on fuzzy modeling is investigated. Takagi-Sugeno (TS) fuzzy model can
be derived by structure and parameter identification, where only the input-output data of the identified
system are available. In the structure identification step, Gustafson-Kessel clustering algorithm (GKCA) is
used to detect clusters of different geometrical shapes in the data set and to obtain the point-wise
membership function of the premise. In the parameter identification step, Unscented Kalman filter (UKF) is
used to estimate the parameters of the premise’s membership function. In the consequence part, Kalman
filter (KF) algorithm is applied as a linear regression to estimate parameters of the TS model using the
input-output data set. Then, the obtained fuzzy model is used to detect the fault. Simulations are provided to
demonstrate the effectiveness of the theoretical results.

This paper addresses a fundamental problem related to the interaction of rule-based autonomous agents in pervasive and intelligent environments. Some rules of behaviour can lead a multi-agent system to display unwanted periodic behaviour, such as networked appliances cycling on and off. We present a framework called Interaction Networks (IN) as a tool to describe and analyse this phenomena. In support of this, and as an aid to the visualisation and understanding of the temporal evolution of agent states, we offer a graphical Multidimensional Model (MDM). We describe an Instability Prevention System (INPRES) based in identifying and locking network nodes. Both IN, MDM and INPRESS enable system designers to identify and prevent such unwanted behaviour. Finally we introduce an Interaction Benchmark (IB) that we use to evaluate, experimentally, the effectiveness of our approach using both simulated and physical implementations.

dans ce papier on étudie les effets d’élimination des valeurs aberrantes et de leurs imputations sur la sélection des
indicateurs statistiques et la classification d’un défaut d’injection. Le défaut est simulé dans un moteur Diesel à combustion interne à six cylindres. Quatre méthodes de classification supervisée sont exploitées pour la discrimination des trois classes de défaut. Les résultats obtenus montrent que l’élimination des valeurs aberrantes et leurs imputations ont amélioré la classification du défaut pour un nombre des variables fini et surtout pour le classificateur k-plus proches voisins.

The local model network (LMN) structure is employed for solving the fault detection and diagnosis problem of plants with unknown parameters.Residual and signals that trigger a decision mechanism are used, which help to conclude for presence, size, and cause of possible faults.The 3-tank benchmark system is used to test the capabilities of the method.

This paper deals with the fault diagnosis problem of plants with unknown description. The modeling technique used is based on the local model network (LMN) which performs local linearization and their structure is similar to the Tagaki-Sugeno fuzzy model. LMN's function as linear estimators
that give satisfactory estimation for plant's output within parts of the operating regime. The method is tested via simulation on the 3-tank system benchmark.

With the increasing use of photovoltaic (PV) systems, the research studies to improve the efficiency of PV systems have gained greater interest in recent years, especially under non-uniform operating conditions. However, there is a little attention on fault diagnosis of PV arrays. This paper develops and demonstrates a method that can efficiently detect the number of open and short circuit faults and discriminate between them and partial shading conditions. This method is based on only the measurement of operating voltage of PV string and ambient temperature. In that manner, the proposed method takes into consideration of the minimum number of sensors to reduce the cost of the system, as one of the main purposes of this study. The simulation and experimental results are presented to demonstrate the effectiveness of the proposed method under both uniform and non-uniform irradiance conditions.

Se presenta el diseño de un sistema de diagnóstico basado en relaciones redundantes y observadores de estado —que funciona manteniendo al sistema de transporte en peración— para la detección, identificación y reconstrucción de fallas en ductos, tales como: anomalías en los instrumentos de medición, bombas y extracciones desconocidas. Los  algoritmos se desarrollaron a partir de ecuaciones diferenciales parciales no lineales que caracterizan el comportamiento del fluido —considerando el principio de conservación de masa y momento— y que fueron discretizadas en el espacio utilizando el Método de Diferencias Finitas. Con la finalidad de discernir entre los diferentes tipos de fallas y de reconstruir el  comportamiento de éstas, el sistema de diagnóstico opera por etapas. La primera etapa, llamada de Detección y Aislamiento de Fallas, tiene como objetivo aislar el síntoma de la falla con ayuda de un conjunto de relaciones redundantes deducidas a partir del modelo nominal del ducto. Esta etapa de aislamiento simplifica la segunda etapa: Reconstrucción de la Falla, en la que algoritmos de observación estiman la evolución temporal  de la falla aislada. El sistema completo de diagnóstico se validó mediante una serie de experimentos realizados en un modelo hidráulico piloto de aproximadamente 200 m de largo, instrumentado y construido específicamente para llevar a cabo pruebas de monitoreo de ductos en el Instituto de Ingeniería de la Universidad Nacional Autónoma de México.