Tesis sobre el tema "Real-Time Fault Detection"

Thesis: S.M., Massachusetts Institute of Technology, Department of Mechanical Engineering, 2019
Thesis: S.M. in Naval Architecture and Marine Engineering, Massachusetts Institute of Technology, Department of Mechanical Engineering, 2019
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 119-122).
Non-intrusive Load Monitoring (NILM) measures power at a central point in an electrical network and disaggregates individual load schedules from the overall power stream. This thesis presents the NILM Dashboard, a data-analysis and user interface tool that provides real-time machinery monitoring and fault diagnostics using NILM data. The Dashboard was developed and deployed for use onboard US Coast Guard Cutters to act as an automatic watchstander and condition-based maintenance aid. The effectiveness of the system is demonstrated on power data collected from electrical panels in the ship's engine room. Case studies are used to evaluate the Dashboard's ability to detect fault conditions in electromechanical systems.
by Thomas John Kane.
S.M.
S.M. in Naval Architecture and Marine Engineering
S.M. Massachusetts Institute of Technology, Department of Mechanical Engineering
S.M.inNavalArchitectureandMarineEngineering Massachusetts Institute of Technology, Department of Mechanical Engineering

This thesis presents techniques for improving building HVAC system performance in existing buildings generated using simulation-based tools and real data. Therefore, one of the aims has been to research the needs and possibilities to assess and improve building HVAC system performance. In addition, this thesis aims at an advanced utilization of building energy management system (BEMS) and the provision of useful information to building operators using simulation tools.

Buildings are becoming more complex systems with many elements, while BEMS provide many data about the building systems. There are, however, many faults and issues in building performance, but there are legislative and cost-benefit forces induced by energy savings. Therefore, both BEMS and the computer-based tools have to be utilized more efficiently to improve building performance.

The thesis consists of four main parts that can be read separately. The first part explains the term commissioning and the commissioning tool work principal based on literature reviews. The second part presents practical experiences and issues introduced through the work on this study. The third part deals with the computer-based tools application in design and operation. This part is divided into two chapters. The first deals with improvement in the design, and the second deals with the improvement in the control strategies. The last part of the thesis gives several rules for fault diagnosis developed using simulation tools. In addition, this part aims at the practical explanation of the faults in the building HVAC systems.

The practical background for the thesis was obtained though two surveys. The first survey was carried out with the aim to find the commissioning targets in Norwegian building facilities. In that way, an overview of the most typical buildings, HVAC equipment, and their related problems was obtained. An on-site survey was carried out on an example building, which was beneficial for introducing the building maintenance structure and the real hydronic heating system faults.

Coupled simulation and optimization programs (EnergyPlus and GenOpt) were utilized for improving the building performances. These tools were used for improving the design and the control strategies in the HVAC systems. Buildings with a hydronic heating system were analyzed for the purpose of improving the design. Since there are issues in using the optimization tool, GenOpt, a few procedures for different practical problems have been suggested. The optimization results show that the choice of the optimization functions influences significantly the design parameters for the hydronic heating system.

Since building construction and equipment characteristics are changing over time, there is a need to find new control strategies which can meet the actual building demand. This problem has been also elaborated on by using EnergyPlus and GenOpt. The control strategies in two different HVAC systems were analyzed, including the hydronic heating system and the ventilation system with the recovery wheel. The developed approach for the strategy optimization includes: involving the optimization variables and the objective function and developing information flow for handling the optimization process.

The real data obtained from BEMS and the additional measurements have been utilized to explain faults in the hydronic heating system. To couple real data and the simple heat balance model, the procedure for the model calibration by use of an optimization algorithm has been developed. Using this model, three operating faults in the hydronic heating system have been elaborated.

Using the simulation tools EnergyPlus and TRNSYS, several fault detection and diagnosis (FDD) rules have been generated. The FDD rules were established in three steps: testing different faults, calculating the performance indices (PI), and classifying the observed PIs. These rules have been established for the air cooling system and the hydronic heating system. The rules can diagnose the control and the component faults. Finally, analyzing the causes and the effects of the tested faults, useful information for the building maintenance has been descriptively explained.

The most important conclusions are related to a practical connection of the real data and simulation-based tools. For a complete understanding of system faults, it is necessary to provide real-life information. Even though BEMS provides many building data, it was proven that BEMS is not completely utilized. Therefore, the control strategies can always be improved and tuned to the actual building demands using the simulation and optimization tools. It was proven that many different FDD rules for HVAC systems can be generated using the simulation tools. Therefore, these FDD rules can be used as manual instructions for the building operators or as a framework for the automated FDD algorithms.

Denne avhandlingen presenterer noen fremgangsmåter for forbedring av ytelser for VVS-tekniske anlegg i eksisterende bygninger basert på bruk av simuleringsverktøy og virkelige måledata. Ett av målene har vært å undersøke behov og muligheter for vurdering og forbedring av ytelser for VVS-anlegg i bygninger. I tillegg har denne avhandlingen hatt som mål å fremme bruk av SD-anlegg samt å fremskaffe nyttig informasjon til driftspersonalet.

Bygninger blir stadig mer kompliserte systemer som inneholder flere og flere komponenter mens SD-anlegg håndterer en stadig større mengde data fra bygningsinstallasjoner. På den ene siden registreres det ofte feil og problemer med hensyn til ytelsene til de VVS-tekniske installasjonene. På den andre siden innføres det stadig strengere lovmessige pålegg og kost-nyttekrav motivert i energieffektiviseringen. SD-anlegg og databaserte verktøy bør derfor brukes mer effektivt for forbedring av ytelsene.

Avhandlingen består av fire hoveddeler hvor hver del kan leses separat. Den første delen, som er basert på literatturstudie, forklarer funksjonskontroll som begrep og prinsipper for oppbygging av verktøy for funksjonskontroll. Den andre delen presenterer praktisk erfaring og problemstillinger utviklet og behandlet i løpet av arbeidet med avhandlingen. Den tredje delen handler om anvendelse av databaserte verktøy for forbedring av ytelsen for VVS-tekniske installasjoner. Den tredje delen er delt opp i to kapitler, hvorav et handler om forbedring av systemløsninger og et om forbedring av styringsstrategier. Den siste delen presenterer flere regler for feilsøking og diagnostisering utviklet gjennom bruk av simuleringsverktøy. I tillegg gir denne delen en praktisk forklaring av feilene i de VVS-anleggene som er behandlet i undersøkelsen.

Det praktiske grunnlaget for avhandlingen er etablert gjennom to undersøkelser. Den første var en spørreundersøkelse som hadde til hensikt å kartlegge målsetninger for funksjonskontroll i norske bygninger. Gjennom dette ble det etablert en oversikt over de mest typiske bygninger med tilhørende VVS-anlegg og de mest forekommende problemene. En dypere undersøkelse ble utført på ett casebygg. Denne undersøkelsen viste seg å være nyttig både for kartlegging av betydningen av organisering av driften av bygningen og for avdekking av de virkelige feilene i det vannbårne oppvarmingssystemet.

En kobling mellom et simulerings- og et optimaliseringsprogram (EnergyPlus og GenOpt) har vært benyttet for forbedring av ytelsene for de VVS-tekniske installasjonene. Disse verktøyene har vært brukt for forbedring av både systemløsningene og styringsstrategiene for VVS-anlegg. Bygninger med vannbåren oppvarmingssystem har vært analysert for å forbedre systemløsningen. På grunn av begrensninger i bruken av optimaliseringsverktøyet GenOpt, har det blitt utviklet egne prosedyrer for håndtering av visse typer problemstillinger hvor denne begrensningen opptrer. Resultatene for optimaliseringen viser at valg av objektfunksjoner påvirker betydelig parametrene i det vannbårne oppvarmingssystemet.

Endringer i egenskapene til både bygningskonstruksjoner og utstyr som skjer på grunn av aldring over tiden, gjør det nødvendig med tilpassning av styringsstrategier slik at det virkelige behovet kan bli dekket. Denne problemstillingen har vært analysert ved bruk av EnergyPlus og GenOpt. Styringsstrategiene for to forskjellige VVS-anlegg, et vannbåret oppvarmingssystem og et ventilasjonsanlegg med varmegjenvinner har blitt behandlet. Den utviklete prosedyren for optimalisering av styringsstrategien består av følgende steg: innføring av optimaliseringsvariabler og objektfunksjon, samt utvikling av informasjonsflyt for behandling av optimaliseringsprosessen.

De virkelige data, både fra SD-anlegg og tilleggsmålinger, har vært benyttet for praktisk forklaring av feilene i oppvarmingssystemet. En prosedyre for modellkalibrering basert på bruk av en optimaliseringsalgoritme som kobler sammen de virkelige data og en enkel varmebalansemodell har blitt foreslått. Tre konkrete driftsfeil i oppvarmingssystemet har blitt belyst gjennom bruk av denne varmebalansemodellen.

Flere regler for feilsøking og diagnostisering har blitt utviklet ved hjelp av simuleringsverktøyene EnergyPlus and TRNSYS. Denne utviklingen har bestått av tre ulike steg: testing av bestemte feil, beregning av ytelsesindikatorer og til slutt klassifisering av de observerte ytelsesindikatorer. Reglene har blitt utviklet for et system av aggregater for luftkjøling og for et vannbåret oppvarmingssystem. Reglene kan diagnostisere både styringsfeil og komponentfeil. Til slutt presenteres informasjon som er nyttig for drift av VVS-tekniske installasjoner i bygninger basert på en analyse av årsakene for og virkningene av de feil som er behandlet.

De viktigste konklusjonene er knyttet til praktisk kombinasjon av virkelige måleverdier og simuleringsverktøy. Informasjon fra det virkelig liv er helt nødvendig for å få en god forståelse av feil som oppstår i anlegg. Det er også vist at potensialet som ligger i alle de data som er tilgjengelige gjennom SD-anlegg, ikke er fullt utnyttet. Gjennom bruk av simuleringsverktøy kan styringsstrategiene alltid bli bedre tilpasset og innjustert til de virkelige behov i bygningen. Simuleringsverktøy kan også brukes for utvikling av prosedyrer for feilsøking og diagnostisering i VVS-tekniske anlegg. Disse prosedyrene kan brukes enten som en veileder for manuell feilsøking og detektering eller som grunnlag for utvikling av automatiserte algoritmer.

Paper II, VI and VII are reprinted with kind permission from Elsevier, sciencedirect.com

Unmanned, as well as manned space missions have seen a high failure rate in the early era of space technology. However, this decreased a lot since technology advanced and engineers learnt from previous experiences and improved critical real time systems with fault detection mechanisms. Fault detection, isolation and recovery, nowadays, is generally available in every flying device. However, the cost of hardware can bottleneck the process of creating such a system that is both robust and responsive. This thesis analyses the possibility of implementing a fault detection, isolation and recovery system inside of a single-threaded, cooperative scheduling operating system. The thesis suggests a cooperative implementation of such a system, where every task is responsible for parts of the fault detection. The analysis is done from both the integration layer, across the operating system and its tasks, as well as from the inside of the detection system, where two key components are implemented and analyzed: debug telemetry and operation modes. Results show that it is possible to implement a fault detection system that is spread across all the components of the satellite and acts cooperatively. Furthermore, the comparison with a traditional, dedicated fault detection system proves that errors can be caught faster with a cooperative mechanism.
Obemannade såväl som bemannade rymduppdrag har sett ett högt misslyckande i rymdteknikens tidiga era. Detta har dock förbättrats mycket sedan ingenjörer började lära sig av sina tidigare erfarenheter och utrustade kritiska realtidssystem med feldetekteringsmekanismer. Idag är alla flygande enheter utrustade med feldetekterings-, isolerings- och återställningsmekanismer. Däremot kan kostnaden för hårdvara vara ett problem för processen att skapa ett sådant system som är både robust och mottagligt. Denna uppsats analyserar möjligheten att implementera ett feldetekterings-, isolerings- och återställningssystem inuti ett enkelgängat samarbetsplaneringssystem. Denna uppsats föreslår ett samarbete för implementering av ett sådant system, där varje uppgift ansvarar för delar av feldetekteringen. Analysen görs från både integrationsskiktet, över operativsystemet och dess uppgifter, samt från insidan av detekteringssystemet, där två nyckelkomponenter implementeras och analyseras. Resultaten visar att det är möjligt att implementera ett feldetekteringssystem som täcker alla satellitkomponenter och som är mottaglig. Dessutom visar jämförelsen med ett traditionellt, dedikerat feldetekteringssystem att fel kan fångas snabbare med en mottagligmekanism.
Misiunile spat,iale cu oameni, atât cât s, i fara oameni, au avut o rata a es, ecurilor destul de ridicata în perioada init,iala a erei tehnologiei spat,iale. În schimb, aceasta a scazut semnificativ odata cu dezvoltarea tehnologiei, dar s, i datorita faptului ca inginerii au învat,at din experient,ele precendente s, i au îmbunatat, it sistemele critice în timp real cu mecanisme de detect,ie a erorilor. Sisteme de detect,ie, izolare s, i recuperare din erori sunt disponibile astazi în aproape toate sistemele spat,iale. Însa, costul echipamentelor poate împiedica crearea unor astfel de sisteme de detect,ie, care sa fie robuste s, i responsive. Aceasta teza analizeaza posibilitatea implementarii unui sistem de detect,ie, izolare s, i recuperare de la erori într-un satelit care este echipat cu un procesor cu un singur fir de execut,ie, care are un sistem de planificare cooperativ în sistemul de operare. Aceasta teza sugereaza o implementare cooperativa a unui astfel de sistem, unde fiecare proces este responsabil de câte o parte din detectarea erorilor. Analiza este realizata atât din perspectiva integrarii în sistemul de operare s, i procesele acestuia, cât s, i din interiorul acestui sistem de detect,ie, unde doua elemente importante sunt implementate s, i analizate: telemetria de depanare s, i modurile de operare. Rezultatele arata faptul ca este posibila implementarea unui sistem de detect,ie care este împart, it în toate componentele sistemului unui satelit s, i se comporta cooperativ. Mai departe, comparat,ia cu un sistem tradit,ional, dedicat, de detect,ie a erorilor arata ca erorile pot fi detectate mai rapid cu un sistem cooperativ.

An estimated 30% of the faults in Induction Machine (IM) are related to its stator. These faults are mostly in the form of an Inter-Turn Short Circuit (ITSC) fault i.e., when two winding inside the stator of IM are shorted due to insulation failure. However, ITSC fault can be avoided by detecting them in advance and then scheduling the maintenance of the IM. This thesis studies two methods for detecting this incipient ITSC fault in a three-phase IM and then estimating the stator resistance unbalance due to the ITSC fault. The first method is based on the asymmetry caused in the IM by the ITSC fault. As a result of this asymmetry, the negative sequence components of the stator voltages and the stator currents are generated inside the IM. A healthy IM also have these negative sequence components due to the manufacturing process and the supply voltage unbalances. The characteristics and the compensation methods of these negative sequence components in a healthy IM are discussed. The results show that after compensating the negative sequence components in a healthy machine, they can be used for detecting an ITSC fault and then to calculate the fault quantities as well as the stator resistance unbalances. The second method for detecting an ITSC fault is based on analysing the stator resistance unbalances. A three-phase drive is used to inject DC voltage in the stationary reference frame. The DC current generated by this DC voltage is measured and then by applying Ohm’s law stator phase resistances are calculated. In a healthy IM, the phase resistances are balanced. However, in case of ITSC fault in any of the phases, the phase resistance of that phase deviates from those of the other two phases which can be utilized for detecting ITSC fault.
Uppskattningsvis 30% av alla fel i induktionsmaskiner (IM) är kopplad till dess stator. Dessa fel är i huvudsak Inter-Turn Short Circuit (ITSC)-fel, dvs. två lindningar inom IM:ens stator blir kortsluta pga. ett isoleringsfel. Emellertid kan man undvika ITSC-fel genom att detektera dem i förhand och planera underhåll. Det här examensarbetet undersöker två metoder för att detektera ett förestående ITSC-fel i en tre-fas IM. Den första metoden är baserad på asymmetrin i IM:er pga. ITSC-felet. Resultatet av den här asymmetrin är att en negativ sekvens genereras i IM:ens statorspänning och statorström. En oskadad IM kan också visa dessa negativa sekvenser pga. tillverksprocessen och statorspänningsobalanser. Egenskaperna och kompensationsmetoderna för dessa negativa sekvenser i en oskadad IM kommer att diskuteras. Resultaten visar att efter kompenseringen av de negativa sekvenserna i en oskadad IM, kan de användas för att detektera ITSC-fel och efteråt för att beräkna felstorheter och även statormotståndobalanser. Den andra metoden för att detektera ITSC-fel är baserad på en undersökning av statormotståndobalanser. Ett tre-fas-drivsystem används för att injektera likspänning i den stationära referensramen. Likströmmen som följer av denna likspänning mäts och statorfasmotstånden beräkna efteråt med Ohms lag. I en oskadad IM är fasmotstånden balanserade. Däremot, när ett ITSC-fel uppstår i en fas, avviker fasmotståndet i den felaktiga fasen från de andra två fasernas, vilket kan användas för att detektera ITSC-fel.

Dans toutes les industries manufacturières, les chambres mises en parallèles sur une même opération de production sont censées donner un résultat similaire et offrir des produits de qualité identique. Ceci n'est toutefois pas toujours le cas dans les unités de production.Le maintien d'une performance stable des chambres parallèles dans l'industrie des semi-conducteurs est un véritable défi car les machines traitent simultanément un grand nombre de produits dans le but de maximiser le rendement et optimiser l'utilisation des machines. Le travail de thèse consiste à proposer une méthodologie permettant de détecter et de corriger en temps réel ces différences de performance en exploitant toutes les données disponibles, utilisées habituellement de façon séparée, pour identifier les causes racines de toute différence significative entre les chambres traitant des produits identiques.L'approche proposée consiste d'abord à détecter les écarts existants entre les chambres parallèles en se référant aux mesures des paramètres physiques. Les données des capteurs sont ensuite analysées pour mettre en évidence les indicateurs causant ces écarts. Ces indicateurs sont ajustés grâce à un mécanisme de contrôle efficace composé de deux parties : La métrologie virtuelle et la régulation. Tout d'abord, l'impact du réglage des paramètres d'entrée des chambres sur la qualité des produits est modélisé. Les modèles créés sont ensuite utilisés pour mettre en œuvre des boucles de régulation dont le but est de faire correspondre les indicateurs source de variabilité, et compenser ainsi l'erreur de sortie
In the manufacturing industries, the machines/chambers placed in parallel on the same production operation are expected to have similar capabilities and, most importantly, to yield identical product quality. However, this is usually not the case in real practice due to the systematic variations accumulated in time. Maintaining stable performance of parallel machines/chambers in the semiconductor industry is a critical challenge given the fact that, in the large-scale production environment, machines/chambers can process a large number of products simultaneously to maximize throughput and optimize machine utilization. Un- surprisingly, after processing very different settings, called recipes, the conditions of parallel machines/chambers will be no longer the same. This thesis proposes a methodology to detect and correct the performance differences in real-time by using all the available data, namely: measurements of physical parameters, data from sensors installed on machines, data from the control loops, and maintenance data. The core idea is to integrate the different sources of data, which are usually used separately, to identify the root causes of any significant differences among the machines/chambers that process identical recipes.The proposed approach starts by detecting existing gaps between parallel machines/ chambers by referring to the measurements of physical parameters since they reflect the quality of manufactured products. The sensor data are then analyzed to highlight the in- dicators that cause these discrepancies. These indicators are adjusted through an effective control mechanism composed of two parts: 1) virtual metrology and 2) process regulation. First, the impact of recipe changes on product quality is quantified by modeling the link between the inputs and outputs of the mismatched machines/chambers. The constructed models are then used to implement the revised control loops to match as much as possible the controllable input factors and compensate for the output errors

Cette thèse de doctorat aborde la problématique de la réalisation d'un système de contrôle actif de détection de défaut et diagnosis (FDD) pour un système de conversion photovoltaïque. Ce type de système de production d'énergie électrique est composé de panneaux solaires, d'un dispositif MPPT, d'un convertisseur de courant DC-DC, d'un onduleur DC-AC et d'une charge. Le système de contrôle actif à tolérance de pannes qui a été développé dans cette thèse est composé de deux étages : * Un étage assurant la fonction de diagnostic et comprenant les fonctions de détection de défauts, la fonction d'isolement de défauts, l'identification de défauts et l'estimation de l'ampleur du/des défaut(s) * Une fonction de reconfiguration du système photovoltaïque. Ce manuscrit est divisé en quatre chapitres : * Introduction au problème et révision de l'état de la technique * Modélisation mathématique du système photovoltaïque avec une validation expérimental de ce dernier effectué sur la plateforme PV de caractérisation du bâtiment réel ADREAM (Laboratoire LAAS-CNRS) * Conception et mise en œuvre du système de diagnostic de pannes du système photovoltaïque comprenant un Système actif à tolérance de pannes * Un système de diagnostic expérimental en cours de développement à l'aide d'un dispositif FPGA
This work contributes by developing an active fault tolerant control (AFTC) for Photovoltaic (PV) systems. The fault detection and diagnosis (FDD) methodology is based on the analysis of a model that compares real-time measurement. We use a high granularity PV array model in the FDD tool to allow faults to be detected in complex conditions. Firstly, the research focuses on fault detection in complex shadow conditions. A real-time approach is presented to emulate the electrical characteristics of PV modules under complex shadow conditions. Using a precise emulators approach is a real challenge to study the high non-linearity and the complexity of PV systems in partial shading. The real-time emulation was validated with simple experimental results under failure conditions to design specific fault-detection algorithms in a first sample. The second part of the research addresses the FDD method for DC/DC and DC/AC power converters that are connected to the grid. Primary results allowed us to validate the system's recovery for normal operating points after a fault with this complete AFTC approach. Emulations based on the simulation of distributed power converters, fault detection methodologies based on a model, and a hybrid diagnostician were then presented

Ce travail de recherche vise à développer des méthodes de détection et diagnostic appropriées et intégrables dans des systèmes soit d'aide à la décision dans le cadre de la gestion temps réel des réseaux d'assainissement, ou dans des procédures d'élaboration automatique de bilans quantitatifs de pollution légalement exigibles. Cette étude méthodologique débouche sur une synthèse et un approfondissement des techniques de détection et diagnostic de dysfonctionnements de réseau d'assainissement, et permet de dégager trois méthodes de détection reposant sur un modèle soit qualitatif soit analytique. L'une de ces méthodes s'applique à construire des comportements attendus du réseau synthétisés dans un modèle qualitatif dont la réponse est croisée avec les mesures. Les deux autres approches s'appuient sur un modèle hydrologique de transformation pluie-débit intégré dans un filtre de Kalman. La détection est alors réalisée soit avec un banc de filtres de Kalman complété par un test multi-hypothèses, soit avec un seul filtre de Kalman associé à un ensemble de tests d'hypothèses. Ces méthodes sont complétées par des analyses de sensibilité des procédures de détection et des exemples d'application pour des pluies de type différent. Enfin, une méthode de validation de mesures hydrauliques acquises en contexte hydrologique de temps sec est proposée. Elle s'appuie sur la création de scénarios de temps sec, comparés aux mesures hydrauliques grâce à des tests d'hypothèses. Cette approche permet alors d'identifier des contextes hydrologiques et hydrauliques caractérisant l'environnement d'un point de mesure

The aim of this thesis is to propose a new method to detect High impedance Faults, which relies on measuring the phase currents on electrical transmission and distribution grids using a common time source for synchronization. The goal is to measure the residual currents both at the transformer side and at the load side of the network. The time synchronization allows synchronized real-time measurements of the two residual currents at the two ends of the grid. In normal operating conditions and in all load conditions (balanced or unbalanced), the residual currents will result almost equal at the two ends. When a High Impedance Fault occurs, the residual currents at the two ends will differ. The difference in magnitude of the residual currents are monitored in real time and used as a pick-up criteria.

Cette thèse représente une contribution aux études sur la disponiblité des dispositifs électrotechnique. L'étude présentée vise à définir la méthode de redondance analytique, basée sur l'estimation d'état, la mieux adaptée au diagnostic des entraînements électriques en considérant les défauts de l'ensemble du convertisseur, de la commande et des capteurs. La méthode retenue doit permettre d'obtenir un modèle de diagnostic implantable en temps réel et sans ajout de capteurs supplémentaires. L'application retenue est un entraînement à courant continu commandé en couple. Le chapitre II compare deux modèles analytiques nommés modèles parallèle et permet de retenir un modèle parallèle "découplé" qui permet une bonne détection et une bonne localisation des défauts d'électronique de puissance ainsi que des défauts capteur. Malheureusement les modèles parallèles sont dépendants des entrées perturbatrices du procédé. Les perturbations génèrent donc des fausses alarmes La batterie d'observateur à entrées inconnues mise au point au chapitre III permet de s'affranchir de l'entrée perturbatrice que constitue dans notre cas le couple de charge. Cette technique est moins dépendante, en terme de découplage, du système car l'injection de sortie grace à la matrice de gain permet de disposer de degrés de liberté supplémentaires qui autorisent un réglage des découplages et des sensibilités. Les observateurs (à entrées inconnues) sont donc, à priori, les modèles de diagnostic les mieux adaptés à la à la détection et la localisation de défauts dans les entrainements électriques à courant continu.

Etude des problèmes théoriques et techniques liés à la réalisation des systèmes à base de connaissances pour assister en temps réel les taches décisionnelles des opérateurs en salle de contrôle des procédés continus : détection de dysfonctionnement, diagnostic et reprise.

Climatic changes are one of the biggest problems that humanity faces and renewable energies are a big weapon to fight this threat. Solar energy is one of the renewable energy sources in current use and to produce this type of energy there are several solar plants placed across the country. These giant plants are made of many sets of solar panels (called arrays) which are responsible for converting solar energy into electricity. One of the critical aspects of these plants' operation is the early detection of solar panel malfunctions. The current methods in use are expensive and consume a lot of time, meaning that, in some cases, the faults are detected only a year later, causing a huge financial impact on the companies responsible for the plants' operation. To cut these losses and to detect the faults as early as possible, this dissertation presents a real-time system capable of detecting malfunctions in a solar panel array. The node should be placed in the array's junction box and detects if an array has a faulty panel. The faults are detected comparing the array's output (voltage and current) with the output of an artificial neural network that models the array's behaviour using the real-time solar irradiance and temperature values. The neural network uses the measured values to carry out an online learning process, improving the network performance. Due to the plant's extension, a low power wide area network (LORAWAN), is used to send the array status and the data collected to the cloud, where they are processed and presented in a dashboard.

Climatic changes are one of the biggest problems that humanity faces and renewable energies are a big weapon to fight this threat. Solar energy is one of the renewable energy sources in current use and to produce this type of energy there are several solar plants placed across the country. These giant plants are made of many sets of solar panels (called arrays) which are responsible for converting solar energy into electricity. One of the critical aspects of these plants' operation is the early detection of solar panel malfunctions. The current methods in use are expensive and consume a lot of time, meaning that, in some cases, the faults are detected only a year later, causing a huge financial impact on the companies responsible for the plants' operation. To cut these losses and to detect the faults as early as possible, this dissertation presents a real-time system capable of detecting malfunctions in a solar panel array. The node should be placed in the array's junction box and detects if an array has a faulty panel. The faults are detected comparing the array's output (voltage and current) with the output of an artificial neural network that models the array's behaviour using the real-time solar irradiance and temperature values. The neural network uses the measured values to carry out an online learning process, improving the network performance. Due to the plant's extension, a low power wide area network (LORAWAN), is used to send the array status and the data collected to the cloud, where they are processed and presented in a dashboard.

碩士
國立交通大學
工學院精密與自動化工程學程
100
In a sensing system with redundant sensors,fault signal sensor was idenfied through the “voting equations”that was composed of relations of redundant sensors.However,while do sensors have noises,the former method will result in identifying inexactly ,nonreal-time estimation,and nonreal-time compensation.According to this issue,the thesis propose one method to solve this question and to reach real-time fault signal identification and correction.Then this algorithm design will take matlab program to stimulate and implement practical experiment to verify responsibility of this algorithm. We propose a progressive method that convert traditional fault identification method to observation nonlinear state estimation.That is,observation can observe every sensors’ fault signals then compensate these errors to reach real-time fault signal correction.To obtain this goal,we can design dynamic equations and output equations that is depicted redundant sensing system.We can scrutinize the observability of system to ensure fault signals that is estimated exactly .We create a state observer to observe fault signal.The state observer which is taken by ours is “kalman filter with fading memory”.After observer appraised the errors of sensors,we will take state feedback to correct fault signal of sensors.That is to say,we make corrected sensors which still can continue to operate in the sensing system. It is to raise output value precision and reliability of system. To simulate the matlab program,the algorithm of this thesis can evaluate instantaneously fault signals and compensate immediately.The point of this thesis design a realistic system to experiment and affirm this algorithm’s responsibility.To sum it up,we design a revolving mechanism of motor and place three accelerators(sensors) on it.The three accelaerators have been placed on different ratio radius(for example:1:1.5:2).Then we use microcontroller to detect accelerators’ A/D(analog to digital) datas. According normal component equation of curvilinear motion ( an=w^2*r),the three normal equations of curvilinear motion have restrainable relations among them. The three restrainable relations can obtain voting equations and then construct a redundant sensing system(observer).Finally,we collect accelerators’ A/D(analog to digital) datas friom microcontroller and take them to matlab program to validate whether the algorithm’s responsibility of this thesis.

碩士
明新科技大學
機械工程系精密機電工程碩士班
105
As the vacuum pumps have to continue operating in a factory, the most used method is the scheduled maintenance. However, this way may disturb the production of the factory, which also causes cost damage. The predictive maintenance provides a way to overcome the weaknesses of scheduled maintenance. When the equipment has no problem, the scheduled maintenance is not necessary. In the other hand, when the equipment releases any error signal, the timely maintenance becomes important. Predictive maintenance usually uses vibration measurement technique to carry out the condition of mechanical components. This research uses the accelerometer to detect the rotating conditions of the vacuum pumps. The collected signal is processed and compared with a golden signal to determine the state of the equipment. This analysis procedure becomes the base of predictive maintenance. Based on the knowledge of measurement technique and predictive maintenance, this research further uses Internet of Things (IOT) technique to construct a remote monitor system for the equipment. To validate this technique, this research also build a rotor system with build-in measurement and communication devices, which is able to transfer the detected signal to a remote server, and provide a real-time computer based monitoring system for the equipment.

碩士
逢甲大學
資訊工程學系
106
Smart Machinery Monitoring Systems have become the most important type of projects for a company to develop smart manufacturing. When smart machinery involves more and more data and periodic updates to the data are necessary, not only the available bandwidth of the internet limits the speed of uploading, but also the reliability of the monitoring platform in the cloud server will decrease and the difficulty of real-time monitoring will increase. In this paper, a feature of production data to set the frequency of uploading, and a way to reduce the amount of data periodically uploaded to a monitoring platform in the cloud server will be introduced. Also, the abnormal message of the built-in data collection method will be analysed to provide decisions for engineers between replacing and maintaining the electronic components, in addition to analysis of sensing component data to reduce the time in which they are out of service.

Faults tend to degrade the performance of HVAC systems by causing occupants' discomfort, increased energy consumption, and shorter life of equipment. Accordingly, it is very important to develop fault detection and diagnosis (FDD) tools for HVAC systems. In this thesis, in consideration of the fact that many faults could be traced back to oversized/undersized equipment, an interactive design tool is first developed to simulate the steady-state performance of a multi-zone variable-air-volume terminal reheat (VAV-TRH) HVAC system. This program is used for verifying HVAC system parameters to ensure that HVAC components are correctly sized. Then, a software program is developed to simulate the dynamic performance of a two-zone variable-air-volume terminal reheat (VAV-TRH) HVAC system. This program runs in two modes: off-line and on-line. Using this program, two control strategies---an optimal control strategy and a reheat control strategy, are studied. Simulation results show that the optimal control strategy leads to more energy savings than the reheat control strategy and is valid most of the operation time. The program is also useful for on-line fault detection and diagnosis for the HVAC system. In order to achieve this goal, an expert rule set for on-line FDD is established using knowledge-based approach. The expert rule set, in which IF-THEN clauses are applied, consists of simple rules that can successfully identify some fairly obvious problems that are often overlooked. The simulation experiments are performed to examine the performance of expert rules using data from real-time simulations. Simulation results show that expert rules are efficient in on-line detection and diagnosis of HVAC faults.

碩士
國立臺灣大學
機械工程學研究所
103
This thesis developed a real-time fault detection techniques for motor drive system of electric vehicles which enhanced the safety of the multiple traction electric vehicles. This research used dSPACE real-time control system simulator as a fault/failure mode analysis platform for motor drive system of electric vehicles. This platform is used to implement the quantitative and qualitative analysis for fault/failure mode of motor drive system and construct the failure mode and effect analysis (DFMEA) in order to execute the research of real-time fault detection and classification of motor drive system. This research proposed a real-time fault detection method based on the wavelet transform technique which combined with model predictive control theory, developing a fault tolerant control techniques for enhancing the safety and system reliability of propulsion system of multiple traction electric vehicles.

This thesis presents a critical appraisal of several differing approaches to the design and testing of fault detection (FD) algorithms monitoring the instrumentation used in the alternated aerobic/anoxic cycles (AC) process for nitrogen removal. Several features are selected as the basis of the FD, involving the slope and the timing of the process measurements of the various nitrogen compounds. Two separate FD algorithms have been developed for the anoxic and the aerobic phases, requiring a separate tuning but sharing the same principles: first some low-level checks are performed on the raw signals, discriminating gross malfunctions like missing data and spikes, then more sophisticated methods are used to investigate the presence of more subtle anomalies that were not detected by the previous screening. The FD problem is treated either in terms of classification problem, testing different algorithms such as binary trees, support vector machines (SVM) and principal component analysis (PCA), or as a forecasting one, using the Bayesian theory to predict the faulty or normal state of the process based on the previous records. An operational data set obtained from a municipal plant was used to first train the algorithm. However, due to the fairly limited information which could be extracted, a more comprehensive data set was created building an AC model based on the standard Benchmark simulation model with improved nitrogen kinetics and seasonal temperature variations. Detailed sensor models were also included, so that the occurrence of faults could be totally controlled, both in kind and timing. The performances of the various methods on either the operational and the synthetic datasets have been assessed comparing the anomalies detected by the methods with those actually observed. While the great majority of the gross faults is successfully detected by the preliminary screening, differing performances of the subsequent finer detection are obtained, depending on both the quality of data set and the detection method used: poorer results are observed using the plant data, in part due to an insufficient characterization of the fault events and in part due to the limited number of signals monitored. The higher availability of measurements provided by the numerical model, instead, enhances the discrimination capabilities of the tested methods, especially the nonlinear SVM, while the PCA-based approach and the Bayesian predictor results less affected by a change in the combination of diagnostic parameters used.

A survey of literature and existing technologies indicated that algorithms and systems for on-board monitoring devices for freight wagons are still not well developed. This was found to be primarily due to the difficulty of developing a practical on-board monitoring system due to the lack of power and electrical communications on standard freight wagons. In addition, large wagon fleet sizes make it imperative that any such system be both robust and cheap. Consequently, it was observed that traditional approaches to freight wagon health monitoring are performed from track side instruments as the wagons pass by. As it could be seen that any freight monitoring system must require only limited instrumentation, and to ensure that the research in this thesis would have practical usefulness the target design was constrained to using just two tri-axial accelerometers. Various approaches to inferring wagon dynamic behaviour and safety from just acceleration measurements were reviewed. These included model-based health monitoring methods and signal-based fault detection and isolation (FDI) methods. Model-based health monitoring methods, such as inverse modelling methods, Kalman Filter based methods, etc., utilise various estimation models to analyse sensor-collected acceleration signals for either real-time wheel-rail dynamic predications or real-time suspension parameter estimation. Conversely, signal-based FDI methods analyse sensor-collected acceleration signals directly via various signal-processing methods to detect the occurring faults directly. From the review it was discovered that linearisation of the models was required by model-based health monitoring methods which affects the accuracy of the results given the highly non-linear nature of railway vehicle suspensions and dynamic responses. The signal-based FDI methods, however, were found to be impeded by a different problem, that of requiring a large prebuilt database to cover all possible fault conditions. Such a database is time-consuming and difficult to build for the proposed wagon application. It was realised that advantage could be taken in the heavy haul train context from the fact that such trains are generally made up of near identical wagons and usually near identical service condition. It was proposed that the problem of a prebuilt database construction could be solved via a “Self-collaborate” system of sharing and comparing data between wagons via local communications. A signal-based method was therefore proposed based on the concept of cross-correlation and comparisons between adjacent heavy haul wagons. To generate data to test the proposed method, simulations were completed using a realistic and detailed MBS model of a typical 40t axle load heavy haul wagon. Simulations were undertaken using the Gensys vehicle dynamics software package. The wagon was modelled with the following detail: all vehicle components, including a carbody, two bolsters, four sideframes and four wheelsets, were given six degrees of freedom (DOFs), and all bolster springs and wedge damper springs were modelled as nonlinear stiffness elements. The track and operational conditions chosen for testing the approach were for a tight curve at prescribed curve speed and straight track at full speed. The FRA Class 4 track irregularity was assumed for the track surface. A method using cross-correlation of acceleration measurements to calculate Fault Indicators (FIs) was developed. This method was tested for various wagon suspension faults including changes to spring stiffness and damping. Two categories of cross-correlation analyses were made, including the cross-correlation analyses of acceleration signals between different directions from same sensors (Category 1) and the cross-correlation analyses of acceleration signals in the same directions between front and rear sensor (Category 2). Analysis results showed that, among all the proposed Fault Indicators, cross-correlation of vertical accelerations between the front and rear sensor was the most sensitive FI to both bolster spring faults and wedge damper faults. Though the considered faults did not cause more than 10% variations in wheel-rail dynamic parameters, namely wheel unloading and derailment index, the proposed FIs indicated over 300% variation, indicating high sensitivity. The sensitivity of these results and the robustness of the proposed method were further explored with variations in track condition and areas of further work recommended.

碩士
崑山科技大學
電子工程研究所
103
This paper is a cloud-based surveillance system, which utilizes Pan-Tile-Zoom camera (PTZC) to develop a moving human detection and behavior analysis under complex foreground for surveillance system development using Storm, a distributed and fault-tolerant real-time computation in the cloud. The system first builds the background image needed for moving human and behavior detection, and then uses the temporal difference to obtain the moving objects. The information of moving human is obtained from a rate algorithm. Finally, the combination of temporal difference and background subtraction methods is used as the foundation for moving human detection and behavior analysis under the complex Foreground. To detect moving human smoothly, the ratio of moving objects was used to differentiate between human and the others. After video streaming was processed by the algorithm, an extra frame was put on the moving human and the sequential actions can also be detected. From the tracking results, the system would generate internal events and send warning messages to users, relative people or organizations through email, SMS or some subsystems coordinated by XMPP protocol. Moreover, this paper will do further experiments on multiple moving humans to show that the system can reach the expectations of multiple moving humans detection and behavior analysis under complex foreground, and the feasibility of the proposed methods.