Submitted by ANDERSON INÁCIO JUNQUEIRA JUNIOR null (anderson.inacio@unirv.edu.br) on 2016-10-07T04:28:58Z No. of bitstreams: 1 Dissertação - Anderson Inácio.pdf: 7113990 bytes, checksum: 9e01964850a125e0dfa6f5661554c9ce (MD5) / Approved for entry into archive by Juliano Benedito Ferreira (julianoferreira@reitoria.unesp.br) on 2016-10-13T20:07:04Z (GMT) No. of bitstreams: 1 junqueirajunior_ai_me_ilha.pdf: 7113990 bytes, checksum: 9e01964850a125e0dfa6f5661554c9ce (MD5) / Made available in DSpace on 2016-10-13T20:07:04Z (GMT). No. of bitstreams: 1 junqueirajunior_ai_me_ilha.pdf: 7113990 bytes, checksum: 9e01964850a125e0dfa6f5661554c9ce (MD5) Previous issue date: 2016-08-04 / O presente trabalho apresenta um projeto e construção de um protótipo para separar partículas de desgaste em óleos lubrificantes. Devido à necessidade da confiabilidade de máquinas e equipamentos para a redução de custos fabris, as indústrias recorreram às ações preditivas de manutenção. Dentre as várias ações preditivas de manutenção pode-se citar a análise de lubrificantes. Lubrificantes são materiais colocados entre duas superfícies interativas afim de preencher as irregularidades superficiais, assim reduzindo o atrito e o desgaste. Os lubrificantes podem ser sólidos, semifluidos e fluidos. O óleo é um lubrificante líquido, partículas de desgaste presentes no óleo podem danificar componentes vitais de máquinas. Um método utilizado para analisar as partículas de desgaste presentes no óleo lubrificante é a ferrografia qualitativa. Com base em fundamentos teóricos de obras renomadas no meio cientifico, o objetivo do presente trabalho é projetar e construir um protótipo separador rotativo de partículas de baixo custo, para obter ferrogramas quantitativos de boa qualidade e comparar com os modelos convencionais encontrados no mercado. Através de cinco combinações diferentes de ímãs permanentes foi possível obter ferrogramas, sendo que as combinações 03, 04 e 05 apresentaram melhor qualidade de formação de anéis de partículas ferrosas. A combinação 04 apresentou apenas dois anéis, estes são mais fidedignos em relação aos anéis do aparelho convencional. / This paper presents a project and construction of a prototype to separate wear particles in lubricating oils. Due to the need for reliability of machines and equipment to reduce manufacturing costs, manufacturers have turned to predictive maintenance actions. Among the various predictive maintenance actions can be cited the analysis of lubricants. Lubricants are interactive material placed between two surfaces in order to fill the surface irregularities, thereby reducing friction and wear. The lubricants may be solid, slurries and fluids. The oil is a liquid lubricant, wear particles in the oil can damage critical components of machinery. A method for analyzing wear particles in the lubricating oil is qualitative ferrography. Based on theoretical foundations of renowned works in the scientific environment, the objective of this work is to project and construct a rotary separator prototype low cost particles for quantitative ferrogramas good quality and compare with conventional models available on the market. Through five different combinations of permanent magnets was possible to obtain ferrogramas, and combinations 03, 04 and 05 showed better quality training rings of ferrous particles. The combination 04 had only two rings, these are more reliable with respect to the rings of the conventional device.

Orientador: Aparecido Carlos Gonçalves / Resumo: O presente trabalho apresenta um projeto e construção de um protótipo para separar partículas de desgaste em óleos lubrificantes. Devido à necessidade da confiabilidade de máquinas e equipamentos para a redução de custos fabris, as indústrias recorreram às ações preditivas de manutenção. Dentre as várias ações preditivas de manutenção pode-se citar a análise de lubrificantes. Lubrificantes são materiais colocados entre duas superfícies interativas afim de preencher as irregularidades superficiais, assim reduzindo o atrito e o desgaste. Os lubrificantes podem ser sólidos, semifluidos e fluidos. O óleo é um lubrificante líquido, partículas de desgaste presentes no óleo podem danificar componentes vitais de máquinas. Um método utilizado para analisar as partículas de desgaste presentes no óleo lubrificante é a ferrografia qualitativa. Com base em fundamentos teóricos de obras renomadas no meio cientifico, o objetivo do presente trabalho é projetar e construir um protótipo separador rotativo de partículas de baixo custo, para obter ferrogramas quantitativos de boa qualidade e comparar com os modelos convencionais encontrados no mercado. Através de cinco combinações diferentes de ímãs permanentes foi possível obter ferrogramas, sendo que as combinações 03, 04 e 05 apresentaram melhor qualidade de formação de anéis de partículas ferrosas. A combinação 04 apresentou apenas dois anéis, estes são mais fidedignos em relação aos anéis do aparelho convencional. / Mestre

Orientador: Mauro Hugo Mathias / Banca: João Zangrandi Filho / Banca: Anselmo Monteiro Ilkiu / Resumo: O procedimento de análise de óleos e graxas vem sendo implementado em diferentes setores, tais como indústria de papel e celulose, usina de álcool, empresas marítimas e empresas aéreas comerciais. As técnicas de análise de óleos lubrificantes permitem o aumento da vida útil de equipamentos, a economia de custos de manutenção. Na área militar esta abordagem é usual desde a década de quarenta. No entanto os países desenvolvidos que vêm participando de grande conflitos bélicos têm explorado muito este recurso com o objetivo de minimizar os custos operacionais e aperfeiçoar o cumprimento de missões. Assim, esta tecnologia envolvida é estratégica, cabendo aos países em desenvolvimento a implementação de programas mais eficazes de forma a ter maior grau de independência, aumento da disponibilidade e confiabilidade de equipamentos de modo a economizar recursos. A proposta da pesquisa em desenvolvimento é complementar o programa de manutenção baseado em análise de óleos e graxas no sentido de explorar e aplicar o potencial destas técnicas, visando as atividades de dificuldades em serviço (problemas operacionais) e investigação de incidentes/acidentes aeronaúticos. O estudo de viabilidade deste tipo de programa poderá gerar um programa de Garantia da Qualidade que permita detectar os modos de falha nos sistemas das aeronaves de defesa. Por todas as considerações acima mencionadas, um programa de análise de óleo é imperativo numa empresa aérea comercial ou mesmo numa Força Aérea de Defesa. As possíveis desvantagens na implementação de um programa de análise de óleo são o seu custo ( logística, recursos humanos, instalações, equipamentos, etc) e o tempo que o citado programa leva para ser consolidado. / Abstract: The procedure of oil and grease analysis has been implemented in different sectors, such as industry of paper and cellulose, maritime, alcohol plants, companies and commercial airlines. The techniques of lubricate oil analysis allow the increase of the useful equipment life, the economy in maintenance costs. In the military field this boarding are usual since the decade of forty, however the developed countries that has been participated in World War II have explored much of this resource with the objective to minimize the operational costs and to optimize the missions fulfillment. Thus this involved technology is strategic, fitting to the developing countries the implementation of more efficient programs of form to have greater degree of independence, increase of the availability an equipment reliability in order to save resources. The proposal of the research carried on is to complement the program of maintenance based on oil and grease analysis and greases in order explore and to apply the potential of these techniques, aiming at to the activities of difficulties in service (operational problems) an inquiry of aeronautical incidents/accidents. The feasibility study this type of program will be able to generate a program of Quality Assurance that allows detecting the failure modes in the systems of the defense aircraft. For all considerations mentioned above, a program of oil analysis is imperative in a commercial airline or even in an Air Force of Defense. The possible disadvantages to the implementation of oil analysis program are the their cost (logistic, human resources, physical place, equipments and so on) and the time to consolidate such kind of program. / Mestre

A crescente implementação pela indústria de técnicas de manutenção preditiva exige cada vez mais o aprimoramento dos procedimentos capazes de fornecer informações sobre o estado de um equipamento. Dentre os procedimentos de análise existentes para máquinas rotativas, a análise de vibração é um dos mais utilizados sendo, atualmente inclusive, presente em larga parcela de setores industriais importantes no Brasil (como o setor de celulose e papel, por exemplo). Isto faz, portanto, cada vez mais importante explorar todas as possibilidades desta técnica. Este trabalho inicia uma investigação sobre as relações entre vibração (assinatura mecânica) e lubrificação de máquinas rotativas e assim, como ponto de partida deste estudo, procura avaliar o efeito da variação da viscosidade do lubrificante no sinal de vibração de caixas de engrenagem. O trabalho apresenta um grande banco de dados experimental, discute diversos métodos de processamento de sinais e apresenta uma característica do sinal de vibração que foi capaz de identificar alterações na viscosidade do óleo lubrificante no caso apresentado. / The crescent implementation, by brazilian industry, of predictive maintenance techniques demands, from vibration analyses processes, more capability for supplying information on the state of equipment. Among the existent analysis procedures for rotative machines, the vibration analysis is one of the more used, being nowadays, present in a wide portion of important industrial sections in Brazil (as the cellulose pulp and paper for instance). This makes, therefore, more and more important to explore all of the possibilities of this method. This work begins an investigation about the relation between vibration (mechanical signature) and lubrication of rotative machines. As a starting point of this study, it tries to evaluate the effect of the variation of the viscosity of the lubricant on the vibration signature of a gear box. The work presents a large experimental database, discusses several methods of signal processing and presents a characteristic of the vibration signal capable to identify alterations in the viscosity of the lubricating oil in the tested equipment.

In the industry, a lot of companies are facing the explosion of big data. With this much information stored, companies want to make sense of the data and use it to help them for better decision making, especially for future prediction. A lot of money can be saved and huge revenue can be generated with the power of big data. When building statistical learning models for prediction, companies in the industry are aiming to build models with efficiency and high accuracy. After the learning models have been developed for production, new data will be generated. With the updated data, the models have to be updated as well. Due to this nature, the model performs best today doesn’t mean it will necessarily perform the same tomorrow. Thus, it is very hard to decide which algorithm should be used to build the learning model. This paper introduces a new method that ensembles the information generated by two different classification statistical learning algorithms together as inputs for another learning model to increase the final prediction power. The dataset used in this paper is NASA’s Turbofan Engine Degradation data. There are 49 numeric features (X) and the response Y is binary with 0 indicating the engine is working properly and 1 indicating engine failure. The model’s purpose is to predict whether the engine is going to pass or fail. The dataset is divided in training set and testing set. First, training set is used twice to build support vector machine (SVM) and neural network models. Second, it used the trained SVM and neural network model taking X of the training set as input to predict Y1 and Y2. Then, it takes Y1 and Y2 as inputs to build the Penalized Logistic Regression model, which is the ensemble model here. Finally, use the testing set follow the same steps to get the final prediction result. The model accuracy is calculated using overall classification accuracy. The result shows that the ensemble model has 92% accuracy. The prediction accuracies of SVM, neural network and ensemble models are compared to prove that the ensemble model successfully captured the power of the two individual learning model.

Failures are the eminent aspect of any machine and so is true for vehicle as it is one of the sophisticated machines of today’s time. Early detection of faults and prioritized maintenance is a necessity of vehicle manufactures as it enables them to reduce maintenance cost and increase customer satisfaction. In our research, we have proposed a method for processing Logged Vehicle Data (LVD) that uses Ant-Miner algorithm which is a Ant Colony Optimization (ACO) based Algorithm. It also utilizes processes like Feature engineering, Data preprocessing. We tried to explore the effectiveness of ACO for solving classification problem in the form of fault detection and prediction of failures which would be used for predictive maintenance by manufacturers. From the seasonal and yearly model that we have created, we have used ACO to successfully predict the time of failure which is the month with highest likelihood of failure in vehicle’s components. Here, we also validated the obtained results. LVD suffers from data imbalance problem and we have implemented balancing techniques to eliminate this issue, however more effective balancing techniques along with feature engineering is required to increase accuracy in prediction.

<div>Equipment health is the root of productivity and profitability in a company; through the use of machine learning and advancements in computing power, a maintenance strategy known as Predictive Maintenance (PdM) has emerged. The predictive maintenance approach utilizes performance and condition data to forecast necessary machine repairs. Predicting maintenance needs reduces the likelihood of operational errors, aids in the avoidance of production failures, and allows for preplanned outages. The PdM strategy is based on machine-specific data, which proves to be a valuable tool. The machine data provides quantitative proof of operation patterns and production while offering machine health insights that may otherwise go unnoticed.</div><div><br> </div><div>Purdue University's Wade Utility Plant is responsible for providing reliable utility services for the campus community. The Wade Utility Plant has invested in an equipment monitoring system for a thirty-megawatt turbine generator. The equipment monitoring system records operational and performance data as the turbine generator supplies campus with electricity and high-pressure steam. Unplanned and surprise maintenance needs in the turbine generator hinder utility production and lessen the dependability of the system.</div><div><br> </div> The work of this study leverages the turbine generator data the Wade Utility Plant records and stores, to justify equipment care and provide early error detection at an in-house level. The research collects and aggregates operational, monitoring and performance-based data for the turbine generator in Microsoft Excel, creating a dashboard which visually displays and statistically monitors variables for discrepancies. The dashboard records ninety days of data, tracked hourly, determining averages, extrema, and alerting the user as data approaches recommended warning levels. Microsoft Excel offers a low-cost and accessible platform for data collection and analysis providing an adaptable and comprehensible collection of data from a turbine generator. The dashboard offers visual trends, simple statistics, and status updates using 90 days of user selected data. This dashboard offers the ability to forecast maintenance needs, plan work outages, and adjust operations while continuing to provide reliable services that meet Purdue University's utility demands. <br>

Background: The gearbox and machinery faults prediction are expensive both in terms of repair and loss output in production. These losses or faults may lead to complete machinery or plant breakdown.  Objective: The goal of this study was to apply advanced machine learning techniques to avoid these losses and faults and replace them with predictive maintenance. To identify and predict the faults in industrial machinery using Machine Learning (ML)  and Deep Learning (DL) approaches.  Methods: Our study was based on two types of datasets which includes gearbox and rotatory machinery dataset. These datasets were analyzed to predict the faults using machine learning and deep neural network models. The performance of the model was evaluated for both the datasets with binary and multi-classification problems using the different machine learning models and their statistics. Results: In the case of the gearbox fault dataset with a binary classification problem, we observed random forest and deep neural network models performed equally well, with the highest F1-score and AUC score of around 0.98 and with the least error rate of 7%.  In addition to this, in the case of the multi-classification rotatory machinery fault prediction dataset, the random forest model outperformed the deep neural network model with an AUC score of 0.98.  Conclusions: In conclusion classification efficiency of the Machine Learning (ML) and Deep Neural Network (DNN) model were tested and evaluated. Our results show Random Forest (RF) and Deep Neural Network (DNN) models have better fault prediction ability to identify the different types of rotatory machinery and gearbox faults as compared to the decision tree and AdaBoost.  Keywords: Machine Learning, Deep Learning, Big Data, Predictive Maintenance, Rotatory Machinery Fault Prediction, Gearbox Fault Prediction, Machinery Fault Database, Internet of Things (IoT), Spectra quest machinery fault simulator, Cloud Computing, Industry 4.0

In recent years, advances in information technology have led to an increasing number of devices (or things) being connected to the internet; the resulting data can be used by applications to acquire new knowledge. The Internet of Things (IoT) (a network of computing devices that have the ability to interact with their environment without requiring user interaction) and big data (a field that deals with the exponentially increasing rate of data creation, which is a challenge for the cloud in its current state and for standard data analysis technologies) have become hot topics. With all this data being produced, new applications such as predictive maintenance are possible. One such application is monitoring a fleet of vehicles in real-time to predict their remaining useful life, which could help companies lower their fleet management costs by reducing their fleet's average vehicle downtime. Consensus self-organized models (COSMO) approach is an example of a predictive maintenance system for a fleet of public transport buses, which attempts to diagnose faulty buses that deviate from the rest of the bus fleet. The present work proposes a novel IoT-based architecture for predictive maintenance that consists of three primary nodes: namely, the vehicle node (VN), the server leader node (SLN), and the root node (RN). The VN represents the vehicle and performs lightweight data acquisition, data analytics, and data storage. The VN is connected to the fleet via its wireless internet connection. The SLN is responsible for managing a region of vehicles, and it performs more heavy-duty data storage, fleet-wide analytics, and networking. The RN is the central point of administration for the entire system. It controls the entire fleet and provides the application interface to the fleet system. A minimally viable prototype (MVP) of the proposed architecture was implemented and deployed to a garage of the Soci\'et\'e de Transport de l'Outaouais (STO), Gatineau, Canada. The VN in the MVP was implemented using a Raspberry Pi, which acquired sensor data from a STO hybrid bus by reading from a J1939 network, the SLN was implemented using a laptop, and the RN was deployed using meshcentral.com. The goal of the MVP was to perform predictive maintenance for the STO to help reduce their fleet management costs. The present work also proposes a fleet-wide unsupervised dynamic sensor selection algorithm, which attempts to improve the sensor selection performed by the COSMO approach. I named this algorithm the improved consensus self-organized models (ICOSMO) approach. To analyze the performance of ICOSMO, a fleet simulation was implemented. The J1939 data gathered from a STO hybrid bus, which was acquired using the MVP, was used to generate synthetic data to simulate vehicles, faults, and repairs. The deviation detection of the COSMO and ICOSMO approach was applied to the synthetic sensor data. The simulation results were used to compare the performance of the COSMO and ICOSMO approach. Results revealed that in general ICOSMO improved the accuracy of COSMO when COSMO was not performing optimally; that is, in the following situations: a) when the histogram distance chosen by COSMO was a poor choice, b) in an environment with relatively high sensor white noise, and c) when COSMO selected poor sensors. On average ICOSMO only rarely reduced the accuracy of COSMO, which is promising since it suggests deploying ICOSMO as a predictive maintenance system should perform just as well or better than COSMO . More experiments are required to better understand the performance of ICOSMO. The goal is to eventually deploy ICOSMO to the MVP.

Cette thèse s’inscrit dans le cadre de la disposition CIFRE entre la société Robert Bosch et l’Université de Caen Normandie au sein du laboratoire GREYC puis LAC. Elle comprend deux volets, l'un étant à caractère fondamental et porte sur la synthèse d'observateurs. Quant à l'autre volet, il est à caractère plus appliqué et porte sur la mise en œuvre d’un réseau de capteurs et assurer l’acheminement des informations provenant de ces derniers. Ce dernier est effectué au moyen du réseau, qu’il soit câblé ou sans fil.Dans le premier volet de l'étude, on s'est intéressé à une classe assez générale de systèmes temps-variant et affines en l'état, dont la sortie est mesurée avec un retard et échantillonnée. La nouveauté dans cette classe de systèmes est double :(i) l'équation d'état est sujette à une injection du signal de sortie et se retrouve de ce fait dépendante des sorties futures qui sont indisponibles;(ii) les sorties futures interviennent, au niveau de l'équation d'état, non seulement sous la forme usuelle d'une fonction indépendante de l'état, mais aussi à travers la matrice d'état ellemême qui, de ce fait, apparaît comme une quantité inconnue du modèle.Ces deux nouveautés du modèle entrainent en fait la perte du caractère « affine en l'état » de ce dernier et font que l'on se retrouve confronté à un problème de synthèse d'observateurs jamais résolu antérieurement. La solution que nous proposons est un observateur de type « Filtre de Kalman » augmenté d'un prédicteur inter-échantillons et d'opérateurs de saturations. Nous analysons la stabilité exponentielle du système d'erreur d'estimation d'état en utilisant le théorème du petit gain et des outils de la stabilité de Lyapunov. L'analyse met en évidence l'existence d'intervalles d'admissibilité dans lesquels doivent se situer les valeurs admissibles du retard et de la période d'échantillonnage, afin de garantir la convergence exponentielle de l'observateur.Dans le deuxième volet, nous nous intéressons à la maintenance prédictive au travers d’applications pratiques via la mise en place d’un réseau de capteurs. Le but de ce réseau est d’effectuer de la maintenance prédictive sur les équipements sensibles. Ce dernier est un composant essentiel à la mise en œuvre d’applications IoT et Industrie 4.0.Des applications de l’IoT et de l’Industrie 4.0 sur le site Robert Bosch de Mondeville sont évoqués ainsi que le développement d’un simulateur de perturbations réseau afin de tester la robustesse de la communication d’un capteur vers un client. / This thesis is part of the CIFRE agreement between the company Robert Bosch and the University of Caen Normandy in the laboratory GREYC then LAC. It consists of two parts, one which is of a fundamental nature and concerns the synthesis of observers. For the other part, it is more applied and concerns the implementation of a sensors network and ensure the routing of information from them. This is done through the network, whether wired or wireless.In the first part of the study, we looked at a fairly general class of time-varying and affine systems as they are, whose output is measured with a delay and sampled. The novelty in this class of systems is twofold:(i) the state equation is subject to an output signal injection and is therefore dependent on future outputs that are unavailable;(ii) future outputs occur at the state equation not only in the usual form of a stateindependent function, but also through the state matrix itself, which fact, appears as an unknown quantity of the model.These two novelties of the model cause in fact the loss of the "affine in the state" character of this last one, and make that one is confronted with a synthesis problem of observers never resolved previously. The solution we propose is a "Kalman filter" type observer augmented by an inter-sample predictor and saturation operators. We analyze the exponential stability of the state estimation error system by using the small gain theorem and tools of Lyapunov stability. The analysis highlights the existence of eligibility intervals in which the allowable values of the delay and the sampling period must be located in order to ensure the exponential convergence of the observer.In the second part, we are interested in predictive maintenance through practical applications via the installation of a sensor network. The purpose of this network is to perform predictive maintenance on sensitive equipment. The latter is an essential component for the implementation of IoT and Industry 4.0 applications.The purpose of this sensor network is to perform predictive maintenance on sensitive equipment. The latter is an essential component for the implementation of IoT and Industry 4.0 applications. Moreover, an observer of sampled data for affine systems in the state with output injection was studied on the basis of observers.Applications of IoT and Industry 4.0 on the Robert Bosch site in Mondeville are discussed as well as the development of a network disturbance simulator to stress the robustness of the communication of a sensor to a client.