Sistema de detecção e isolamento de falhas em sistemas dinâmicos baseado em identificação paramétrica

The present research aims at contributing to the area of detection and diagnosis of failure through the proposal of a new system architecture of detection and isolation of failures (FDI, Fault Detection and Isolation). The proposed architecture presents innovations related to the way the physical values monitored are linked to the FDI system and, as a consequence, the way the failures are detected, isolated and classified. A search for mathematical tools able to satisfy the objectives of the proposed architecture has pointed at the use of the Kalman Filter and its derivatives EKF (Extended Kalman Filter) and UKF (Unscented Kalman Filter). The use of the first one is efficient when the monitored process presents a linear relation among its physical values to be monitored and its out-put. The other two are proficient in case this dynamics is no-linear. After that, a short comparative of features and abilities in the context of failure detection concludes that the UFK system is a better alternative than the EKF one to compose the architecture of the FDI system proposed in case of processes of no-linear dynamics. The results shown in the end of the research refer to the linear and no-linear industrial processes. The efficiency of the proposed architecture may be observed since it has been applied to simulated and real processes. To conclude, the contributions of this thesis are found in the end of the text

Detecção e isolamento de falhas em sistemas dinâmicos baseados em redes neurais

This master dissertation presents the development of a fault detection and isolation system based in neural network. The system is composed of two parts: an identification subsystem and a classification subsystem. Both of the subsystems use neural network techniques with multilayer perceptron training algorithm. Two approaches for identifica-tion stage were analyzed. The fault classifier uses only residue signals from the identification subsystem. To validate the proposal we have done simulation and real experiments in a level system with two water reservoirs. Several faults were generated above this plant and the proposed fault detection system presented very acceptable behavior. In the end of this work we highlight the main difficulties found in real tests that do not exist when it works only with simulation environments

Sistema híbrido para detecção e diagnóstico de falhas em sistemas dinâmicos

The industries are getting more and more rigorous, when security is in question, no matter is to avoid financial damages due to accidents and low productivity, or when it s related to the environment protection. It was thinking about great world accidents around the world involving aircrafts and industrial process (nuclear, petrochemical and so on) that we decided to invest in systems that could detect fault and diagnosis (FDD) them. The FDD systems can avoid eventual fault helping man on the maintenance and exchange of defective equipments. Nowadays, the issues that involve detection, isolation, diagnose and the controlling of tolerance fault are gathering strength in the academic and industrial environment. It is based on this fact, in this work, we discuss the importance of techniques that can assist in the development of systems for Fault Detection and Diagnosis (FDD) and propose a hybrid method for FDD in dynamic systems. We present a brief history to contextualize the techniques used in working environments. The detection of fault in the proposed system is based on state observers in conjunction with other statistical techniques. The principal idea is to use the observer himself, in addition to serving as an analytical redundancy, in allowing the creation of a residue. This residue is used in FDD. A signature database assists in the identification of system faults, which based on the signatures derived from trend analysis of the residue signal and its difference, performs the classification of the faults based purely on a decision tree. This FDD system is tested and validated in two plants: a simulated plant with coupled tanks and didactic plant with industrial instrumentation. All collected results of those tests will be discussed

Critério de correntropia no treinamento de redes fuzzy wavelet neural networks para identificação de sistemas dinâmicos não lineares

The great interest in nonlinear system identification is mainly due to the fact that a large amount of real systems are complex and need to have their nonlinearities considered so that their models can be successfully used in applications of control, prediction, inference, among others. This work evaluates the application of Fuzzy Wavelet Neural Networks (FWNN) to identify nonlinear dynamical systems subjected to noise and outliers. Generally, these elements cause negative effects on the identification procedure, resulting in erroneous interpretations regarding the dynamical behavior of the system. The FWNN combines in a single structure the ability to deal with uncertainties of fuzzy logic, the multiresolution characteristics of wavelet theory and learning and generalization abilities of the artificial neural networks. Usually, the learning procedure of these neural networks is realized by a gradient based method, which uses the mean squared error as its cost function. This work proposes the replacement of this traditional function by an Information Theoretic Learning similarity measure, called correntropy. With the use of this similarity measure, higher order statistics can be considered during the FWNN training process. For this reason, this measure is more suitable for non-Gaussian error distributions and makes the training less sensitive to the presence of outliers. In order to evaluate this replacement, FWNN models are obtained in two identification case studies: a real nonlinear system, consisting of a multisection tank, and a simulated system based on a model of the human knee joint. The results demonstrate that the application of correntropy as the error backpropagation algorithm cost function makes the identification procedure using FWNN models more robust to outliers. However, this is only achieved if the gaussian kernel width of correntropy is properly adjusted.

Identificação não linear usando uma rede fuzzy wavelet neural network modificada

In last decades, neural networks have been established as a major tool for the identification of nonlinear systems. Among the various types of networks used in identification, one that can be highlighted is the wavelet neural network (WNN). This network combines the characteristics of wavelet multiresolution theory with learning ability and generalization of neural networks usually, providing more accurate models than those ones obtained by traditional networks. An extension of WNN networks is to combine the neuro-fuzzy ANFIS (Adaptive Network Based Fuzzy Inference System) structure with wavelets, leading to generate the Fuzzy Wavelet Neural Network - FWNN structure. This network is very similar to ANFIS networks, with the difference that traditional polynomials present in consequent of this network are replaced by WNN networks. This paper proposes the identification of nonlinear dynamical systems from a network FWNN modified. In the proposed structure, functions only wavelets are used in the consequent. Thus, it is possible to obtain a simplification of the structure, reducing the number of adjustable parameters of the network. To evaluate the performance of network FWNN with this modification, an analysis of network performance is made, verifying advantages, disadvantages and cost effectiveness when compared to other existing FWNN structures in literature. The evaluations are carried out via the identification of two simulated systems traditionally found in the literature and a real nonlinear system, consisting of a nonlinear multi section tank. Finally, the network is used to infer values of temperature and humidity inside of a neonatal incubator. The execution of such analyzes is based on various criteria, like: mean squared error, number of training epochs, number of adjustable parameters, the variation of the mean square error, among others. The results found show the generalization ability of the modified structure, despite the simplification performed

Modelagem e controle adaptativo de uma planta didática de nível com instrumentação industrial

The control, automation and optimization areas help to improve the processes used by industry. They contribute to a fast production line, improving the products quality and reducing the manufacturing costs. Didatic plants are good tools for research in these areas, providing a direct contact with some industrial equipaments. Given these capabilities, the main goal of this work is to model and control a didactic plant, which is a level and flow process control system with an industrial instrumentation. With a model it is possible to build a simulator for the plant that allows studies about its behaviour, without any of the real processes operational costs, like experiments with controllers. They can be tested several times before its application in a real process. Among the several types of controllers, it was used adaptive controllers, mainly the Direct Self-Tuning Regulators (DSTR) with Integral Action and the Gain Scheduling (GS). The DSTR was based on Pole-Placement design and use the Recursive Least Square to calculate the controller parameters. The characteristics of an adaptive system was very worth to guarantee a good performance when the controller was applied to the plant

Metodologia de análise e diagnóstico de fratura em um componente automotivo sujeito a cargas cíclicas alternadas

The competitiveness of the trade generated by the higher availability of products with lower quality and cost promoted a new reality of industrial production with small clearances. Track deviations at the production are not discarded, uncertainties can statistically occur. The world consumer and the Brazilian one are supported by the consumer protection code, in lawsuits against the products poor quality. An automobile is composed of various systems and thousands of constituent parts, increasing the likelihood of failure. The dynamic and security systems are critical in relation to the consequences of possible failures. The investigation of the failure gives us the possibility of learning and contributing to various improvements. Our main purpose in this work is to develop a systematic, specific methodology by investigating the root cause of the flaw occurred on an axle end of the front suspension of an automobile, and to perform comparative data analyses between the fractured part and the project information. Our research was based on a flaw generated in an automotive suspension system involved in a mechanical judicial cause, resulting in property and personal damages. In the investigations concerning the analysis of mechanical flaws, knowledge on materials engineering plays a crucial role in the process, since it enables applying techniques for characterizing materials, relating the technical attributes required from a respective part with its structure of manufacturing material, thus providing a greater scientific contribution to the work. The specific methodology developed follows its own flowchart. In the early phase, the data in the records and information on the involved ones were collected. The following laboratory analyses were performed: macrography of the fracture, micrography with SEM (Scanning Electron Microscope) of the initial and final fracture, phase analysis with optical microscopy, Brinell hardness and Vickers microhardness analyses, quantitative and qualitative chemical analysis, by using X-ray fluorescence and optical spectroscopy for carbon analysis, qualitative study on the state of tension was done. Field data were also collected. In the analyses data of the values resulting from the fractured stock parts and the design values were compared. After the investigation, one concluded that: the developed methodology systematized the investigation and enabled crossing data, thus minimizing diagnostic error probability, the morphology of the fracture indicates failure by the fatigue mechanism in a geometrically propitious location, a tension hub, the part was subjected to low tensions by the sectional area of the final fracture, the manufacturing material of the fractured part has low ductility, the component fractured in an earlier moment than the one recommended by the manufacturer, the percentages of C, Si, Mn and Cr of the fractured part present values which differ from the design ones, the hardness value of the superior limit of the fractured part is higher than that of the design, and there is no manufacturing uniformity between stock and fractured part. The work will contribute to optimizing the guidance of the actions in a mechanical engineering judicial expertise

Computabilidade no espaço dos intervalos reais: um modelo BSS intervalar

A matemática intervalar é uma teoria matemática originada na década de 60 com o objetivo de responder questões de exatidão e eficiência que surgem na prática da computação científica e na resolução de problemas numéricos. As abordagens clássicas para teoria da computabilidade tratam com problemas discretos (por exemplo, sobre os números naturais, números inteiros, strings sobre um alfabeto finito, grafos, etc.). No entanto, campos da matemática pura e aplicada tratam com problemas envolvendo números reais e números complexos. Isto acontece, por exemplo, em análise numérica, sistemas dinâmicos, geometria computacional e teoria da otimização. Assim, uma abordagem computacional para problemas contínuos é desejável, ou ainda necessária, para tratar formalmente com computações analógicas e computações científicas em geral. Na literatura existem diferentes abordagens para a computabilidade nos números reais, mas, uma importante diferença entre estas abordagens está na maneira como é representado o número real. Existem basicamente duas linhas de estudo da computabilidade no contínuo. Na primeira delas uma aproximação da saída com precisão arbitrária é computada a partir de uma aproximação razoável da entrada [Bra95]. A outra linha de pesquisa para computabilidade real foi desenvolvida por Blum, Shub e Smale [BSS89]. Nesta aproximação, as chamadas máquinas BSS, um número real é visto como uma entidade acabada e as funções computáveis são geradas a partir de uma classe de funções básicas (numa maneira similar às funções parciais recursivas). Nesta dissertação estudaremos o modelo BSS, usado para se caracterizar uma teoria da computabilidade sobre os números reais e estenderemos este para se modelar a computabilidade no espaço dos intervalos reais. Assim, aqui veremos uma aproximação para computabilidade intervalar epistemologicamente diferente da estudada por Bedregal e Acióly [Bed96, BA97a, BA97b], na qual um intervalo real é visto como o limite de intervalos racionais, e a computabilidade de uma função intervalar real depende da computabilidade de uma função sobre os intervalos racionais

Processo de difusão com agregação e reorganização espontânea em uma rede 2D

Difusive processes are extremely common in Nature. Many complex systems, such as microbial colonies, colloidal aggregates, difusion of fluids, and migration of populations, involve a large number of similar units that form fractal structures. A new model of difusive agregation was proposed recently by Filoche and Sapoval [68]. Based on their work, we develop a model called Difusion with Aggregation and Spontaneous Reorganization . This model consists of a set of particles with excluded volume interactions, which perform random walks on a square lattice. Initially, the lattice is occupied with a density p = N/L2 of particles occupying distinct, randomly chosen positions. One of the particles is selected at random as the active particle. This particle executes a random walk until it visits a site occupied by another particle, j. When this happens, the active particle is rejected back to its previous position (neighboring particle j), and a new active particle is selected at random from the set of N particles. Following an initial transient, the system attains a stationary regime. In this work we study the stationary regime, focusing on scaling properties of the particle distribution, as characterized by the pair correlation function ø(r). The latter is calculated by averaging over a long sequence of configurations generated in the stationary regime, using systems of size 50, 75, 100, 150, . . . , 700. The pair correlation function exhibits distinct behaviors in three diferent density ranges, which we term subcritical, critical, and supercritical. We show that in the subcritical regime, the particle distribution is characterized by a fractal dimension. We also analyze the decay of temporal correlations

Método híbrido para detecção e diagnóstico de falhas baseado em resíduos

The detection and diagnosis of faults, ie., find out how , where and why failures occur is an important area of study since man came to be replaced by machines. However, no technique studied to date can solve definitively the problem. Differences in dynamic systems, whether linear, nonlinear, variant or invariant in time, with physical or analytical redundancy, hamper research in order to obtain a unique solution . In this paper, a technique for fault detection and diagnosis (FDD) will be presented in dynamic systems using state observers in conjunction with other tools in order to create a hybrid FDD. A modified state observer is used to create a residue that allows also the detection and diagnosis of faults. A bank of faults signatures will be created using statistical tools and finally an approach using mean squared error ( MSE ) will assist in the study of the behavior of fault diagnosis even in the presence of noise . This methodology is then applied to an educational plant with coupled tanks and other with industrial instrumentation to validate the system.

Utilizando Pesos estáticos e dinâmicos em sistemas multi-classificadores com diferentes níveis de diversidade

Although some individual techniques of supervised Machine Learning (ML), also known as classifiers, or algorithms of classification, to supply solutions that, most of the time, are considered efficient, have experimental results gotten with the use of large sets of pattern and/or that they have a expressive amount of irrelevant data or incomplete characteristic, that show a decrease in the efficiency of the precision of these techniques. In other words, such techniques can t do an recognition of patterns of an efficient form in complex problems. With the intention to get better performance and efficiency of these ML techniques, were thought about the idea to using some types of LM algorithms work jointly, thus origin to the term Multi-Classifier System (MCS). The MCS s presents, as component, different of LM algorithms, called of base classifiers, and realized a combination of results gotten for these algorithms to reach the final result. So that the MCS has a better performance that the base classifiers, the results gotten for each base classifier must present an certain diversity, in other words, a difference between the results gotten for each classifier that compose the system. It can be said that it does not make signification to have MCS s whose base classifiers have identical answers to the sames patterns. Although the MCS s present better results that the individually systems, has always the search to improve the results gotten for this type of system. Aim at this improvement and a better consistency in the results, as well as a larger diversity of the classifiers of a MCS, comes being recently searched methodologies that present as characteristic the use of weights, or confidence values. These weights can describe the importance that certain classifier supplied when associating with each pattern to a determined class. These weights still are used, in associate with the exits of the classifiers, during the process of recognition (use) of the MCS s. Exist different ways of calculating these weights and can be divided in two categories: the static weights and the dynamic weights. The first category of weights is characterizes for not having the modification of its values during the classification process, different it occurs with the second category, where the values suffers modifications during the classification process. In this work an analysis will be made to verify if the use of the weights, statics as much as dynamics, they can increase the perfomance of the MCS s in comparison with the individually systems. Moreover, will be made an analysis in the diversity gotten for the MCS s, for this mode verify if it has some relation between the use of the weights in the MCS s with different levels of diversity

Aspectos dinâmicos da água versus álcool e acetona por técnicas de congelamento

Water still represents, on its critical properties and phase transitions, a problem of current scientific interest, as a consequence of the countless open questions and of the inadequacy of the existent theoretical models, mainly related to the different solid and liquid phases that this substance possesses. For example, there are 13 known crystalline forms of water, and also amorphous phases. One of them, the amorphous ice of very high density (VHDA), was just recently observed. Other example is the anomalous behavior in the macroscopic density, which presents a maximum at the temperature of 277 K. In order to experimentally investigate the behavior of one of the liquid-solid phase transitions, the anomaly in its density and also the metastability, we used three different cooling techniques and, as comparison systems, we made use of the solvents: acetone and ethyl alcohol. The first studied cooling system employ a Peltier plate, a device recently developed, which makes use of small cubes made up of semiconductors to change heat among two surfaces; the second system is a commercial refrigerator, similar to the residential ones. Finally, the liquid nitrogen technique, which is used to refrigerate the samples in a container, in two ways: a very fast and other one, almost static. In those three systems, three Beckers of aluminum were used (with a volume of 80 ml, each), containing water, alcohol and acetone. They were closed and maintained at atmospheric pressure. Inside of each Becker were installed three thermocouples, disposed along the vertical axis of the Beckers, one close to the inferior surface, other to the medium level and the last one close the superior surface. A system of data acquisition was built via virtual instrumentation using as a central equipment a Data-Acquisition board. The temperature data were collected by the three thermocouples in the three Beckers, simultaneously, in function of freezing time. We will present the behavior of temperature versus freezing time for the three substances. The results show the characterization of the transitions of the liquid

Algoritmo modificado de PSO matricial aplicado a identificação de sistemas com análise de convergência

Recentemente diversas técnicas de computação evolucionárias têm sido utilizadas em áreas como estimação de parâmetros de processos dinâmicos lineares e não lineares ou até sujeitos a incertezas. Isso motiva a utilização de algoritmos como o otimizador por nuvem de partículas (PSO) nas referidas áreas do conhecimento. Porém, pouco se sabe sobre a convergência desse algoritmo e, principalmente, as análises e estudos realizados têm se concentrado em resultados experimentais. Por isso, é objetivo deste trabalho propor uma nova estrutura para o PSO que permita analisar melhor a convergência do algoritmo de forma analítica. Para isso, o PSO é reestruturado para assumir uma forma matricial e reformulado como um sistema linear por partes. As partes serão analisadas de forma separada e será proposta a inserção de um fator de esquecimento que garante que a parte mais significativa deste sistema possua autovalores dentro do círculo de raio unitário. Também será realizada a análise da convergência do algoritmo como um todo, utilizando um critério de convergência quase certa, aplicável a sistemas chaveados. Na sequência, serão realizados testes experimentais de maneira a verificar o comportamento dos autovalores após a inserção do fator de esquecimento. Posteriormente, os algoritmos de identificação de parâmetros tradicionais serão combinados com o PSO matricial, de maneira a tornar os resultados da identificação tão bons ou melhores que a identificação apenas com o PSO ou, apenas com os algoritmos tradicionais. Os resultados mostram a convergência das partículas em uma região delimitada e que as funções obtidas após a combinação do algoritmo PSO matricial com os algoritmos convencionais, apresentam maior generalização para o sistema apresentado. As conclusões a que se chega é que a hibridização, apesar de limitar a busca por uma partícula mais apta do PSO, permite um desempenho mínimo para o algoritmo e ainda possibilita melhorar o resultado obtido com os algoritmos tradicionais, permitindo a representação do sistema aproximado em quantidades maiores de frequências.