| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 23.13 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Um engenheiro eletrotécnico, principalmente ao longo do seu primeiro ano académico, é introduzido à análise de circuitos elétricos. Esta análise é iniciada com a apresentação aos componentes elétricos e eletrónicos e aos seus respetivos símbolos. De seguida, após este primeiro contacto, o aluno começa a associar alguns sím
bolos e a formar os seus primeiros circuitos elétricos. Tradicionalmente, o aluno começa por desenhar o circuito em papel. Após isto, é feita a primeira análise e os cálculos necessários. A simulação do circuito é, por vezes, uma forma complementar de validar e analisar outros parâmetros, como a tensão e a corrente. A simulação é um passo importante na aprendizagem, pois possibilita a validação dos cálculos realizados em papel e flexibilidade a nível de projeto e pré-validação de circuitos elétricos e eletrónicos. No estudo, e em forma de treino, os alunos analisam vários circuitos que, por vezes, não apresentam qualquer solução que permita validar os resultados. Por norma, são então introduzidas ferramentas de simulação de modo a obter a validação. Esta é uma tarefa árdua e que consome bastante tempo de estudo dos alunos. De modo a ultrapassar esta dificuldade, foi criada a plataforma U=RIsolve [1] que, num dos seus futuros modos de operação, permitirá ao aluno economizar o seu tempo na tarefa de transferir o circuito para o simulador. Para isto é utilizada uma fotografia ou digitalização do esquema do circuito para introduzir do circuito no
simulador. Primeiramente, a imagem é submetida num algoritmo de visão computacional. Este é capaz de interpretar a topologia de um esquema elétrico e construir o seu modelo a partir da análise dessa imagem. Este módulo de software, desenvolvido pelo colega Hugo Barbosa [2] no âmbito do seu projeto de tese, permite a deteção e extração dos elementos presentes no circuito. Estes são, por exemplo, componentes,
os nós, as interligações entre elementos, os valores e os seus identificadores. Assim sendo, os outputs deste módulo são um conjunto de ficheiros com a informação extraída da imagem. Nos excertos da imagem original, foram introduzidas pequenas caixas delimitadoras à volta dos elementos dos circuitos. Deste modo, estes excertos serão utilizados por outro módulo de software capaz de os classificar. Consequentemente, este projeto/tese consistiu em criar um algoritmo de inteligência artificial que utiliza os outputs do módulo anterior e classifique cada elemento do circuito. Para isto, serão utilizadas técnicas de Machine Learning, como Deep Learning, e redes neuronais. No entanto, devido à pouca oferta de conjuntos de imagem, ou seja, datasets, que preenchessem todos os requisitos propostos, foi necessária a recolha de amostras de símbolos para treinar o modelo de Machine Learning. Em suma, a finalidade deste modulo de software é o reconhecimento dos caracteres alfanuméricos contidos nos identificadores de cada componente do circuito. Deste modo,
comtodas estas partes interligadas integradas na framework U=RIsolve, o aluno conseguirá validar os seus resultados utilizando esta ferramenta de auto-aprendizagem na análise de circuitos elétricos(segundo diferentes métodos).
Anelectrical engineering student, especially during their first academic year, is introduced to the analysis of electrical circuits. This analysis begins with the introduction to electrical and electronic components and their respective symbols. Next, after this initial contact, the student starts to associate some symbols and form their first electrical circuits. Traditionally, the student begins by drawing the circuit on paper. After this, the first analysis and necessary calculations are made. Circuit simulation is sometimes used as a complementary way to validate and analyze other parameters, such as voltage and current. Simulation is an important step in learning, as it allows for the validation of calculations done on paper and offers flexibility in project design and pre-validation of electrical and electronic circuits. In the study, and as a form of practice, students analyze various circuits that sometimes do not provide any solution to validate the results. Normally, simulation tools are then introduced to obtain validation. This is a tedious task that consumes a lot of the students’ study time. To overcome this difficulty, the platform U=RIsolve [1] was created, which, in one of its future modes of operation, will allow the student to save time on the task of transferring the circuit to the simulator. For this, a photograph or scan of the circuit diagram is used to input the circuit into the simulator. First, the image is submitted to a computer vision algorithm. This algorithm can interpret the topology of an electrical diagram and build its model from the analysis of that image. This software module, developed by colleague Hugo Barbosa [2] as part of his thesis project, allows the detection and extraction of elements present in the circuit. These elements are, for example, components, nodes, interconnections between elements, values, and their identifiers. Thus, the outputs of this module are a set of files with the information extracted from the image. In the excerpts of the original image, small bounding boxes were introduced around the circuit elements. In this way, these excerpts will be used by another software module capable of classifying them. Consequently, this project/thesis consisted of creating an artificial intelligence algorithm that uses the outputs of the previous module and classifies each element of the circuit. For this, Machine Learning techniques, such as Deep Learning, and neural networks will be used. However, due to the lack of image sets, or datasets, that met all the proposed requirements, it was necessary to collect samples of symbols to train the Machine Learning model. In summary, the purpose of this software module is to recognize the alphanumeric characters contained in the identifiers of each circuit component. Thus, with all these interconnected parts integrated into the U=RIsolve framework, the student will be able to validate their results using this self-learning tool in circuit analysis (according to different methods).
Anelectrical engineering student, especially during their first academic year, is introduced to the analysis of electrical circuits. This analysis begins with the introduction to electrical and electronic components and their respective symbols. Next, after this initial contact, the student starts to associate some symbols and form their first electrical circuits. Traditionally, the student begins by drawing the circuit on paper. After this, the first analysis and necessary calculations are made. Circuit simulation is sometimes used as a complementary way to validate and analyze other parameters, such as voltage and current. Simulation is an important step in learning, as it allows for the validation of calculations done on paper and offers flexibility in project design and pre-validation of electrical and electronic circuits. In the study, and as a form of practice, students analyze various circuits that sometimes do not provide any solution to validate the results. Normally, simulation tools are then introduced to obtain validation. This is a tedious task that consumes a lot of the students’ study time. To overcome this difficulty, the platform U=RIsolve [1] was created, which, in one of its future modes of operation, will allow the student to save time on the task of transferring the circuit to the simulator. For this, a photograph or scan of the circuit diagram is used to input the circuit into the simulator. First, the image is submitted to a computer vision algorithm. This algorithm can interpret the topology of an electrical diagram and build its model from the analysis of that image. This software module, developed by colleague Hugo Barbosa [2] as part of his thesis project, allows the detection and extraction of elements present in the circuit. These elements are, for example, components, nodes, interconnections between elements, values, and their identifiers. Thus, the outputs of this module are a set of files with the information extracted from the image. In the excerpts of the original image, small bounding boxes were introduced around the circuit elements. In this way, these excerpts will be used by another software module capable of classifying them. Consequently, this project/thesis consisted of creating an artificial intelligence algorithm that uses the outputs of the previous module and classifies each element of the circuit. For this, Machine Learning techniques, such as Deep Learning, and neural networks will be used. However, due to the lack of image sets, or datasets, that met all the proposed requirements, it was necessary to collect samples of symbols to train the Machine Learning model. In summary, the purpose of this software module is to recognize the alphanumeric characters contained in the identifiers of each circuit component. Thus, with all these interconnected parts integrated into the U=RIsolve framework, the student will be able to validate their results using this self-learning tool in circuit analysis (according to different methods).
Description
Keywords
Circuit analysis Image processing Machine learning Deep learning Neural networks e-learning Image classification Electrical diagrams U = RIsolve Análise circuitos Processamento imagem Redes neuronais Classificação imagens Esquemas elétricos