Diagnóstico de falhas em circuitos analógicos utilizando inteligência de enxame

• Autor(a) principal: Galindo, Jalber Dinelli Luna
• Data de Publicação: 2022
• Outros Autores: [e-mail]
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/19360
• topic: Electronic engineering; Computer system failures; Expert systems (Computer); Linear integrated circuits; Collective intelligence; Engenharia eletrônica; Falhas de sistemas de computação; Sistemas especialistas (Computação); Circuitos integrados lineares; Inteligência coletiva; ENGENHARIAS::ENGENHARIA ELETRICA::ELETRONICA INDUSTRIAL, SISTEMAS E CONTROLES ELETRONICOS

Open-circuit or short-circuit faults, as well as in discrete parameters are the most used models in simulation methods before testing. Since the response of an analog circuit to an input signal is continuous, failures in any particular circuit component may not characterize all possible component failures. There are three important features in failure diagnostic of analog circuits: identifying faulty components, determining faulty element values, and circuit tolerance constraints. To solve this problem, two methodologies are proposed and implemented, which are based on optimization using swarm intelligence, for the diagnosis of failures: Particle Swarm Optimization (PSO); and bat behavior inspired optimization, termed Bat Algorithm (BA). The non-linear equations of the circuit under test are used to calculate its parameters. The objective is to identify the circuit componente that has the potential to present a failure by comparing the responses measured the real circuit and that obtained by the optimization process. Two circuits are used as case studies to evaluate the performance of the proposed implementations: the Tow-Thomas Biquad filter (circuit 1) and the ButterWorth filter (circuit 2). The proposed methodologies are able to identify or, at least, reduce the number of possibly failing components. The four main performance metrics used are extracted: accuracy, precision, sensitivity and specificity. The BA technique offers a better performance, using the combination of the maximum accessible nodes of the circuit under test, with the considered metric values 95.84%, 81.45%, 82.16% and 97.66%, respectively for circuit 1. For circuit 2, the obtained metric values are 95.13%, 74.87%, 73.30% and 97.42%, respectively. The BA technique is more efficient regarding the execution time. For circuit 1, there was an average reduction of 7.95% of the time when compared to the average time of the PSO for the circuit without failures, and of 8.12% for the cases with failure. For circuit 2, there was an average reduction of 12.2% of the time when compared to the average time of the PSO for the circuit without failures, and of 11.2% for the cases with failure.