Deteção de Falha de Ignição

• Autor(a) principal: Santos, Francisco José Maravilha
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/18279
• topic: Misfire; Machine Learning; Deep Learning; MATLAB; Falha de Ignição

This project analyzes some models of Machine Learning and Deep Learning in the context of misfire detection. Initially, the themes of misfire and Machine Learning are contextualized. The main objective of this project is to replace the existing ignition failure detection algorithm with a Machine Learning model. For this, the process was divided into several phases: problem formulation, data exploration, preparation and pre-processing of the data, building the model, and exporting it. For this, two versions of Matlab were used, 2020a and 2016b. The 2020a version was used to carry out all steps up to the export of the model. The 2016b version was used to perform the comparison with the detection algorithm already developed. Furthermore, dSpace TargetLink was used to generate the C code. The construction of several models allows, through different metrics such as accuracy, precision, recall, and F1 score, to analyze and compare them and determine which model is the best. With the completion of this project, we learned about the ignition failure event, but mainly about Machine Learning. All the necessary steps were learned, both in terms of data preparation and programming for the construction of the model and calculation of the respective metrics to evaluate the models. With this type of work, it was highlighted that Machine Learning is an iterative process, it can be present in the most diverse areas and with many different purposes. Machine Learning is already present in many industries and applications of our daily lives, but it is estimated that in the future its presence will be almost the majority.