/img alt="Imagem da capa" class="recordcover" src="""/>
Trabalho de Conclusão de Curso - Graduação
Análise de dados de manutenção usando aprendizado de máquina: estudo de caso em uma peneira vibratória do segmento de mineração
In search of a recurring problem in a large mining company, that could be studied and approached, research and analysis were carried out on supervisory and data acquisition systems refered to equipment that are inserted in the ore beneficiation process, and so a Screen was selected for this study, t...
Autor principal: | LACERDA, Rickelle Moraes |
---|---|
Grau: | Trabalho de Conclusão de Curso - Graduação |
Idioma: | por |
Publicado em: |
2022
|
Assuntos: | |
Acesso em linha: |
https://bdm.ufpa.br:8443/jspui/handle/prefix/3941 |
Resumo: |
---|
In search of a recurring problem in a large mining company, that could be studied and approached, research and analysis were carried out on supervisory and data acquisition systems refered to equipment that are inserted in the ore beneficiation process, and so a Screen was selected for this study, that presented 109 failure events in the departments of Electrical, Instrumentation and Automation throughout 2019. From now on, quantitative and qualitative analyzes, investigations and field research were made to seek justifications for the high number of events that generated maintenance. A Losses Profile, which consists of stratifying the losses of the production process by using Pareto charts, in order to identify which are the greatest gain opportunities in the equipment was elaborated. As well as the use of the FMEA tool (Failure Mode and Effect Analysis) that pointed out which failures are most critical and which actions with a certain priority must occur in order to achieve greater Reliability of the asset. In this context, it was understood the importance to, trough Artificial Intelligence and Machine Learning, create a code in MATLAB, using the techniques: PCA and Clustering. Which was able to, from samples of data regarding the functioning of the Screen, to analyze the behavior of these data and find out what would be the ones that present anomalies, fitting in with defective data class groups, and those that do not have anomalies in non defective data class groups. So that it is possible to diagnose the equipment and in the future, improving the code, even predict faults in this equipment and its similars, causing losses in the process to be mitigated. |