logo-ri

Repositório Institucional da Produção Científica da Marinha do Brasil (RI-MB)

Please use this identifier to cite or link to this item: https://www.repositorio.mar.mil.br/handle/ripcmb/845686
Title: Exploring the use of machine learning for improving the efficiency of coating performance evaluation.
Authors: Farias, João
metadata.dc.contributor.advisor: Dr Michele Curioni
Keywords: Corrosao
Espectroscopia de impedancia eletroquimica
aprendizado de maquinas
DGPM knowledge areas: Engenharia Naval
Issue Date: 2022
Publisher: University of Manchester
Description: The use of machine learning was explored in the context of electrochemical impedance spectroscopy (EIS), with the purpose of overcoming some of its inherent complexities and increasing the efficiency in its use for coating performance evaluation . For this project, EIS and visual inspection data from marine coatings exposed to accelerated corrosion tests for approximately 2.5 years were applied to machine learning techniques to acquire a prototyped classification-type algorithm. Also, electrochemical tests–including EIS and alternative methods–were applied to coated samples immersed in 5 wt.% NaClto generate datafor the training, testing and validation of fitting-type machine learning algorithms, “prepared as a proof of concept”. An experimental setup using automatablecomponentswas adopted, which surpassed the necessity of preparing and performing each electrochemical test one-by-one.Overall, the results have shown that machine learning approaches have the potential to overcome several complexities related to EIS, and this combination of knowledges should be further exploited.
Abstract: The use of machine learning was explored in the context of electrochemical impedance spectroscopy (EIS), with the purpose of overcoming some of its inherent complexities and increasing the efficiency in its use for coating performance evaluation . For this project, EIS and visual inspection data from marine coatings exposed to accelerated corrosion tests for approximately 2.5 years were applied to machine learning techniques to acquire a prototyped classification-type algorithm. Also, electrochemical tests–including EIS and alternative methods–were applied to coated samples immersed in 5 wt.% NaClto generate datafor the training, testing and validation of fitting-type machine learning algorithms, “prepared as a proof of concept”. An experimental setup using automatablecomponentswas adopted, which surpassed the necessity of preparing and performing each electrochemical test one-by-one.Overall, the results have shown that machine learning approaches have the potential to overcome several complexities related to EIS, and this combination of knowledges should be further exploited.
Access: Open access
URI: https://www.repositorio.mar.mil.br/handle/ripcmb/845686
Type: Master thesis
Appears in Collections:Engenharia Naval: Coleção de Dissertações



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.