abstract
Organic corrosion inhibitors are playing a crucial role to substitute traditional protective technologies, which have acute toxicity problems associated. However, why some organic compounds inhibit corrosion and others do not is still not well understood. Therefore, we tested different machine learning (ML) methods to distinguish efficient corrosion inhibitors for aluminum alloys commonly used in aeronautical applications. In this work, we have obtained information that can greatly contribute to automate the search for new and more efficient protective solutions in the future: (i) a ML algorithm was selected that is able to classify correctly efficient inhibitors (i.e., with more than 50% efficiency) and non-inhibitors (i.e., with lower or equal to 50% efficiency), even when information about different alloys at different pHs is included in the same data set, which can significantly increase the information available to train the model; (ii) new descriptors related to the self-association of the molecules were evaluated, but improvements to the predictive power of the models are limited; (iii) average differences concerning the descriptors in this work were identified for inhibitors and non-inhibitors, having the potential to serve as guidelines to select potentially inhibitive molecular systems. This work demonstrates that ML can significantly accelerate research in the field by serving as a tool to perform an initial virtual screen of the molecules.
keywords
MODELING CORROSION; DESCRIPTORS
subject category
Chemistry; Science & Technology - Other Topics; Materials Science
authors
Galvao, TLP; Novell-Leruth, G; Kuznetsova, A; Tedim, J; Gomes, JRB
our authors
João Tedim
Associate Professor
José R. B. Gomes
Principal Researcher
Tiago Luís Pereira Galvão
Junior ResearcherGroups
G3 - Electrochemical Materials, Interfaces and Coatings
G6 - Virtual Materials and Artificial Intelligence
Projects
Dados Inteligentes para Desenhar Inibidores de Corrosão (DataCor)
Unveiling the self-healing mechanisms associated with smart nanocontainers (SELMA)
acknowledgements
This work was developed within the scope of the project CICECO-Aveiro Institute of Materials, UIDB/50011/2020 and UIDP/50011/2020, financed by national funds through the FCT/MEC and when appropriate cofinanced by FEDER under the PT2020 Partnership Agreement. It was also financed in the framework of the project DataCor (refs POCI-01-0145-FEDER-030256 and PTDC/QUI-QFI/30256/2017) and SELMA (PTDC/QEQ-QFI/4719/2014), Project 3599 Promover a Producao Cientifica e Desenvolvimento Tecnologico e a Constituicao de Redes Tematicas (3599-PPCDT) and FEDER funds through COMPETE 2020, Programa Operacional Competitividade e Internacionalizacao (POCI).