Utilização de machine learning e algoritmo genético no design de ligas de titânio para aplicações biomédicas

• Autor(a) principal: Nogueira, Luís Guilherme Santagnelo
• Data de Publicação: 2025
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://hdl.handle.net/20.500.14289/21740
• topic: Liga de titânio beta; Módulo elástico; Algoritmo genético; Machine learning; ENGENHARIAS::ENGENHARIA DE MATERIAIS E METALURGICA::METALURGIA FISICA; ENGENHARIAS::ENGENHARIA DE MATERIAIS E METALURGICA::METALURGIA FISICA::ESTRUTURA DOS METAIS E LIGAS

The production of metastable beta titanium alloys for biomedical applications has grown over the last decade, aiming to meet the demand for materials with an elastic modulus more compatible with bone tissue. This contributes positively to biocompatibility and improves mechanical performance. Alloys such as Ti–29Nb–13Ta–4.6Zr (TNZT), Ti–12Mo–6Zr–2Fe (TMZF), and Ti–35Nb–7Zr–5Ta (TiOsteum) are widely used for their properties, such as high specific mechanical strength and corrosion resistance. These alloys also incorporate beta-phase stabilizing and biocompatible elements, such as Zr, Ta, Nb, and Mo, avoiding elements toxic to humans, such as V, Co, Cr, and Cu. A fundamental challenge in using these alloys for orthopedic implants is reducing the difference between their elastic modulus and that of human bone, minimizing the phenomenon of stress shielding, which can cause bone fragility over time. This discrepancy is governed by the final microstructure formed during alloy processing and heat treatment, which includes parameters such as average grain size, alpha phase volume fraction, martensite volume fraction, and omega phase volume fraction, all of which affect mechanical properties. The integration of machine learning, especially through genetic algorithms, can optimize the development of these alloys. This methodology helps identify compositions that minimize the elastic modulus by adjusting parameters and identifying more efficient combinations. Thus, this study aims to apply machine learning combined with genetic algorithms to obtain chemical compositions of titanium alloys with predominantly beta-phase microstructures and the lowest possible elastic modulus. To this end, both experimental data described in the literature and mathematical models were used to optimize parameters such as E, Moeq (molybdenum equivalent), Bo (bond order), and Md (mean d orbital energy level). Using genetic algorithm codes, a list of chemical compositions with potentially desirable microstructures and properties was obtained.