A Data-Driven Approach for Mass-Spring Model Parametrization Based on Continuous Models

• Ano de defesa: 2015
• Autor(a) principal: Silva, Josildo Pereira da
• Orientador(a): Apolinário Júnior, Antonio Lopes
• Banca de defesa: Apolinário Júnior, Antonio Lopes, Giraldi, Gilson Antonio, Souza Neto, Eduardo Alberto de, Medeiros Filho, Esdras Soares de, Silva, Rodrigo Luis de Souza da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Instituto de Matemática
• Programa de Pós-Graduação: Programa Multiinstitucional de Pós-graduação em Ciência da Computação, UFBA-UNIFACS-UEFS
• País: brasil
• Palavras-chave em Português: Modelos Deformáveis; Modelos Baseados em Física; Parametrização de Modelos Massa Mola; Modelos de Elementos Finitos
• Área do conhecimento CNPq: Sistemas Computacionais
• Link de acesso: http://repositorio.ufba.br/ri/handle/ri/22848

Nowadays, the behavior simulation of deformable objects plays important roles in several fields such as computer graphics, computer aided design, computer aided surgery And robotics. The two main categories of deformable models are: based on continuum mechanics, like Finite Element Model (FEM) or Isogeometric Analysis (IGA); and using discrete representations, as a Mass - Spring Model (MSM). FEM methods are known for their high computational cost and precision, while MSM methods, although simple and affordable for real-time applications, are di cult to parameterize. There is no general physically based or systematic method in the literature to determine the mesh topology or MSM parameters from a known material. Therefore, in this thesis, we proposea methodology to parametrize the MSM based on continuous models with focus on the simulation of deformable objects in real-time for application in virtual environments. We developed two data-driven approaches to the parametrization of the MSM by using FEM and IGA models as reference of derivation with higher order elements. Based on experimental results, the precision achieved by these new methodologies is higher than other approaches in literature. In particular, our proposal achieves excellent results in the parametrization of the MSM with higher order elements which does not occur with other methodologies