Artificial intelligence based topology evaluation for kinematic interference assessment in gear skiving.
Data
2024-09-02
Autores
Machado, Lucas Martins 
Orientadores
Reis, Danieli Aparecida Pereira
Tipo
Trabalho de conclusão de curso
Título da Revista
ISSN da Revista
Título de Volume
Resumo
As engrenagens são componentes indispensáveis em várias aplicações industriais, inclusive automotivas, aeroespaciais e robóticas. Para atender à crescente demanda por vários tipos de engrenagens, os fabricantes estão recorrendo ao skiving de engrenagens, um processo de usinagem versátil, devido à sua capacidade de produzir engrenagens retas e helicoidais, externas e internas, tudo em uma única máquina. No entanto, a obtenção da qualidade desejada para as engrenagens é dificultada pela topografia complexa da superfície, caracterizada por perfis de dentes complexos e características microgeometricas. Essas características são essenciais para a otimização da densidade de potência, do padrão de contato e da vida útil à fadiga, que são fatores importantes, especialmente quando se lida com os crescentes requisitos de mobilidade elétrica, como velocidades de rotação e gradientes de torques mais ingrimes. No contexto da Quarta Revolução Industrial, caracterizada pela integração de tecnologias digitais e abordagens orientadas por dados, a Inteligência Artificial (IA) surge como uma solução transformadora. A avaliação da topografia baseada em IA aproveita os algoritmos de aprendizado de máquina para decifrar as características da superfície da engrenagem e reconhecer melhor os padrões de desvio cinemáticos. Este projeto investigou a integração de I.A no processo de mediçoes de fabricadas por skiving, atraves da criação de uma rede neural artificial que foi capaz de predizer com 80% precisão o eixo de desvio cinematico dada uma topografia de um dente com desvios cinemáticos. para criar a I.A 6000 simulações foram configuradas no SPARTApro, para criar os dados de treino, durantes as simulações mudanças foram feitas nos eixos X, Y, Z e suas rotações, da ferramenta e da engrenagem, tais foram feitas atraves do uso de arquivos free kinematic, as mudanças feitas por estes arquivos representam desvio cinematicos. Programas auxiliares foram desensenvolvidos para este trabalho em python para que seja possivel rodar continuamente as simulações e pre processar os dados, o trabalho também usou uma high performance computing cluster dada a quantidade e tempo de simulações, esse trabalho foi feito no laboratorio de maquinas ferramenta e engenharia de produção (WZL) da RWTH Aachen University (RWTH) em colaboração com o Laboratorio de Comportamento Mecanico de Metais (LCMM) da UNIFESP.
Gears are indispensable components in various industrial applications, including automotive, aerospace, and robotics. To meet the increasing demand for diverse types of gears, manufacturers are turning to gear skiving, a versatile machining process, given its ability to make straight and helical tooth, external and internal gears all in one machine. Nevertheless, achieving the desired gear quality is hindered by complex surface topography, characterized by intricate tooth profiles and microgeometric features. These features are essential for the optimization of power density, contact pattern and fatigue life, which are important factors especially when dealing with the rising electro mobility (e-Mobility) requirements such as higher speeds of rotations and steeper torques. In the context of the Fourth Industrial Revolution, characterized by the integration of digital technologies and data-driven approaches, Artificial Intelligence (AI) emerges as a transformative solution. AI-based topography evaluation leverages machine learning algorithms to decipher gear surface characteristics and better recognize kinematic deviation patterns. This project investigated the integration of AI into the gear skiving measurement processes, by creating an A.I, namely an artificial neural network that is capable of predicting with close to 80% accuracy the deviation axis given the topography of a gear tooth with a kinematic deviation. To create the A.I 6000 simulations were set up on SPARTApro to create the training data for the artificial neural network, during these simulations changes were made to the X,Y,Z axis coordinates and its rotations, of the tool and the workpiece and were achieved through the use of a free kinematic file, these changes represent possible kinematic deviations. Auxiliary programs were developed for this work in python so as to be able to do continuous simulation and data pre-process, it also used a high performance computing cluster given the amount of simulations and time to run, this work was done in the Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University (RWTH) in collaboration with the Laboratorio de Comportamento Mecanico de Metais (LCMM) da UNIFESP.
Gears are indispensable components in various industrial applications, including automotive, aerospace, and robotics. To meet the increasing demand for diverse types of gears, manufacturers are turning to gear skiving, a versatile machining process, given its ability to make straight and helical tooth, external and internal gears all in one machine. Nevertheless, achieving the desired gear quality is hindered by complex surface topography, characterized by intricate tooth profiles and microgeometric features. These features are essential for the optimization of power density, contact pattern and fatigue life, which are important factors especially when dealing with the rising electro mobility (e-Mobility) requirements such as higher speeds of rotations and steeper torques. In the context of the Fourth Industrial Revolution, characterized by the integration of digital technologies and data-driven approaches, Artificial Intelligence (AI) emerges as a transformative solution. AI-based topography evaluation leverages machine learning algorithms to decipher gear surface characteristics and better recognize kinematic deviation patterns. This project investigated the integration of AI into the gear skiving measurement processes, by creating an A.I, namely an artificial neural network that is capable of predicting with close to 80% accuracy the deviation axis given the topography of a gear tooth with a kinematic deviation. To create the A.I 6000 simulations were set up on SPARTApro to create the training data for the artificial neural network, during these simulations changes were made to the X,Y,Z axis coordinates and its rotations, of the tool and the workpiece and were achieved through the use of a free kinematic file, these changes represent possible kinematic deviations. Auxiliary programs were developed for this work in python so as to be able to do continuous simulation and data pre-process, it also used a high performance computing cluster given the amount of simulations and time to run, this work was done in the Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University (RWTH) in collaboration with the Laboratorio de Comportamento Mecanico de Metais (LCMM) da UNIFESP.
Descrição
Citação
MACHADO, Lucas M. Artificial intelligence based topology evaluation for kinematic interference assessment in gear skiving. 2024. 39 f. Trabalho de Conclusão de Curso (Graduação em Engenharia de Materiais) – Universidade Federal de São Paulo, São José dos Campos, 2024.