You are in the accessibility menu

Please use this identifier to cite or link to this item: http://acervodigital.unesp.br/handle/11449/94540
Title: 
Redes neurais: aplicação no monitoramento da vida de ferramentas de corte
Author(s): 
Alexandre, Rogério Pinto
Institution: 
Universidade Estadual Paulista (UNESP)
Sponsorship: 
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
Abstract: 
  • Este trabalho discute a utilização de Redes Neurais Artificiais na avaliação da evolução do desgaste da ferramenta de corte no processo de torneamento a partir dos níveis de vibração do sistema porta-ferramenta - ferramenta - peça. Os níveis de desgaste foram definidos em termos da rugosidade superficial da peça e desgaste de flanco da ferramenta e posteriormente, os valores de vibração medidos no porta ferramenta, foram correlacionados com esses níveis de desgaste. Foram realizados diversos ensaios utilizando o aço ABNT 1045, com ferramentas de metal duro sem cobertura e com cobertura de nitreto de titânio. Os testes foram efetuados utilizando as rotações 630, 800, 1000 e 1250 rpm, variando-se a velocidade de corte entre 100 e 200 m/min. Os sinais de vibração foram processados e analisados utilizando valores RMS (Root Mean Square) e a Transformada Wavelet, sendo que neste caso, foram extraídos os valores RMS dos coeficientes wavelet. Os dados obtidos foram utilizados nas fases de treinamento e validação das redes neurais empregadas, utilizando o algoritmo Backpropagation. Os dados foram agrupados em quatro estágios que classificam o nível de desgaste da ferramenta em estágio inicial, estágio intermediário, estágio avançado e estágio crítico de desgaste. Os testes mostraram que a utilização dos valores RMS do sinal de vibração pode levar a resultados satisfatórios, entretanto, houve algumas situações de insucesso. Alternativamente, a rede neural quando treinada com os valores RMS dos coeficientes wavelet apresentou uma melhor capacidade de identificação, com um percentual de acerto maior do que quando treinada apenas com os valores RMS dos sinais de vibração. Os resultados mostram que o monitoramento da vibração da ferramenta e a utilização de redes neurais artificiais para identificação dos diferentes... .
  • This work discusses the use of Artificial Neural Networks (ANN) for the evaluation of tool wear in turning operations by using the vibration of the tool holder - tool - workpiece system. The wear levels were defined in terms of the surface roughness and flank wear of the cutting tool and later on, the vibration measured in the tool holder were correlated with those wear levels. Several experiments were carried out at different cutting conditions using ABNT 1045 steel as the workpiece material, the surface roughness were measured for carbide cutting tools without coating and with coating of titanium nitride. The tests were made using the rotations 630, 800, 1000 and 1250 rpm, being varied the cutting speed between 100 and 200 m/min. The vibration signals were processed and analyzed using RMS (Root Mean Square) values and the Wavelet Transform, in the later case; the RMS values were extracted from the wavelet coefficients. The obtained data were used in the training phases and validation of the neural network, using the Backpropagation algorithm. The data were settled in four groups that classify the level of the tool wear, initial, intermediary, advanced and critical. The experiments showed that the use of RMS values of the vibration signals can bring to satisfactory results, however, the neural network trained with the RMS values from the wavelet coefficients presented a better identification capacity, with a larger percentile of success, than that when just RMS values of the vibration signals were taken. The results show that the monitoring of the vibration of the tool holder-tool-workpiece system and the use of artificial neural network for identification of the different state of the tool, they can be used to settle the end of the tool life in the turning process. It was also evidenced the potentiality of Transformed Wavelet to be used... (Complete abstract click electronic address below).
Issue Date: 
28-Jan-2005
Citation: 
ALEXANDRE, Rogério Pinto. Redes neurais: aplicação no monitoramento da vida de ferramentas de corte. 2005. 105 f. Dissertação (mestrado) - Universidade Estadual Paulista, Faculdade de Engenharia de Ilha Solteira, 2005.
Time Duration: 
105 f. : il.
Publisher: 
Universidade Estadual Paulista (UNESP)
Keywords: 
  • Redes neurais (Computação)
  • Wavelets (Matematica)
  • Usinagem
  • Ferramentas para cortar metais - Deterioração
  • Tornearia
  • Desgaste
  • Neural network
  • Wavelet
URI: 
Access Rights: 
Acesso aberto
Type: 
outro
Source:
http://repositorio.unesp.br/handle/11449/94540
Appears in Collections:Artigos, TCCs, Teses e Dissertações da Unesp

There are no files associated with this item.
 

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