Pattern recognition approaches for diagnosis of cutting tool wear condition uri icon

Abstract

  • The cutting tool condition is an important factor in all metal cutting process. However, direct monitoring systems are not easily implemented because they require ingenious measuring methods. This paper proposes an indirect monitoring approach based on vibration measurements. Vibration signals were characterized by Mel Frequency Cepstrum Coefficients and associated with the cutting tool condition. Afterwards, a diagnosis system was implemented where three approaches are compared: Hidden Markov Models, Artificial Neural Networks, and Learning Vector Quantization. Early results show the feasibility of the proposal.

Publication date

  • August 22, 2007