Fast fourier transformation of emitted noises from welding machines and their classification with acoustic method

Gokmen G., Akgun O., Akıncı T. Ç., Şeker Ş. S.

MECHANIKA, vol.23, no.4, pp.588-593, 2017 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 23 Issue: 4
  • Publication Date: 2017
  • Doi Number: 10.5755/j01.mech.23.4.14876
  • Journal Name: MECHANIKA
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.588-593
  • Istanbul Technical University Affiliated: Yes


In this study, a method that determines the welding machine types using acoustic method and Fast Fourier Transformation (FFT) and Artificial Neural Networks (ANN) has been suggested. FFT was used in order to bring out the characteristics of welding machines and ANN to classify them. To this end, the sounds of three arc, gas metal arc and spot weld machines were transferred to a computer during welding process via a microphone and recorded separately and then, by applying FFT, discrete frequency components were ascertained. The selected 500 frequency components were normalized and used as an input of an ANN model. It was observed that ANN model could classify welding machine types following training, validation and test stages, through the recorded sounds with a great success.