Laser arc sound signal processing and welding status recognition based on geometric learning

DOI

10.2991/icmse-15.2015.251How to use a DOI?

Keywords

wavelet threshold; Laser arc welding; multi-weight neural network; Feature extraction

Abstract

Laser arc sound signal hides welding status information. It is a great significance to control the B&K company and four channels input sound and vibration analyzer. Using wavelet base which is db4, combining different methods to decompose laser arc sound signal. These methods include hard threshold, soft threshold and double threshold double factors. The results show that choosing double threshold double factor has the highest SNR. After processing, 1024 consecutive arc sound signal sampling points are selected as a sample, and features are extracted in time domain and frequency domain. Thirty samples are respectively selected from three welding conditions including complete penetration, incomplete penetration and welding wear, which constitute training samples. Twenty samples are respectively selected from three welding conditions, which constitute test samples. test samples are respectively identified by probabilistic neural network(PNN) and multi-weights neural network(MWNN). Results show that the whole recognition rate of multi-weight neural network is higher than the whole recognition rate of the probability neural network.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Liang Hua
AU  - Chang-Wei Zheng
AU  - Ju-Ping Gu
AU  - Yu-Qing Liu
PY  - 2015/12
DA  - 2015/12
TI  - Laser arc sound signal processing and welding status recognition based on geometric learning
BT  - Proceedings of the 2015 6th International Conference on Manufacturing Science and Engineering
PB  - Atlantis Press
SP  - 1383
EP  - 1393
SN  - 2352-5401
UR  - https://doi.org/10.2991/icmse-15.2015.251
DO  - 10.2991/icmse-15.2015.251
ID  - Hua2015/12
ER  -