Feature Extraction and Dimensionality Reduction of Arc Sound under Typical Penetration Status in Metal Inert Gas Welding
Lijun Liu,Hu Lan,Hongyan Zheng,Xiaoxia Jian
DOI: https://doi.org/10.3901/cjme.2012.02.293
2012-01-01
Abstract:Arc sound is well known as the potential and available resource for monitoring and controlling of the weld penetration status,which is very important to the welding process quality control,so any attentions have been paid to the relationships between the arc sound and welding parameters.Some non-linear mapping models correlating the arc sound to welding parameters have been established with the help of neural networks.However,the research of utilizing arc sound to monitor and diagnose welding process is still in its infancy.A self-made real-time sensing system is applied to make a study of arc sound under typical penetration status,including partial penetration,unstable penetration,full penetration and excessive penetration,in metal inert-gas（MIG） flat tailored welding with spray transfer.Arc sound is pretreated by using wavelet de-noising and short-time windowing technologies,and its characteristics,characterizing weld penetration status,of time-domain,frequency-domain,cepstrum-domain and geometric-domain are extracted.Subsequently,high-dimensional eigenvector is constructed and feature-level parameters are successfully fused utilizing the concept of primary principal component analysis（PCA）.Ultimately,60-demensional eigenvector is replaced by the synthesis of 8-demensional vector,which achieves compression for feature space and provides technical supports for pattern classification of typical penetration status with the help of arc sound in MIG welding in the future.