Identification of Weld Defects in GMAW Based on Arc Sensing

CO₂ gas shielded arc welding is widely used in automatic and robotic welding.The automatic monitoring of weld quality is a problem that needs urgently solved in industry.Through-the-arc sensing,due to its advantages in practical applications,gets more and more concerns in recent years.This paper presents a monitoring method of weld defects for CO₂ gas shielded arc welding process.It is based on the feature extraction for the arc signals by the classification of signals' histogram in the welding process using Self-Organize feature Map (SOM) neural networks.Experiments show that this strategy realizes effectively the identification of weld defects and can be used in the on-line monitoring of welding process.It is very important for the welding process in achieving the aim of "zero defect" products.