Abstract: A real time monitoring system is developed to conduct data processing and feature extracting for robotic CO₂ welding of T-joint.A three-dimensional feature vector consisting of standard deviation,kurtosis,and variation coefficient of transient welding current values is constructed to describe the difference between normal welding cases and perturbed ones.Taking the three-dimensional feature as input variable of the fuzzy Kohonen clustering network,the recognition and classification for different welding conditions are carried out,which lay the foundation for real time monitoring of robotic CO₂ weldinhg of T-joint.