It is a key issue to establish an appropriate model of the heat source in the simulation of the keyhole plasma arc welding (PAW). It requires that the model account for the keyhole effect and have the characteristic of volumetric distribution along the direction of the plate thickness. For available heat source models, neither Gaussian nor double ellipsoidal modes of heat source is applicable to keyhole PAW process. With considering the force of the high speed plasma jet and the associated strong momentum, a modified three-dimensional conical heat source model is proposed as the basis for the numerical analysis of temperature fields in keyhole PAW process. Further, a new heat source model for quasi-steady state temperature field in keyhole PAW is developed to consider the “bugle-like” configuration of keyhole and the decay of heat intensity distribution of the plasma arc along the direction of the workpiece thickness. Based on this heat source model, finite-element analysis of temperature profile in keyhole PAW is conducted and the weld geometry is determined. The results show that the predicted location and locus of the melt-line in the PAW weld cross section are in good agreement with experimental measurements.

This paper introduces an intelligent system for monitoring and recognition of process disturbances during short-short-Cirutiong gas-metal arc welding. It is based on the measured and statisticlly processed dataof welding electrical parameters. A 12-dimensional array of process features is designed to describevarious welding conditions and is employed as input vector of the intelligent system. Three methods, Such as fuzzy c-means, neural netowrk and fuzzy Kohonen clustering network areused to conduct process monitoring and automatic recognition. The correct recoginition rates of these three methods are compared.