Obtaining weld pool vision information during aluminium alloy TIG welding

Abstract

An image sensing system for the TIG (tungsten inert-gas arc) welding process of aluminium alloy was established. The relationships between the image sensing system and the characteristic of welding current were discussed in detail. Front and back images of the weld pool were obtained with different welding parameters. In order to process the image, the characteristics of an aluminium alloy were analysed. Image processing and pattern recognition were first used to obtain information from the TIG welding process for aluminium alloy. The image of the weld pool was pre-processed using a series of methods: a weighted median filter, a statistical threshold by expectation and the projection method. A neural network method was used to extract the edge of the images of the weld pool. The result of detecting the edge with a BP neural network were excellent. The symmetry of the weld pool for aluminium alloy was studied when the welding current is large. The edge of the image of the whole weld pool is obtained in a single side image and an accurate method for measuring the weld pool geometry parameter is provided. Experiments show that using the image sensing to control the TIG weld width for aluminium alloy is an effective method.