Optimised penetration for fillet welding

There is no economic technology available to measure the fillet weld geometry on a regular basis. External aspects of a fillet weld can be easily measured using specifically designed gauges, but the internal characteristic of penetration cannot. To guarantee the structural integrity of a fillet weld it is important that there is enough penetration to ensure that the axis between the bar and the plate is effectively ‘cut’, see figure 1.
A random study of the parameters set by welders showed wide ranging differences, some of which led to lack of penetration in the fillet weld.
As a result of this finding, a programme of work has been put in place which relies on an Artificial Neural Network (ANN) approach as a method to develop the optimum set of fillet welding parameters.
Initial work on this showed that when welding vertically up, there were very few issues in developing a set of parameters. However, there was significant variation in the data when welding in the downhand position. It was concluded from this that the number of potential variables being measured and controlled involved would have to be increased.in order to produce a reliable ANN.
This paper shows how the approach was developed to tackle the issue on welding in the downhand position. The use of a mechanised welding tractor was built into the project, and variations in torch angle, push-pull angle and stand-off were among the variables incorporated.
Other potential variables such as the gap between the plate and the bar, the bar geometry and the gas type and gas flow rate have currently been kept constant.
The progress to date will be presented.