Prediction of Welding Penetration in Gas Metal Arc Welding Process using an Artificial Neural Network

Abstract

Gas metal arc welding (GMAW) process is an important component in many industrial operations. The research on controlling GMAW metal transfer modes is essential to high quality welding procedures. The GMAW parameters are the most important factors affecting the quality, productivity and cost of welding joint. The present work describes the development of an ANN based on back propagation (BP) of error for prediction of the gas metal arc welding parameters. The input parameters of the model consist of arc current, voltage and welding speed whereas the output of the model is the depth of the penetration. The effect of network parameters on the mean square error (MSE) of prediction is studied. In present work an attempt has been made to use Neuro solution 4.3 for the ANN applying to gas metal arc welding process. The ANN was subsequently trained with experimental data. Testing of the ANN is carried out using experimental data not used during training. The results showed that the outcomes of the calculation were in good agreement with the experimental data; this indicates that the developed neural network can be used as an alternative way for calculating the process parameters.