Abstract
There is a growing need to automate the gas tungsten arc welding process for fabrication and repair of nuclear components due to an increasing shortage of experienced welders. Therefore, a collaborative effort has been performed in this study to develop a fully autonomous gas tungsten arc welding system with adaptive capabilities. The system employs the application of two neural networks that have been presented in. The first utilizes a vision based convolutional neural network to perform real time control of the filler wire entry position into the weld pool. The second predicts optimal weld parameters and torch positioning for each weld pass deposited within a multi-pass groove. A commercialization path for the technology is in-progress, with the artificial intelligent algorithms currently being incorporated and tested on commercially available equipment.