Several developments have been observed in the field of materials processing and welding is a vital metal joining process that has potential industrial applications. Friction welding of tube to tube plate using an external tool (FWTPET) is a relatively newer solid state welding process used for joining tube to tube plate of either similar or dissimilar materials with enhanced mechanical and metallurgical properties. Generally, welding is a multi-input and multi-output process in which there exists a close relationship between the quality of joints and the welding parameters. In the present study, Artificial Neural Network (ANN) has been used to predict the strength behavior of FWTPET process. Several Neural Network architectures have been subjected to analysis and the optimal architecture has been determined. The optimal architecture has been used to predict the output process parameter. The predicted output and input parameters have been optimized using Genetic Algorithm (GA). GA optimized and experimentally determined process parameters were compared and the deviation is found to be minimal. Besides, the most influential process parameter has been determined using statistical analysis of variance (ANOVA). © 2010 Elsevier B.V. All rights reserved.