Backpropagated Neural Network Model for TTT-TIC Welding Process

Abstract

The TTT-TIC welding process is a newly introduced hybrid welding technology in which Transient Thermal Tensioning (TTT) is followed by Trailing Intensive Cooling (TIC) during the process of welding. This novel method was found effective to reduce welding induced residual stress and distortion. Here, two symmetrical heat sources are moving ahead and one cooling source is trailing behind the welding heat source. In this paper, numerical models were developed for this new welding process and validated against the data from literature. Backpropagated Neural Network (BPNN) model was developed from the simulated results of numerical models. A full factorial design and analysis conducted with the help of BPNN model could reveal the curvature effects of input parameters on maximum value of longitudinal residual stress. The Neural Network Toolbox in MATLAB was used to develop, optimize and simulate the neural network model.