Electron Beam Welding (EBW) of Aerospace Alloy (Inconel 825): Optimization and Modeling of Weld Bead Area

This study investigates the optimum weld area on a popular aerospace alloy (i.e., Inconel 825) made by the electron beam welding technique. Welding speed (S), beam current (I), accelerating voltage (V), and beam oscillation (O) are considered as process parameters to study the weld bead area (WA) of the weldments. An instructive study on multiple non-linear neural regression analyses has been done as a basic introduction to neuro regression modeling with artificial neural network (ANN) philosophy. To do this, the experimental prediction has been modeled with 14 predictive functional structures using fundamental regression modal types to test the accuracy of their predictions. To train the program with the chosen model R^2_training, test it R^2_testing, verify the accuracy R^2_validation is used, and check whether the values are within the engineering limits. Optimization algorithms with three different scenarios have been applied. Only one of the 14 models gave realistic results. It has been seen that the scenario types, selection of different constraints, and different models for design variables affect the optimization results.