A combined wavelet packet and Hilbert-Huang transform for defect detection and modelling of weld strength in friction stir welding process

The objectives of the current research work is two folded. First, to develop effective methodology for identification of internal defects in friction stir welded samples. Identification of tunnel type of defect is attempted through the analyses of real-time force signals and analyzing the signals with a proposed combination of wavelet packet transform with Hilbert-Huang transform. The analyses results in two features namely instantaneous phase and instantaneous frequency that efficiently describe the presence of defects in the welded samples. The study presented for tunnel defect can as well be implemented in detection of other types of defects in friction stir welded samples. The second objective of this work is to develop a data driven model for accurate estimation of ultimate tensile strength of the welded samples. For this purpose support vector machine learning based support vector regression model is developed. The model parameters are optimized using the grid search method to control the overfitting during training of the model. The accuracy of the developed model leads to the impression that it can be further modified for real-time welding operations for accurate modelling of ultimate tensile strength of the joints.