Neural Networks and Acoustic Emission for Modelling and Characterization of the Friction Stir Welding Process

This paper presents an analysis of the correlation between acoustic emission (AE) signals and the main parameters of friction stir welding (FSW) process, based on artificial neural networks (ANN). The AE signals have been acquired by the data acquisition instrument NI USB-9234, applied during the welding process carried out on plates of 3mm thick of aluminium AA1050 alloy. Statistical and temporal parameters of discomposed EA signals using Wavelet Transform (WT) have been used as input of the ANN, while the outputs of model include the welding parameters: tool rotation speed and travel speed, as well as the tool profile. A multilayer feed-forward ANN has been selected and trained, using different algorithms and network architectures. The parameters provided by the ANN constitute the model and the characterization of the FSW process; fmally an analysis of the comparison between the measured and the calculated data is presented, validating the results. The model obtained can be used to develop the automatic control of the parameters of the FSW process, based on vibro-acoustic signals, which constitutes the following step in this research line.