Friction Stir Welding of Two Dissimilar Metals: Weld Quality Characterization Using the Wavelet Transform Approach

The present work attempted to track weld quality by analysing torque and force signals for Al/Cu dissimilar friction stir welding for two different pin geometries. The captured torque and forces often possess relevant information, which is not always distinguishable from the raw data. Thus, a need for further processing of that raw information arises. Here signals have been analysed by discrete wavelet transformation. 'Maximum energy to entropy criteria' has been employed to select the best mother wavelet function. Original signals, approximate signals, and the sum of detail signals for all decomposed levels have been evaluated and analysed for possible relation with different weld properties. It has been found that the sum of detail signals best reflects the bead surface texture and approximate components possess a good match for macrostructures of weld formation. In addition, all the welds are subjected to microstructural analysis for better correlation of weld quality assessment done by wavelet analysis. As per wavelet analysis, the samples with poor weld quality have unwanted microstructural properties.