AA2195-T8 Al–Li alloy plates were welded by friction stir welding (FSW) at tool rotational speed of 1,000 rpm and tool traverse speeds (TS) of 100–400 mm·min−1 under three types of butting surface conditions, i.e., (1) without butting surface treatment, (2) butting surface milled, and (3) bead-on-plate welding. The effect of welding heat input and butting surface condition on joint line remnant (JLR) and mechanical properties of friction stir welded 2195-T8 Al–Li alloy was investigated comprehensively. In the stir zone of 2195-T8 FSW joints, there exists JLR composed of alumina-particle arrays and microcracks generated from the initial butting surface, and the morphology of JLR would evolve from smooth to serrate as TS increases. Moreover, as TS increases (i.e., the welding heat input decreases), JLR deteriorates the tensile strength of the 2195-T8 FSW joints, with joints prematurely fracturing along JLR. The fracture mode of 2195-T8 FSW joints was considered to be determined by the lower one between strength of JLR (S JLR) and strength of the lowest hardness zone (S LHZ), and JLR tends to be the fracture path at lower welding heat input. Furthermore, butting surface treatment (milling off oxide layer prior to welding) was found to be able to make the JLR in the 2195-T8 FSW joints less distinct and thus improve S JLR, while fracture along JLR could not be avoided.

No abstract

In this study, a stable real-time monitoring system was established to monitor electrical and mechanical signals during resistance spot welding process. The sudden decrease of electrode voltage signal, the fluctuation of dynamic electrode force and the obvious decline of dynamic resistance could be used for recognizing expulsion phenomena during resistance spot welding which would reduce welding quality and should be avoided as much as possible. In order to research the welding quality estimation methods, four estimation models were built based on regression analysis and back-propagation neural network. The results showed that the estimation accuracy of back-propagation neural network was higher than the model of regression analysis, and the characteristic values of dynamic signals during resistance spot welding process could improve the estimation accuracy significantly.

No abstract