Abstract
Coaxial one-side resistance spot welding (COS-RSW) is a newly developed process for joining metals and composites. In the present study, AL5052 and carbon-fiber-reinforced plastic (CFRP) lap joints were fabricated via COS-RSW. The welding process was modeled numerically using an in-house finite element code called JWRIAN. Single lap shear tests were performed to evaluate the joining strength. The molten zone diameter was defined and measured experimentally to verify the numerical model. An artificial neural network (ANN) was established based on multitask learning, and its training dataset was prepared via finite element analysis (FEA). The well-trained ANN was employed to generate a process window for the COS-RSW. Results demonstrated that the FEA could accurately reproduce the COS-RSW process, which served as an efficient tool for generating a process dataset without performing experiments. The ANN performed multitask learning well and predicted the welding output effectively. Furthermore, an index representing the average value of the maximum temperature in the molten interface of CFRP, was adopted to evaluate the contribution of the integral interface temperature field to the bonding strength qualitatively. An optimal value, which was close to the CFRP decomposition temperature of 340 °C, was obtained, and it exhibited an excellent correlation with higher bonding strengths. The process window provided welding parameters directly to yield the desired results.
3-D modelling of the coaxial one-side resistance spot welding of AL5052/CFRP dissimilar material