<p>For friction stir welding (FSW) of dissimilar components, the underlying correlation between in-process thermal-material flow and intermixing behavior and joint performance is crucial for process optimization and application in industry, but it is still far from being fully understood. In this study, a 3D model based on computational fluid dynamics (CFD) approach with a novel heat source and a self-adaptive boundary condition was proposed to simulate the multi-physics coupling fields during FSW of Al/Cu dissimilar alloys. Then, the numerical results were combined with the experimental observations to elucidate the significance of dissimilar Al/Cu flow and intermixing on joint formation mechanism and intermetallic compounds (IMCs) distribution. First, it was shown that both heat input and temperature increased with changing tool offset from Al-AS to Cu-RS, but decreased with increasing welding speed. Second, the dissimilar Al/Cu flow pattern around the pin was varied completely since the flow channel for Al-AS at the RS was changed from opening to closing, so that the dissimilar Al/Cu intermixing and interlocking was enhanced with changing tool offset from Al-AS to Cu-RS. However, there were limited variations of both flow pattern and mechanical interlocking of Al/Cu with increasing welding speed. Third, as an integrated result of the thermal-material flow behavior during Al/Cu FSW, the IMCs thickness at the bonding interface decreased with changing tool offset from Al-AS to Cu-RS, and also with increasing welding speed. Finally, it was demonstrated that numerical simulation well predicted the coupled multi-physics characteristics in FSW of Al/Cu with various welding conditions, and the significance of dissimilar material flow and intermixing on the FSWed Al/Cu joint performance was particularly elucidated.</p>