Fractal interface evolution and turbulent mixing behavior induced by nonlinear oscillation of the keyhole in laser penetration welding of dissimilar metals

This study presents an innovative multi-physics framework integrating continuum thermodynamics-based simulation with machine learning to elucidate the fractal interface evolution, turbulent mixing behavior, and nonlinear-vibration-chaos-turbulence coupling mechanisms during laser penetration welding of dissimilar metals. A thermal-fluid-electromagnetic multi-field coupling model is established, employing the level set method combined with fractal theory to capture keyhole dynamics and interface complexity. For the first time, the periodic keyhole oscillations at heterogeneous interfaces, driven by thermal property mismatch between dissimilar steels, fundamentally govern the fractal morphology formation, while Marangoni-driven dual-layer vortex structures generate intense shear flow that enhances solute mixing, resulting in fractal interfaces with diffusion coefficients three times higher than smooth interfaces. Machine learning models trained on 200 datasets achieve exceptional predictive accuracy with R 2 values of 0.94 and 0.92 for mixing zone thickness and interface fractal dimension, respectively, identifying heat input as the dominant parameter (r ~ 0.82). Critically, this work discovers strong coupling among nonlinear vibration, chaos, and turbulence within the welding pool, with a critical energy density threshold at 5 × 10 4 W/m 2 marking the transition from chaotic to stable regimes. Distinct process windows are identified: high power (>2500W)-low speed (<20 mm/s) for maximum mixing thickness (1200-1400 μm), and moderate power-speed combinations (2000-2500W, 20-35 mm/s) for optimal fractal dimension (18-20). This work provides a quantitative framework for intelligent process control in dissimilar metal laser welding, bridging multi-scale physics understanding with data-driven optimization strategies.