Machine Learning-Assisted Burst Femtosecond Laser Polishing of Invar Alloy: Process Optimization and Performance Enhancement

As a key low-expansion material for high-end equipment such as aerospace and precision instruments, the surface quality of Invar alloy directly determines the operational performance of devices. To fill the research gap in the multi-parameter synergy and mechanism of Invar alloy laser polishing, this study performs polishing experiments on Invar alloy using a burst-mode femtosecond laser, with a repetition rate of 1 MHz and four sub-pulses per burst. The results indicate that energy density plays a dominant role in the polishing effect: with the increase in energy density, the surface roughness first decreases and then increases. A stable molten pool is formed under medium energy density (0.47–0.64 J/cm2), and under the optimal parameter conditions, the surface roughness is reduced to 394 ± 50 nm, representing a 52% reduction compared to the original surface (821 nm). Scanning speed and scanning pitch affect the polishing effect by synergistically regulating energy input: increasing scanning speed under high energy density can inhibit the rise in roughness, while a small scanning pitch can lower the threshold of optimal energy density. Amplitude spectrum analysis reveals that the medium-scale surface undulations are significantly improved after polishing. A four-layer Fully Connected Neural Network (FCNN) model is established to achieve high-precision prediction of polishing effects with a coefficient of determination R2 = 0.92, which enables rapid prediction of unknown polishing parameter combinations and provides a new solution path for the optimization of polishing effects. This study clarifies the interaction mechanism between a burst-mode laser and Invar alloy, proposes an efficient ultra-precision polishing method for Invar alloy, and lays a theoretical foundation for its application in the field of high-end manufacturing.