Data-driven modeling and optimization of a robotized multi-needle ultrasonic peen-forming process for 2024-T3 aluminum alloy

The uniformity and efficiency of ultrasonic needle peen-forming (UNPF) cannot be achieved by a handheld single-needle peening device. Thus, a novel robotized UNPF system is designed in this paper with a multi-needle peening device. By taking the aluminum alloy 2024-T3 specimens as an example, the peening device’s travel velocity and standoff distance are first determined to balance processing efficiency and safety. Thereafter, the resulting arc height for Almen-size specimens with respect to other manipulated variables (MVs), such as the vibration amplitude, the air pressure, and the number of peening passes, are fitted by the data-driven models. As the radius of curvature at the saturation state is identical for small and large specimens with similar shapes, its relationship with the arc height and at the saturation state is represented by a function of elliptic paraboloid for understanding the limit of UNPF capability. It is shown that specimens of different sizes but the same aspect ratio have a consistent radius of curvatures after being treated with the same MVs. Notably, by increasing the vibration amplitude from 21 to 35  \(\mathrm {\mu m}\) $\mu \mathrm{m}$ , the ratios of peening time required for achieving full coverage to its counterpart for achieving saturation state reduced from 76 to 8.3%, showing a significant improvement in forming productivity and capability. Furthermore, increasing the number of peening passes, from 1 to 32 significantly reduces surface roughness by up to 86.4%, approaching the as-received specimen’s smoothness. Moreover, the maximum compressive residual stress after UNPF is  \({-}\) $-$ 500 MPa at a depth of 1.2 mm, and the hardness is drastically increased by 45.4% for specimens of thickness 1.5 mm, improving their fatigue life.