Ultrasound-assisted direct energy deposition (UADED) attracts increasing attention due to its capability to tailor the grain structure. However, the involved molten pool dynamics, particularly the complex interaction of ultrasound-flow-solidification, remain unclear to date, which hinders quantitative prediction and regulation of the microstructures and mechanical properties of UADED components. Here, in situ high-speed imaging and high-fidelity multi-physics modeling are leveraged to investigate flow characteristics and liquid-to-solid transformation in UADED for Inconel 718. The inertial force activated by ultrasound is revealed to drive the molten pool to flow forward and backward along the vibration direction, resulting in poor surface quality. A hybrid deposition strategy is developed to minimize ultrasound-induced defects and produce superior microstructure with alternating coarse- and fine- grains. Such a layered microstructure results in 28% and 15% improvement in the yield strength and ultimate tensile strength compared to the counterpart by additive manufacturing without ultrasound. This work provides unprecedented understanding into the molten pool dynamics in the UADED process as well as valuable guidance to manipulate molten pool flow.

In ultrasound-aided laser melting processes such as additive manufacturing, it is generally believed that acoustic cavitation is essential for grain refinement during solidification while acoustic streaming plays a negligible role. We propose a non-contact ultrasound approach to provide low-intensity ultrasound, i.e., below the melt cavitation threshold, ensuring a pure acoustic streaming regime. Without cavitation, it is found that fine equiaxed grains still can be achieved. This is attributed to the combined effects of acoustic streaming and Marangoni force, which create a high-frequency-shaking type melt flow in the melt pool, leading to fatigue fracture of dendrites and thus grain refinement. Moreover, low-intensity ultrasound can offer stable melt pool modulation throughout layer-by-layer processing, enabling uniform grain refinement in large-scale samples, which is a challenge for the current direct-contact ultrasound approach.

Cathode spot motion influences the physical characteristics of arc plasma and the related macroparticles (MPs) in resultant films; these MPs limit the application of arc ion plating (AIP). In this paper, a scanning radial magnetic field (SRMF) was applied to the cathode surface to control the cathode spot motion and reduce the MP contamination in the deposited films. It was shown that film surface morphologies prepared using SRMF were better than those using a static radial magnetic field (RMF). The improvement was greater with increased scanning range and frequency. Using SRMF, cathode spot motion was confined to a spiral trajectory on the cathode surface and the spots moved over a large area and at a fast-moving velocity. Both the large moving area and the fast velocity decreased the temperature on the cathode surface and thus reduced the emission of the MPs.

Artificial intelligence brings novel perspectives to laser metal additive manufacturing (LMAM), enhancing precision, efficiency, structural and process optimization. Machine learning-assisted process-structure-property correlation in additive manufacturing (ML-PSP-AM) presents an effective pathway for structural innovation and performance optimization, leveraging automation and intelligence to address the growing processing demands across industries. This review differs from existing literature by presenting a multi-scale, PSP-centered analysis of ML applications in LMAM, integrating discussions that span from processing-driven macro-scale formation to meso/micro-scale defect prediction and microstructure-property relationships. By evaluating state-of-the-art ML applications across various AM stages, we identify current limitations, propose targeted strategies, and outline opportunities to improve accuracy, minimize defects, and enhance mechanical properties such as strength and fatigue life. Looking ahead, the advancement of ML-assisted AM should focus on breakthroughs from “0 to 1” in application and innovations from “1 to ∞” in algorithms. The realization of ML-PSP-AM represents a transformative yet disruptive integration of manufacturing engineering, artificial intelligence, and materials science, driving significant progress in modern manufacturing technologies.