Melt pool geometry governs porosity formation, inter-layer bonding, microstructural evolution, and residual stress development in Laser Powder Bed Fusion (L-PBF), making its accurate prediction central to process qualification and process-window development. However, experimental qualification is resource-intensive, high-fidelity multi-physics simulations remain computationally expensive for large-scale parameter exploration, and reduced-order analytical models often lose accuracy near melting-regime transitions. Existing machine learning (ML) approaches provide deterministic predictions but do not explicitly address uncertainty or systematic prediction bias, limiting their reliability across different operating conditions and experimental datasets. This study presents an uncertainty-aware hybrid ML framework for melt pool geometry prediction and process qualification in L-PBF of Inconel 718 (IN718). Single-track and multi-track specimens were fabricated over a broad range of laser powers, scan speeds, hatch spacings, and layer counts and characterized using optical and scanning electron microscopy. The proposed framework combines k-fold ensemble regression with physics-informed directional classification, where the ensemble standard deviation serves as a physically interpretable uncertainty signal and is subsequently converted into a targeted bias-correction mechanism. The melt pool depth-to-width ratio (D/W) was adopted as a unifying descriptor linking melt pool morphology, thermal-gradient-driven grain growth direction, melting-regime transitions, and process-map development. The framework was trained using 68 experimentally measured melt pools and evaluated using an external dataset of 54 measurements compiled from published L-PBF IN718 studies, including validation and independent test datasets. Compared with established analytical models and polynomial curve-fitting baselines, the proposed approach improved predictive accuracy, enhanced cross-study generalization, and provided useful ensemble-based uncertainty estimates. Multi-track experiments further validated the maximum hatch spacing criterion for predicting lack-of-fusion onset from single-track measurements, achieving prediction errors below 6% and enabling the identification of defect-free processing windows. The results demonstrate that ensemble prediction uncertainty contains information related to regime-dependent model sensitivity and can be exploited to improve prediction reliability near conduction-transition-keyhole boundaries. Beyond the specific IN718 case, the proposed framework provides a potentially transferable strategy for uncertainty-aware process qualification and rapid process-window development and offers a computationally efficient foundation for digital twin development, in-situ monitoring, and future closed-loop control of metal additive manufacturing (AM).
Magnesium alloys exhibit high specific strength and good machinability, and are promising materials for high-precision micro-components. However, their hexagonal close-packed structure provides limited slip systems, which restricts their plastic deformation during conventional laser shock forming (LSF) and leads to insufficient forming depth and non-uniform wall thickness. To address these limitations, this study proposes an electro-pulse–assisted laser shock forming (EP-LSF) process that integrates pulse current with laser shock loading to regulate the transient electro–thermo–mechanical responses of AZ31 magnesium foils. A multi-physics numerical model was developed using ABAQUS and combined with experimental investigations to analyse the dynamic response and deformation behaviour under high-strain-rate conditions. Experimentally, the forming depth under EP-LSF exhibited a non-monotonic dependence on the pulse-current density, reaching 217.7 μm at 20A·mm⁻², which was 95.5% higher than that achieved under conventional LSF. At this optimum current density, the wall-thickness uniformity improved, local thinning decreased, and component hardness increased. Numerical simulations validated this optimum current-density response: the displacement in the thickness direction agreed well with the measured forming depth, while the equivalent plastic-strain distribution and residual stress field on the laser-shocked surface indicated more homogeneous plastic flow, an approximately four-fold increase in the surface residual compressive stress, and reduced tensile stress at the die fillet. In terms of microstructural evolution, the synergy between Joule heating and the non-thermal effects of pulse current promoted atomic diffusion and re-crystallisation, leading to grain refinement, an increased fraction of high-angle grain boundaries, and weakened texture, thereby enhancing the ductility. These results indicated that a moderate pulse-current density could provide an optimum electro–thermo–mechanical coupling condition for high-strain-rate micro-forming of magnesium alloys. Overall, this study provides fundamental insight into how pulse current acts as an external field to regulate the deformation stability and microstructural evolution during high-strain-rate micro-forming of magnesium alloys, offering guidance for the forming and manufacturing of lightweight complex micro-components.
The roll forming process effectively reduces sheet metal springback through multi-pass local plastic strain accumulation and residual stress reconstruction. This paper extends previous work that developed a generalized anisotropic distortional hardening (G-ADH) model based on the associated flow rule. Considering the relationship of shear stress and uniaxial tension stress for 1.8GPa steel, the analytical Poly6-I is used. To accurately predict springback in ultra-high strength sheet in nonlinear loading path, the G-ADH model with Young's modulus degradation is used. For comparison, the Chaboche model is also used to predict the loading-reverse loading curves. The results show that the G‑ADH model offers greater flexibility and higher accuracy than the Chaboche model. To further verify the validity of these models, the springback obtained from cold roll‑forming experiments on U‑shaped sections of 1.8GPa steel is employed. The experimental and numerical results demonstrate that, regardless of the presence of side rolls, the “G-ADH + Poly6-I” model considering Young's modulus exhibits high consistency with experimental data in predicting springback with relative errors below 1.5%, which is attributed to simultaneous description of reverse-loading hardening and degradation of the unloading modulus. The springback differs at different profiles perpendicular to the sheet length direction, which originates from the inconsistency in the extent of the local plastic deformation zones. Lower springback is observed during the forming process that includes side rolls. This study reveals the correlation between the springback amplitude and the accumulation of equivalent plastic strains under low stress during U rolling forming of 1.8GPa steel, providing an accurate numerical tool for predicting and reducing springback of ultra-high strength alloys under bending-unloading-reverse bending deformation paths.
Joining thin-walled Al/Cu tubes is challenging because of the large differences in material properties and the low structural rigidity of thin-walled tubes, which often result in severe deformation and poor weld quality during magnetic pulse welding (MPW). To overcome these limitations, this study proposes a novel magnetic pulse welding with granular support (MPW-GS), in which dynamically compacted granular media provide transient radial confinement to regulate tube deformation during high-velocity impact. Experiments demonstrate that AA1060/T2 joints can be produced under a discharge energy of 32kJ. Compared with conventional MPW, the proposed process suppresses the radial collapse of the inner tube by 65.67%, while achieving a maximum tensile load of 1.35 kN, exceeding the strength of the AA1060 base material. Typical wavy interfaces with the largest amplitude of 7 μm and a transition zone were observed, confirming an effective metallurgical bond at the welded interface. The FEM-DEM coupled simulations reveal that GS enhanced stress transfer and collision stability through dynamic densification, thereby improving the quality of the welded interface. Based on these findings, a stress-wave propagation model is established to explain the underlying welding mechanism. The analysis demonstrates that the dynamically compacted granular support enhances stress-wave reflection at the inner tube interface, thereby promoting interfacial wave formation and establishing a positive feedback mechanism between granular densification and stress-wave amplification. These works provide new insight into support material, deformation control and interface evolution in magnetic pulse welding of thin-walled dissimilar tubes.
Reinforcement loss during laser powder bed fusion remains insufficiently understood because of complex particle behaviour within transient melt pools. Using GH3536–5 wt% microscale TiC as a model system, this study clarifies TiC loss mechanisms and links particle retention to microstructural and mechanical evolution. Single-track experiments and specimen fabrication were combined with multiscale characterisation, melt-pool simulations, thermodynamic/kinetic calculations, and spatter analysis. The TiC loss rate decreased by approximately 50% as scanning speed increased from 400 to 1400 mm/s; high-loss and low-loss specimens exhibited loss rates of 24% and 12%, respectively. No direct TiC loss by evaporation was identified. Partial TiC dissolution released Ti and C, subsequently forming nanoscale (Ti, Mo)C precipitates. Meanwhile, GH3536 matrix evaporation promoted spatter generation, and the resulting droplets entrained and ejected TiC particles floating on the melt-pool surface. Thus, dissolution–reprecipitation transformed TiC within the build, whereas spattering provided the direct pathway for material removal. Higher scanning speed and greater TiC retention jointly refined the grain size from 13.7 to 9.8 μm, while texture weakening was mainly associated with greater TiC retention. The low-loss specimen achieved ultimate tensile strengths of 1258 MPa at 25 °C and 331 MPa at 900 °C, 23% and 19% higher than those of the high-loss specimen, respectively. Fracture shifted from TiC cracking at 25 °C to matrix–TiC interfacial debonding at 900 °C. This work provides a transferable framework for controlling reinforcement retention in laser-processed metal-matrix composites.
Highly doped GaN substrates are essential for low-resistance vertical power devices, but their intrinsic carrier-concentration inhomogeneity causes spatially nonuniform oxidation and consequently nonuniform removal in electrochemical mechanical polishing, which limits surface quality and process controllability. This study investigated the oxidation behaviour of OVPE-grown n-type GaN under chemical, electric-field-assisted, and photo-assisted reaction fields and proposed a slurry-less photoelectrochemical mechanical polishing (PECMP) process for uniform finishing of doping-inhomogeneous substrates. The results show that anodic oxidation and UV-assisted oxidation exhibit opposite carrier-concentration dependences: anodic oxidation was accelerated in high-carrier regions because of enhanced current density and hole supply, whereas UV-assisted oxidation was suppressed in high-carrier regions because intensified recombination shortens the effective lifetime of photogenerated holes. By balancing these opposite responses, spatially uniform surface modification was achieved across the substrate surface. Based on this mechanism, a slurry-less PECMP process using a soft fixed CeO2 abrasive was developed to selectively remove the modified layer without damaging the GaN substrate. Under an electric-field-dominant condition, the material removal rate reached 3.81µmh⁻¹, but carrier-dependent topography remained. Under the balanced condition, the formation of characteristic floral-like topography was suppressed and an ultra-smooth surface with Sa = 0.397nm was obtained. Stress mapping using Raman spectroscopy further confirmed the elimination of scratch-related residual-stress features. The results demonstrated that uniform and damage-free polishing of highly doped GaN was achieved not by simply increasing the oxidation rate, as in conventional PECMP, but by controlling the relative contributions of reaction fields with opposite carrier-concentration responses. This process design provides an effective route for uniform, efficient, and slurry-less finishing of semiconductors and other hard materials with spatially inhomogeneous reactivity.
Joining N36 zirconium alloy to Inconel 718 (IN718) is attractive for nuclear-related structural integration, but fusion welding is restricted by rapid Zr-Ni reaction and brittle intermetallic cracking. The key question addressed here is whether laser energy can be redistributed in space and time so that a refractory Nb interlayer remains a solid barrier while both adjacent alloys still form metallurgical bonds. We therefore developed full domain power modulation (FDPM) laser welding, in which one oscillation cycle is divided into 36 phase segments and the laser power is assigned according to the instantaneous beam position over the N36, Nb, and IN718 regions. This phase-resolved strategy produced a relative "cold-center, hot-sides" thermal field: the central Nb region stayed below the Nb melting point, whereas the two sides received sufficient heat for localized wetting, reaction, and bonding. The preserved Nb-rich band blocked long-range Zr/Ni intermixing and reconstructed the joint into a ductile β-(Zr, Nb) solid-solution band at the N36/Nb side and a Laves/μ-containing reaction layer at the Nb/IN718 side. The optimized joint reached 84 MPa tensile strength and about 11% elongation, substantially higher than direct laser welding but still far below the base-metal strengths because brittle TCP phases remain at the Nb/IN718 interface. These findings demonstrate a viable strategy for retaining a diffusion barrier in select incompatible systems. Although strength remains limited by brittle interfacial phases, this concept lays a foundation for further interface optimization.
