Liwei Wang

Abstract Spinodoid architected materials have drawn significant attention due to their unique nature in stochasticity, aperiodicity, and bi-continuity. Compared to classic periodic truss-, beam-, and plate-based lattice architectures, spinodoids are insensitive to manufacturing defects, scalable for high-throughput production, functionally graded by tunable local properties, and material failure resistant due to low-curvature morphology. However, the design of spinodoids is often hindered by the curse of dimensionality with an extremely large design space of spinodoid types, material density, orientation, continuity, and anisotropy. From a design optimization perspective, while genetic algorithms are often beyond the reach of computing capacity, gradient-based topology optimization is challenged by the intricate mathematical derivation of gradient fields with respect to various spinodoid parameters. To address such challenges, we propose a data-driven multiscale topology optimization framework. Our framework reformulates the design variables of spinodoid materials as the parameters of neural networks, enabling automated computation of topological gradients. Additionally, it incorporates a Gaussian Process surrogate for spinodoid constitutive models, eliminating the need for repeated computational homogenization and enhancing the scalability of multiscale topology optimization. Compared to ‘black-box’ deep learning approaches, the proposed framework provides clear physical insights into material distribution. It explicitly reveals why anisotropic spinodoids with tailored orientations are favored in certain regions, while isotropic spinodoids are more suitable elsewhere. This interpretability helps to bridge the gap between data-driven design with mechanistic understanding. To this end, we test our design framework on several numerical experiments. We find our multiscale spinodoid designs with controllable anisotropy achieve better performance than single-scale isotropic counterparts, with clear physics interpretations.

Spatiotemporal video grounding aims to localize target entities in videos based on textual queries. While existing research has made significant progress in exocentric videos, the egocentric setting remains relatively underexplored, despite its growing importance in applications such as augmented reality and robotics. In this work, we conduct a systematic analysis of the discrepancies between egocentric and exocentric videos, revealing key challenges such as shorter object durations, sparser trajectories, smaller object sizes, and larger positional shifts. To address these challenges, we introduce EgoMask, the first pixel-level benchmark for fine-grained spatiotemporal grounding in egocentric videos. It is constructed by our proposed automatic annotation pipeline, which annotates referring expressions and object masks across short-, medium-, and long-term videos. Additionally, we create EgoMask-Train, a large-scale training dataset to facilitate model development. Experiments demonstrate that the state-of-the-art spatiotemporal grounding models perform poorly on our benchmark EgoMask, but fine-tuning on EgoMask-Train yields significant improvements, while preserving performance on exocentric datasets. Our work thus provides essential resources and insights for advancing egocentric video understanding. Our code is available at https://github.com/LaVi-Lab/EgoMask .

Abstract Centrifugal impellers are critical components in turbomachinery, where performance and reliability are challenged by complex multiphysics interactions under transonic conditions. Existing design methods often neglect thermo–structural interactions, resulting in compromised accuracy and safety. This study addresses these limitations by proposing a multiparameter optimization framework integrating bidirectional thermofluid–structural interaction (BTFSI) analysis to simultaneously enhance aerodynamic efficiency and structural integrity. A BTFSI numerical model was developed, incorporating blade heat transfer and deformation feedback, and applied to optimize five key parameters: blade material, thickness, count, tip clearance, and profile. Results demonstrate that optimized carbon fiber blades reduce maximum stress by 74% and deformation by 61.7% compared to traditional stainless steel, while achieving a 1.1% increase in polytropic efficiency and 0.76% higher pressure ratio. The framework resolves interdisciplinary contradictions between aerodynamic performance and structural safety, offering a validated methodology for transonic impeller design. This work advances high-speed turbomachinery optimization by balancing multiphysics interactions, providing critical insights for material selection and performance enhancement in energy-intensive applications.

The microstructure and impact toughness of the weld were examined after welding X80 pipeline steel using the pulsed MAG(Metal-Active Gas) method, with TiO2 nanoparticles (20 nm) added to the weld. This analysis employed various material characterization techniques. The results show that the weld Microstructure primarily consists of acicular ferrite (AF), granular bainite (GB), polygonal ferrite (PF). The addition of nano-TiO 2 to the weld increased the acicular ferrite content and led to the formation of oxides with Ti-Mn-O-Si-Al as the main constituents. These oxides feature an Al 2 O 3 outer layer coated with MnTi 2 O4-Ti 2 O 3 -type oxides, which effectively promote AF nucleation and refine the weld microstructure. Nano-TiO 2 oxide inclusions were shown to facilitate AF nucleation via two mechanisms: the Mn-depletion zone (MDZ) mechanism and the low lattice mismatch mechanism. However, not all oxide inclusions were observed to induce nucleation through the low lattice mismatch between AF and oxide inclusions. The average impact energy measured in the weld at a temperature of −40 °C was increased by 22.06 % with the addition of nano-TiO 2 (83J) compared to the weld without nano-TiO 2 (68 J). This improvement is attributed to the higher energy required for crack propagation in AF compared to GB. As such, the increased AF content in the weld effectively hinders crack propagation, thereby enhancing the weld’s impact toughness. Furthermore, the reduction in average weld grain size from 7.14 μm to 5.14 μm and the apparent increase in misorientation angles of 45° or higher corroborate the observed improvement in low-temperature impact toughness.

Abstract Additive manufacturing has enabled the fabrication of functionally graded materials (FGMs), such as compositionally graded alloys (CGAs), offering unprecedented flexibility in structural design. CGAs hold significant potential for thermal-elastic applications, yet existing design methods often overlook temperature-dependent material properties due to the complexity of coupled physics, design-dependent temperature fields, and local constraints. To address these challenges, we propose a topology optimization (TO) framework that concurrently designs geometry and graded material composition while accounting for temperature-dependent material behaviors and nonlinear thermal analysis. Our method employs a radial basis function (RBF)-based interpolation scheme to model material properties as functions of both temperature and material composition. Additionally, we leverage automatic differentiation and adjoint sensitivity analysis for computational efficiency and extensibility to GPU acceleration. Numerical examples demonstrate the effectiveness of our approach, underscoring (1) the critical role of temperature-dependent material properties in thermal-elastic structure optimization and (2) the benefits of continuous material grading in enhancing structural performance.

Spinodoid architected materials have drawn significant attention due to their unique nature in stochasticity, aperiodicity, and bi-continuity. Compared to classic periodic truss-, beam- and plate-based lattice architectures, spinodoids are insensitive to manufacturing defects, scalable for high throughput production, functionally graded by tunable local properties, and material failure resistant due to low-curvature morphology. However, the design of spinodoids is often hindered by the curse of dimensionality with extremely large design space of spinodoid types, material density, orientation, continuity, and anisotropy. From a design optimization perspective, while genetic algorithms are often beyond the reach of computing capacity, gradient-based topology optimization is challenged by the intricate mathematical derivation of gradient fields with respect to various spinodoid parameters. To address such challenges, we propose a data-driven multiscale topology optimization framework. Our framework reformulates the design variables of spinodoid materials as the parameters of neural networks, enabling automated computation of topological gradients. Additionally, it incorporates a Gaussian Process surrogate for spinodoid constitutive models, eliminating the need for repeated computational homogenization and enhancing the scalability of multiscale topology optimization. Compared to 'black-box' deep learning approaches, the proposed framework provides clear physical insights into material distribution. It explicitly reveals why anisotropic spinodoids with tailored orientations are favored in certain regions, while isotropic spinodoids are more suitable elsewhere. This interpretability helps to bridge the gap between data-driven design with mechanistic understanding.

Power-law precursory acceleration of observable quantities has been accepted as an effective way to predict time to failure in both materials and structures. However, the form of the power-law exponent is seldom known a priori and is a key challenge in blind prediction. We report a linear relation with respect to time t of the estimated failure times t_{*} that are calculated step by step using the most recent updates of the monitored quantity. Our findings indicate that the monitored quantity can be defined as any power of the inverse rate. All projections of t_{*} for any exponent universally intersect with the straight line of t=t_{*}, with the intersection uniquely defining the failure time. The method is validated against synthetic data, laboratory experiments (materials failure), and volcanic eruption data (structural failure). Our work provides the basis for a significant improvement in time to failure forecasting where the controlling power-law exponent is not known in advance.

A novel bypass-gas metal arc directed energy deposition (bypass-GMA DED) was proposed to prepare high strength Al–Cu–Li-based components. An Al–5.51Cu–0.48Li alloy was successfully prepared, followed by subsequent T6 treatment. The microstructure evolution and improvement of mechanical properties after the T6 state were benchmarked against the as-deposited condition. The component fabricated by bypass-GMA DED has obvious periodic distribution characteristics, with a large number of fine equiaxed grains distributed near the interlayer fusion boundary and coarse columnar grains within the intralayer region. After heat treatment, the Al–5.51Cu–0.48Li alloy is mainly composed of four types of strengthening phases, θ′, θ″, δ′ and T1, which contrasts with the sole presence of θ phase in the deposited condition. These phases are conducive to the improvement of mechanical properties. The strength of the prepared alloy is higher than those additively manufactured Al–Cu–Li alloys, reaching values similar to those of conventional third generation Al–Cu–Li alloys.

BACKGROUND: Cardiometabolic risk factors (CMRFs) related to metabolic dysfunction-associated steatotic liver disease (MASLD) comprised overweight/obesity, impaired glucose metabolism, hypertension, hypertriglyceridemia and low high-density lipoprotein cholesterol. We aimed to describe the overlap prevalence and synergistic interaction of the five CMRFs on MASLD and liver fibrosis. METHODS: Using data of 2017-2020 National Health and Nutrition Examination Survey, we included non-pregnant participants aged ≥ 20 years who completed vibration-controlled transient elastography examinations and had sufficient information to determine their metabolic status. Logistic and generalized linear regression models were performed to assess synergistic interaction between CMRFs on MASLD and identify the contributions to liver fibrosis. RESULTS: The overall estimated prevalence of MASLD was about 33.1%. More than 80% of patients had three or more CMRFs. For MASLD, synergistic interaction between pairs of overweight/obesity and other four CMRFs were higher than it between other CMRFs' pairs [attributable proportion(AP): 40-50% vs 20-30%]. For liver fibrosis, overweight/obesity and impaired glucose metabolism or hypertension had significant synergistic interactions (AP: 50% or 30%, respectively). We identified 27 out of 31 possible CMRF combinations. Combinations including dyslipidemia were more frequent in men than women (77% vs 59%). Combinations including hypertension were less in Mexican Americans than other ethnicities (25% vs 45-57%). Most combinations with three or more CMRFs, regardless of overlap type, had significant associations with elevated liver stiffness value. CONCLUSIONS: CMRF overlap was quite common and had additive interaction in patients with MASLD. Overlapping number may be more important than combination type in liver fibrosis development.

The morphologies and quantities of the LPSO phases in Mg97Zn1Y2 and Mg96.6Zn1Y2.2Zr0.2 (at. %) alloys were regulated by different heat treatments. The two alloys show similar morphologies after the same heat treatments and are composed of three kinks of LPSO phases, namely, LPSO blocks, LPSO laths, and LPSO lamellae. Moreover, the volume fractions of the LPSO blocks and the thickness distributions of the LPSO lamellae are statistically analyzed. In addition, the strengthening effects of the LPSO phases with different morphologies on the alloy’s hardnesses were systematically measured and summarized. The results indicate that both the block-shaped LPSO phases and LPSO laths can significantly strengthen the grain interiors and intergranular regions of the Mg alloys, while the strengthening effect of LPSO lamellae is negligible. These results may shed light on the correspondence between microstructures and heat treatment processes and would be conductive to obtain Mg alloys with better mechanical properties.

ER5356, ER2319 and Zn wires were synchronously fed using a multi-wire gas metal arc directed energy deposition (DED) system. This allowed the successfully fabrication of Al-6.0Zn-2.5Mg-1.5Cu alloy. The microstructure and mechanical properties of the fabricated parts were studied. It is shown that the main precipitated phases are η and nanoscale η'. The upper and lower parts of the deposited parts are composed by coarse equiaxed grains and columnar dendrite structures, respectively. The average microhardness of the fabricated alloy was 118.5 HV. The ultimate tensile strength and elongations in the perpendicular to the building direction (BD) and parallel to the BD were 267 MPa and 2.7 %, 238 MPa and 2.3 %, respectively. The average tensile strength is higher than that of 7075-O alloy. The fracture mode of the samples composed brittle and ductile features, although the former dominated. This multi-wire arc DED approach provides a new choice for the in-situ synthesis of Al-Zn-Mg-Cu alloys. • Al-6.0Zn-2.5Mg-1.5Cu parts were successfully fabricated by a multi-wire gas metal arc (GMA) directed energy deposition (DED). • The Al-6.0Zn-2.5Mg-1.5Cu alloy fabricated with GMA-DED matched the composition and phase structure of the target alloy. • A higher material strength than of as-cast 7075 aluminum alloy shows the potential of GMA-DED technique to fabricate high strength Al alloys in the as-deposited state.

Lotus-type porous metals, characterized by low densities, large surface areas, and directional properties, are contemporarily utilized as lightweight, catalytic, and energy-damping materials; heat sinks; etc. In this study, the effects of dimensionless working parameters on the morphology of lotus-type pores in metals during unidirectional solidification were extensively investigated via general algebraic expressions. The independent dimensionless parameters include metallurgical, transport, and geometrical parameters such as Sieverts' law constant, a partition coefficient, the solidification rate, a mass transfer coefficient, the imposed mole fraction of a solute gas, the total pressure at the top free surface, hydrostatic pressure, a solute transport parameter, inter-pore spacing, and initial contact angle. This model accounts for transient gas pressure in the pore, affected by the solute transfer, gas, capillary, and hydrostatic pressures, and Sieverts' laws at the bubble cap and top free surface. Solute transport across the cap accounts for solute convection at the cap and the amount of solute rejected by the solidification front into the pore. The shape of lotus-type pores can be described using a proposed fifth-degree polynomial approximation, which captures the major portions between the initial contact angle and the maximum radius at a contact angle of 90 degrees, obtained by conserving the total solute content in the system. The proposed polynomial approximation, along with its working parameters, offers profound insights into the formation and shape of lotus-type pores in metals. It systematically provides deep insights into mechanisms that may not be easily revealed with experimental studies. The prediction of a lotus-type pore shape is thus algebraically achieved in good agreement with the available experimental data and previous analytical results.

In this work, we demonstrate the feasibility of applying electric pulses to heal surface cracks on nickel-based alloy GH4169 under compressive load. There exists an incubation time for the onset of the healing of a crack, which is associated with local temperature at the crack tip. The crack size decreases with increasing the pulsing time at a constant healing rate prior to complete healing of the crack. Increasing compressive load accelerates the healing process. The electric pulsing leads to the formation of an influential zone surrounding the crack with the finest grain sizes in the healed crack and the coarsest grain sizes away from the influential zone. The indentation hardness increases with the increase the distance to the crack tip of the healed crack. A model of viscous flow in the crack channel is proposed for the crack healing. The resultant force on the crack faces due to the crack healing increases with the increase of the healing time and the decrease of the crack width.

Exponential growth in the number of parameters used to train deep neural network (DNN)/machine learning (ML) models for artificial intelligence (AI) training/ inference applications requires extensive compute resources like CPUs, GPUs, and memory, interconnected at extremely high bandwidth. Heterogeneous integration via chiplet architectures is key to enabling economically feasible growth of power efficient computing, given the slowdown in Moore’s law. In this paper, we summarize innovative advanced packaging technologies that directly enabled the heterogenous integration of multiple chiplets including CPUs, GPUs, IO die, high bandwidth memory (HBM) die, and passive components in the largest, most complex, and high power (750 W) MI300X Instinct™ accelerator package built by AMD. Three key technologies are described: direct Cu-Cu hybrid bonding, 2.5D integration on a large silicon interposer, and metal thermal interface (TIM)-based cooling solution. The resulting 3.5D packaging technology is described and package-level reliability results are presented.

During Directed Energy Deposition-Arc (DED-Arc), the droplet transfer, molten dynamics and temperature characteristics greatly impact the morphology and microstructure of the as-built component. In the study, a commonly used computational fluid dynamics (CFD) model, which ignores the droplet growth (CFD c model), and an improved CFD model considering the droplet growth (CFD i model) were coupled to balance the calculation efficiency and accuracy. A NiTi shape memory alloy is used as a model alloy. The CFD i model can inherit the thermophysical data of the molten pool in a quasi-steady state calculated by CFD c model and continue the subsequent calculations. With the coupled CFD c -CFD i model, the metal transfer phenomena in the liquid bridge transfer (LBT) and free droplet transfer (FDT) modes were well compared. The numerical results are in excellent agreement with the experimental data. It is shown that the metal transfer in LBT mode is more stable than in the FDT mode, resulting in a more stable molten pool and better forming quality. Besides, the LBT mode is also superior to FDT mode due to the narrower phase transformation range and better mechanical properties of the as-built samples. The present findings lay foundations for optimizing the DED-arc process for any engineering metallic alloy.

Aiming to decouple the inherent relationship between mass transfer and heat transfer in traditional arc-based directed energy deposition, a novel heterogeneous multi-wire indirect arc directed energy deposition (DED) has been developed for in-situ synthesis of Al-Zn-Mg-Cu alloy components. Multi-wires (Al-Cu and Al-Mg) with a bypassing Zn wire have been used to replace the traditional homogeneous twin-wires. The process, microstructure and mechanical properties of the deposited Al-Zn-Mg-Cu alloy components obtained by multi-wire indirect arc DED were investigated. The results indicate that the wire feeding speed, current and angle between the two wires have a significant influence on the multi-wire indirect arc DED process. When the current was 200 A, the different wire feeding speeds could be used for both wires and the angle between them was 90°. The resulting indirect arc presented a ‘heart’ shape and allowed to obtain an Al-5.7Zn-3.4Mg-1.6Cu (wt%) alloy with a high deposition rate of 5.1 kg/h. The Al-5.7Zn-3.4Mg-1.6Cu alloy is mainly composed of α-Al, S (Al2CuMg), η (Mg (Al, Zn, Cu)2) and η′ phases. The composition and phases are in accordance with the 7xxx series aluminum alloys. The microstructure is dominated by columnar and equiaxed grains, and it has obvious periodic distribution along the building direction, which is related to the process thermal cycle. Fine second phases η′ are observed to precipitate during the manufacturing process. Furthermore, the average hardness, ultimate tensile strength and elongation of the fabricated material are 98.6 HV, 243.9 MPa and 5.9%, respectively. These mechanical properties are higher than those of as-cast 7050 aluminum alloy, thus showing the potential of this new process variant to fabricate high strength Al alloys in the as-deposited state. The fracture morphology exhibit features mainly associated to a ductile-like fracture, accompanied by some transgranular and partial cleavage fracture characteristics. This novel multi-wire indirect arc DED provides a new choice for arc-based directed energy deposition of Al-Zn-Mg-Cu alloys and shows great potential for the in-situ synthesis of other high-performance alloys.

Studies successfully exploit electric current treatment including electropulsing solution and direct-current (DC) aging to thoroughly substitute post-weld heat treatment and realize artificial controlling of precipitation as well as the improvement of mechanical property in 6N01 Al alloy friction-stir-welding (FSW) joints . Utilizing electropulsing solution, β ' can be dissolved completely within 3 s to form super-saturated solid solution . The subsequent DC two-stage aging promotes the formation of high-density βʺ within 5 h and increases the microhardness by 21.8% and strength by about 15%. This study extends perspectives on post-weld treatment of precipitation-hardening Al alloy joints and expand the potential function of artificial controlling of precipitation.