Wide-bandgap (WBG) semiconductors such as SiC and GaN require interconnect materials that can operate reliably at elevated temperatures, driving interest in low-temperature sintered metal pastes for advanced packaging. To address the high cost and electromigration issues of nano-Ag pastes, this study investigates a micro-nano Cu paste system through molecular dynamics (MD) simulations and experiments.A multiparticle atomistic Cu/Ag interfacial sintering model was established to reveal the coupled evolution of atomic diffusion, local crystal structure, and dislocation behavior. In parallel, experiments were conducted to evaluate the effects of temperature, pressure, and holding time on interfacial microstructure and shear strength. The results show that increasing temperature enhances atomic mobility and Cu–Ag interfacial diffusion, thereby accelerating neck growth, reducing porosity, and improving interfacial continuity. Extending the holding time provides a longer kinetic window for diffusion, structural rearrangement, and defect interaction, although its incremental effect gradually diminishes once the primary bonding framework is established. Among the selected process variables, pressure shows the strongest influence on sintering quality because it directly improves contact intimacy, shortens the effective diffusion distance, and promotes stress-assisted plastic accommodation. Common neighbor analysis and dislocation extraction further indicate that the increase in HCP-like local structures and dislocation activity is closely associated with defect-assisted densification and interfacial stabilization. Both simulation and experimental orthogonal analyses yield the same significance order of processing variables within the selected ranges: pressure > temperature > time. Under optimized condition of 280 °C, 15 MPa, and 20 min, the joint achieved a maximum shear strength of 31.4 MPa.

The pursuit of higher power and density in wide bandgap power modules makes thermal management a critical challenge. Thermal interface materials (TIMs) play a critical role by filling microscopic air gaps and thereby enhancing heat transfer for power electronics. Combining the recent development of nano metal sintering technology and exceptional intrinsic thermal conductivity of certain carbon allotropes, micro-nano Cu sintering with carbon fiber (CF) reinforcement was promoted as potential high performance and cost-effective TIM for power modules. In this work, a Cu/CF composite paste was synthesized and sintered by the thermal-compressing process, followed by the detailed investigation and simulation on its sintered interfacial microstructure and mechanical and thermal properties. This work demonstrated a significant improvement of thermal and mechanical properties through the optimization of CF doping ratio. The addition of 5 wt. % CF increased the average shear strength of Cu sintering joints by 65.3% to 44.8 MPa, which is attributed to enhanced interface adhesion from the nano Ni particle coating on the CF surface. Thermal conductivity rose from 77.1 to 87.4 W/(m K) and 107 W/(m K) with 5 and 10 wt. % CF doped, respectively. Furthermore, adding 10 wt. % CF increased thermal diffusivity by 27.4%, which helps alleviate transient thermal loads. This work highlights the strong potential of Cu/CF composites as future TIM for high-density power modules.

A hybrid action deformation control method based on hybrid proximal policy optimization （HPPO） is proposed for titanium alloy structural components. Existing reinforcement learning algorithms are generally confined to either discrete or continuous action spaces， and thus cannot simultaneously optimize machining sequence and allowance. The proposed method unifies both decision variables—machining sequence as discrete actions and machining allowance as continuous parameters—into a single parameterized hybrid action space. Online deformation force monitoring data serve as state feedback to enable adaptive control under dynamic machining conditions. A dual-layer reward mechanism combining process-level deformation force uniformity with terminal deformation convergence is designed to guide the agent toward synchronized suppression of both local and global deformations. Experimental validation on a Ti6Al4V aviation structural component demonstrates that the proposed method reduces average machining deformation from 0.103 mm to 0.054 mm， with RMSE decreasing from 0.119 mm to 0.071 mm， representing a 47.57% reduction relative to the uncontrolled case. These results confirm the accuracy and effectiveness of the proposed method in real manufacturing environments.

Additive manufacturing enables the fabrication of patient-specific self-expanding Nitinol stents. However, the relationships among processing conditions, microstructural evolution, multi-scale mechanical behaviour, and in vitro deployment performance remain poorly understood. This study carried out microstructural, mechanical and functional assessments for personalised Nitinol stents produced by micro-laser powder bed fusion (μ-LPBF). The as-printed Nitinol exhibited a predominantly equiaxed microstructure with low porosity. The austenite finish temperature remained below body temperature, indicating stable austenitic behaviour under physiological conditions. Mechanical testing further revealed a measurable superelastic response with ~3% recoverable tensile strain. Electrochemical polishing transformed a particle-covered surface into a glossy finish for the as-printed stents, with the arithmetic mean roughness (Ra) being reduced to an average value of 1.89 ± 0.60 μm. The personalised μ-LPBF stents exhibited enhanced luminal restoration during in-vitro deployment, with local expansion exceeding that of the conventional design by up to 16.84%. These results demonstrate that the personalised μ-LPBF Nitinol stents achieve improved in-vitro luminal restoration compared with the conventional geometry.

Through-silicon via (TSV)-integrated embedded microchannel is a promising thermal management strategy for high-power-density packaging. However, the associated leakage risk and long-term reliability under coupled multiphysics conditions remain insufficiently understood. In this study, a finite element method (FEM) was developed to investigate the leakage risk introduced by TSV-integrated microchannels and evaluate their effects on long-term thermo-mechanical reliability. A reliability-oriented optimization strategy was proposed to enhance packaging performance. The results show that the potential leakage-risk regions are mainly located at the inletside pin-fin structures and the lower edge of the chip, and The microbump is the most failure-prone interconnect component. The optimized design increases the Nusselt number (Nu) by 22.12% while maintaining a favorable balance between heat transfer enhancement and fluid energy loss, with all comprehensive performance factor (PEC) values exceeding 1. Moreover, both optimized design and increased Reynolds number (Re) reduce the viscoplastic dissipation energy density of the microbumps, thereby delaying irreversible damage accumulation. As a result, the fatigue life of critical bump is extended by a factor of 9.27 at Re = 200. In addition to redistributing the temperature and stress fields, microchannel integration alters the fatigue-life distribution of the microbumps. These findings provide a theoretical basis and design guidance for improving the long-term reliability of TSV-integrated microchannel cooling in heterogeneous packaging.

Residual stress fields significantly affect the structural integrity and service performance of large-scale components. Accurate quantification of the residual stress field is essential for reliable manufacturing and structural evaluation. However, traditional inference techniques typically rely on predefined basis functions or assumed stress distributions, which require accurate prior knowledge and often fail in the presence of complex geometries or heterogeneous stress patterns. To address these limitations, this study proposes a residual stress inference method based on spectral clustering of the deformation force-residual stress relationship tensor, which eliminates the need for explicit prior assumptions. The core idea is to partition the global residual stress field into mechanically consistent regions, thereby reducing the number of unknowns while preserving predictive accuracy. The method is validated using a forged aluminum alloy component with a complex geometry. In numerical simulations, the full-field stress inference achieves a mean absolute error (MAE) of 4.33 MPa. In machining experiments, the inferred stress field enables accurate predictions of deformation forces and displacements, with MAE of 10.48 N and 0.18 mm, respectively. These results demonstrate the feasibility and effectiveness of the proposed approach for residual stress field inference in large-scale components, offering valuable support for stress-aware manufacturing and assembly in aerospace, energy, and related engineering domains.

Predictive learning for spatio-temporal processes (PL-STP) on complex spatial domains plays a critical role in various scientific and engineering fields, with its essence being the construction of operators between infinite-dimensional function spaces. This paper focuses on the unequal-domain mappings in PL-STP and categorising them into increase-domain and decrease-domain mapping. Recent advances in deep learning have revealed the great potential of neural operators (NOs) to learn operators directly from observational data. However, existing NOs require input space and output space to be the same domain, which pose challenges in ensuring predictive accuracy and stability for unequal-domain mappings. To this end, this study presents a general reduced-order neural operator named Reduced-Order Neural Operator on Riemannian Manifolds (RO-NORM), which consists of two parts: the unequal-domain encoder/decoder and the same-domain approximator. Motivated by the variable separation in classical modal decomposition, the unequal-domain encoder/decoder uses the pre-computed bases to reformulate the spatio-temporal function as a sum of products between spatial (or temporal) bases and corresponding temporally (or spatially) distributed weight functions, thus the original unequal-domain mapping can be converted into a same-domain mapping. Consequently, the same-domain approximator NORM is applied to model the transformed mapping. The performance of our proposed method has been evaluated on six benchmark cases, including parametric PDEs, engineering and biomedical applications, and compared with four baseline algorithms: DeepONet, POD-DeepONet, PCA-Net, and vanilla NORM. The experimental results demonstrate the superiority of RO-NORM in prediction accuracy and training efficiency for PL-STP.

Controlling machining deformations resulting from unbalanced stress fields inside structural components is a significant challenge in the manufacturing industry. Prediction of machining deformation fields is fundamental for deformation control and requires numerous iterations to optimize the machining process. Conventional prediction methods such as numerical analysis are tailored to a fixed geometry, making them time-consuming and inefficient for components with various geometries. In this study, a general data-driven model is proposed for predicting machining deformation fields in components with varying geometries and stress fields. This model is based on a geometry-oriented neural operator that incorporates global geometry information into the function space, modeling the relationship between the input function (stress fields) and the output function (deformation fields). Global geometric information is extracted using a graph neural network applied to a geometric graph and embedded into the input and output function space through an encoder-query framework. The proposed model achieved low root-mean-squared errors ranging from 0.001 to 0.016 mm, with maximum prediction errors between 0.003and 0.047 mm across different types of components, including beams and frames. The main contribution of this research is the significant advancement in the application of neural operators to the development of general models for predicting machining deformation. The underlying principles of the proposed model provide an important reference for wider applications related to the control of machining deformation in the context of digital and intelligent manufacturing.

Machining deformation caused by residual stress seriously affects the machining accuracy of thin-walled rotary parts, which plays a pivotal role in aerospace manufacturing. It is still a challenge for deformation control due to sufficient information absence, especially for stable process control. In order to address this issue, a new reinforcement learning framework by fusing data, mechanism and causality is proposed, where high control accuracy and stability are expected due to the utilization of multiple source information, especially for the incorporation of causality information, which is quite beneficial for stable decision. The effect and stability of the deformation control method are verified through simulation and actual machining experiments. In the actual experimental verification, the deformation control effect improved by 95.7% with the uncontrolled deformation of the part; the standard deviation of machining error and the peak-valley difference of the proposed method decreased by 67.16% and 46.53%, respectively, which indicates that the dimensions of the parts are more stable by using the method presented in this article.

The residual stress within material is one of the main reasons for machining deformation, and it is more severe for die-forged blanks. A challenge is posed for residual stress field inference by the extremely complex distribution of the residual stress field in die-forged blanks. To address these challenges, this paper proposes a method for inferring the residual stress field in die-forged blanks based on multi-source information fusion. The proposed method is facilitated by fusing local residual stress measurement data near surface layer, simulation residual stress data in bulk component, as well as in-situ deformation force monitoring data dues to residual stress relaxation during machining process. Precise inference of 3D residual stress is realized by introducing the constraints of residual stress solution with in-situ deformation force as well as the relationship among relative variables, which are cooperated in the proposed neural network structure. This method has been validated on 7075 aluminum alloy die-forged blank, in the simulation environment, the mean absolute error of all test samples is as follows: X-direction: 13.89 MPa; Y-direction: 7.37 MPa; Z-direction: 17.11 MPa, and with mean absolute error of 0.09 mm using predicting deformation as evaluation index under the actual machining conditions.

Predicting part machining deformation is vital for optimizing design and manufacturing processes, thereby enhancing the quality and performance of heavy machinery parts. Traditional numerical methods, such as the finite element method, are limited by their computational inefficiency. Furthermore, recent data-driven approaches for predicting machining deformation face challenges due to the complex features and variable geometries of parts throughout design iterations and machining processes. To this end, this paper proposes a method, Voxel-FNO, which rapidly predicts machining deformation for parts with variable feature geometry. This method utilizes the Fourier neural operator to capture the underlying mechanistic relationship between residual stress and machining deformation of parts. Both stress and geometry are sampled by voxel into standard domain before being input into the neural network model. This approach ensures efficiency and applicability, even as part geometries change. The proposed method is verified in both simulation and real environment, demonstrating its accuracy, stability, and generalization capability for varying part geometries, compared to the accurate results from the finite element method. It shows prediction max errors of 0.003 mm, 0.002 mm, and 0.018 mm, and RMSE of 0.0003 mm, 0.0002 mm, and 0.0013 mm for deformations in X, Y, and Z directions, respectively, compared with FEM results.

Solving partial differential equations (PDEs) across varying geometric domains and parameters represents a significant challenge in fields such as materials science, engineering, design and medical imaging, primarily due to the high computing cost associated with recomputing the solution for every change in geometry or parameters. This paper presents a neural operator learning framework for solving PDEs with various domains and parameters, named diffeomorphism neural operator (DNO). The framework transforms the problem of operator learning on varying domains into learning on a generic domain through a diffeomorphic mapping. The efficiency and effectiveness of DNO are validated in experiments that rapidly provide solutions to various PDEs across different domains and parameters. Our method obtains solutions multiple orders of magnitude faster when adapted to changes in shape and size. DNO offers advangates for a broad spectrum of scientific and engineering applications that require dynamic domain and parameter handling. The authors introduce the Diffeomorphism Neural Operator (DNO) to solve partial differential equations across varying domains and parameters by diffeomorphism mapping various physical domains to a generic domain. Experiments demonstrate its efficiency and accuracy, with strong generalization across different shapes and sizes.

Monitoring tool breakage during computer numerical control machining is essential to ensure machining quality and equipment safety. In consideration of the low cost in long-term use and the non-invasiveness to workspace, using servo signals of machine tools to monitor tool breakage has been viewed as the solution that has great potential to be applied in real industry. However, because machine tool servo signals can only partially and indirectly reflect tool conditions, the accuracy and reliability of existing methods still need to be improved. To overcome this challenge, a novel two-step data-driven tool breakage monitoring method using spindle servo signals is proposed. Since spindle cutting torque is acknowledged as one of the most effective and reliable physical signals for detecting tool breakage, it is introduced as the key intermediate variable from spindle servo signals to tool conditions. The monitored spindle servo signals are used to predict the spindle cutting torque in real time based on a long short-term memory neural network, and then the predicted spindle cutting torque is used to detect tool breakage based on a one-dimensional convolutional neural network. The experimental results show that the proposed method can accurately predict the spindle cutting torque for normal tools and broken tools. Compared with the tool breakage monitoring methods that directly use spindle servo signals, the proposed method has higher detection accuracy and more reliable detection results, and the performance is more stable when increasing the detection frequency and decreasing training data.

The residual stress inside a component is one of the main factors affecting its material property and manufacturing quality, such as geometric stability and fatigue life. It is important to understand the characteristics of the residual stress across the volume of a component, referred to as its residual stress field. However, the existing destructive and nondestructive methods cannot accurately and efficiently measure/model the residual stress across the whole component material due to insufficient measurement data. This article presents a novel method for modeling the residual stress field by fusing both local and global measurement data. It is based on the heteroscedastic latent Gaussian process (HLGP), where different measurement variances and the heteroscedasticity of the residual stress field can be addressed. A residual stress field can be modeled as a latent Gaussian process along the global material space, and its different means and variances are controlled by two latent Gaussian processes. This method takes the advantages of existing measurement and inferencing methods to model the residual stress field more accurately and reliably, which was approved by both simulation and actual experimental results on typical structural components, compared with the latent Gaussian process and homoscedastic latent Gaussian process currently used for modeling the residual stress field.

Accurate prediction of machining deformation in structural components is essential for ensuring dimensional precision and reliability. Such deformation often originates from residual stress fields, whose distribution and influence vary significantly with geometric complexity. Conventional numerical methods for modeling the coupling between residual stresses and deformation are computationally expensive, particularly when diverse geometries are considered. Neural operators have recently emerged as a powerful paradigm for efficiently solving partial differential equations, offering notable advantages in accelerating residual stress-deformation analysis. However, their direct application across changing geometric domains faces theoretical and practical limitations. To address this challenge, a novel framework based on diffeomorphic embedding neural operators named neural diffeomorphic-neural operator (NDNO) is introduced. Complex three-dimensional geometries are explicitly mapped to a common reference domain through a diffeomorphic neural network constrained by smoothness and invertibility. The neural operator is then trained on this reference domain, enabling efficient learning of deformation fields induced by residual stresses. Once trained, both the diffeomorphic neural network and the neural operator demonstrate efficient prediction capabilities, allowing rapid adaptation to varying geometries. The proposed method thus provides an effective and computationally efficient solution for deformation prediction in structural components subject to varying geometries. The proposed method is validated to predict both main-direction and multi-direction deformation fields, achieving high accuracy and efficiency across parts with diverse geometries including component types, dimensions and features.

The emergence of wide-bandgap (WBG) semiconductors such as silicon carbide (SiC) and gallium nitride (GaN) has paved the way for a new generation of power electronic devices that can operate at temperatures above 250 degrees C in the sectors such as automotive, aerospace, energy industry. To enable the products embracing such devices high-temperature joining methods will be key to ensure the electrical, thermal and mechanical interconnections. Current solutions such as solder attachment and nano-sintering are finding less attractive as reliable, cost-effective and lead-free alternatives, demanding new materials and processes to meet more stringent requirements. We have herewith successfully fabricated a Cu/Ag multilayered composite preform through the accumulative roll bonding (ARB) for SiC/multilayer/Cu die-attach. With such preform as an inserted interlayer, high performance and reliable interconnections are achievable through bonding at 300 degrees C in ambient atmosphere, where bonding shear strength can be as high as 34.33 MPa. It is found that the formation of Ag hillocks on the surface of the Cu/Ag multilayers primarily contributed to the elevated interfacial adhesion with SiC die and Cu substrate, owing to the stress migration bonding (SMB) process mechanism. By correlating interfacial characteristics of bonded cross-section with performance such shear strength it is possible to identify the optimal bonding temperature and pressure for further reliability enhancements. We firmly believe this has provided a unique solution to enable advanced die-attach interconnects in power module, hence providing a new costeffective manufacturing route in the uptake of WBG devices.

Die-forging structural parts are widely used in the main load-bearing components of aircrafts because of their excellent mechanical properties and fatigue resistance. However, the forming and heat treatment processes of die-forging structural parts are complex, leading to high levels of internal stress and a complex distribution of residual stress fields (RSFs), which affect the deformation, fatigue life, and failure of structural parts throughout their lifecycles. Hence, the global RSF can provide the basis for process control. The existing RSF inference method based on deformation force data can utilize monitoring data to infer the global RSF of a regular part. However, owing to the irregular geometry of die-forging structural parts and the complexity of the RSF, it is challenging to solve ill-conditioned problems during the inference process, which makes it difficult to obtain the RSF accurately. This paper presents a global RSF inference method for the die-forging structural parts based on the fusion of monitoring data and distribution prior. Prior knowledge was derived from the RSF distribution trends obtained through finite element analysis. This enables the low-dimensional characterization of the RSF, reducing the number of parameters required to solve the equations. The effectiveness of this method was validated in both simulation and actual environments.

This review provides a comprehensive analysis of interfacial reactions and the impact of surface metallization in high-temperature die-attach, which is critical for ensuring the reliability of interconnects and joints in power electronic module packaging and integration. With the emergence of high-temperature filler materials, distinctive features in interfacial interactions and microstructural evolution arise, necessitating detailed examination to select suitable surface finishes based on the filler metals and specific applications. Metallization does not always enhance joint quality and reliability, so cost-effectiveness and manufacturability must also be considered when metallization is deemed viable. The formation of intermetallic compounds (IMCs) during interfacial reactions is particularly important, although solid solution formation at interfaces also warrants attention. This review evaluates five commonly used high-temperature metal solder fillers—high-Pb solder, Au-based solder, Bi-Ag solder, Zn-Al solder, and nano Ag paste—focusing on their interactions with various metallized surfaces in die-attach bonding. The effects of metallization on interfacial reactions and bond formation are discussed, leading to recommendations for cost-effective and reliable metallizations suitable for these applications.