The wind power prediction contributes significantly to the reliable operation of offshore wind turbine systems and the efficient scheduling of integrated power grids. However, the intrinsic characteristics of offshore wind energy make the precise prediction of wind power remain a complex issue, and existing approaches still show some disadvantages when facing large-scale data. To cope with this problem, a novel deep learning (DL) algorithm called Informer is applied to the ultra-short-term offshore wind power prediction in this research. The Informer can extract features and capture the sequence dependency more effectively from long time-series data. In addition, the discrete wavelet transform (DWT) and improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) are adopted for data preprocessing to achieve more precise prediction. The performance of the Informer-based approach is analyzed and contrasted with that of long short-term memory (LSTM)-based, DLinear-based, and Transformer-based benchmarks. The comparison is conducted for varying lengths of input sequence prediction and multi-time steps ahead prediction. Results show that the proposed approach is superior to the benchmarks for all diverse instances studied in this research.
The growing demand for energy-efficient transportation has significantly impacted the global automotive industry. Consequently, the traditional internal combustion engine vehicle market is shifting towards energy-efficient vehicles such as electric vehicles (EVs), fuel cell electric vehicles (FCEVs), and hybrid electric vehicles (HEVs). To minimize energy consumption and develop optimal driving strategies, eco-friendly vehicles must maximize energy efficiency while maintaining the same travel time for the same route. Cruise control, which maintains a constant driver-set speed, is not suitable for achieving optimal energy performance in various driving environments. To address this, we conducted a study on training an eco-driving control using Reinforcement Learning (RL) with a vehicle simulator. The EV was modeled and simulated in MATLAB/Simulink based on longitudinal dynamics, considering road gradients. RL agents were trained to minimize energy consumption by optimizing the vehicle driving policy. Especially, both model-free reinforcement learning, including the Deep Q-network (DQN) and the Deep Deterministic Policy Gradient (DDPG), and Model-based Reinforcement Learning (MBRL) based on Q-learning, were applied to evaluate and compare their performance. Furthermore, a time-based comparative evaluation of model-free reinforcement learning and MBRL was conducted to analyze how differences in learning stability and convergence behavior affect the energy efficiency of the EV simulator. The results show that both RL methods improve energy efficiency compared to cruise control, with MBRL achieving the highest performance by utilizing an accurate environment model, while model-free reinforcement learning demonstrates practical effectiveness even without such a model.
This work examines how joint design–controlled dilution governs microstructural evolution and titanium carbide morphology in GTAW cladded hardfacing layers produced using recycled Ti6Al4V chips and graphite. Hardfacing coatings were deposited into V-groove geometries of varying depths alongside flat surface configurations to investigate the effect of joint geometry on deposit composition and macrostructure. To account for concave bead profiles, a refined mathematical model tailored to V-groove cross-sections was implemented, providing an effective practical alternative to relate joint preparation to dilution. Microstructural and phase characterizations confirmed a nearly complete conversion of titanium into TiC within a ferritic/martensitic matrix, alongside cementite formation in the synthesized composite hardfacing layer. Elemental analysis revealed that vanadium dissolves into the carbide lattice forming a (Ti, V)C solid solution, whereas aluminium partitions entirely into the metallic matrix as a solid solution element. Without secondary phase formation at these concentration levels, the ternary Fe–Ti–C phase diagram remains a valid thermodynamic framework for accurately predicting TiC morphological evolution during solidification. Increasing flux mass reduced dilution and altered weld pool chemistry, driving a clear morphological transition from fine, dispersed TiC particles at high dilution to coarse dendritic structures under low dilution. Elevated titanium concentrations detected within the matrix under low-dilution conditions confirm high solute availability during solidification, which actively promoted lateral growth above the system liquidus line.
Ultra‑precision machining (UPM) is essential for producing components with sub‑micron accuracy and superior surface finish, yet its industrial adoption is constrained by low efficiency, complex stability management, and insufficient intelligent monitoring. To address these challenges, this study develops a holistic digital twin framework for UPM. The architecture is organized into three layers: physical, communication, and service. MTConnect provides standardized data exchange across these layers, while mechanistic modelling is combined with machine learning to enable real‑time monitoring, adaptive control, and intelligent process planning. The intelligent stability management strategy incorporates a transfer‑learning‑based chatter detection model with an Elastic Weight Consolidation (EWC)-based continuous learning algorithm to adaptively update the stability boundary (SLD) to reflect evolving machining states. Experimental validation on a micro‑milling platform confirmed the effectiveness of the framework, with the chatter detection model achieving 96.6
In this article, an intelligent method for optimizing rotation speed and output power of a micro pneumatic turbine is proposed by combining radial basis function neural network (RBFNN), non-dominated sorting genetic algorithm Ⅱ (NSGA Ⅱ) and transfer learning (TL). Firstly, design of experiments and Latin hypercube sampling are adopted to generate structure parameters of micro pneumatic turbine. Then, actual conditions driving simulation (ACDS) is used to calculate turbine performance for constructing dataset. Thirdly, RBFNN is trained to predict turbine performance rapidly at one supply pressure. The optimal turbine structure parameters for the maximum rotation speed and output power are obtained by using NSGA Ⅱ based on the RBFNN. Furthermore, TL is used to extend the trained RBFNN (TL RBFNN) for obtaining the optimal turbine parameters at other supply pressures. The results demonstrate that rotation speed and output power of the optimized micro pneumatic turbine increase by 14.82
Laser direct writing has emerged as a scalable and maskless fabrication strategy for flexible transparent electrodes (FTEs). However, in conventional single-step laser patterning, the optical transparency of metal-mesh electrodes is fixed once the mesh pitch is defined, and attempts to locally increase transparency through laser ablation (LA) often disrupt the conductive percolation network, resulting in electrical degradation. Here, we present a laser dual patterning (LDuP) approach that sequentially integrates laser-induced reductive sintering (LRS) and laser ablation (LA) to enable spatially programmable transparency modulation while preserving electrical continuity. A 532 nm continuous-wave laser induces photothermal reduction and sintering of solution-processed NiOx nanoparticles under ambient conditions to form continuous Ni mesh electrodes. Subsequently, a 355 nm nanosecond pulsed laser introduces controlled sub-line-width perforations within the preformed Ni electrodes. By maintaining the ablation diameter smaller than the electrode line width, the global conductive percolation network remains intact while the effective open-area fraction increases. Systematic investigation of the LA behavior reveals distinct ablation mechanisms between the NiOx nanoparticle film and the reductively sintered Ni electrodes, arising from differences in optical absorption, thermal transport, and microstructure. These material-dependent interactions enable selective perforation of Ni electrodes without complete disruption of the conductive network. The perforated electrodes show improved optical transparency with a moderate increase in sheet resistance and demonstrate the robust retention of global electrical connectivity through preserved conductive pathways. The entirely maskless and vacuum-free process provides a precision-engineered and environmentally favorable route for fabricating Ni-based FTEs with tunable optical functionality and maintained global electrical connectivity.
In material extrusion (MEX) based additive manufacturing process, weak bonding between infill lines and layers poses a significant challenge, often leading to structural defects. In this study, we present a novel method for layer-by-layer defect detection and selective repair within the MEX process using object detection model. This method involves training the object detection model on images of printed surfaces using an in-situ image acquisition system. Using the trained model, This paper presents an algorithm to generate compensation G-code that directly targets detected defects during the printing process, which can reduce compensation time and material consumption. While conventional MEX processes frequently result in numerous unfilled gap defects when printing composite, the developed compensation strategy significantly enhanced the gap filling by automatically detecting and selectively repairing the defective areas. Furthermore, data augmentation through simple rotation of the training data enabled the recognition of defects at diverse raster angles, demonstrating the adaptability of our system across various printing geometries and paths. Comparative analysis of mechanical properties under identical conditions revealed that the structures printed with our compensation method exhibited a notable increase in maximum tensile strength by approximately 45
The titanium alloy tri-flow impeller is widely employed in aerospace, energy, and other high-performance applications; however, its machining process faces challenges such as high energy consumption, rapid tool wear, and low efficiency. For this reason, this study develops a multi-objective prediction model and parameter-adaptive optimization strategy for titanium alloy tri-flow impeller milling, incorporating error compensation for varying tool wear conditions. First, the factors influencing high-efficiency and low-carbon machining under multi-factor coupling were analyzed, resulting in the establishment of mechanistic models that describes energy consumption, tool wear, and machining efficiency based on multiple process parameters. Second, a data-driven model was constructed using a backpropagation neural network (BPNN) to capture discrepancies between experimental and mechanistic model predictions for energy consumption, tool wear, and machining efficiency, achieving prediction accuracies of 97.18
Titanium alloys are widely employed in aerospace and medical applications due to their excellent mechanical properties. Surface microstructuring has been demonstrated as an effective approach to enhance its functional performance. Fabrication of microstructure surfaces on TC6 alloy remains challenging due to the limitations in efficiency and precision. In this work, surface micro-structuring of TC6 based on laser structured wheel was proposed to address the problems. Initially, a nanosecond pulsed laser was employed to precisely fabricate the abrasive units on the CBN wheel. Subsequently, a microstructure simulation model integrating the topographies of laser structured wheel and dynamic contact mechanics was developed to simulate the generated surface morphologies and grinding forces. In the theoretical model, the wheel topography was reconstructed by convolution operations to account for grain randomness and wear evolution, and the grain-workpiece interactions, including rubbing, plowing, chip formation and pile-up effects were further analyzed to achieve accurate prediction of microstructure topographies and grinding forces. Finally, the wear of the structured wheel and its influence were quantitatively investigated. The theoretical model and experimental results provide guidelines for designing structured wheels and optimizing grinding parameters in high efficiency and precision manufacturing of microstructures.
Wound-rotor synchronous motors (WRSMs) have attracted renewed attention as rare-earth-free alternatives to permanent-magnet motors, offering flexible flux control and robust thermal performance. In particular, air-cooled WRSMs are promising for compact and energy-efficient drive systems; however, their thermal behavior under natural convection has not been sufficiently investigated through experimentally validated studies. This paper presents a comprehensive thermal analysis and experimental validation of a 1.5 kW open-type WRSM operating under natural convection conditions. A dual modeling framework combining a lumped-parameter thermal equivalent circuit (TEC) and three-dimensional transient finite element method (FEM) was developed. The experimentally measured core and stray losses were applied as heat sources in both models to ensure consistent boundary conditions. Temperature measurements were conducted over a 7200-s operating period using K-type thermocouples installed at representative locations on a fabricated prototype to enable direct comparison with numerical predictions. Both TEC and FEM accurately demonstrated the transient and steady-state thermal behavior of the WRSM. The FEM provides higher accuracy and detailed spatial temperature distributions, while the TEC model achieves comparable prediction accuracy with substantially reduced computational effort. Experimental validation confirmed that natural air cooling is sufficient to maintain winding temperatures below the insulation class limit at the investigated power level. These rotor-side temperatures represent model-based estimates anchored by validated measurements at the stator and housing boundaries. The proposed modeling approach provides practical guidance for selecting appropriate thermal analysis tools at different design stages and supports the development of simplified, resource-efficient, and environmentally conscious electric drive systems based on air-cooled WRSMs.
This study presents the design and preliminary evaluation of an ultra-lightweight carbon fiber reinforced polymer (CFRP) wheel prototype fabricated using optimized continuous fiber winding paths. Seven winding-path configurations were first evaluated through compression tests using single trapezoidal specimens. Based on the maximum load and load-to-weight ratio, Path F was selected and applied to a full-scale CFRP wheel prototype. The final prototype weighed 1.6 kg, including the CFRP structure, residual PLA support molds, milled-carbon-fiber bolt-circle region, epoxy resin, and EVA contact layer. The full wheel sustained a maximum static compression load of 7155.9 N and supported a stationary compact vehicle without visible structural failure. The results demonstrate the preliminary feasibility of the proposed CFRP wheel concept under static loading conditions. These results establish a promising foundation for further development and validation of the proposed CFRP wheel under representative automotive loading and environmental conditions.
Energy efficiency is now a critical constraint in precision engineering, where control-induced dissipation affects power draw, thermal loading, and mechanical wear. This review reframes the sliding mode control (SMC) robustness–precision trade-off as a structural consequence of how dissipation is localized in closed-loop dynamics. By interpreting sliding actions as dissipation-shaping mechanisms, classical, finite-time, and high-order schemes are unified as nonlinear damping realizations with distinct dissipation geometries. This framework identifies an intermediate gap in current SMC formulations where robustness and precision frequently conflict. Beyond reinterpretation, this paper discusses design considerations for reconciling these objectives by treating mechanical, electrical, and virtual damping as integrated control resources. These insights provide a roadmap for tailoring dissipation profiles, enabling physically consistent and energy-aware robust control for sustainable manufacturing systems.
This study proposes a frequency-dependent modeling framework to analyze tension-induced thickness deviations and their viscous–capillary attenuation behavior in roll-to-roll (R2R) slot-die coating processes, with emphasis on process stability and material sustainability. A lubrication-theory-based governing equation is formulated to describe the evolution of thickness deviations generated by periodic web tension disturbances. Linear stability analysis reveals that the attenuation rate scales with the fourth power of the disturbance wave number, leading to rapid suppression of high-frequency variations while low-frequency disturbances persist along downstream web transport. From a sustainability perspective, the results indicate that uncontrolled low-frequency tension disturbances can significantly increase coating nonuniformity, thereby elevating material waste and reprocessing energy consumption. Experimental validation using an industrial R2R platform confirms the predicted frequency-dependent attenuation behavior and demonstrates that disturbance frequency management can effectively reduce thickness variability. The proposed frequency-aware framework provides quantitative insight into disturbance attenuation dynamics and offers a systematic basis for designing energy-efficient and waste-minimizing R2R coating operations.
Injection molding is one of the most widely used polymer manufacturing processes and is associated with substantial energy consumption, estimated at approximately 30 billion kilowatt-hours (kWh) annually in the United States. Improving the ability to predict and monitor energy usage is therefore an important step toward more sustainable manufacturing. This study investigates three predictive modeling approaches, namely physics-based (PB), data-driven (DD), and hybrid models, for estimating energy consumption in the injection molding process. The PB model captures thermodynamic and mechanical energy contributions, while the DD model learns empirical relationships from experimental data. Building on the complementary strengths of these approaches, a hybrid modeling framework is introduced that integrates PB and DD predictions via a combination of a hidden Markov model (HMM) and a probability-informed rule-based decision-making (PRBDM) mechanism. This framework is introduced as a general methodology for energy prediction in manufacturing systems with uncertainty and variability, in which predictions switch between DD and PB results according to specific rules or probabilistic latent-state inference of the process. To demonstrate the framework, experimental data from a comprehensive set of injection molding experiments involving multiple materials, processing conditions, and equipment configurations were used to train and evaluate the models. On a held-out test partition, the hybrid framework achieved a coefficient of determination (R²) of 0.6639 and a prediction accuracy of 85.51
Ice accumulation on surfaces remains a challenge across various outdoor infrastructures. However, many structure-based anti-icing surfaces still suffer from limited durability and poor de-icing efficiency after complete ice formation. Here, we present a scalable anti-icing surface based on a Carbon–PDMS composite layer incorporating 100 μm microhole patterns infused with butter as a confined phase-change lubricant. In this design, carbon serves as a photothermal layer enabling rapid solar-to-thermal conversion under 1 sun irradiation, while the microhole structures provide superhydrophobicity and confined lubricant reservoirs. Butter acts as a phase-change lubricant that destabilizes ice nucleation and forms a slippery interfacial layer upon melting, reducing ice–substrate interactions. As a result, the surface exhibited delayed ice formation and accelerated de-icing under both no-light and light conditions. Under no-light conditions, the freezing delay reached 209 s, while the de-icing time decreased to 196 s. Under 1 sun irradiation, photothermal activation further extended the freezing delay to 437 s and reduced the de-icing time to 132 s. In addition, the butter-infused surface exhibited a low ice adhesion strength of 1.7 kPa, indicating effective mitigation of interfacial anchoring. The enhanced performance is attributed to the combined effects of air entrapment, suppressed heterogeneous nucleation, photothermal heating, and phase-change lubrication. Moreover, the microhole structures retained butter within the patterned regions after repeated melting–solidification cycles, supporting stable operation. Overall, this work provides a practical design strategy for multifunctional anti-icing surfaces integrating structural control, photothermal de-icing, and slippery lubrication.
Thermal damage susceptibility in titanium alloys during abrasive waterjet machining (AWJM) significantly compromises surface integrity. This study systematically investigates the pivotal role of initial β phase content in regulating thermal response and microstructural evolution during AWJM. Through a comparative analysis of Ti-6Al-4 V and TC18 alloys, real-time thermal monitoring and multi-scale characterization were employed to reveal distinct thermomechanical behaviours. Results demonstrate that the β-rich TC18 alloy exhibited a 14.11
Lithium-ion batteries offer high energy density and stability, but suffer performance degradation under varying operating conditions. The solid electrolyte interphase (SEI), formed by electrolyte decomposition, plays a critical role in such degradation. Although SEI formation predominantly occurs during charging, the impact of discharge conditions remains unclear. In this work, we quantify the discharge-rate dependence of SEI resistance by electrochemical impedance spectroscopy (EIS) and assess its effect with single-cycle simulations. Capacity fade was most pronounced at 1 C discharge, showing a 32
X22CrMoV12-1 refractory heat-resistant high-strength stainless steel is primarily used in the machining and manufacturing of heavy-duty gas turbine compressor blades. To address the demand for green finishing of blades, this paper conducted cutting experiments on X22CrMoV12-1 steel under dry cutting, minimum quantity lubrication (MQL), cryogenic CO₂, and the Supercritical CO2 based minimum quantity lubrication (CMQL) process. Compared to dry cutting, CMQL demonstrated 52.2
In advanced heterogeneous packaging, interposers serve as key components, integrating multiple chiplets onto a single package substrate through high-density signal redistribution. However, incorporating high-CTE metallic interconnects within low-CTE dielectrics and substrates inevitably induces significant warpage during thermal cycling. This deformation poses a severe threat to mechanical reliability, potentially leading to interconnect failure and yield loss, particularly as package dimensions increase while bump pitches and silicon die thicknesses scale down. While accurate warpage prediction is crucial for robust design, conventional finite element analysis (FEA) faces prohibitive computational challenges due to the extreme geometric scale mismatch between fine routing features and the global package structure. To address this challenge, this study proposes a deep learning–accelerated multiscale framework for rapid and accurate warpage prediction of interposer routing interconnects. The proposed approach establishes a warpage prediction pipeline by representing the spatial heterogeneity of complex routing patterns through equivalent properties. Validation against full routing-level simulations confirmed that the proposed method precisely predicts warpage behavior under optimal partitioning conditions. Furthermore, a deep learning surrogate model was employed to replace the iterative FEA-based homogenization process. The trained model enables rapid property estimation while maintaining high physical consistency with routing geometries. Consequently, the integrated methodology achieves a 2,643-fold speedup in analysis time while preserving the accuracy of full routing-level simulations. Ultimately, this framework provides a scalable and resource-efficient simulation strategy that minimizes computational overhead for the precision design of advanced packaging.
The fatigue reliability of plate heat exchangers (PHEs) is increasingly governed by brazed joints as ultra-thin stainless steel (STS) plates are adopted to enhance efficiency. However, the peel-mode fatigue behavior of copper-brazed joints, particularly under conditions combining ultra-thin geometry, micro-scale fillet features, and Mode-I loading, remains largely unexplored. This study presents a systematic characterization of peel fatigue in 316 L (50 μm)–Cu (12 μm)–316 L (50 μm) brazed joints through an integrated experimental and finite element framework. Two distinct fatigue regimes were identified: high-cycle fatigue (50–100 N; 104-105 cycles) characterized by stable crack growth within the copper layer, and low-cycle fatigue (150–200 N; 103-104 cycles) dominated by plastic deformation and accelerated crack initiation. Three crack propagation modes were observed, with copper-confined propagation dominating at lower loads. Finite element analysis considering fillet radii of 4–9 μm revealed a strong geometry-dependent stress response, where local stress increased from 205 to 217 MPa (6
