This study examines the porosity behavior during the directed energy deposition (DED) of dissimilar metals S45C and H13. We analyzed the effects of deposition parameters, including laser power, feed rate, and powder characteristics, on pore formation, taking into account the unique properties of these metals. Our findings indicate that laser power is the primary factor influencing porosity. At a low power of 200 W, insufficient energy input, along with differences in thermal conductivity and chemical composition between S45C and H13, led to incomplete melting and lack-of-fusion, resulting in high porosity. As the laser power increased to 400-600 W, the melt pool stabilized, enhancing interfacial bonding and significantly reducing porosity. However, at an excessive power of 800 W, rapid melting and solidification of the powder caused gas entrapment and pore formation, which increased porosity, particularly due to the differing thermal conductivities of S45C and H13. Therefore, our results suggest that maintaining an adequate laser power of 400-600 W is essential for achieving a stable melt pool and minimizing porosity in the DED process for dissimilar S45C and H13 metals.

This paper examines the role of generative AI and large language models (LLMs) in advancing intelligent manufacturing as we transition from Industry 4.0 to Industry 5.0. We begin by analyzing the current limitations of rule-based and manufacturing data systems in facilitating flexible, human-centric production. Next, we categorize LLM utilization strategies into three methodological axes: fine-tuning domain-specific models, employing general-purpose models through prompt engineering, and utilizing retrieval-augmented generation (RAG), which includes multimodal RAG that integrates sensor and text data. For each strategy, we present representative case studies across key application areas such as asset management, maintenance intelligence, quality control, process optimization, and knowledge- and document-centric support systems. Concurrently, we explore how information modeling and ontology-based knowledge graphs can be integrated with LLMs to enhance structured manufacturing semantics, improve source traceability, and minimize hallucinations. Finally, we summarize the advantages and limitations of each approach and propose future research directions for human-centric manufacturing, including the development of trustworthy LLM pipelines, standardized data schemas, and closer integration between digital twins and LLM-based decision support systems.

Digital twin technologies in manufacturing have evolved into dynamic, data-synchronized systems that facilitate real-time monitoring and control. Given that machining involves closely interconnected multi-physics behaviors, the effectiveness of a digital twin largely relies on the accuracy and reliability of its underlying process models. This review systematically evaluates three primary paradigms for machining process modeling in digital twins: physics-based, data-driven, and hybrid approaches. Physics-based models provide interpretability and physical consistency but are hindered by high computational costs and limited adaptability to changing conditions. In contrast, data-driven models offer real-time capabilities and adaptive learning but face challenges related to data scarcity and black-box behavior. Hybrid modeling has emerged as the most promising approach, combining physical laws with machine learning through techniques such as parameter correction, physics-guided learning, and state-estimation-based intelligent control. Recent research demonstrates significant advancements in predictive performance, adaptability, and computational efficiency across various machining applications, underscoring the effectiveness of new process modeling strategies for digital twins. However, challenges remain, including multi-physics integration, model reduction for real-time deployment, and autonomous self-updating in data-limited scenarios. The review concludes that hybrid models present the most viable pathway to achieving high-fidelity, self-adaptive, and trustworthy digital twins for autonomous manufacturing.

This study examines a 2kW photovoltaic (PV) support structure, highlighting the vulnerability of conventional metal frames to corrosion and strength degradation in harsh environmental conditions. To overcome these challenges, we propose using pultruded fiber-reinforced polymer (PFRP) members as an alternative structural material. An optimal design framework is established to identify efficient PFRP cross-sections. The study aims to determine lightweight cross-sectional dimensions for box sections (columns and girders) and C-sections (purlins) while maintaining structural safety. We evaluate structural performance using the allowable stress design (ASD) method, incorporating safety factors recommended by the American Association of State Highway and Transportation Officials (AASHTO). Finite element analysis (FEA) assesses critical design constraints, including buckling, material failure, and serviceability deflection limits. From the feasible designs, we select the lightest cross-sectional configuration that meets all safety requirements. The results demonstrate that PFRP members can significantly reduce weight while ensuring structural safety, thus validating their potential as an alternative to conventional metal photovoltaic support structures.

Flexible electronics are becoming the next generation of devices due to their advantages, such as mechanical flexibility, eco-friendliness, large-area applicability, and scalability for mass production. However, solution-based manufacturing processes are prone to defects like discontinuities and local smudging, which can significantly degrade both device quality and yield. To tackle these challenges, rapid and accurate defect classification is crucial for real-time diagnosis during manufacturing. This study investigates the impact of data scale and key training hyperparameters on the performance of deep learning–based defect diagnosis models, using a dataset of conductive pattern defects in flexible electronics. We specifically examine how the number of training images affects model accuracy and generalization, and we analyze how adjustments to hyperparameters—such as L2 regularization and dropout—influence model performance in data-limited scenarios. Our findings offer insights into optimal training strategies tailored to different data scales and learning constraints, providing practical guidelines for designing and developing AI-based defect diagnosis models for flexible electronic devices.

This paper presents a tiltable cable-suspended aerial manipulation (SAM) system designed to improve the utility of aerial manipulators in industrial settings. Although drone-robot arm systems have shown promise, suspended configurations encounter notable stability challenges, particularly during inclined operations. To tackle these challenges, we performed simulation-based analyses focusing on the system's kinematics, dynamic response, and thrust requirements under tilted conditions. We utilized Monte Carlo sampling and forward kinematics to assess the workspace and manipulability. The findings indicated that each propeller needs to generate over 32 N of thrust to maintain stable control. Additionally, simulation experiments showed that the system can uphold its attitude and execute end-effector motions effectively, even in the presence of disturbances. This study establishes a foundational verification step toward developing a physical SAM system capable of safe and robust operation in inclined scenarios.

Military shelters contain various electronic devices that generate significant heat during operation due to their high power output. This heat buildup can degrade the performance of the equipment and shorten its operational lifespan. In high-temperature environments, overheating can lead to serious malfunctions in communication systems or information management platforms, jeopardizing the efficiency and reliability of military operations. Conversely, in low-temperature or high-humidity conditions, condensation may form inside the shelter, increasing the risk of physical damage to electronic components. Such damage can significantly compromise the reliability and durability of the equipment, raising the likelihood of system failure. This study proposes using various environmental control systems, including heating, ventilation, and air conditioning (HVAC) units and air ducts, to mitigate the adverse effects of temperature and humidity fluctuations within military shelters. To achieve this, thermal analysis models were utilized to evaluate and verify the performance of these systems. The analysis specifically examined the heat output of individual devices to determine if the proposed control systems could effectively maintain optimal operating temperatures within the shelter. The results of this study aim to provide a valuable foundation for designing environmental control systems that ensure thermal stability in military shelters.

In the field of gimbal targeting systems, image error tracking plays a crucial role in various applications, including object detection, enemy surveillance, and aircraft inspection. Enhancing image tracking performance presents a significant challenge due to singularity issues at the azimuth and elevation joints. To tackle this problem, this paper proposes a rotation-matrix-based tracking error compensation method centered on real-time object tracking. Specifically, our approach involves creating a virtual rotation frame that aligns the visual tracking frame with the gimbal base frame. Using our method, a gimbal with two degrees of freedom (DOF) can achieve superior tracking performance near the ±90° joint positions compared to conventional gimbal tracking methods. We compare the proposed method with existing approaches in the literature by assessing initial pose RMS error and singular pose RMS error through MATLAB Simscape simulations. The experimental results demonstrate that our method can reduce the line of sight RMS error by 89% in the azimuth position and by 99% in the elevation position, respectively.

The design of the extrusion die significantly affects both the extrusion process and the quality of multi-lumen tubes. Traditional design methods that rely on trial and error tend to increase manufacturing time and costs while diminishing product quality. This study utilizes inverse extrusion simulation and optimization to design the extrusion die without the need for trial and error. The inverse extrusion simulation generates the die profile necessary to achieve the desired extrudate shape. Subsequently, direct extrusion simulations are conducted to predict the extrudate profile based on the derived die. The optimal volumetric flow rates of air within the lumens are also identified to ensure the extrudate meets the target profile. The results from the direct extrusion simulation, combined with optimization, confirm that the designed extrusion die can successfully produce the target profile. Using the derived die, the multi-lumen tube with the desired specifications is successfully extruded. This design and manufacturing approach enhances both the quality and productivity of multi-lumen tubes.

The practical application of Raman spectroscopy is often constrained by its low signal sensitivity, particularly for low-concentration liquid samples. This study introduces a straightforward platform that enhances Raman signals by physically concentrating analytes, providing an alternative to complex substrate fabrication and chemical treatments. We employed a femtosecond pulse laser to create functional micro-grid patterns on a silicon (Si) substrate. This laser process induces localized ablation and simultaneous oxidation, resulting in three-dimensional, hydrophilic microstructures of nonstoichiometric silicon oxide (SiO_2-x). These grid structures effectively confine aqueous sample droplets through a pinning effect, functioning as a microwell array that traps and concentrates suspended polystyrene (PS) particles. This physical concentration mechanism achieved a notable signal enhancement, with a maximum factor of 5.2 for PS particles, without the need for sample dehydration. This work presents a simple, cost-effective, and highly reproducible alternative to conventional SERS for analyzing low-concentration liquid samples, demonstrating strong potential for integration into microfluidic systems.