Large models, exemplified by ChatGPT, have reached the pinnacle of contemporary artificial intelligence (AI). However, they are plagued by three inherent drawbacks: excessive training data and computing power consumption, susceptibility to catastrophic forgetting, and a deficiency in logical reasoning capabilities within black-box models. To address these challenges, we draw insights from human memory mechanisms to introduce “machine memory,” which we define as a storage structure formed by encoding external information into a machine-representable and computable format. Centered on machine memory, we propose the brand-new machine memory intelligence (M2I) framework, which encompasses representation, learning, and reasoning modules and loops. We explore the key issues and recent advances in the four core aspects of M2I, including neural mechanisms, associative representation, continual learning, and collaborative reasoning within machine memory. M2I aims to liberate machine intelligence from the confines of data-centric neural networks and fundamentally break through the limitations of existing large models, driving a qualitative leap from weak to strong AI.
As pivotal supporting technologies for smart manufacturing and digital engineering, model-based and data-driven methods have been widely applied in many industrial fields, such as product design, process monitoring, and smart maintenance. While promising, both methods have issues that need to be addressed. For example, model-based methods are limited by low computational accuracy and a high computational burden, and data-driven methods always suffer from poor interpretability and redundant features. To address these issues, the concept of data-model fusion (DMF) emerges as a promising solution. DMF involves integrating model-based methods with data-driven methods by incorporating big data into model-based methods or embedding relevant domain knowledge into data-driven methods. Despite growing efforts in the field of DMF, a unanimous definition of DMF remains elusive, and a general framework of DMF has been rarely discussed. This paper aims to address this gap by providing a thorough overview and categorization of both data-driven methods and model-based methods. Subsequently, this paper also presents the definition and categorization of DMF and discusses the general framework of DMF. Moreover, the primary seven applications of DMF are reviewed within the context of smart manufacturing and digital engineering. Finally, this paper directs the future directions of DMF.
Machining high-performance engineering materials, faces challenges including low machining efficiency, poor workpiece surface integrity, and rapid tool wear, which restrict high quality and efficient machining. Ultra-high-speed machining (UHSM) has been expected to address these issues. However, the material removal mechanisms involved in UHSM remain unclear and need further exploration. This paper reviews the criteria for evaluating the ductile and brittle behaviors of high-performance materials subjected to machining, as well as the developmental history of the material’s ductile-brittle transition induced by machining, proposing the concept of relativization of ductile-brittle property. Additionally, it further summarizes three typical material removal mechanisms: ductile-mode removal based on shear stress, brittle-mode removal based on tensile stress, and extrusion removal based on compressive stress, clarifying the universality of the brittle-mode removal. On this basis, this paper focuses on the discussion of the material removal mechanisms in UHSM, including high strain-rate-induced material embrittlement, UHSM-induced skin effect of damage, and the thermal effect in UHSM. Furthermore, it provides a detailed description of the typical characteristics of chip morphology in the ductile-brittle transition region (DBTR) under the high strain rate condition and, for the first time, elucidates the material removal mechanisms in the DBTR from a microstructural dislocation perspective, enriching the basic theory of UHSM. In the discussion section, it standardizes the definition for the UHSM, and explores the dislocation movement at high strain rates and the crack propagation in the UHSM. Finally, based on the current status of the UHSM technology, it summarizes the relevant research hotspots. For the first time, this paper brings up the brittle-mode removal mechanism under ultra-high-speed conditions, which is helpful to promote the UHSM for industrial applications.
Milling force is key to the understanding of cutting mechanism and the control of machining process. Traditional milling force models have limited prediction accuracy due to their simplified conditions and incomplete knowledge contained for model construction. On the other hand, due to the lack of guidance from physics, the data-driven models lack interpretability, making them challenging to generalize to practical applications. To meet these difficulties, a deep network model guided by milling dynamics is proposed in this study to predict the instantaneous milling force and spindle vibration under varying cutting conditions. The model uses a milling dynamics model to generate data sets to pre-train the deep network and then integrates the experimental data for fine-tuning to improve the model’s generalization and accuracy. Additionally, the vibration equation is incorporated into the loss function as the physical constraint, enhancing the model’s interpretability. A milling experiment is conducted to validate the effectiveness of the proposed model, and the results indicate that the physics incorporated could improve the network learning capability and interpretability. The predicted results are in good agreement with the measured values, with an average error as low as 2.6705%. The prediction accuracy is increased by 24.4367% compared to the pure data-driven model.
Gecko-inspired van der Waals force-based adhesion technology demonstrates significant potential for robotic operations. While superior adhesion is achieved under parallel contact during testing, engineering operations often involve non-parallel contact, weakening adhesion, and compromising task stability and efficiency. Stable attachment under such non-parallel contacts remains challenging. Inspired by the soft muscle and rigid bone in the gecko’s sole, this study proposes a self-adaptive core-shell dry adhesive by embedding a thin, rigid piece into a soft, thick elastomer comprising a top adhesion tip with a mushroom-like geometry for interfacial adhesion based on the van der Waals force and a bottom core-shell configuration for interface stress regulation. Unlike traditional core-shell structures with a fixed “dead core,” the proposed “live core” rotates within the soft shell, mimicking skeletal joints. This enables stress equalization at the interface and facilitates adaptive contact to macroscopic interfacial angle errors. This innovative core-shell configuration demonstrates an adhesion strength 100 times higher than conventional homogeneous structures under non-parallel contact and offers anti-overturning ability by mitigating torsional effects. The proposed strategy can advance the development of gecko-inspired adhesion-based devices and systems.
Human skin exhibits a remarkable capability to perceive contact forces and environmental temperatures, providing complex information that is essential for its subtle control. Despite recent advancements in soft tactile sensors, accurately decoupling signals—specifically separating forces from directional orientation and temperature—remains a challenge thus resulting in failure to meet the advanced application requirements of robots. This study proposes, F3T, a multilayer soft sensor unit designed to achieve isolated measurements and mathematical decoupling of normal pressure, omnidirectional tangential forces, and temperature. We developed a circular coaxial magnetic film featuring a floating mount multilayer capacitor that facilitated the physical decoupling of normal and tangential forces in all directions. Additionally, we incorporated an ion gel-based temperature-sensing film into the tactile sensor. The proposed sensor was resilient to external pressures and deformations, and could measure temperature and significantly eliminate capacitor errors induced by environmental temperature changes. In conclusion, our novel design allowed for the decoupled measurement of multiple signals, laying the foundation for advancements in high-level robotic motion control, autonomous decision-making, and task planning.
Extreme environments are unstructured and change rapidly, making human exploration in unfamiliar areas difficult. Construction robotics can help reduce risks to human safety and property in these environments by integrating digital technology and artificial intelligence. This technology has the potential to significantly improve the quality and efficiency of construction, making it a key area for future research. Extreme environments include hazardous work sites, polluted areas, and harsh natural conditions. Our review of construction robotics in these settings highlights several knowledge gaps. We focused on four main areas: mechanism design, perception, planning, and control. Our analysis reveals challenges in practical applications, such as creating adaptable mechanisms, accurately perceiving changing environments, planning for unstructured sites, and optimizing control models. Future research should explore: biomimetic designs inspired by nature, multimodal data fusion for perception, adaptive planning strategies, and hybrid control models that combine data-driven and mechanism-based approaches.
Drug research and development (R&D) plays a crucial role in supporting public health. However, the traditional drug-discovery paradigm is hindered by significant drawbacks, including high costs, lengthy development timelines, high failure rates, and limited output of new drugs. Recent advances in micro/nanotechnology, along with progress in computer science, have positioned microfluidics and artificial intelligence (AI) as promising transformative tools for drug development. Microfluidics offers miniaturized, multiplexed, and versatile platforms for high-dimensional data acquisition, while AI enables the rapid processing of complex, large-scale microfluidic data; together, they are accelerating a paradigm shift in the drug-discovery process. This paper first outlines the mainstream microfluidic strategies and AI models used in drug R&D. It then summarizes and discusses real-world applications of the integrated use of these technologies across various stages of drug discovery, including early drug discovery, drug screening, drug evaluation, drug manufacturing, and drug delivery systems. Finally, the paper examines the main limitations of microfluidics and AI in drug R&D and offers an outlook on the future convergence of these technologies.
Clean hydrogen (H2) is highly desirable for the sustainable development of society in the era of carbon neutrality. However, the current capability of water electrolysis and steam methane (CH4) reforming to produce green and blue H2 is very limited, mainly due to the high production cost, difficult scale-up technology, or operational risk. Here, we propose the direct catalytic decomposition of diesel using a nano-Fe-based catalyst to produce the so-called “jadeite H2,” while simultaneously fixing the carbon from the diesel in the form of carbon nanotubes (CNTs). Efforts are made to understand the suppression mechanism of the CH4 byproduct, such as by tuning the catalyst type, space velocity, and reaction time. The optimal green index (GI)—that is, the molar ratio of H2/carbon in a gaseous state—of the proposed technology exceeds 42, which is far higher than those of any previously reported chemical vapor deposition (CVD) method. Moreover, the carbon footprint (CFP) of the proposed technology is far lower than those of grey H2, blue H2, and other dehydrogenation technologies. Compared with most of the technologies mentioned above, the energy consumption (per mole of H2) and reactor amplification of the proposed technology validate its high efficiency and great practical feasibility.
A promising way to realize controlled nuclear fusion involves the use of magnetic fields to control and confine the hot plasma configuration. This approach requires superconductor magnets operating above 15 T for the next generation of fusion devices. Due to their high in-field transport current capacity, rare-Earth barium copper oxide (REBCO) coated conductors are promising materials for manufacturing of cable-in-conduit conductors (CICCs) for fusion. However, the high-aspect-ratio geometry makes it difficult to find a multi-tape CICC configuration that fulfills the high engineering current density requirements while retaining enough flexibility for winding large-scale magnets. Moreover, the multilayer structure and inherent brittleness make the REBCO tapes susceptible to degradation during CICC manufacturing and operation. For more than a decade, the development of a reliable REBCO-based CICC that can sustain the huge combined mechanical, thermal, and Lorentz loads without degradation has been ongoing, albeit with limited progress. In this paper, we report on a prototype REBCO CICC that can withstand an applied cyclic Lorentz load of at least 830 kN·m−1, corresponding to a transport current of 80 kA at 10.85 T and 4.5 K. To our knowledge, this is the highest load achieved to date. The CICC uses 288 tapes wound into six strengthened sub-cables, making it capable of having a current sharing temperature, Tcs, of around 39 and 20 K when operated under 10.85 T with a current of 40 and 80 kA, respectively. Scaled to a 20-T peak field and 46.5-kA transport current, this provides a temperature margin of over 10 K with respect to an operating temperature of 4.5 K. In addition, no perceptible transport current performance degradation was observed after cyclic Lorentz loading, cyclic warm-up/cool-down (WUCD), and quench campaigns. The proposed REBCO CICC is a milestone in the development of high-temperature superconductors for large-scale and high-field magnet applications.
Polyacrylonitrile-based commercial carbon fibers (CFs) have garnered significant attention in mechanical applications because of their exceptional mechanical properties. However, their functional versatility relies heavily on the structural intricacies of duplex carbon layers. Current modification approaches, though effective, are encumbered by complexity and cost, limiting widespread adoption across diverse fields. We herein present a straightforward modification strategy centered on regulating carbon layers to unlock the multifunctional potential of CFs. Our method leverages two common anions, Cl− and SO42−, to facilitate oxidation reactions in CFs under robust alkali and high voltage conditions. Cl− effectively activates carbon layers, while SO42− facilitates layer movement. The electrocatalytic activities of the resultant CFs are enhanced, with state-of-the-art performance as supercapacitors and exceptional stability. Moreover, our approach achieves a groundbreaking milestone by bending and fusing CFs without using binders. This breakthrough can reduce the manufacturing costs of CF-based products. It also facilitates the development of novel microelectronic devices.
Fiber-based strain sensors have emerged as revolutionary components in flexible electronics owing to their intrinsic compliance and textile compatibility, particularly in human-centric applications ranging from health diagnostics to motion tracking. While substantial progress has been achieved, a critical challenge persists in reconciling the contradictory demands of ultrahigh sensitivity and stable signal transmission through rational structural design. Herein, we develop dual-structure silver (Ag)/polyurethane (PU) fiber-based strain sensors (Ag@PUx) via an integrated wet spinning and interfacial metal ion deposition (IMID) strategy. Notably, we propose a mechanical pre-stretching strategy that enables precise regulation of strain sensitivity and sensing range through controlled substrate deformation. Systematic characterization reveals that pre-stretched PU fibers form ordered microscale conductive networks, exhibiting exceptional electrical stability (conductivity (σ) = 1.9 × 105 S·m−1; the change in resistance value under external tensile force (ΔR)/the initial resistance of the sensor (R0) < 0.03 under 360° torsional deformation) with a high quality factor (Q) of 10.1 at 50% strain. In contrast, non-prestretched counterparts develop microcrack-dominated architectures, achieving a high sensitivity (gauge factor (GF) = 7.7) through strain-induced crack propagation and a fracture strain exceeding 660%. A systematic investigation elucidates the underlying mechanisms behind these distinct sensing performances. The Ag@PUx fiber-based electronics are capable of adapting to various tasks including human motion monitoring, voice recognition, and gesture recognition. Importantly, we developed the Ag@PUx fiber-based electronics to monitor motion states while stably transmitting electrical signals. Ultimately, the Ag@PUx show great promise in applications such as motion monitoring, waist rehabilitation, thermal management, electromagnetic shielding, and antibacterial deodorization.
The contamination of wastewater with organic pollutants and nitrogen compounds poses significant environmental challenges. The primary objective of wastewater treatment is the simultaneous denitrification and decarbonization of ammonia nitrogen and organics into harmless by-products. This study presents a novel method for the directional generation of chlorine radical species like ·ClO and ·Cl using electro-reactive membranes (EMs) known as RuO2@PbO2-M, which were fabricated using an electro-deposition coupled template approach. This method facilitates the rapid and efficient conversion of ammonia to nitrogen and concurrently reduces the chemical oxygen demand in the effluent. Our system achieved ultra-efficient simultaneous denitrification and decarbonization with minimal energy consumption in single-filtration mode, thereby eliminating the need for chemical precursors. We elucidate the formation pathway of ·ClO and ·Cl during the electrochemical oxidation process involving RuO2@PbO2-M, where ·Cl generated from RuO2 reacts with ·OH from PbO2 under hypochlorous acid conditions, thereby enhancing nitrogen and carbon removal. These findings highlight a novel electro-filtration and an innovative reactive membrane design for ·ClO synthesis, which provides a new research framework for the concurrent removal of nitrogen and carbon, and offers a promising solution to enhance wastewater treatment efficiency.
The increasing population and continuous urbanization make food security a key consideration in sustainable development. Efficient farming strategies with low environmental footprints are thus increasingly required to meet food demands. This study presents a design for environmentally friendly, economical, and modular vertical farming systems, in which vegetables are cultivated in a carbon dioxide (CO2)-enriched atmosphere enabled by direct air capture (DAC) and subjected to artificial light exposure. We established a vertical farming setup and conducted experiments to identify productive cultivation strategies by regulating lighting, CO2 concentration, biochar application, and plant species. Additionally, a self-developed DAC rotary adsorber was utilized to achieve stable and efficient CO2 enrichment. Compared with the control group, the fresh weight of the vegetables in the experimental groups increased by up to 57.5%. Furthermore, we performed a comprehensive evaluation of the design and demonstrated that integrating photovoltaic-thermal (PVT) and DAC units increased the system’s net present value (NPV) by 157% compared with a conventional design without these units. Importantly, we found it possible to maintain the low carbon footprint of the system (0.468 kg-CO2 equivalent·kg−1 (CO2eq·kg−1)-vegetable) in the production process. Parametric studies and an application analysis on a global scale reveal the wide adaptability of this strategy to diverse conditions. These findings, together with the modular characteristics of vertical farming systems, highlight the promising potential of this design to increase food security and foster sustainable agriculture.
As a natural alternative to antibiotics, probiotics have considerable potential for use in livestock farming. However, the current use of probiotics in livestock poses potential public health risks due to inadequate regulations, including issues such as the inferior quality and dissemination of antibiotic resistance. In this study, 95 non-duplicate commercial probiotic products for livestock were collected from different regions of China. Our findings revealed that the labeling compliance rate for Lactobacillus was the lowest, at just 11%, and approximately 33.3% of the products were contaminated with opportunistic pathogens containing various virulence and antibiotic-resistance genes (ARGs). Isolates of Bacillus and Enterococcus from the products exhibited diverse clonal types and geographical dispersion, whereas certain Enterococcus exhibited close phylogenetic relationships to clones associated with human infectious diseases. Compared with Bacillus and Lactobacillus, Enterococcus exhibited a higher prevalence of ARGs. Specifically, the oxazolidine-resistance gene optrA, which is located on novel transferable plasmids, was found in one isolate of Enterococcus faecium (E. faecium). Using chicken models, we observed that the optrA-positive E. faecium disrupts the normal intestinal microbiota in chickens and alters the abundance of intestinal resistome and mobile genetic elements (MGEs). Furthermore, metagenomic analysis revealed that the optrA gene can be transferred via transposon IS1216E to commensal intestinal bacteria, including Enterococcus cecorum (E. cecorum), Enterococcus gallinarum (E. gallinarum), and Lactobacillus crispatus (L. crispatus) species. In summary, our study confirms that the probiotic products used in Chinese livestock production present problems such as non-compliance with good manufacturing practice (GMP) production standards and insufficient elucidation of the molecular genetic background of probiotic strains. The widespread use of low-quality Enterococcus strains containing various ARGs as probiotics could disrupt intestinal homeostasis and serve as a reservoir and source of ARGs. We emphasize the importance of carefully evaluating the use of Enterococcus strains as probiotics to avoid potential negative effects on livestock and human health.
Radiation-induced lung injury (RILI) is a common complication of cancer radiotherapy, yet effective treatments remain elusive. Compound Kushen injection (CKI), a traditional Chinese medicine (TCM) formula, is widely used in clinical practice for treating radiation-related diseases and as an adjunct therapy for cancer and has demonstrated some effectiveness. However, the mechanisms underlying CKI intervention in RILI and its role in cancer adjunctive therapy remain unclear. In this study, we refined previous statistical approaches and successfully integrated quantitative data on the compounds in CKI. We constructed a network-based holistic target model and developed modular biological networks to explore the modular regulatory effects of CKI in RILI. Through this network-based analysis, we identified specific alkaloid components of CKI that contribute to its therapeutic effect in alleviating RILI. Furthermore, through transcriptomic analysis, we confirmed that oxidative stress plays a central role in the treatment of RILI by CKI. The modular regulatory effects of CKI have been validated in animal models of irradiation, demonstrating the ability of CKI to alleviate oxidative stress, reduce inflammation, regulate immune responses, and inhibit apoptosis. In addition, we demonstrated that nuclear factor erythroid 2-related factor 2 (NRF2) serves as a key mediator of the antioxidant effects of CKI. Matrine and sophoridine, representative alkaloids in CKI, exhibit binding interactions with NRF2. CKI promotes the nuclear translocation of NRF2, and NRF2 activates its downstream targets, such as heme oxygenase-1 (HO-1) and NAD(P)H quinone dehydrogenase 1 (NQO1), to suppress oxidative stress in RILI. This, in turn, inhibits the expression of inflammatory molecules, including interleukin (IL)-6, tumor necrosis factor (TNF)-α, and inducible nitric oxide synthase (iNOS), while promoting the activity of antioxidants such as superoxide dismutase (SOD) and glutathione peroxidase-4 (GPX-4), thereby exerting therapeutic effects on RILI.
Neonatal hypoxic-ischemic encephalopathy (HIE), resulting from perinatal asphyxia-induced hypoxic-ischemic brain damage (HIBD), is a severe neurological disorder that impairs neurodevelopment, and no definitive therapies are available. The polyphenolic natural compound salvianolic acid C (SAC) exhibits antioxidant, anti-inflammatory, and antiapoptotic properties. In this study, we evaluated the efficacy of SAC in treating HIE via animal and human brain organoid experiments. Human brain organoids served as a translational platform for assessing natural product efficacy and clinical effect prediction. Rat brain tissues were harvested at two time points (24 h and 7 d after HIBD and SAC administration) for single-nucleus RNA sequencing. In vitro and in vivo experiments, including microarrays and gene silencing, were employed to confirm the sequencing findings. Our findings demonstrated that during the acute phase of HIBD, SAC suppressed signal transducer and activator of transcription 3+ (Stat3+) astrocyte-driven acute neuroinflammation, decreased inflammatory factor release, and maintained glial-immune homeostasis. During the subacute phase, SAC promoted oligodendrocyte differentiation and facilitated crosstalk between anti-inflammatory microglia and myelinating oligodendrocytes, establishing a regenerative microenvironment and enhancing neuregulin 3 (NRG3)-receptor tyrosine-protein kinase erbB-4 (ErbB4) signaling axis activity. These coordinated mechanisms highlight the dual capacity of SAC in mitigating early injury and driving structural repair in the later stages. This study revealed the pathophysiology of HIE and the multitarget neuroprotective effects of SAC against this disorder at single-cell resolution, advancing the mechanistic foundations for SAC-based therapies in neonatal brain injury.
Considering the development of urban freight transport, this paper presents an operational strategy for freight transport based on the urban metro system. To improve the alignment between service capacity and transport demand under passenger and freight co-transportation (PFCT), a mixed-integer nonlinear programming model (MINLP) is developed to simultaneously optimize the train timetable (TT) and rolling stock circulation plan (RSCP), with particular consideration of flexible train composition mode and skip-stop strategies. Moreover, by introducing allocation rules for passengers and freight, the tripartite interests of operators, passengers, and freight agents are synergistically considered in the proposed model. To facilitate the model solution, a variable neighborhood search (VNS) algorithm is designed for the generation of high-quality solutions in a reasonable computational time. Finally, based on a simplified example and empirical data from the Beijing Metro Yizhuang Line, several sets of numerical examples are implemented to validate the applicability and effectiveness of the model and the approach.