Search Results (16,568)

Low-Rank Adaptation (LoRA) has become a widely adopted parameter-efficient fine-tuning (PEFT) technique for large language models (LLMs). LoRA’s benefits stem from its light weight and modular adapters. Standard LoRA applies adapters uniformly across all Transformer layers, implicitly assuming that each layer contributes equally to task adaptation. However, LLMs are found to have internal substructures that contribute in a disproportionate manner. In this work, we provide a theoretical analysis of how LoRA weight updates are influenced by a layer’s activation magnitude. We propose Act-LoRA, a simple activation-guided layer selection strategy for selective Low-Rank Adaptation. We evaluate this strategy for both encoder-only and decoder-only architectures using the GLUE benchmark. Our method achieved a 20% GPUh saving with a 1% drop in GLUE score using DeBERTaV3-Base on a single-instance GPU with 50% less LoRA parameters. It also achieved 2% GPUh savings with a less than 0.15% drop in GLUE score with the Llama-3.1-8B model in Distributed Data Parallel mode with 25% fewer LoRA parameters. Our experiments and analysis show that the compute and memory requirements of LoRA adapters increase linearly with the number of selected layers. We further compare activation-guided selection against gradient-guided importance metrics and show that activation norms yield more stable and reproducible layer rankings across seeds and datasets. Overall, our results demonstrate that activation-guided layer selection is a practical and effective way to improve the efficiency of LoRA fine-tuning, making it immediately compatible with some existing PEFT techniques and distributed training frameworks. Full article

Thermal excursions during post-weld heat treatment (PWHT) and on-site fabrication frequently compromise the integrity of Grade 92 steel. While hardness fluctuations are documented, the correlation between initial properties and long-term creep stability remains controversial. This study aims to evaluate the relationship between thermal history and subsequent creep performance. Heat treatments of T92 steel across a wide temperature range (760–1000 °C) were performed, followed by creep tests at 600 °C/130 MPa and microstructural characterization. Results reveal a non-monotonic evolution of hardness and strength, reaching a minimum at 850 °C due to martensitic lath recovery into ferrite, but nearly doubling the as-received (AR) values above 900 °C due to fresh martensite formation. Creep life drops to a minimum at 850 °C and recovers to the AR level at 950 °C. A significant “decoupling” occurs at 1000 °C, where the sample possesses the highest hardness but only exhibits one-fourth the life of the 950 °C sample. Superior performance stems from the retained M₂₃C₆ and its dynamic precipitation, which pins dislocations to form micro-lath structures. Conversely, 1000 °C facilitates full carbide dissolution, accelerating dislocation recovery. These findings emphasize precise PWHT control and demonstrate that a 950 °C rejuvenation treatment can restore over-tempered or damaged components. Full article

Beer is an alcoholic beverage made primarily from malted cereals, water, hops, and yeast. Although barley is the most common cereal in brewing, corn malts are also used to produce beer in different countries. However, research on their production, physicochemical properties, and aging evolution is limited. In the present study, the evolution of various physicochemical features during the aging of six lager- and ale-fermented corn beers was investigated. Results after 18 months of aging showed decreases in most of the measured properties: total phenolics between 16 and 20%, antioxidant capacity between 17 and 23% by DPPH assay and 23–41% by ABTS assay, free anthocyanins between 38 and 55%, bitterness units between 32 and 42%, and SRM color and color intensity only dropped in lager beers, while in ale beers these properties increased. Finally, tonality increased in lager beers and one ale beer. This study enabled a more in-depth analysis of corn beer, focusing on the evolution of physicochemical properties during aging that are relevant to beer quality. Full article

To investigate the influence of soil parameters on the cushioning performance of landing airbags, a landing airbag cushioning dynamics model considering soil characteristics was established based on the control volume method and a crushable foam model. Experimental validation was conducted for both the airbag cushioning model and the soil impact model, respectively, with good consistency between simulated and experimental results. Based on the established model, the influence of soil on the cushioning performance of landing airbags was analyzed. The analysis results indicate that soil absorbs energy through compressive deformation during the cushioning process, thereby exhibiting a certain degree of cushioning performance. Softer soil absorbs more energy, and the payload is less prone to rebound. However, excessively soft soil causes the airbag to sink into the soil, hindering the venting of gas outward and resulting in hard landings for payloads. Therefore, three indicators—airbag peak pressure, payload maximum acceleration, and maximum drop height—are used to comprehensively evaluate the cushioning performance of airbags, and the influence laws of soil parameters are quantitatively researched. The research shows that the soil density, shear modulus, and yield parameters A₁ and A₂ significantly influence cushioning performance. Specifically, the shear modulus and yield parameter A₁ exhibit logarithmic growth relationships with the three cushioning performance indicators, while the yield parameter A₂ and soil density show linear growth relationships with the three cushioning performance indicators. Full article

Hydrogen combustion is known to be fast compared to traditional hydrocarbon fuels. The fast combustion leads to a higher thermal efficiency. In this research a 600 cc single cylinder hydrogen engine was tested at 1250 rpm, lambda = 2 and 3, and three load levels (load was represented by Manifold Absolute Pressure (MAP); MAPs tested were 75, 95 and 120 kPa) and compared to operation with gasoline and propane. The fast burn duration (Mass Fraction Burnt MFB10% to MFB90%) and the MFB 50% were determined and analyzed. The hydrogen MFB50% location for Minimum Timing for Best Torque (MBT) was found to occur at around the typical 8 Crank Angle Degrees (CADs) After Top Dead Center (ATDC). Measurements of ignition delay based on the fast data direct measurement of spark ignition coil current drop to the change in polarity of net heat release are presented. With shifts towards direct injection and higher injection pressures, consideration was given to the hydrogen pressurization penalty, where it was calculated that pressurizing hydrogen to 100 bar at the flow required for lambda = 2 operation is 2.3 bar, i.e., higher than the Friction Mean Effective Pressure (FMEP)! Furthermore, hydrogen is widely cited to have a higher heat loss than typical hydrocarbon fuels. In this paper, detailed analyses at lambda 2 and lambda 3 showed that hydrogen in fact has lower heat losses. Full article

Integrating cover crops (CCs) into crop rotations has gained interest in the Southeastern United States due to the benefits that CCs offer, which improve soil health for agricultural production. However, more information is needed on how CCs may affect the development of soil water repellency (SWR), which can negatively impact soil hydrology. The development of SWR threatens crop yields, food security, and farmer livelihoods. To address this knowledge gap, a field experiment measured the water repellency (WR) of four common CC species and a fallow treatment. CC samples were oven-dried, ground, and analyzed for WR using the water drop penetration time (WDPT) test. The mean WDPTs of the CC residues collected at termination and four weeks post-termination ranged from 49 to 4174 and 8 to 2627 s, respectively. Large WDPTs (>5 s) indicate that CC residues can potentially influence the development of SWR. All CC residues exhibited WR. The results suggest that farmers may need to consider alternative CC species depending on when they plant their cash crops in relation to CC termination. Considering the effects of CCs on SWR will enable farmers to make informed management decisions to mitigate SWR development and maintain soil health in a changing climate. Full article

At the current stage, water resource shortages and significant regional disparities in resource distribution severely restrict China’s food security. Existing research primarily focuses on resource use efficiency, while lacking a systematic framework to distinguish between equality and equity in the coupled distribution of irrigation water, grain production, and nitrogen pollution across major river basins. The core objective of this study is to utilize the Concentration Index (CI) to construct a unified equity assessment framework, quantify the evolution of equality and equity in irrigation water use, grain production, and nitrogen loss to surface water in different river basins in China from 1992 to 2017, and determine the key influencing factors. For positive production resources, a distribution that benefits low-income groups is equity, while for pollution burdens, this distribution pattern is inequity. The results show that water shortages in Northern China have intensified, and higher income groups have obtained excessive benefits. The distribution of grain production has shifted from favoring higher income groups to favoring low-income groups, with the Concentration Index changing from 0.214 to −0.052, indicating an enhancement in equity. Irrigation water use has shown a certain degree of improvement, with the CI dropping from 0.023 to 0.017. However, nitrogen loss to surface water has exacerbated environmental inequality, with the CI dropping from 0.10 to 0.03, indicating that pollution burdens have shifted to low-income groups. Changes in equity across the country are driven by a small number of high-intensity grain production areas, and the key influencing factors include food security policies, urbanization, population size, and nitrogen fertilizer application. An asymmetric coupling relationship exists between water resource shortages and equity, and the regional economic foundation determines the formation of synergy or trade-offs. The findings underscore the necessity of transitioning from efficiency-focused to equity-focused agricultural governance in China. Targeted policies should include cross-basin ecological compensation mechanisms, differentiated technology promotion strategies, and integrated water–food-pollution management systems to balance food security, environmental protection, and social justice. Full article

Background/Objectives: The aim of this study was to report first experiences with dynamic intraligamentary stabilization (DIS) technique for anterior cruciate ligament (ACL) rupture in children and adolescents. Methods: A case series of 22 children and adolescents with a mean age of 13.3 years underwent primary repair of an ACL rupture using the DIS technique as an off-label use in skeletally immature patients. Patients were evaluated for laxity, strength, range of motion (ROM), and functional tests, as well as Tegner, Lysholm, International Knee Documentation Committee (IKDC), and PedsQL scores after 3 years. A follow up after 11 years was conducted to analyze long-term results, rerupture rates and reinterventions. Results: Three years after surgery, there was no significant difference in laxity, strength, ROM, and in the functional tests comparing the injured to the contralateral knee. The Tegner Index after surgery showed a slight drop of 0.8 points, from 7.1 preoperatively to 6.3. Mean IKDC, Lysholm, and peds-QL scores were 91.17 (range 62.64–98.85, median 94.25), 88.27 (range 58–100, median 93), and 88.78 (range 58.15–100, median 91.30). Overall failure rate of the DIS-repaired knees was 55% (12 of 22 patients). In ten patients, reruptures happened at an average time of 3.2 years after initial surgery; additionally, two patients with chronic instability had to undergo revision ACL reconstruction. Conclusions: DIS repair might help ACL healing with satisfactory functional outcomes. However, given the high failure and reintervention rates, further studies need to show non-inferiority of the DIS technique in children and adolescents before being considered a valid treatment option. Full article

This study developed and evaluated an integrated experimental–computational framework to quantify coconut-oil transport and marbling stabilization in soy protein concentrate (SPC) during static holding and co-extrusion with a cooling die. Temperature-sweep rheology and Differential Scanning Calorimetry (DSC) identified the main gelation transition at 65–78 °C, with oil shifting gelation to higher temperatures and increasing enthalpy, supporting an exit/cooling target of 70–75 °C. Static drop tests at 100 °C for 60 s were analyzed by depth-resolved imaging and coupled with a single-phase CFD model to inversely calibrate an effective diffusion coefficient for coconut oil in SPC (D_ref = 4.86 × 10⁻¹⁸ m²/s). A viscosity-coupled fractional Stokes–Einstein relationship then gave temperature-dependent effective diffusivities of 1.89 × 10⁻¹⁸ to 4.86 × 10⁻¹⁸ m²/s over 60–100 °C, indicating reduced oil mobility during cooling. Additional static time-temperature comparisons suggested limited redistribution beyond ~50 s. Co-extrusion simulations and product imaging further indicated that staged hot-zone residence followed by rapid cooling can help stabilize oil domains into marbling-like structures. The framework can support selection of cooling-die temperatures, residence times, and oil-injection conditions. Future work should extend the framework by linking marbling microstructure with sensory performance, oxidative stability, and sensitivity analysis of key transport parameters. Full article

Background: Load carriage is an essential part of the occupational work of many jobs, yet there is little research on the hemodynamic responses to load carriage. It is known that front load carriage elicits larger increases in arterial stiffness than load carried on the side of the body. However, the hemodynamic forward and reflected pulse wave responses to load carriage are unknown and could relate to cardiac risk. Methods: We compared responses to 30 s front load carriage between 45 females and 23 males, with pre- and post-carry hemodynamics assessed using pulse wave analysis. Results: We found increases (p < 0.001) in arterial stiffness (24.8% females; 32.4% males), forward pulse wave (5.8 mmHg females; 5.7 mmHg males), and reflected pulse wave (6.8 mmHg females; 9.9 mmHg males). Pre- and post-carriage forward and reflected pulse waves were lower in females (p < 0.05). Compared to males, females overall had more relationships between the change in vascular measurements and other variables. We found an inverse relationship between changes in myocardial supply–demand (SEVR) and changes in forward pulse wave in females (r = −0.37, p < 0.001) but not males. Also, a direct relationship between changes in SEVR and changes in aortic DBP (r = 0.30, p = 0.04) and changes in resting DBP (r = 0.35, p = 0.02) existed in females. Conclusions: The data suggest that sex-related differences in hemodynamic responses exist. Females may experience a larger drop in estimated myocardial supply–demand balance accompanied by lower diastolic filling. Employers should be aware of these inherent cardiac risks with load carriage in their female employees. Full article

Efficient ortho–para hydrogen conversion is essential to suppress spontaneous heat release and boil-off losses during cryogenic liquid hydrogen storage and pre-liquefaction processes. In this study, a novel catalyst-filled wavy plate-fin heat exchanger (CFHE) is proposed to simultaneously enhance heat transfer and ortho–para hydrogen conversion under cryogenic conditions. Compared with conventional straight-fin configurations, the wavy-fin structure introduces controlled flow perturbations and increased specific surface area, thereby intensifying transport processes. Three-dimensional computational fluid dynamics (CFD) simulations, using the SST k–ω turbulence model, coupled with an ortho–para hydrogen conversion kinetic model were performed to quantitatively investigate the effects of key geometric parameters and catalyst loading on hydrogen conversion, heat transfer, and pressure drop within a Reynolds number range of 941–1577 and a temperature range of 35–20 K. Within the same CFHE configuration, the para-hydrogen fraction remains nearly unchanged without catalyst but increases significantly with catalyst loading. However, the catalyst reduces the global average Colburn j-factor by about 25%. Despite higher friction losses, the outlet–inlet temperature difference decreases to about 0.866 times that of the non-catalyst case, indicating improved temperature uniformity. A comprehensive performance index e, integrating heat transfer enhancement, flow resistance, and conversion efficiency, was introduced and optimized using a genetic algorithm. The optimized CFHE achieves an outlet para-hydrogen fraction exceeding 95% of the thermodynamic equilibrium value while maintaining hydrogen entirely in the gaseous phase to avoid catalyst deactivation. Overall, the catalyst-packed wavy channel configuration demonstrates superior conversion efficiency, enhanced thermal uniformity, and improved overall performance compared with straight-fin structures, providing quantitative design guidance for high-performance heat exchangers in cryogenic hydrogen liquefaction systems. Full article

Reservoir quality in shallow lacustrine-mixed siliciclastic–carbonate systems is commonly governed by mineral assemblages and diagenetic modification. Here we investigate the Neogene Youshashan Formation (Oil Groups III–V) in the Nanyishan area, western Qaidam Basin, to quantify mineralogical and diagenetic controls on pore systems and flow. We integrate whole-rock XRD and log-derived mineral profiles with thin-section/SEM petrography, NMR T2 spectra, mercury injection capillary pressure (MICP), and a water-drop test. Dissolution-related pores and dolomitization-related intercrystalline pores dominate the pore space, whereas cementation and clay-related filling/coating locally restrict pore throats and connectivity. Algal limestones (average porosity 23.17% and permeability 54.3 mD; MICP r50 = 0.085 μm) show better reservoir quality than dolomitic rocks (average porosity 17.24% and permeability 15.13 mD; MICP r50 = 0.039 μm), consistent with more effective pore throat networks. In Oil Group III (Well NQ2-6-2), higher dolomite content is generally associated with higher porosity but shows no systematic relationship with permeability, highlighting the primacy of connected pore throats. Water-drop behaviors (beading, semi-beading, infiltration) provide a rapid, semi-quantitative screening indicator when interpreted together with pore throat metrics, and support a four-class reservoir-typing scheme (Types I–III and non-reservoir) for sweet-spot identification in mixed lacustrine reservoirs. Full article

Multiple sclerosis (MS) is a highly disabling chronic autoimmune disease of the central nervous system with neuroinflammatory and neurodegenerative alterations found in the white and grey matter of the brain. The pathogenesis of MS is complex and not fully understood. Mitochondrial dysfunctions are suspected to play an important role. The visual system is often affected in MS. Optic neuritis is a frequent symptom, but also the retina itself, including retinal synapses appear compromised in MS independent from demyelination of the optic nerve. A previous study demonstrated synapse-specific alterations of mitochondria in photoreceptor synapses in the Experimental Autoimmune Encephalomyelitis (EAE) mouse model of MS at day 9 after injection, an early time point in pre-clinical EAE. In the present study, we analysed even earlier stages of pre-clinical EAE for possible alterations of synaptic mitochondria. For this purpose, we performed qualitative and quantitative immunolabelling analyses of the mitochondrial cristae organising protein MIC60 at retinal synapses and functional analyses by measuring synaptic mitochondrial membrane potential (during rest and depolarisation-induced exocytosis) and visually guided behaviour (optometry analyses). At day 3 after injection, morphological and functional data were indistinguishable between MOG/CFA-injected EAE mice and CFA-injected control mice. But already on day 5 after injection, we observed a decreased expression of the mitochondrial MIC60 protein at synaptic mitochondria, a decreased synaptic mitochondrial membrane potential at rest, an enhanced drop of mitochondrial membrane potential during stimulated exocytosis and a decreased visual performance of the respective EAE mice. These data argue that synaptic pathology in the EAE retina begins as early as day 5 after injection. Our data propose that dysfunctions of mitochondria play an important role already at the very early stages of synaptic pathology in EAE. Full article

To reduce the carbon footprint, electric vehicles (EVs) are considered an alternative transportation choice. However, increased use of EVs could lead to overloading the existing power network when accounting for all installed chargers. With the increasing deployment of EV chargers, universities are potential locations for the oversized power network issue. This paper applies reinforcement learning (RL) to optimize for EV charging infrastructure at the university scale using real-world data, directly contributing to sustainable energy management by reducing grid burden and increasing renewable energy utilization. The RL-based charger aims to reduce the burden on the grid while increasing renewable energy utilization. This study investigated practical relevance in real-world systems, considering three demand scenarios: random, stochastic historical demand from Qatar University, and actual online data from Caltech University. Three RL algorithms—Deep Q-Network (DQN), Advantage Actor–Critic (A2C), and Proximal Policy Optimization (PPO)—are applied. While training, the historical stochastic data requires more tuning of the RL framework than the random demand, emphasizing the importance of realistic demand profiles. The performance of the RL approach depends on the type of demand. The results show that the proposed RL approach can efficiently mitigate the peak charging currents. For the Qatar University historical demand scenario, the PPO algorithm minimized the peak charging currents by 50% relative to uncontrolled charging (160 A to 80 A) and Model Predictive Control maintained the energy transfer capability at 99.710%. For the random demand type, the peak charging currents are minimized by 38.3% as compared to uncontrolled charging (128 A to 79 A), with a nominal reduction in energy transfer capability to 95.89%. Scalability is tested by integrating the model into the IEEE-33 bus network. Without solar integration, the proposed RL-based EV charging management model improves the voltage drop by 0.05 p.u., leading to reduction in the line losses by 17% as compared to the MPC benchmark method and by 32% as compared to the uncontrolled charging scheme. Further, the proposed RL approach leads to a 9% reduction in line current during peak hours in the IEEE-33 bus system. With solar integration into the IEEE-bus system, the proposed framework of the RL approach improved the sustainability of the charging infrastructures by enhancing solar energy utilization by 42.5%. These findings validate the applicability of the proposed model used for optimizing the sustainable EV charging infrastructure while managing the charging coordination problem. Full article

Slaughterhouse by-products are promising feedstocks for anaerobic digestion due to their high lipid and protein content. However, their complex structures often limit hydrolysis, and excessive pretreatment can induce inhibitory conditions. This study evaluates the effects of ball-milling (BM), ball-milling with water (BM + water), and combined thermal hydrolysis and ball-milling (THP + BM) on the digestion performance of a mixed substrate of slaughterhouse and agricultural wastes. The results demonstrate that all BM-based pretreatments significantly improved digestion kinetics, reducing the lag phase by 26–66% and shortening the T₅₀ values by approximately 40% compared to the untreated substrate. While no statistically significant differences were observed in the ultimate methane yield, the onset of methanogenesis was markedly accelerated in the BM and BM + water treatments. In contrast, despite achieving superior solubilization, the THP + BM treatment failed to provide proportional kinetic enhancements. This was attributed to a severe initial metabolic imbalance—characterized by a pH drop below the inhibitory threshold (6.33)—which induced physiological stress and delayed the functional recovery of methanogens. These findings indicate that while ball-milling effectively facilitates digestion initiation by enhancing physical accessibility, the intensity of combined thermal-mechanical processes must be strategically optimized. For high-strength organic biomass, managing pretreatment severity is crucial to prevent initial acid stress and maximize process efficiency. Full article