Search Results (3,092)

In recent years, the use of UAVs has become increasingly widespread, and the public safety risks posed by unauthorized UAV flights have become increasingly prominent, creating an urgent need for effective detection and identification of UAV targets. However, such targets are small in size, have low contrast, and exhibit an extremely low signal-to-noise ratio; conventional detection methods generally suffer from insufficient feature discrimination, missed detections, and false alarms in complex backgrounds. To address these challenges, this paper proposes a Frequency-Guided Cross-scale Refinement Network (FGCR-Net). Based on an encoder-decoder architecture, this network achieves end-to-end collaborative optimization through cross-layer feature fusion, side-channel prediction refinement, and frequency-domain background suppression. First, a multi-path selective cross-layer fusion module (SCFM) is designed. This module employs coordinated modeling via both channel and spatial paths, supplemented by adaptive weighting with learnable coefficients, to perform differentiated selective fusion of the encoder’s fine-grained features and the decoder’s semantic features, thereby bridging the semantic gap at jump connections; Second, we designed a Cross-Scale Adaptive Fusion Enhancement Attention Module (CAFEM), which cascades multi-receptive-field hollow convolutions, strip pooling, and a bidirectional semantic guidance mechanism to perform cross-scale refinement on the side outputs of each decoder layer, thereby alleviating the issues of blurred boundaries and false alarms caused by inconsistent quality of multi-scale prediction maps and insufficient cross-layer consistency; finally, we design a Frequency-Guided Semantic Enhancement Module (FGSEM), which uses the Fast Fourier Transform (FFT) to decouple encoder features into the frequency domain. By leveraging low-frequency energy to predict the background confidence map and applying spatially selective suppression to high-frequency components, this module distinguishes, from a frequency-domain perspective, the high-frequency responses of complex backgrounds and targets that are highly similar in the spatial domain. Experiments on MSDS-UAV, a self-built multi-scenario UAV dataset for small targets, demonstrate that our method consistently outperforms existing state-of-the-art methods across multiple performance metrics, with Pixel Accuracy, Mean Intersection over Union, and Probability of Detection reaching 92.76%, 70.91%, and 92.69%, respectively; Compared to the baseline model, these three metrics improved by 1.90, 3.20, and 3.76 percentage points, respectively, fully validating the effectiveness and superiority of the proposed method. Full article

Reinforcement learning (RL) has become a powerful tool for quadrupedal locomotion, and a sim-to-real approach is widely adopted to avoid hardware damage during training. However, the “sim-to-real gap” remains a critical challenge, particularly for robots driven by high-gear-ratio actuators, in which nonlinear friction effects are strongly amplified. Conventional methods, such as actuator networks or heuristic domain randomization, often require specialized sensors or extensive trial-and-error to tune appropriate randomization ranges. Building on a recent system-identification framework for actuator dynamics, we extend it with an augmented friction model that incorporates the Stribeck effect to capture the low-velocity nonlinearities characteristic of high-gear-ratio actuators. The physical parameters are identified from real-robot trajectory data using an evolutionary algorithm, and the resulting simulation is used to train a locomotion policy that is transferred zero-shot to a 55 kg quadruped without additional fine-tuning or base- or controller-level dynamics randomization. On our platform, adding the Stribeck term lowers the actuator identification error by 16% relative to a Coulomb–Viscous model on the trajectory used for identification, and this advantage generalizes to an unseen trajectory not used for identification. It also lowers the simulation-to-reality mean-velocity degradation from 40.2% and 26.8% for the Coulomb–Viscous model to 27.2% and 21.6% for our method at the 0.3 and 1.0 m/s commands, respectively. The trained policy achieves stable locomotion on flat ground as well as rough terrain including steps and stairs. These results indicate that explicitly modeling low-velocity friction is beneficial for high-fidelity sim-to-real transfer in high-reduction systems. Full article

Micro electrical discharge deposition (micro-EDD) enables maskless direct-write metallic tracks on conductive substrates, but reported implementations rely on purpose-built machines with programmable pulse generators and gap servos. This paper reports a low-cost micro-EDD platform retrofitted from a desktop CNC router driven by GRBL firmware, extended with an Android control application for path definition and process supervision. Discharge is produced by an RC relaxation circuit with no pulse generator and no gap feedback. Copper, aluminium, nickel, and titanium electrodes were deposited onto silicon substrates in an argon atmosphere, maintaining a 10 μm separation. The working point, pd ≈ 0.76 Torr·cm, lies within 20% of the argon Paschen minimum. Track width rises linearly with supply voltage over 330–400 V (w = 0.570 V − 99.2 μm, R² = 0.994), extrapolating to a deposition threshold of 174 V, 37 V above the argon breakdown minimum—indicating an energy threshold distinct from breakdown. Across electrode materials, width shows a trend consistent with thermal diffusivity as $w\propto {\alpha }^{0.46}$ (R² = 0.69, N = 4), resolving the inconsistency that copper, the best conductor, gives the widest track. Polarity reversal changes width by factors of 4.6 (Cu) and 7.5 (Al), consistent with anode-dominated energy partition. Péclet numbers remain below 3 × 10⁻⁴ and overlap ratios above 10³, excluding thermal advection and insufficient overlap as causes of the feed-rate collapse. A five-axis kinematic extension is derived, with Z travel identified as the bounding constraint. Full article

While the individual effects of sustainable fillers on cementitious systems are widely known, their combined tribological and fracture behaviors within a polymer concrete matrix remain largely unexplored. Addressing this gap, this study experimentally investigates the physical, mechanical, and tribological characteristics of polyester-based polymer concrete (PC). In addition to reference specimens produced with polyester resin and silica sand, modified mixtures were developed by replacing the silica sand with ground granulated blast-furnace slag (GGBS) and recycled waste concrete aggregate (WC) at various substitution ratios (0%, 5%, 10%, 15%, 20%, and 25%). To evaluate the performance of the developed PCs, parameters including unit weight, water absorption capacity, flexural and compressive strengths, Shore D hardness, surface roughness, acid resistance, Bohme abrasion resistance, and fracture energy were analyzed. Furthermore, temperature variations on the friction surfaces were monitored in real time using a thermal camera during the Bohme abrasion tests. To elucidate the fracture mechanisms, the fractured surfaces were examined via digital microscopy. The quantitative findings indicate that a 25% GGBS replacement optimizes mechanical performance, increasing the compressive and flexural strengths by 21.2% (108.30 MPa) and 30.6% (35.29 MPa), respectively, alongside a 27% improvement in Bohme abrasion resistance. However, this modification significantly increases material brittleness, reducing the fracture energy by 53.3% compared to the reference. Conversely, although incorporating WC offers sustainability advantages, it limits mechanical performance, leading to decreases of up to 9.9% (80.46 MPa) in compressive strength and 15% (22.97 MPa) in flexural strength at a 20% substitution rate. Regarding fracture energy, while the W20 series absorbed more energy than the B25 series, it still remained 47.8% lower than the reference. Additionally, the GGBS-incorporated series demonstrated higher susceptibility to sulfuric acid attack compared to the WC-incorporated series. Full article

Background/Objectives: Epidemic infectious diseases remain a persistent global threat, but whether declining outbreak frequency translates into reduced mortality is unclear. This gap is critical for emergency care systems, where timely intervention determines survival. This study aims to characterize global epidemic trends and to assess whether reductions in outbreak frequency have been accompanied by reductions in epidemic mortality. Methods: We performed a retrospective global analysis of epidemic infectious disease events recorded in the Emergency Events Database (EM-DAT) from 2000 to 2025, following the STROBE reporting guideline. Data on disease type, geographic distribution, deaths, and affected populations were extracted. Completeness of mortality and affected-population reporting was quantified by region and disease type. Crude mortality ratios (CMRs) were compared across regions and pathogens using Kruskal–Wallis tests, and temporal trends were assessed with Mann–Kendall analysis and Sen’s slope estimation. Results: We identified 893 epidemic events across 131 countries, resulting in 125,262 deaths and over 12.7 million affected individuals. Mortality data were available for 773 events (86.6%), with completeness ranging from 92.0% in Africa to 23.5% in Europe. Burden was overwhelmingly concentrated in sub-Saharan Africa, which accounted for 80.2% of deaths. Cholera was the most frequent pathogen, while Ebola showed extreme lethality (CMR 42.71%). Despite a significant decline in annual epidemic frequency (τ = −0.668, p < 0.001), the greater part of which preceded the COVID-19 pandemic, neither annual mortality (τ = −0.237, p = 0.094) nor the annual mortality ratio (τ = −0.114, p = 0.427) showed a commensurate decline. Marked heterogeneity across regions and diseases underscores the dominant role of healthcare system capacity. Conclusions: Global progress in reducing epidemic frequency has not been matched by gains in survival. The persistence of high mortality in regions with constrained emergency care capacity highlights a critical failure in translating outbreak control into lives saved. Strengthening emergency care systems, improving access to timely treatment, and addressing structural inequities are essential. Fewer epidemics are not enough; fewer deaths when they occur is what is needed. Full article

Background and Objectives: Impaired autophagy has been implicated in neurodegeneration, but circulating autophagy-related proteins have shown inconsistent associations with cognitive impairment. We compared serum autophagy-related protein 5 (ATG5) among older adults with normal cognition, mild cognitive impairment (MCI), and clinically diagnosed Alzheimer’s disease (AD), and examined its relationship with global cognitive performance. Materials and Methods: This single-centre cross-sectional study enrolled 164 older adults (55 with normal cognition, 55 with MCI, and 54 with clinically diagnosed AD). Cognitive status was determined through integrated clinical assessment rather than a single test cut-off. An enzyme-linked immunosorbent assay was employed to quantify serum ATG5, yielding quantifiable values for 156 participants, right-censored observations above the highest calibrator for six, and unmeasurable results for two. Since the two 96-well plates, drawn from one ELISA kit lot and processed on the same day, exhibited a substantial discrepancy in measurement scale across runs, they were analysed as distinct strata, with the principal group contrast estimated via a right-censored Tobit regression of log-transformed ATG5 within each plate, controlling for cognitive group, age and sex; a pooled plate-adjusted model also provided support. Supportive analyses used within-plate z-standardised ATG5. Associations with clinical variables were assessed using age- and sex-adjusted partial Spearman correlations with false discovery rate correction. Results: Across both analytical strata (plate 1 omnibus p = 0.293; plate 2 p = 0.264) and within the supportive pooled model (p = 0.143; AD vs. normal geometric mean ratio 0.89, 95% CI 0.73–1.09), serum ATG5 concentrations did not differ significantly between cognitive groups. Concentrations on plate 2 were 47% lower (GMR 0.53, 95% CI 0.45–0.62), a gap exceeding the inter-assay imprecision declared by the manufacturer and unexplained by the calibrator values from plate 2; since relative dispersion matched across both plates (likelihood ratio p = 0.730), the pattern suggests a multiplicative scale shift of unknown origin. Serum ATG5 levels were weakly associated with global cognitive performance assessed via the S-MMSE (partial ρ = 0.283, p < 0.001, FDR p = 0.010), an association that persisted after adjusting for education, albumin and folate and retained a similar magnitude within a linear model (β = 0.044, p = 0.014). Incorporating ATG5 into a model that included age, sex and education failed to enhance discrimination (ΔAUC +0.012 and +0.000). Conclusions: Serum ATG5 did not discriminate normal cognition, MCI, and clinically diagnosed AD. A weak rank-based association with S-MMSE was observed, but the effect was small. These findings do not support serum ATG5 as a diagnostic discriminator for clinically defined cognitive status. Full article

During the study of premixed NH₃/CH₄ fuel mixtures, flame separation into two reaction zones was observed upon flashback into a Bunsen burner. Stable combustion regimes were obtained, with a gap between the burner rim and the upper luminous region, the size of which depended on the mixture composition and the position of the combustion zone inside the burner. The work examines combustion regimes of NH₃/CH₄ mixtures with an ammonia content of 10–60% in the fuel blend. The equivalence ratio was varied in the range of 0.95–1.1. The Reynolds number was maintained in the range of 270–300, ensuring laminar flow conditions. At the burner exit, temperature and gas composition profiles were measured. Flame emission was recorded at the chemiluminescence bands of OH* (308 nm) and CH* (431 nm), and emission spectra were acquired both inside and outside the burner. Spectra in the range of 190–1080 nm were recorded. Spectral analysis revealed a band corresponding to NO₂ emission in the upper part of the luminous region. It is most likely that nitrogen dioxide is formed through low-temperature reactions occurring when the combustion products mix with atmospheric air. Full article

Nucleosome assembly protein 1 (NAP1) is highly conserved across eukaryotes, yet its biochemical function, particularly its proposed role as a histone chaperone, remains unresolved. Dinoflagellates, including Karenia brevis (Kb) and Crypthecodinium cohnii (Cc), lack architectural nucleosomes and express core histones at unusually low levels, yet retain abundant NAP1 transcripts encoding two to three distinct homologs. Confocal immunolocalization of K. brevis showed strong nuclear signals for both homologs, between chromosomes and at the nucleolus, with higher nuclear-to-cytoplasmic ratios in G₂ than in G₁ (the two gap phases of the cell cycle); labeling at chromosome-territory margins and at the nuclear cortex is consistent with an association with the telomeric nucleosomes anchored to the nuclear envelope. Recombinant KbNAP1Bp reproduced the canonical yeast NAP1 fold, recovered H2A-immunoreactive material in immunoprecipitation, and preferentially retarded larger DNA fragments in gel mobility assays, whereas KbNAP1Ap did none of these under the conditions tested. In C. cohnii, the NAP1 protein peaked at S–G₂, and exposure to a CcNAP1.1-antisense oligodeoxynucleotide (ODN) was associated with an S–G₂/M delay. These findings uncouple NAP1 abundance from the availability of a canonical nucleosomal substrate and suggest evolutionary repurposing toward non-nucleosomal roles in chromosome-territory organization. They help define the minimal functional core of this conserved chaperone family and caution against treating NAP1 abundance as a proxy for nucleosome assembly activity. Full article

Battery state-of-health (SOH) benchmarks often mix cycles from the same cell across training and test sets. We examined whether predictor structure changes the gap between within-cell interpolation and cross-cell transfer. Seven regressors and five predictor sets were evaluated in four NASA cells (636 cycles) using ten random 80:20 splits and leave-one-cell-out (LOCO) testing repeated over ten model seeds. Across-model median random split RMSE fell from 5.00 SOH points with age alone to 1.52 with diagnostics plus age, whereas median LOCO RMSE fell only from 5.83 to 4.61; the absolute protocol gap therefore increased from 0.83 to 3.09 points, and the median cell-paired LOCO/random ratio increased from 1.24 to 4.53. With all diagnostics, model-level ratios ranged from 1.65 to 5.16, and random-versus-LOCO ranks were uncorrelated (descriptive Spearman ρ = 0.00). Holding the number of training cycles constant, increasing the number of source cells from one to three reduced median LOCO RMSE from 5.25 to 4.73 points, but 32.1% of cell–model pairs did not improve. Standardised sliced Wasserstein shift was associated with the validation ratio (descriptive ρ = 0.73) but not with absolute LOCO RMSE (ρ = −0.01). Rated capacity normalisation preserved the predictor-dependent gradient. An eight-cell Oxford analysis was limited to an age-only boundary check and was not a feature-matched replication. These four-cell laboratory results show that richer discharge diagnostics can improve represented cell interpolation much more than unseen cell transfer; claims should therefore report absolute and relative errors with complete cell-held-out validation. Full article

AI-driven credit scoring is supervised as a high-risk application in banking and insurance, yet unfairness is rarely operationalized as a measurable category of model, conduct, legal, and reputational risk. Using 20,000 anonymized applications from a Southern European digital lender (15.2% twelve-month default rate), we estimate three model families—a regularized logistic regression, a gradient-boosting machine, and a multi-layer perceptron—under a fully crossed design in which each family is evaluated without mitigation and under pre-processing (reweighing), in-processing (an exponentiated-gradient reduction, applicable to any base learner, together with adversarial debiasing where gradient-based training permits it), and post-processing (reject-option) interventions, so that the mitigation effect is no longer confounded with the choice of estimator. No sensitive-group field enters any estimated specification; group membership is used exclusively for auditing. Predictive performance (AUC-ROC, Brier score and Brier skill score relative to the base-rate forecast, F1 on the default class, Gini, and the Kolmogorov–Smirnov statistic) is reported jointly with group fairness (demographic-parity and equal-opportunity differences, disparate-impact ratio, Theil index) and with group-conditional calibration, at an explicitly stated and economically justified decision threshold. Every fairness quantity is accompanied by stratified-bootstrap confidence intervals and, for stochastic learners, by seed-level dispersion. The interpretable benchmark attains an AUC of 0.780 and a Brier score of 0.104 against 0.129 for the constant base-rate forecast, and the high-capacity models improve on it by under one AUC point. Disparity is present but is located geographically rather than in the composite group label: the disparate-impact ratio is 0.724 [0.693, 0.754] for the lowest socio-economic neighborhood cluster, excluding the four-fifths screening value, against 0.809 [0.776, 0.840] for the ethno-socioeconomic proxy, whose interval contains it, and no measurable gender disparity. Group membership is recoverable from the neutral feature set at an AUC of 0.654, and 42% of the group gap in predicted risk travels through the bureau credit score alone, so feature deletion cannot close the channel. Feature attributions and an auxiliary group-recoverability test locate the proxy pathways through which disparity arises, and a misclassification-sensitivity analysis bounds the effect of error in the group proxy, which attenuates measured disparity toward parity. We map the results onto Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, the GDPR as interpreted in SCHUFA Holding, Directive (EU) 2023/2225, EBA loan-origination guidance, and Solvency II, EIOPA, and IAIS expectations, and propose fairness-risk controls organized around impact assessment, independent validation, and three lines of defense governance. Because the evidence comes from credit origination at a single lender, the insurance argument is developed at the level of regulatory and governance architecture rather than as an empirical transfer of estimates. Full article

Fixed full-boundary headlands create redundant non-working space in irregular fields, while working direction and swath offset jointly affect coverage quality and inter-swath turning cost. This study proposes a dynamic-headland-aware coverage path planning framework that allocates headland space according to swath endpoints and turning demand and generates parallel working swaths and Bézier U-turns for each direction–offset candidate. Field-specific empirical cumulative distribution function (ECDF)-midrank normalization converts turning distance, coverage error, and headland ratio into relative quality scores, and DQN-SHADE searches the resulting non-smooth discrete evaluation landscape. In geometric simulations on 24 actual field boundaries, DQN-SHADE achieved the lowest mean $\mathrm{gap}$ to the discrete reference optimum (0.0034) and the highest threshold success rate, ${\mathrm{SR}}_{5\times {10}^{-3}}$, of 80.83% among six stochastic optimizers under a common candidate-evaluation budget. Relative to fixed full-boundary headlands, dynamic headland allocation reduced the mean headland ratio from 13.30% to 7.52%, increased retained working area by 16,378.23 m² per field, and maintained 98.16% mean coverage. The proposed framework improves field-space utilization while maintaining coverage quality and geometric feasibility under the evaluated simulation conditions. Full article

Magnetorheological brakes (MRBs) use field-sensitive magnetorheological fluids (MRFs) to provide fast and controllable braking, but conventional electromagnet (EM)-based designs suffer from high power use, heat generation, and limited torque density, which limit their practical efficiency. These limitations highlight the need for improved MRB designs that enhance performance while reducing energy use and heat, as current studies have not fully achieved a balanced improvement in magnetic efficiency and thermal stability within a single system. To overcome these challenges, this study proposes a hybrid dual-disk MRB incorporating a segmented Halbach configuration of six permanent magnets (PMs) and six electromagnets. This design aims to improve magnetic-field distribution and overall system efficiency. Finite-element analysis shows that the proposed configuration effectively concentrates magnetic flux within the MR gap while reducing leakage, achieving a peak flux density of 0.78 T, approximately 55% higher than conventional EM-only designs. The system delivers a maximum braking torque of 16.93 Nm at 2 A, with a high torque-to-volume ratio. In general, the proposed hybrid Halbach MRB improves energy efficiency and thermal behavior while providing faster response and stable performance, with an essential fail-safe capability due to zero-field torque, making it suitable for advanced braking and actuation applications. Full article

Seedling growth is an essential stage for the future growth of trees, and Paulownia is treated as a rapidly growing hardwood tree species with wide industrial application. There is a knowledge gap in the modeling and analysis of the seedling growth pattern of Paulownia hybrid clones under field nursery conditions. A total of 300 saplings belonging to 25 Paulownia hybrid clones (named PT1–PT25, 12 saplings per clone) were observed 16 times after flat stubble treatment, yielding 4800 records. After accounting for mortality under field nursery conditions, 118 saplings survived, resulting in 1888 valid records for model construction. Seven theoretical growth equations with clone-specific dummy variables were fitted using nonlinear mixed-effects models (NLMM) with an AR(1) correlation structure to account for the hierarchical structure of the repeated measurements. Among the seven candidate models, Hossfeld IV was selected as the optimal model for both ground diameter (AIC = 8250.62, RMSE = 1.47 mm, conditional R² = 0.9892) and tree height (AIC = 15586.26, RMSE = 10.48 cm, conditional R² = 0.9916). However, because the Hossfeld IV model lacks an analytical third derivative, the standard logistic model was adopted to calculate the biologically meaningful fast-growing period boundaries. PT24 exhibited the highest growth potential among all clones, with asymptotic maximum values of 86.76 mm for ground diameter and 490.00 cm for tree height. However, its low survival rate (25%, n = 3) limits its practical recommendation. In addition, height-to-diameter (H/D) ratio analysis revealed considerable variation among clones, ranging from 3.89 cm∙mm⁻¹ (PT22) to 8.06 cm∙mm⁻¹ (PT3). The H/D ratio showed no significant correlation with ground diameter A (r = 0.286, p = 0.166), but a significant positive correlation with tree height A (r = 0.660, p < 0.001). PT24 maintained a moderate H/D ratio (6.92) despite its superior growth, indicating favorable stem stability. PT6 (91.67% survival, moderate growth) represents a more robust alternative for breeding programs prioritizing establishment success. This study demonstrates that a dual-model strategy provides a robust framework for evaluating clone-specific growth potential and supports informed breeding decisions in Paulownia improvement programs. Full article

Blockchain-based electronic voting has yet to reach the electoral environments that could benefit from it most. Existing systems are designed for well-connected, grid-powered urban settings and fail precisely where digital voting could widen participation the most; this gap motivates a framework engineered from the ground up for low-infrastructure conditions rather than one adapted to them. Across much of Africa, four deployment barriers stand in the way: intermittent connectivity that excludes an estimated 43% of the population, identity documents that remain valid for up to ten years and degrade conventional face recognition from 99% to below 80%, power demands of 200–250 W that rule out battery-powered operation, and cryptographic designs locked to a single algorithm regardless of operating context. This paper presents AdaptVote, a five-layer framework built on the premise that these barriers must be removed jointly rather than in isolation. At its core, the AI-FOLM algorithm anchors INT8-quantised ArcFace embeddings to age-stable craniofacial ratios on a Xilinx Zynq-7020 FPGA, reaching a 96.7% True Accept Rate at 0.1% False Accept Rate over 6–10 year age gaps—2.2 percentage points beyond the state of the art. A nullifier-based protocol allows ballots to be cast entirely offline with Merkle-tree integrity, an adaptive ECDSA/EdDSA/BLS layer cuts Ethereum settlement costs by 60–86%, and the complete polling station runs at 4.2 W peak power and 28 Wh per 12 h session, a 99.3% energy reduction over GPU baselines. A twelve-hour field deployment with 300 voters at Mohammed First University, Nador, Morocco, conducted under 35% network availability, recorded a 95.7% first-attempt authentication rate (95% CI: 92.7–97.5%), 99.2% system uptime, and a 53.3% reduction in average voting time relative to paper ballots ($p<0.001$). Together, these results suggest that trustworthy electronic voting can be engineered for, rather than merely adapted to, low-infrastructure electoral settings. Full article

The strip width in ultra-wide-strip tandem cold rolling changes not only the total rolling load, but also the transverse span over which the work roll (WR) is loaded. This study quantifies the isolated width effect on strip crown at 40 mm from the edge ($C_{40}$), flatness and WR elastic deformation in a 2180 mm CVC-6 tandem cold mill. A three-dimensional multi-stand elastic–plastic finite element (EPFE) model was established for five representative widths of 900, 1200, 1500, 1800 and 2100 mm, corresponding to contact-span ratios of 0.413–0.963. The results show that the strip width increased from 900 mm to 2100 mm, $C_{40}$ decreased from 20~80 μm to −50~−280 μm, and 1800 mm was the transition point from the positive crown to the negative crown. At the same time, the quadratic flatness component increased toward a center-wave mode, whereas the quartic component decreased toward an edge–center coupled-wave mode. Mechanistically, the relative WR axis deflection at the strip edge increased much faster than the local WR flattening compensation, producing an edge-open loaded roll gap. The findings indicate that strip width should be treated as an independent preset variable for WR bending, intermediate-roll bending and intermediate-roll shifting in ultra-wide cold rolling. Full article