Hard Region Research Articles

Towards explainable evaluation: Explaining predicted performance using local performance regions

Identifying which instances in a learning problem are difficult to be predicted by a model is important to avoid critical errors at deployment time as well as to plan how to learn an improved model (e.g., by training data cleaning or augmentation). Previous works have been mainly devoted on measuring instance hardness or developing meta-learners (e.g., assessors) to predict a base model performance based on the instances’ features while neglecting interpretability. In this paper, we propose a method to explain the performance of learned models in a problem based on the induction of meta-rules. Each meta-rule identifies a local region of instances, called Local Performance Region (LPR), where the base model has a predictable performance. The meta-rules are induced using a reduced number of attributes, in such a way that each LPR can be more easily inspected (e.g., by an attribute plot). The proposed method combines assessors, data augmentation and rule induction procedures. Initially, given a dataset of interest and a base model, we build an assessor model, which will be able to predict the base model’s performance for new instances. The assessor is trained based on the test results obtained when the base model is evaluated, thus generalizing the observed errors across instances in the dataset. In our work, we built an assessor in a case study to predict the probability of incorrect classifications of a Random Forest (RF) base model, achieving a mean absolute error of 0.05 in a hold out experiment. Once learned, the assessor is used to predict the model’s errors for new instances in an augmented dataset, covering a variety of features. Finally, meta-rules are learned to approximate the assessor’s predictions in local regions of instances. Experiments show the usefulness of the proposal by finding 18 local regions of bad RF performance, demonstrating a special case of LPRs, called Local Hard Regions (LHRs). By explaining the (in)correctness of model predictions, LPRs constitute a novel application in explainable AI, but focusing on explaining model performance, which can be adopted to different ML contexts.

Strength-ductility mechanism of combined heterostructured CoCrFeMnNi high-entropy alloy prepared by laser heat-treatment process

Two types of typical combined heterostructured CoCrFeMnNi high-entropy alloys with superior mechanical properties that evade the strength-ductility trade-off are fabricated using laser heat-treatment process. The lamellar grain structure with relatively high microhardness can hardly be observed due to the short time annealing at low temperature (873 K for 5 min) before laser scanning. The hetero-deformation induced strengthening, as well as the excellent deformation ability of the bottom hard region which can enhance the interaction between nanotwin-dislocation and maintains the large microhardness difference among the soft-hard layers, are the main factors that improve the yield strength of CoCrFeMnNi high-entropy alloy. Moreover, the combined heterostructured structure without lamellar deformed bands is demonstrated to have a relatively better uniform elongation, which is attributed to the stronger deformation ability of the non-recrystallized grains in the bottom hard region and the effective delay of the crack growth. Consequently, the combined heterostructured structure without lamellae is a promising way to simultaneously improve the strength and ductility, which can be efficiently prepared by the laser surface heat-treatment technique.

Microstructural evolution and interfaces in cold sprayed hybrid aluminum alloy powders of different hardness

The development of thick cold sprayed deposits of hard amorphous/nanocrystalline Aluminum High Entropy Alloys (Al HEA) is limited by their poor plastic deformability. To address this, Al HEA powder (Al90.05Ni4.3Y4.4Co0.9Sc0.35 (at. %)) is blended with soft Al 6061 in a 1:1 wt% ratio and cold sprayed. The microstructure of the composite deposits reveals that while Al HEA splats undergo limited deformation, Al 6061 splats deform extensively. Extreme deformation of Al 6061 captures the hard Al HEA particles, resulting in a composite deposit of 5 mm thickness, compared to 0.2 mm for pure Al HEA. Within the composite, Al HEA particles deform through shear band propagation and viscous flow, whereas Al 6061 splats deform via dislocation slip motion. These deformation mechanisms induce interface nanocrystallization, leading to a 123 % increase in lattice strain, a 67 % reduction in crystallite size, and an 800 % increase in dislocation density compared to the initial feedstock. This microstructural evolution, combined with the presence of hard Al HEA regions, results in a mean hardness of 2.56 GPa and an elastic modulus of 76 GPa. These values represent a 47 % and 6 % improvement over conventional polycrystalline cold-sprayed aluminum deposits. Thus, this study provides insights into the microstructural and interface evolution and its effect on indentation characteristics for making high-strength Al HEA deposits.

A novel Zr-based metallic glass composite and its plasticizing mechanism

A novel Zr-based metallic glass composite was prepared utilizing a die casting process in this study. This composite utilizes high-porosity open-pored foam metal as the second phase and takes advantage of the unique distribution characteristics of this phase to achieve structural characteristics resembling hard regions surrounded by soft regions. This facilitates the formation and interleaving of numerous shear bands during the deformation process of the composite, resulting in work-hardening capability and favorable compressive plasticity. The results indicate that the compressive strength of the metallic glass composites containing 2.72 % foamed Ni and 3.71 % foamed Cu respectively does not decrease significantly compared to single-phase metallic glass, while their compressive plasticity increases from 1.60 ± 0.46 % to 12.21 ± 1.04 % and 13.01 ± 1.15 %, respectively. This study not only uncovers the preparation method and plasticizing mechanism of a novel metallic glass composite, but also offers a fresh perspective for the design of high-plasticity metallic glass composites.

Interface characteristics and performance of additively manufactured steel-titanium bimetals prepared through composition gradient layers

In this study, laser-directed energy deposition (LDED), a powder-feeding laser additive manufacturing technique, was utilized to fabricate bimetals comprising high-strength steel (HSS) and Ti6Al4V (TC4). Through meticulous compositional control, four distinct regions (labeled Areas 1 through 4) comprising Fe-Ti intermetallics with varying sizes and morphologies were obtained at the LDEDed HSS/TC4 interface. These zones exhibited thicknesses of approximately 190 μm, 210 μm, 525 μm, and 205 μm, respectively. The effect of these intermetallics on the performance of LDEDed HSS/TC4 bimetals was investigated. The results indicated that most Fe-Ti intermetallics, characterized by reticular/granular morphologies or fine precipitates, were dispersed at the bonded interfaces (Areas 1, 3, 4). Nevertheless, irregular and long-striped C14_Laves phases were observed in Area 2, indicating the presence of a pure Fe2Ti region. Performance testing revealed that fractures predominantly occurred within a hard and brittle Fe2Ti region (Hardness: 891.5HV) during mechanical grinding or tensile clamping, highlighting the formidable challenge of achieving high-quality LDEDed HSS/TC4 bimetals through a component-regulated building strategy. The difficulty arises from the high Fe content in HSS and high Ti content in TC4, resulting in the formation of a pure Fe2Ti region at the interface. Thermodynamic calculations suggest that the composition-gradient-layer strategy may not be optimal for the integrated laser additive manufacturing of HSS and TC4. These findings provide valuable insights into the design of steel-titanium functional gradient materials and other multi-material components.

Sea Cucumber-Inspired Polyurethane Demonstrating Record-Breaking Mechanical Properties in Room-Temperature Self-Healing Ionogels.

Practical applications of existing self-healing ionogels are often hindered by the trade-off between their mechanical robustness, ionic conductivity, and temperature requirements for their self-healing ability. Herein, this challenge is addressed by drawing inspiration from sea cucumber. A polyurethane containing multiple hydrogen-bond donors and acceptors is synthesized and used to fabricate room-temperature self-healing ionogels with excellent mechanical properties, high ionic conductivity, puncture resistance, and impact resistance. The hard segments of polyurethane, driven by multiple hydrogen bonds, coalesce into hard phase regions, which can efficiently dissipate energy through the reversible disruption and reformation of multiple hydrogen bonds. Consequently, the resulting ionogels exhibit record-high tensile strength and toughness compared to other room-temperature self-healing ionogels. Furthermore, the inherent reversibility of multiple hydrogen bonds within the hard phase regions allows the ionogels to spontaneously and efficiently self-heal damaged mechanical properties and ionic conductivity multiple times at room temperature. To underscore their application potential, these ionogels are employed as electrolytes in the fabrication of electrochromic devices, which exhibit excellent and stable electrochromic performance, repeatable healing ability, and satisfactory impact resistance. This study presents a novel strategy for the fabrication of ionogels with exceptional mechanical properties and room-temperature self-healing capability.

Partially recrystallized heterostructure via Mo-doping endows CoCrNi medium-entropy alloy with exceptional strength and ductility

CoCrNi medium-entropy alloy (MEA) has excellent ductility, but the low strength at room temperature limits its practical usage. This study proposes an effective strategy to enhance the strength of CoCrNi MEA without apparent loss of ductility via partially recrystallized heterostructures resulted from doping of large atomic radius element, which was substantiated in a Mo-doping (CoCrNi)98Mo2 MEA. Compared with CoCrNi, the strength of (CoCrNi)98Mo2 is remarkably improved, maintaining the high uniform elongation of ∼24%. The yield strength and ultimate tensile strength increased from 950 MPa, 1287 MPa to 1131 MPa, 1564 MPa, respectively. The difference in deformation between the soft recrystallized region (SRR) and the hard non-recrystallized region (HNRR) leads to an excellent hetero-deformation induced (HDI) strengthening. Additionally, it is easier to reach the critical stress of twinning caused by high HDI stress at the interface of SRR/HNRR, which contributes to activation of high-density twins and produce dynamic Hall-Petch strengthening effect.

Comparative Analyses of Electro-Palatoraphic Features in Uighur and Chinese Fricative /s/

This article examines the Uighur and Chinese fricative /s/ with specific vowels by using Electropalatograph. By investigating similarities and differences in the alveolar region, blocking segments, and synergistic articulation, it proves that Uighur and Chinese fricatives /s/ are both lingual anterior consonants due to the same or similar articulatory places. During articulation, they share some commonality. Firstly, the contact between tongue and palate in the target frame is reduced when fricative /s/ is pronounced in both languages. Secondly, the tongue-palate contact primarily occurs in the gingival area. However, it has a larger Anterior Contact than Posterior Contact. Additionally, without complete obstructions between the tongue and palate, there is a gap between the tip of the tongue and the gingiva, allowing an airflow to articulate the fricative. Most important of all, a comparative analysis reveals distinctions between Uighur and Chinese fricatives /s/ from the aspect of synergistic articulation. The Uighur /s/ has tongue-palate contact with the soft palate in the hard palate region, which is not obvious in Chinese fricative /s/.

Геолого-геофизическое опробование железных руд гамма-альбедным методом

The standard method of testing iron ores, due to its high labor intensity and low representativeness, cannot serve as a basis for controlling the quality of ores in the process of their extraction and processing. The proposed spectrometric modification of the gamma-albedo method makes it possible to exclude labor-intensive operations for grinding ores, preparing analytical samples (~0.1 mm) and directly chemical analysis for iron. Studies of the spectral distribution of the scattered gamma radiation of the radioisotope source Se-75 (~220 keV) on ores with different iron content revealed critical energies and zones of their displacement depending on the quality of the ore and the energy of the scattered gamma radiation. The found critical energies in the spectral distribution of the albedo of gamma radiation made it possible to choose the optimal energy intervals in the hard (above the critical energy) and soft (below the critical energy) regions of the spectrum in terms of the maximum contrast ratio Ψ of the scattered gamma radiation intensities measured in the selected energy intervals, with a change in the concentration of iron. In the process of testing the proposed modification of the method for monitoring the iron content in partially prepared samples weighing ~ 60 kg, with a particle size of 40 mm, a root-mean-square error of 1.32% abs was achieved.

Super-resolution reconstruction of ultrasound image using a modified diffusion model

Objective. This study aims to perform super-resolution (SR) reconstruction of ultrasound images using a modified diffusion model, designated as the diffusion model for ultrasound image super-resolution (DMUISR). SR involves converting low-resolution images to high-resolution ones, and the proposed model is designed to enhance the suitability of diffusion models for this task in the context of ultrasound imaging. Approach. DMUISR incorporates a multi-layer self-attention (MLSA) mechanism and a wavelet-transform based low-resolution image (WTLR) encoder to enhance its suitability for ultrasound image SR tasks. The model takes interpolated and magnified images as input and outputs high-quality, detailed SR images. The study utilized 1,334 ultrasound images from the public fetal head-circumference dataset (HC18) for evaluation. Main results. Experiments were conducted at 2× , 4× , and 8× magnification factors. DMUISR outperformed mainstream ultrasound SR methods (Bicubic, VDSR, DECUSR, DRCN, REDNet, SRGAN) across all scales, providing high-quality images with clear structures and rich detailed textures in both hard and soft tissue regions. DMUISR successfully accomplished multiscale SR reconstruction while suppressing over-smoothing and mode collapse problems. Quantitative results showed that DMUISR achieved the best performance in terms of learned perceptual image patch similarity, with a significant decrease of over 50% at all three magnification factors (2× , 4× , and 8× ), as well as improvements in peak signal-to-noise ratio and structural similarity index measure. Ablation experiments validated the effectiveness of the MLSA mechanism and WTLR encoder in improving DMUISR’s SR performance. Furthermore, by reducing the number of diffusion steps, the computational time of DMUISR was shortened to nearly one-tenth of its original while maintaining image quality without significant degradation. Significance. This study demonstrates that the modified diffusion model, DMUISR, provides superior performance for SR reconstruction of ultrasound images and has potential in improving imaging quality in the medical ultrasound field.

Effect of heat treatments on inhomogeneous deformation of the melt-pool structure of Al–Si alloy manufactured via laser powder bed fusion

Laser powder bed fusion (L-PBF) is an additive manufacturing technology widely applied to the manufacture of metallic parts. The L-PBF processed aluminum-silicon (Al–Si) cast-type alloy components exhibit an anisotropic tensile ductility. In this study, we investigated the influence of heat treatments on the microstructural features of an Al–12%Si binary alloy fabricated via L-PBF and the associated inhomogeneous deformation of its melt-pool structure, which contributes to the resulting anisotropic tensile ductility. The mechanical inhomogeneity of the melt-pool structure and the change induced by annealing at various temperatures (300 and 530 °C) were characterized using microscale digital-image correlation analyses of scanning electron microscopy images obtained in situ during tensile deformation and nanoindentation hardness mapping. In the specimen fabricated via L-PBF, the high strain was concentrated at the locally coarsened microstructure (the soft region) along the boundaries of the melt pools rather than at the refined solidification microstructure (the hard region) within the melt pools. The localized strain varied depending on the geometrical relation between the orientation of the melt-pool boundary and the tensile direction, thus contributing to the anisotropic tensile ductility. This tendency appeared less pronounced in the specimen annealed at 300 °C, which exhibited a slightly homogenized melt-pool structure. The reduced strain localization is associated with a reduced difference in local strength between the melt-pool boundary (soft region) and the melt-pool interior (hard region). The slight homogenization of the melt-pool structure emphasized the effect of grain morphologies in the α-Al matrix on the inhomogeneous deformation within melt pools. In the specimen annealed at 530 °C, which exhibited a homogenized α-Al/Si two-phase microstructure, the uniformly distributed Si phase would be responsible for the homogenous deformation, resulting in an isotropic tensile ductility. This study advances our understanding of the correlation between the melt-pool structure and the deformation behavior of Al–Si alloys processed using L-PBF, which provides new insights for controlling the ductility by post-processing.

High Harmonic Generation Seeding Echo-Enabled Harmonic Generation toward a Storage Ring-Based Tender and Hard X-ray-Free Electron Laser

To align with the global trend of integrating synchrotron light source (SLS) and free electron laser (FEL) facilities on one site, in line with examples such as SPring-8 and SACLA in Japan and ELETTRA and FERMI in Italy, we actively explore FEL options leveraging the ultralow-emittance electron beam of the NSLS-II upgrade. These options show promising potential for synergy with storage ring (SR) operations, thereby significantly enhancing our facility’s capabilities. Echo-enabled harmonic generation (EEHG) is well-suited to SR-based FELs, and has already been demonstrated with the capability of generating extremely narrow bandwidth as well as high brightness, realized using diffraction-limited short pulses in transverse planes and Fourier transform-limited bandwidth in the soft X-ray spectrum. However, regarding a conventional EEHG scheme, the combination of the shortest seed laser wavelength (256 nm) and highest harmonic (200) sets the short wavelength limit to λ = 1.28 nm. To further extend the short wavelength limit down to the tender and hard X-ray region, a vital option is to shorten the seed laser wavelength. Thanks to recent advances in high harmonic generation (HHG), packing 109 photons at one harmonic within a few-femtosecond pulse could turn such a novel HHG source into an ideal seeding for EEHG. Thus, compared to the cascaded EEHG, the HHG seeding option could not only lower the cost, but also free the SR space for accommodating more user beamlines. Moreover, to mitigate the SASE background noise on the sample and detector, we combine the HHG seeding EEHG with the crab cavity short pulse scheme for maximum benefit.

Adversarial Attacks on Video Object Segmentation With Hard Region Discovery

Adversarial Attacks on Video Object Segmentation With Hard Region Discovery

Demarcation of groundwater potential zones in Ponnai sub-basin, Southern India using remote sensing, GIS and fuzzy-AHP approaches

ABSTRACT Groundwater is an essential source for household, irrigation and industry needs. The amount of groundwater and its quality are rapidly declining because of urbanization and industrialization. Groundwater resources in hard rock regions should be handled properly to maximize their utilization. This work aims to demarcate the Groundwater Potential Zones (GWPZ) in a part of the Palar River Basin (Ponnai sub-basin) by combining Remote Sensing (RS), Geographic Information System (GIS) and Fuzzy-Analytical Hierarchy Process (FAHP). The factors which influence the groundwater like geology, land use/land cover (LULC), geomorphology, slope, soil, rainfall, lineament density (LD), drainage density (DD), normalized difference vegetation index (NDVI), topographic wetness index (TWI) and topographic position index (TPI) are considered for mapping of GWPZ. The weightage is allocated to these parameters on the basis of the FAHP technique and the obtained groundwater potential map was classified as, very poor (13.5%), poor (21.76%), moderate (31.5%), good (20.66%) and very good (8.54%). The GWPZ map was validated with the actual well yield of the Ponnai sub-basin. The research work will serve as a guideline for the effective use of groundwater and long-term sustainable management of the aquifer in the Ponnai sub-basin.

Bifocal photon sieve imaging in the hard x-ray region.

Hard x-rays are widely used for plasma diagnosis, nondestructive inspection, and high-resolution x-ray imaging. A typical x-ray source is a tabletop micro-focus x-ray source. Here, a bifocal photon sieve (PS) with the smallest diameter of 59.6 nm was designed and fabricated by electron-beam lithography to focus hard x-rays on variable-resolution array images. An imaging experiment at 8.39 keV demonstrates that the designed and fabricated PS has two different focal lengths. The numerous pinholes that can be optimized provide richer degrees of freedom to realize considerably more functionalities. A multi-focal PS provides the possibility of splitting x-rays and further extends interferometry from visible light to hard x-rays.

Microstructure and Properties of TiCp/Fe Hierarchical Composites Prepared by a New Pressure Infiltration Method.

TiCp/steel composites are conventionally produced via powder metallurgy. In this paper, a liquid pressure infiltration method was developed to prepare a kind of spherical hierarchical architectured composite, in which spherical TiCp-rich hard phase regions were uniformly dispersed in TiCp-free soft phase region. The microstructure and mechanical properties of the architectured composites were carefully studied and compared with the common composite, as well as the effect of TiCp fraction on the properties. The results show that architecturual design can effectively improve both the toughness and strength of the composites. With TiCp content increasing from 30% to 50%, both the bending strength and the impact toughness of the architectured composites first increase, then decrease, and reach the highest at 40% TiCp. The highest impact toughness reaches 21.2 J/cm2, being 6.2 times that of the common composite and the highest strength being 67% higher. The pressure infiltration method possesses adaptability to varying shapes and sizes of the products, allowing for large-scale preparation. Therefore, for the first time, the combination of pressure infiltration preparation and architectural design was applied to TiCp/steel composites.

Influence of HAZ microstructure on RCF under twin-disc test of a flash-butt welded rail

A new twin-disc test was performed using discs containing all microstructures of the heat-affected zone (HAZ) of flash-butt welded rails. It was observed that the lower hardness regions (spheroidized cementite in a ferritic matrix) at the HAZ boundary underwent a local deformation, generating valleys and reproducing a common defect type observed in rail welds: double squat-like defect. Longer crack lengths or crack clusters were observed in the HAZ boundary and central weld regions. The cross-section metallography showed that these cracks are directly related to the microstructure of spheroidized cementite at the HAZ edges or pro-eutectoid ferrite at the central region. The decrease in surface roughness before the cracks, the presence of oxide on the bottom crack surface, and superior spheroidized cementite microstructure in the upper crack region suggest that these larger/cluster cracks are predominated by extrusion of the softer material above the harder material (pearlitic/base metal).

Functional matching and renormalization group equations at two-loop order

We present a systematic method for determining the two-loop effective Lagrangian resulting from integrating out a set of heavy particles in an ultraviolet scalar theory. We prove that the matching coefficients are entirely determined from the (double-)hard region of the loop integrals and present a master formula for matching, applicable to both diagrammatic and functional approaches. We further employ functional methods to determine compact expressions for the effective Lagrangian that do not rely on any previous knowledge of its structure or symmetries. The same methods are also applicable to the computation of renormalization group equations. We demonstrate the application of the functional approach by computing the two-loop matching coefficients and renormalization group equations in a scalar toy model.

MicroSegNet: A deep learning approach for prostate segmentation on micro-ultrasound images

Micro-ultrasound (micro-US) is a novel 29-MHz ultrasound technique that provides 3-4 times higher resolution than traditional ultrasound, potentially enabling low-cost, accurate diagnosis of prostate cancer. Accurate prostate segmentation is crucial for prostate volume measurement, cancer diagnosis, prostate biopsy, and treatment planning. However, prostate segmentation on micro-US is challenging due to artifacts and indistinct borders between the prostate, bladder, and urethra in the midline. This paper presents MicroSegNet, a multi-scale annotation-guided transformer UNet model designed specifically to tackle these challenges. During the training process, MicroSegNet focuses more on regions that are hard to segment (hard regions), characterized by discrepancies between expert and non-expert annotations. We achieve this by proposing an annotation-guided binary cross entropy (AG-BCE) loss that assigns a larger weight to prediction errors in hard regions and a lower weight to prediction errors in easy regions. The AG-BCE loss was seamlessly integrated into the training process through the utilization of multi-scale deep supervision, enabling MicroSegNet to capture global contextual dependencies and local information at various scales. We trained our model using micro-US images from 55 patients, followed by evaluation on 20 patients. Our MicroSegNet model achieved a Dice coefficient of 0.939 and a Hausdorff distance of 2.02 mm, outperforming several state-of-the-art segmentation methods, as well as three human annotators with different experience levels. Our code is publicly available at https://github.com/mirthAI/MicroSegNet and our dataset is publicly available at https://zenodo.org/records/10475293.

Less Emphasis on Hard Regions: Curriculum Learning of PINNs for Singularly Perturbed Convection-Diffusion-Reaction Problems

Less Emphasis on Hard Regions: Curriculum Learning of PINNs for Singularly Perturbed Convection-Diffusion-Reaction Problems