Refinement Strategy Research Articles

Fully unsupervised wear anomaly assessment of aero-bearings enhanced by multi-representation learning of deep features

The assessment of wear anomalies using vision-based techniques is crucial for automated inspections of machine health conditions. Given the diversity of worn anomalies and the time-consuming nature of labeling, recent advancements in unsupervised anomaly detection have focused on comparing extracted features with those of normal images. However, this approach faces significant challenges in constructing a comprehensive comparison sample database and extracting distinctive features from worn surfaces with varying degrees of severity. To overcome these limitations, we introduce a fully unsupervised anomaly assessment method that detects anomalies solely based on single-bearing images, without the need for unworn samples. This approach eliminates the requirement for a comparison sample database by constructing an implicit feature database from the single-bearing surfaces using pre-trained convolutional neural networks and dimensionality reduction strategies. Additionally, to address the limited differentiability of wear characteristics, we develop a feature refinement strategy that incorporates multi-representation learning and topographical distribution characteristics of worn surfaces. By integrating the statistics of the refined feature database with uncertainty distribution-based normal feature weights, we construct an anomaly distribution map. This map enables us to assess the anomaly degree of a single-bearing effectively. To validate our method, we conducted experiments using real aero-engine bearings. The results demonstrate that our proposed approach can accurately assess anomaly degrees without pre-prepared comparison samples, achieving a detection performance of 90%, which fulfills the requirements for wear anomaly assessment.

Adaptive Deep Fourier Residual method via overlapping domain decomposition

The Deep Fourier Residual (DFR) method is a specific type of variational physics-informed neural network (VPINN). It provides a robust neural network-based solution to partial differential equations (PDEs). The DFR strategy is based on minimizing the dual norm of the weak residual of a PDE, which is equivalent to minimizing the energy norm of the error. To compute the dual norm of the weak residual, the DFR method employs an orthonormal spectral basis of the test space, known for rectangles or cuboids for multiple function spaces.In this work, we extend the DFR method with ideas of traditional domain decomposition (DD). This enables two improvements: (a) to solve problems in more general polygonal domains and (b) to develop an adaptive refinement technique in the test space using a Döfler marking algorithm. In the former case, we retain the desirable equivalence between the employed loss function and the H1 error under non-restrictive assumptions, numerically demonstrating adherence to explicit bounds in the case of the L-shaped domain problem. In the latter, we show how refinement strategies lead to potentially significant improvements against a reference, classical DFR implementation with a test function space of significantly lower dimensionality, allowing us to better approximate singular solutions at a more reasonable computational cost.

Multistage charging facility planning on the expressway coordinated with the power structure transformation

AbstractThis study presents a novel multistage expressway fast charging station (EFCS) planning problem coordinated with the dynamic regional power structure (PS) transformation. Under the prerequisite of the EFCS network's sustainable operation, network accessibility, and orderly construction, a three‐step planning method oriented to the enhancement of energy saving and emission reduction (ESER) benefits and rationalization of facility utilization is developed: (i) EV‐expanded network, (ii) multiagent‐based dynamic traffic assignment (MA‐DTA), and (iii) deployment refinement. Embedding the MA‐DTA and customized refinement strategy into the iterative planning structure enables the integration of operations and planning of the EFCS network. A numerical experiment and an empirical study in the Shandong Peninsula urban agglomeration are conducted. It demonstrates that the method can find a high‐quality solution within acceptable computation time and is applicable to realistic large‐scale EFCS planning. The planning method can effectively play the role of economy and facility in inducing EV users' charging demands, thus enhancing the overall ESER benefits. The integration of operation and planning is proven effective in reasonably matching the supply and demand of facility service and charging loads in a full‐time period. Further, the multistage EFCS planning schemes during 2025–2045 are explored, and some insightful policy implications are revealed.

Irregular text block recognition via decoupling visual, linguistic, and positional information

Scene text recognition has made great progress in regular formats, and the recent research has focused on irregular text recognition. In this work, we investigate a new challenge problem of recognizing a text block instance with irregular arrangement of characters, which is referred to as irregular text block recognition (ITBR). This problem is prevalent in daily scenarios, especially with the increasing use of rich text designs in signboards, logos, posters, and other mediums. The primary challenge arises from the weakened position clues and the highly complex reading order, which can often only be deciphered by a heavy reliance on understanding the intrinsic linguistic information. Hence, conventional recognition methods that employ inflexible character grouping rules, coupled with positional information, or constrained by vocabulary reliance, may struggle with the ITBR problem. To this end, we propose a progressive layout reasoning network (PLRN) to recognize the irregular text block by decoupling visual, linguistic, and positional information. PLRN comprises a character spotting module that recognizes the character set based solely on visual features with a new TopK-rank decoding mechanism, and a linkage reasoning module to interpret the character relationships within this set with a progressive refinement strategy. The linkages are initially reasoned by linguistic information and then progressively refined through the incorporation of proximity and tendency information, allowing for explicit decoupling and improved reasoning accuracy. To assess the effectiveness of the proposed method, we construct a new dataset called TextBlock600. This dataset consists of 600 images of irregular text blocks, each with complete sequence annotations. Experimental results demonstrate that PLRN shows promising performance in ITBR, opening up possibilities for further research in this field. Code and datasets will be available at https://github.com/eezyli/PLRN.

Grain refinement induced unusually large shape memory effect in lightweight titanium alloy

High-performance and lightweight shape memory alloys (SMAs) play a pivotal role in a variety of cutting-edge fields like aerospace, which however, are still quite difficult to be obtained until now. In the present work, we meticulously develop a novel lightweight Ti-12.5V-3.5Al-1Hf-0.8Fe-0.01B (at.%) SMA, which has a low mass density of ∼4.73 g/cm3. Prominently, distinct from the conventional wisdom in previous studies that is mainly based on eliminating grain boundaries or developing desired texture, we utilize the grain refinement strategy introduced by elaborate thermal-mechanical training, which effectively enables us to achieve exceptional shape memory response and strength-ductility combination simultaneously. It was revealed that a dominant β phase structure is achieved in this newly designed fine-grain SMA, accompanied by some α″ variants penetrating the grain boundaries. Compared to the coarse-grained counterpart, the deformation process becomes more intricate, involving stress-induced martensitic transformation and additional reorientations, which ultimately results in an unusually large recoverable strain up to 9.3 %. Furthermore, the fine-grain structure also contributes to an enhanced specific yield strength up to ∼159 MPa·m³/kg, while maintaining a large tensile ductility of ∼30 %. In light of its impressive structure-function integration and cost efficiency, this new-type SMA would emerge as a promising candidate for various advanced lightweight applications in a broad spectrum of fields.

Mapped material point method for large deformation problems with sharp gradients and its application to soil‐structure interactions

AbstractThe material point method (MPM) is often applied to large deformation problems that involve sharp gradients in the solution field. Representative examples in geomechanics are interactions between soils and various “structures” such as foundations, penetrometers, and machines, where the displacement fields exhibit sharp gradients around the soil‐structure interfaces. Such sharp gradients should be captured properly in the MPM discretization to ensure that the numerical solution is sufficiently accurate. In the MPM literature, several types of locally refined discretizations have been developed and used for this purpose. However, these local refinement schemes are not only quite complicated but also restricted to certain types of basis functions or update schemes. In this work, we propose a new MPM formulation, called the mapped MPM, that can efficiently capture sharp gradients with a uniform background grid compatible with every standard MPM basis function and scheme. The mapped MPM is built on the method of auxiliary mapping that reparameterizes the given problem in a different domain whereby sharp gradients become much smoother. Because the reparameterized problem is free of undesirably sharp gradients, it can be well solved with the standard MPM ingredients including a uniform background grid. We verify and demonstrate the mapped MPM through several numerical examples, with particular attention to soil‐structure interaction problems.

An adaptive continuous–discontinuous approach for the analysis of phase field fracture using mesh refinement and coarsening schemes and octree-based trimmed hexahedral meshes

An adaptive continuous–discontinuous approach for the analysis of phase field fracture using mesh refinement and coarsening schemes and octree-based trimmed hexahedral meshes

DIAGNOSTICS OF THE TRANSFORMATION OF INTELLECTUAL CAPITAL PROPERTIES IN UKRAINE

Abstract. The focus of this study is on the analysis of transformations of the nation's intellectual capital during the period preceding the conflict. It identifies and organizes modern methods of assessing intellectual capital, classifies indicators considering their impact on the total volume of intellectual capital, and implements the normalization of these indicators for creating radial diagrams and determining the state of intellectual capital in Ukraine. The effect of changes in intellectual capital on the fluctuations of the country's Gross Domestic Product (GDP) is also examined. Methodology. To achieve the objectives of the study, various methods were employed: dialectical method and synthesis for analyzing the evolution of human, organizational, and structural capitals; graphic analysis for the visual representation of data; and comprehensive and structural analysis with expert assessments to evaluate the impact of intellectual capital transformations on economic indicators. Scientific and Practical Significance. The work contributes to understanding the impact of changes in the structure of intellectual capital on GDP dynamics, which is key to developing an effective strategy for managing the country's intellectual resources. From a practical standpoint, the research aids in establishing criteria for evaluating intellectual capital and developing motivational mechanisms based on digital innovations and artificial intelligence for optimizing the structure of intellectual capital, paving the way for the development of a refinement strategy.

A study on the hydroacoustic characterisation of a cavitating propeller by dynamic adaptive mesh refinement technique

Underwater radiated noise (URN) of a marine propeller has received significant interest in recent decades due to its implications on marine fauna. Therefore, an accurate prediction of URN at an early stage of the propeller design is becoming imperative. This study presents a numerical investigation into the noise prediction of a marine propeller, including cavitation and a comparison with experimental test results obtained from the URN database from the King's College D (KCD) standard propeller series. Amongst the propellers tested in the series, the member KCD-193 was chosen to scrutinise in this study due to the significant variance of the cavitation types experienced by this propeller member and consequent characteristic variations observed in its URN spectral response. Numerical URN predictions of different flow conditions, represented by the advance coefficient and cavitation number, were conducted to investigate their effects on the noise spectrum. These predictions were compared with the experimental results to enable interpretation of the impact of various aspects of the simulation on URN prediction accuracy. In this investigation, one of the most prominent noise sources, tip-vortex cavitation (TVC), was identified as a critical aspect that needs to be captured by the numerical simulations for accurate URN predictions using CFD simulations. The influence of TVC on the spectrum was observed to be significant. The inception and stable presence of TVC dominated the frequency response of the broadband hump. In order to address this, a systematic adaptive mesh refinement strategy was implemented based on the vortex criterion to solve the flow characteristics in the propeller slipstream accurately. To further complement this task, a correlation between the cavitation bubble growth and collapse phenomenon by the sensitivity of the broadband hump on the spectrum was established based on the experimental results. The central frequency of the broadband hump was observed to vary with the advance coefficient and cavitation number. The reduction in the cavitation number resulted in a shift of this hump towards lower frequencies. The URN level of the hump decreased slightly in the high frequency by the reduction in the advance coefficient and the developing cavitation, demonstrating the cushioning effect on the spectrum. An accurate assessment of the noise spectrum, as far as numerical predictions are concerned, particularly on the broadband hump frequency bandwidth, was directly associated with the resolution of the TVC.

A Posteriori Error Estimation for Parabolic Problems with Dynamic Boundary Conditions

This paper is concerned with adaptive mesh refinement strategies for the spatial discretization of parabolic problems with dynamic boundary conditions. This includes the characterization of inf–sup stable discretization schemes for a stationary model problem as a preliminary step. Based on an alternative formulation of the system as a partial differential–algebraic equation, we introduce a posteriori error estimators which allow local refinements as well as a special treatment of the boundary. We prove reliability and efficiency of the estimators and illustrate their performance in several numerical experiments.

Dynamic crack growth in orthotropic brittle materials using an adaptive phase-field modeling with variable-node elements

In this paper, crack growth in orthotropic materials subjected to dynamic loading is numerically studied using an adaptive phase-field method. The study starts with a coarse structured mesh and the adaptive refinement strategy based on a user-defined threshold on the phase-field variable is proposed for computational efficiency, and variable-node elements are employed to treat the hanging nodes as a result of local adaptive refinement. The Hughes-Hilbert-Taylor (HHT) time integration scheme is adopted for temporal discretization. The directionality of orthotropic materials is represented by a penalized second-order structural matrix, which is incorporated in the crack face energy density. Through numerical examples, the influence of the material orientation on the dynamic crack growth in orthotropic materials is studied and the reliability of the proposed framework is validated.

Towards Efficient Diffusion-Based Image Editing with Instant Attention Masks

Diffusion-based Image Editing (DIE) is an emerging research hot-spot, which often applies a semantic mask to control the target area for diffusion-based editing. However, most existing solutions obtain these masks via manual operations or off-line processing, greatly reducing their efficiency. In this paper, we propose a novel and efficient image editing method for Text-to-Image (T2I) diffusion models, termed Instant Diffusion Editing (InstDiffEdit). In particular, InstDiffEdit aims to employ the cross-modal attention ability of existing diffusion models to achieve instant mask guidance during the diffusion steps. To reduce the noise of attention maps and realize the full automatics, we equip InstDiffEdit with a training-free refinement scheme to adaptively aggregate the attention distributions for the automatic yet accurate mask generation. Meanwhile, to supplement the existing evaluations of DIE, we propose a new benchmark called Editing-Mask to examine the mask accuracy and local editing ability of existing methods. To validate InstDiffEdit, we also conduct extensive experiments on ImageNet and Imagen, and compare it with a bunch of the SOTA methods. The experimental results show that InstDiffEdit not only outperforms the SOTA methods in both image quality and editing results, but also has a much faster inference speed, i.e., +5 to +6 times. Our code available at https://anonymous.4open.science/r/InstDiffEdit-C306

Cross-Gate MLP with Protein Complex Invariant Embedding Is a One-Shot Antibody Designer

Antibodies are crucial proteins produced by the immune system in response to foreign substances or antigens. The specificity of an antibody is determined by its complementarity-determining regions (CDRs), which are located in the variable domains of the antibody chains and form the antigen-binding site. Previous studies have utilized complex techniques to generate CDRs, but they suffer from inadequate geometric modeling. Moreover, the common iterative refinement strategies lead to an inefficient inference. In this paper, we propose a simple yet effective model that can co-design 1D sequences and 3D structures of CDRs in a one-shot manner. To achieve this, we decouple the antibody CDR design problem into two stages: (i) geometric modeling of protein complex structures and (ii) sequence-structure co-learning. We develop a novel macromolecular structure invariant embedding, typically for protein complexes, that captures both intra- and inter-component interactions among the backbone atoms, including Calpha, N, C, and O atoms, to achieve comprehensive geometric modeling. Then, we introduce a simple cross-gate MLP for sequence-structure co-learning, allowing sequence and structure representations to implicitly refine each other. This enables our model to design desired sequences and structures in a one-shot manner. Extensive experiments are conducted to evaluate our results at both the sequence and structure level, which demonstrate that our model achieves superior performance compared to the state-of-the-art antibody CDR design methods.

SD-MVS: Segmentation-Driven Deformation Multi-View Stereo with Spherical Refinement and EM Optimization

In this paper, we introduce Segmentation-Driven Deformation Multi-View Stereo (SD-MVS), a method that can effectively tackle challenges in 3D reconstruction of textureless areas. We are the first to adopt the Segment Anything Model (SAM) to distinguish semantic instances in scenes and further leverage these constraints for pixelwise patch deformation on both matching cost and propagation. Concurrently, we propose a unique refinement strategy that combines spherical coordinates and gradient descent on normals and pixelwise search interval on depths, significantly improving the completeness of reconstructed 3D model. Furthermore, we adopt the Expectation-Maximization (EM) algorithm to alternately optimize the aggregate matching cost and hyperparameters, effectively mitigating the problem of parameters being excessively dependent on empirical tuning. Evaluations on the ETH3D high-resolution multi-view stereo benchmark and the Tanks and Temples dataset demonstrate that our method can achieve state-of-the-art results with less time consumption.

An isogeometric approach to coupled thermomechanics in 3D via hierarchical adaptivity

In the present work, a complete numerical scheme to perform adaptive refinement and coarsening in the context of isogeometric analysis applied to thermomechanics is introduced. In particular, truncated hierarchical B-splines (THB-splines) are considered as a basis for the analysis and suitably graded isogeometric meshes are constructed to guarantee admissible configurations at each refinement and coarsening iteration. The algorithms are developed and implemented in a full three-dimensional framework and applied to solve coupled thermoelastic problems. The proposed numerical framework is verified on two relevant problems with respect to overkill solutions, obtained either with a commercial finite element software or with uniform isogeometric meshes. Different refinement and coarsening strategies are considered and compared. The obtained results demonstrate the ability of the THB-spline-based adaptive scheme to deliver robust discretizations guaranteeing a suitable trade-off between computational accuracy and number of degrees of freedom.

Low-cycle fatigue crack growth in brittle materials: Adaptive phase-field modeling with variable-node elements

This work presents an adaptive phase field framework for the simulation of crack nucleation and propagation in brittle materials subject to low-cycle loading. We introduce a fatigue history strain parameter within the phase-field framework to capture the fatigue effect. The resulting coupled differential equations are solved using a staggered iteration scheme. In order to improve the computational efficiency of the phase-field model, an adaptive refinement scheme is introduced. This scheme utilizes an artificial threshold that takes into consideration of both phase-field variables and cumulative fatigue history field variables to determine the elements to be refined. The handling of the hanging nodes resulting from mesh refinement is accomplished through the application of the variable-node element technique, offering a flexible and efficient mesh integration technique. We validate our proposed numerical scheme through five representative numerical examples: single-edge cracked specimen, compact tension specimen, double-edge cracked panel, plate with a hole and plate with multi-cracks and hole. The results demonstrate that the adaptive mesh refinement scheme significantly reduces computational costs without compromising the accuracy of the numerical predictions.

Estimation of possible locations for green roofs and bioswales and analysis of the effect of their implementation on stormwater runoff control

IntroductionThe objectives of this study are twofold. The first is to identify potential green infrastructure construction sites by building rooftops and sidewalks. The second is to analyze internal flooding for a wide range of drainage areas and to quantitatively evaluate the effectiveness of stormwater runoff control.MethodsThe target area has approximately 600 ha with a runoff coefficient of 0.71. Using Arc GIS Desktop 10.8.1, this study has identified green roofs and bioswales that would be highly beneficial in capturing large amounts of rainfall. In addition, Info Works ICM was used for the inundation analysis, which can simultaneously calculate the flow in sewer pipelines and above-ground inundation flow. Runoff coefficients were calculated for each land use using the urban land use subdivision mesh data with 100 m unit. This study targeted a 10-year probability rainfall (total rainfall: 86.3 mm, maximum hourly rainfall: 52.3 mm/h, duration: 3 h) with a middle concentrated rainfall waveform obtained from past experiments in the d4PDF database of ensemble climate prediction contributing to global warming.ResultsThe amount of land availability for green roofs and bioswales was about 1 and 0.1% of the drainage area, respectively. The runoff coefficients for green roofs only, bioswales only, with and without introduction of both green roofs and bioswales were 70.34, 70.87, 70.28, and 70.93%, respectively. The difference in runoff coefficients was about 0.65 percentage points even when both were constructed. As a result of inundation analysis, the reduction was 2.5% for the maximum waterlogged area, 1.5% for the flooded area, and 0.7% for the average depth of waterlogging divided by the maximum waterlogged area. The construction of green roofs and bioswales in the same area or downstream of the area shows little mitigation effect when flooding occurs in an area near the downstream end of the sewer network.DiscussionAlthough this study has mainly discussed the stormwater runoff control aspect, the most important feature of green infrastructure is its multifunctionality. In terms of utilizing and promoting green infrastructure, it is important to visualize its multifaceted effects and share them with many stakeholders.

New Adaptive Mesh Refinement Strategy for Entry Guidance via Sequential Convex Programming

This study focuses on the entry guidance problem, which is a highly nonlinear and constrained optimal control problem. The entry guidance problem is formulated as a nonlinear trajectory optimization problem with six-degrees-of-freedom dynamics, path constraints, and boundary constraints. The formulated problem is discretized and solved using sequential convex programming (SCP) with successive linearization. However, during the discretization process, conventional collocation methods encounter a challenge with high-frequency jittering in the control and bank angle profiles near active path constraints. Because the considered problem especially involves tight path constraints, the solution jittering issue becomes severe. Therefore, this study proposes a novel mesh refinement strategy to address this problem. Unlike conventional mesh refinement strategies based on interpolation error, the proposed approach resolves the convergence and solution jittering issue of the SCP framework by linking the direct and indirect methods with the costate estimation under successive linearization. Along with the development, we introduce a novel costate estimation method and provide proof of its validity. We present numerical simulation results of the proposed method and conduct a comparison with conventional approaches. The results obtained demonstrate successful mitigation of the solution jittering issue and a significant improvement in solution accuracy regarding compliance with the necessary conditions of optimality.

Assessment of various isogeometric contact surface refinement strategies

Assessment of various isogeometric contact surface refinement strategies

Survey on Stock Price Forecasting

This research project delves into the challenging domain of stock price prediction by harnessing the capabilities of neural networks. In an ever-evolving financial landscape, the project seeks to develop a robust model that can analyze historical stock data, identifying intricate patterns and trends to inform future price movements. Through the convergence of data science and financial analysis, this endeavor aims to provide investors with valuable insights, aiding them in navigating the complexities of the stock market. Employing machine learning algorithms, particularly neural networks, the model scrutinizes extensive sets of historical stock data, considering factors such as past price trends, trading volumes, and relevant financial indicators. Techniques such as regression, time-series analysis, and neural networks are leveraged to unveil complex relationships within the data, enhancing the model's predictive capabilities. The model undergoes meticulous training on historical data to glean insights from past market behaviors. Subsequently, it undergoes rigorous testing on unseen data to evaluate its predictive accuracy. Continuous refinement and optimization strategies are implemented to ensure the model's adaptability to the ever-changing dynamics of the market. Ultimately, this project aspires to furnish investors with a powerful tool for making informed decisions in the face of the unpredictable nature of financial markets. By integrating the precision of neural networks with the intricacies of stock market analysis, the research aims to contribute to the advancement of stock price forecasting methodologies.