Journal of Computational Science最新文献

{"title":"A Galerkin algorithm leveraging Bernoulli polynomials for accurate solutions of time-fractional diffusion-wave equations","authors":"R.M. Hafez ,&nbsp;M.A. Abdelkawy ,&nbsp;A. Biswas ,&nbsp;H.M. Ahmed","doi":"10.1016/j.jocs.2025.102607","DOIUrl":"10.1016/j.jocs.2025.102607","url":null,"abstract":"<div><div>This study presents a modified Galerkin technique utilizing Bernoulli polynomials for time-fractional diffusion-wave equations (TFDWEs). The proposed approach combines fractional calculus, namely Caputo derivatives, with a semi-discrete approach to achieve a high numerical accuracy. By utilizing Bernoulli polynomials as an efficient basis to approximate the solution, the algorithm transforms the governing equations into very sparse linear systems that can be solved computationally efficiently. Detailed numerical investigations, including applications to fractional wave equations and fourth-order diffusion-wave equations, demonstrate the method’s ability to achieve reduced errors and better computational efficiency. The results underline the stability and accuracy of the proposed technique, which turns out to be particularly suitable for simulating complex physical systems characterized by memory effects and anomalous diffusion.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102607"},"PeriodicalIF":3.1,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144306890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Realistic wildfire growth simulations applying a differentiable parametric representation of the fire front based on Composite Bézier curves","authors":"Irene González,&nbsp;Carlos Carrillo,&nbsp;Ana Cortés,&nbsp;Tomàs Margalef","doi":"10.1016/j.jocs.2025.102640","DOIUrl":"10.1016/j.jocs.2025.102640","url":null,"abstract":"<div><div>Modelling the evolution of a forest fire in Wildland Urban Interface (<em>WUI</em>) areas is still a major challenge in the field of forest fire simulation. Most existing forest fire spread simulators are based on polygonal representations of the fire perimeter, which often fail to capture the complexities of fire behaviour in these areas. Elliptical Wave Propagation (<em>EWP</em>) based simulators rely on this type of forest fire perimeter representation, that is, they represent the fire perimeter as a series of points connected by straight lines where the evolution of the fire front is performed by evaluating the spread of each perimeter point using as the spread direction the direction of the normal vector at each of them. To this end, <em>EWP</em>-based simulators have been built on top of the Richard model, which uses a differentiable parametric representation of the fire front. However, due to the polygonal representation used by <em>EWP</em>-based simulators, these cannot exploit the mathematical potential of using a parametric representation of the fire perimeter, which could compromise the accuracy of the simulations.</div><div>To address these limitations, we propose a novel parametric representation of the fire front using <em>Composite Bézier Curves</em> (CBC). The proposed wildfire perimeter representation improves the realism of the fire shapes being smooth and rounded. The first implementation of this proposal was done keeping the original method of normal vector calculation. This approach has been called <em>Composite Bézier Curves</em> using Neighbours (<em>CBCN</em>). However, an improved methodology has also been proposed where a more accurate method for calculating the normal vector directions is used, which is aligned with the curvatures of the fire front, thereby improving the overall modelling of fire dynamics. This advanced proposal has been called <em>Composite Bézier Curves</em> using Differentials (<em>CBCD</em>). Both proposed methodologies have been integrated into FARSITE, a well-known <em>EWP</em>-based forest fire spread simulator. Traditional polygonal representation (<em>LIN</em>) and the new CBC-based approach (<em>CBCN</em> and <em>CBCD</em>) were tested in ideal scenarios and two real cases. The obtained results show that any CBC-based representation generates more realistic fire shapes and they also enhance the simulator’s ability to model fire spread in <em>WUI</em> areas, with <em>CBCD</em> being the proposal that obtains the best results.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102640"},"PeriodicalIF":3.1,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144307164","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"PINN-parafoil: A physics-informed neural network method for complex parafoil dynamics simulating","authors":"Yu Yan ,&nbsp;Yufeng Fan ,&nbsp;Lulu Ning ,&nbsp;Caixia Su ,&nbsp;Pengju Wang ,&nbsp;Yongfeng Cao","doi":"10.1016/j.jocs.2025.102639","DOIUrl":"10.1016/j.jocs.2025.102639","url":null,"abstract":"<div><div>Accurately solving complex parafoil dynamics is essential for simulating para-foil system behavior. However, traditional numerical integration methods struggle with computational efficiency due to the high complexity of these models. This paper introduces a physics-informed neural network approach (PINN-Parafoil) that efficiently estimates numerical solutions for complex parafoil dynamics. By leveraging the superior function approximation capabilities of neural networks, PINN-Parafoil delivers near closed-form solutions, overcoming the computational challenges associated with conventional integration techniques. Unlike standard neural network methods, PINN-Parafoil incorporates the governing physical laws of parafoil dynamics as prior constraints, ensuring that the model outputs align with both training data and underlying physical principles. To validate this approach, the PINN-Parafoil model was trained and tested against the traditional Runge–Kutta solver for the 9-degree-of-freedom (DOF) parafoil model. Experimental results show that PINN-Parafoil achieves 25 times greater computational efficiency compared to traditional methods, while maintaining high accuracy with negligible numerical differences from true values. The resulting motion curves exhibit consistent dynamic characteristics with reference trajectories. Additionally, ablation studies highlight the critical role of physical constraints in enhancing model accuracy and stability. PINN-Parafoil offers a fast, accurate, and reliable proxy for simulating complex parafoil dynamics. Its efficiency and effectiveness make it a promising tool for various applications, including parafoil system design, trajectory planning, and homing control. This method provides robust technical support for both research and practical implementations in these fields, setting a foundation for further exploration and refinement of physics-informed neural network methodologies.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102639"},"PeriodicalIF":3.1,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291025","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A computational approach for solving the Gierer-Meinhardt (G-M) model in the context of biological pattern formation","authors":"Nek Muhammad Katbar ,&nbsp;Shengjun Liu ,&nbsp;Hongjuan Liu","doi":"10.1016/j.jocs.2025.102651","DOIUrl":"10.1016/j.jocs.2025.102651","url":null,"abstract":"<div><div>A mathematical framework with a particular focus on developmental biology can be attained from the Gierer-Meinhardt model, which explains the emergence of spatial patterns within biological systems. These patterns emerge when different chemical substances interact in a complicated manner, following a structured mathematical framework (Gierer-Meinhardt model), which helps explain how these patterns develop over time. The production of animal stripes on the skin and the organization of embryonic development are biological processes that usually involve these patterns. The present study conducts a detailed mathematical analysis of the Gierer-Meinhardt model by incorporating activation function such as radial basis function. The findings of the present study indicate that the radial basis function neural network is an effective tool for analyzing such complex mathematical models. By correlating the well-established biological models with computational tools like the ANN-RBF networks, new opportunities are created for examining the intricacy of living systems, and the foundation for further research in developmental biology and other fields.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102651"},"PeriodicalIF":3.1,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144307165","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unconditionally stable method for the high-order Allen–Cahn equation","authors":"Seungyoon Kang,&nbsp;Youngjin Hwang,&nbsp;Junseok Kim","doi":"10.1016/j.jocs.2025.102636","DOIUrl":"10.1016/j.jocs.2025.102636","url":null,"abstract":"<div><div>We propose an unconditionally stable algorithm for the Allen–Cahn (AC) equation that incorporates a high-order free energy. The high-order AC equation improves the preservation of interfacial dynamics and suppresses noise. The proposed method guarantees unconditional stability, which is essential for precise phase transition modeling and preserving detailed characteristics. To effectively solve the governing equation, it is divided into two subproblems, each of which is solved separately. The nonlinear operator is handled using a frozen coefficient method, followed by a closed-form solution. The linear operator is solved by applying the discrete cosine transform. To verify the effectiveness of the proposed algorithm, we carried out various computational simulations in two- and three-dimensional space. The proposed method ensures unconditional stability, and therefore allows stable solutions even with relatively large time steps. Moreover, we investigate the notable characteristics of the high-order AC equation, particularly its enhanced capability to effectively handle phase separation phenomena in the presence of significant noise and complex phase interfaces.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102636"},"PeriodicalIF":3.1,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144241866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DP-PINN+: A Dual-Phase PINN learning with automated phase division","authors":"Da Yan,&nbsp;Ligang He","doi":"10.1016/j.jocs.2025.102637","DOIUrl":"10.1016/j.jocs.2025.102637","url":null,"abstract":"<div><div>Physics-Informed Neural Networks (PINNs) are a promising application of deep neural networks for the numerical solution of nonlinear partial differential equations (PDEs). However, it has been observed that standard PINNs may not be able to accurately fit all types of PDEs, leading to poor predictions for specific regions in the domain. A common solution is to partition the domain by time and train each time interval separately. However, this approach leads to the prediction errors being accumulated over time, which is especially the case when solving “stiff” PDEs. To address these issues, we propose a new PINN training scheme, called DP-PINN+ (Dual-Phase PINN+). DP-PINN+ divides the training into two phases based on a carefully chosen time point <span><math><msub><mrow><mi>t</mi></mrow><mrow><mi>s</mi></mrow></msub></math></span>. The phase-1 training aims to generate the accurate solution at <span><math><msub><mrow><mi>t</mi></mrow><mrow><mi>s</mi></mrow></msub></math></span>, which will serve as the additional intermediate condition for the phase-2 training. The method for determining the optimized value of <span><math><msub><mrow><mi>t</mi></mrow><mrow><mi>s</mi></mrow></msub></math></span> is proposed in this paper. Further,new sampling strategies are proposed to enhance the training process. These design considerations improve the prediction accuracy significantly. We have conducted the experiments to evaluate DP-PINN+ with both “stiff” and non-stiff PDEs, including 1D Burger’s Equation, 1D Allen–Cahn Equation, 2D and 3D Navier–Stokes Equations (i.e., 2D cylinder wake and 3D unsteady Beltrami flow). We compared DP-PINN+ with the state-of-the-art PINNs in literature, including Time Adaptive PINN, SA-PINN, bc-PINN and NSFNets.The results show that the solutions predicted by DP-PINN+ exhibit significantly higher accuracy. This paper is extended from our conference paper published in ICCS2024 Yan and He (2024) <span><span>[1]</span></span>.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102637"},"PeriodicalIF":3.1,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144322595","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Physics-informed neural networks for compliant robotic manipulators dynamic modeling","authors":"Zhiming Li ,&nbsp;Shuangshuang Wu ,&nbsp;Wenbai Chen ,&nbsp;Fuchun Sun","doi":"10.1016/j.jocs.2025.102633","DOIUrl":"10.1016/j.jocs.2025.102633","url":null,"abstract":"<div><div>Deep learning is widely used in robotics, yet often overlooks key physical principles in dynamic modeling, leading to a lack of interpretability and generalization. To address this issue, recent innovations have introduced physics-informed neural networks (PINNs), which integrate fundamental physics into deep learning and offer significant advantages in modeling rigid-body dynamics. This study focuses on the application of PINNs to model compliant robotic manipulators. This requires extending PINNs to handle complex compliant dynamics. We propose an augmented PINN model capable of comprehensively learning manipulator dynamics, including compliant components. The model is tested on dynamic modeling of two physical compliant manipulators and a simulated manipulator. The results highlight its exceptional precision and generalization across a wide range of robotic systems, from purely rigid to compliant structures.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102633"},"PeriodicalIF":3.1,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144261851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Convolutional neural networks can diagnose schizophrenia","authors":"Murside Degirmenci ,&nbsp;Murat Surucu ,&nbsp;Matjaž Perc ,&nbsp;Yalcin Isler","doi":"10.1016/j.jocs.2025.102634","DOIUrl":"10.1016/j.jocs.2025.102634","url":null,"abstract":"<div><div>Schizophrenia is a severe mental disorder that affects how individuals think, perceive, and behave, often making accurate and timely diagnosis a significant challenge for clinicians. Traditional diagnostic approaches, such as interviews and psychological tests, have limitations in capturing the complex neurological underpinnings of the condition. In recent years, machine learning and deep learning techniques have shown promise in improving diagnostic accuracy across a variety of medical domains. However, relatively few studies have applied these methods to schizophrenia diagnosis, despite their potential. In this study, we investigate whether convolutional neural networks can effectively diagnose schizophrenia using publicly available EEG data. We achieved classification accuracies of 98.26% in subject-independent settings and 91.21% in subject-dependent settings on the test data, using a fully connected layer based on a Multi-Layer Perceptron classifier. These results appear promising when compared to the current state of the art.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102634"},"PeriodicalIF":3.1,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144221841","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A MATLAB code for finding the kernel of a simple polygon","authors":"Annamaria Mazzia","doi":"10.1016/j.jocs.2025.102616","DOIUrl":"10.1016/j.jocs.2025.102616","url":null,"abstract":"<div><div>This paper presents an algorithm for determining the kernel of a simple polygon. Traditional algorithms typically define the kernel by intersecting carefully chosen half-planes. In contrast, we explore a less-used approach as described in Zhao and Wang, (2010) , that handles concave vertices of the polygon as part of the kernel computation. This approach leverages two key techniques. First, it intersects the polygon’s interior with lines passing through edges adjacent to concave vertices. Second, it analyzes the orientations of two specific triangles identified by the sequence of vertices defining the line segment’s endpoints. This method for ray-line segment intersection plays a crucial role in efficiently determining the kernel. While the original approach effectively determines the kernel for a subset of simple polygons, it has limitations in handling all possible cases. This paper addresses these limitations by presenting a refined algorithm that expands the applicability of the method. The enhanced algorithm is implemented in MATLAB and validated through extensive testing to ensure its accuracy and efficiency.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102616"},"PeriodicalIF":3.1,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144205426","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhanced metabolite-disease associations prediction via Neighborhood Aggregation Graph Transformer with Kolmogorov–Arnold Networks","authors":"Pengli Lu ,&nbsp;Jian Zhang ,&nbsp;Wenzhi Liu ,&nbsp;Fentang Gao","doi":"10.1016/j.jocs.2025.102629","DOIUrl":"10.1016/j.jocs.2025.102629","url":null,"abstract":"<div><div>Metabolites are essential products of cellular chemical reactions, critical for sustaining life and reproduction. Research shows that metabolite concentrations in patients differ from those in healthy individuals, making metabolite-based disease prediction crucial for diagnosis and treatment. To address the limitations of current computational methods in accuracy and interpretability, we propose a novel Neighborhood Aggregation Graph Transformer method (AGKphormer). This method enhances link relationships by optimizing the minimum nuclear norm using the Alternating Direction Method of Multipliers (ADMM) and incorporates Fast Kolmogorov–Arnold Networks (FastKAN) to improve both accuracy and interpretability. We first construct a heterogeneous network based on the correlation and similarity between metabolites and diseases. Then, we utilize the ADMM algorithm to enhance link relationships by solving the minimum nuclear norm, reducing sparse relationships between nodes and providing richer features for neural network learning. For the features learned by the graph convolutional network (GCN), we employ a Graph Transformer augmented with FastKAN to learn long-range dependencies. This approach enables global feature embedding and addresses GCN’s smoothness issue while enhancing interpretability. Through five-fold cross-validation, AGKphormer achieved average AUC and AUPR values of 97.32% and 97.34%, respectively, outperforming most methods and demonstrating its effectiveness in predicting disease-related metabolites. Additionally, case studies further confirm that AGKphormer is a reliable tool for discovering potential metabolites.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"90 ","pages":"Article 102629"},"PeriodicalIF":3.1,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144241865","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}