Tracking the information about your manuscript
Communicate with the editorial office
Query manuscript payment status Edit officeCollecting, editing, reviewing and other affairs offices
Managing manuscripts
Managing author information and external review Expert Information Expert officeOnline Review
Online Communication with the Editorial Department
Visited
Page visits today: 31
About the Journal
Governed by: Jiangsu Education Department
Sponsored by: Nantong University
Published by: Editorial Office of Journal of Nantong University (Natural Science Edition)
Issues per year: 4
ISSN: 1673-2340
CN: 32-1755/N
Segmentation of diabetic macular edema based on federated learning
CHEN Qiong;SUN Jingbo;LI Junlin;SHU Jiachen;DENG Yunjun;CHENG Xi;CHEN ZongcunDeep learning technology plays a crucial role in the segmentation of spectral domain optical coherence tomography(SD-OCT) images for diabetic macular edema(DME). A DME segmentation algorithm based on federated learning(DMESA-FL) is proposed to address key challenges such as data privacy protection, computational cost control, and uncertainty quantification. Initially, a scale-aware pyramid fusion module and global pyramid guidance modules are incorporated into the convolutional neural network(CNN) to capture multi-scale contextual information and fuse the global contextual information flow with the features of the decoding path. Subsequently, the improved CNN is employed as the prediction model within the federated learning framework, and sequential training is adopted to update the global model, thereby enhancing data security. Ultimately, a feature discretization preprocessing module is introduced for all clients to reduce the computational burden of CNN and improve its generalization capability. During the feature discretization process, a fitness function based on rough sets is constructed to assess data uncertainty, and a genetic algorithm(GA) is utilized to search for the optimal breakpoints in SD-OCT images(the optimal feature discretization scheme for SD-OCT images). Additionally, an uncertainty constraint term is introduced into the loss function of the network for effectively integrating the average approximation precision of rough sets as prior knowledge into CNN. The comparative results between DMESA-FL and the state-of-the-art SD-OCT fundus image segmentation algorithms demonstrate that DMESA-FL can efficiently train models across different clients without data sharing,thereby achieving precise segmentation of DME.
Transformation methods from semi-three-way decision spaces to three-way decision spaces based on copula functions and their applications
WANG Yiding;QIAO Junsheng;LI Tengbiao;DING WeipingIn recent years, three-way decisions have achieved rapid development both in practical applications and theoretical research. In particular, as a distinctive and valuable extension of three-way decisions, three-way decision spaces have become one of the current research focuses. At present, research on three-way decision spaces mainly focuses on two aspects: 1) the transformation methods from semi-three-way decision spaces to three-way decision spaces based on common aggregation functions; 2) the construction methods of(semi-) three-way decision spaces based on fuzzy sets and their derived sets. Meanwhile, as a vital class of aggregation functions, copula functions have been widely applied in fields such as finance and insurance, yet they have not been introduced into three-way decision spaces. Therefore, given the significant role of common aggregation functions in advancing three-way decision spaces,this paper aims to conduct extended research on the aforementioned research topics in three-way decision spaces using copula functions. Specifically, firstly, three transformation methods from semi-three-way decision spaces to three-way decision spaces based on copula functions are proposed. Secondly, using copula functions, some construction methods of(semi-) three-way decision spaces based on fuzzy sets are proposed. Finally, comprehensive numerical experiments are conducted, including dataset experiments, comparative analysis, and sensitivity analysis of the proposed methods.The experimental results show that the proposed methods are feasible and effective, and the parameterin the proposed methods is also effective and stable.
A fuzzy concept-cognitive learning method integrating fuzzy coverage
WU Yuqing;LIN Yidong;LIANG TaojuConcept-cognitive learning is an emerging interdisciplinary research area that aims to continuously learn new knowledge by imitating the human cognitive process. However, existing concept-cognitive learning models usually ignore the local variability of objects in concepts, the redundancy of concept space, and the interpretability of concepts, which leads to model cognitive bias and underutilization of valid information. Therefore, a fuzzy concept-cognitive learning model integrating membership degree and coverage is proposed in this paper. Firstly, to enhance the representation capability of the concept extension, a membership function with an offset threshold is introduced to explore the correlation between objects and the concept. A membership matrix is then constructed to further transform the concept space into a fuzzy coverage. Secondly, high-correlation objects are filtered through the fuzzy β cut set, and the importance of different concepts is explored through coverage rates. This enables the construction of a core concept space, which effectively reduces the redundancy of the concept space and enhances cognitive learning efficiency.Finally, the proposed model is compared with four machine learning algorithms and two concept-cognition algorithms using the ten-fold cross validation method. The experimental results demonstrate that the model achieves higher average accuracy than the other comparative algorithms across 14 datasets, with the smallest performance fluctuation range across different datasets. Moreover, it maintains a leading position in terms of precision, recall, and F1 score,fully validating the feasibility and effectiveness of the proposed model.
Argument mining method in financial domain based on large language models
DING Fei;KANG XinIn recent years, financial text analysis has been shifting from coarse-grained sentiment polarity classification to fine-grained inference tasks focusing on logical relations between sentence pairs. To address limitations in existing approaches—such as the inability to capture semantic relations like ″Support″ ″Attack″ and ″Unrelated″ the lack of reasoning interpretability, and severe class imbalance—this study proposes a structure-enhanced Financial Prompt-based Argumentation Network(FinPromptNet). Built upon the LLaMA3-8B-Instruct large language model, FinPromptNet integrates structurally explicit prompt templates, chain-of-thought(CoT) reasoning guidance, partial parameter fine-tuning, and a joint strategy combining weighted sampling and cost-sensitive optimization. Experiments conducted on the NTCIR-17 FinArg-1 financial sentence-pair classification dataset demonstrate that FinPromptNet outperforms state-of-the-art baselines including FinBERT and T5 in terms of accuracy, micro-F1, and macro-F1. Specifically, it achieves a macro-F1 score of 67.1%, outperforming T5 by 7.3 percentage points, and yields over 10 percentage points of improvement in F1 for the underrepresented ″Attack″ class. These results highlight the effectiveness of structureaware prompt design and imbalance-aware learning in improving both model performance and interpretability for financial logical reasoning tasks.
4D radar-camera fusion algorithm for intelligent navigation of inland unmanned vessels
YAO Shanliang;GUAN Runwei;DING Weiping;YUE Yong;ZHU Xiaohui;CHEN SurongInland unmanned vessels demonstrate significant application value in environmental monitoring, water rescue, and transportation. However, their perception systems face critical challenges including water surface reflections,adverse illumination, and variable weather conditions. Current aquatic perception research primarily relies on camera or low-resolution radar data, which fails to meet multimodal sensing requirements in complex scenarios. To address these challenges, this research proposes a 4D radar-camera fusion solution. We first construct a multidimensional feature representation system for 4D radar that dynamically extracts key features including distance, azimuth, velocity,and reflection intensity. Subsequently, we design a dynamic scene-adaptive cross-modal fusion mechanism that employs attention-based weighting to effectively integrate different modal features, enabling real-time adaptive algorithms to handle environmental variations. The heterogeneous sensors are then deeply coupled at the feature level through a carefully designed deep learning model. Experimental validation demonstrates significant improvements in environmental adaptability, with the proposed fusion solution achieving target detection accuracy improvements of3.4 and 3.9 percentage points over traditional vision systems under poor lighting and harsh weather conditions,respectively. This research not only provides an effective technical solution for intelligent perception in complex aquatic environments, but also advances the development of inland unmanned vessels toward intelligent and autonomous operation.
A Hybrid CNN-LSTM based Method for Classifying Cardiac Arrhythmias from ECG
JU Xiaolin;ZHOU Xin;WANG Haochen;GAO Zhan;Cardiovascular disease is a leading cause of death， and electrocardiogram (ECG)， as a non-invasive examination， is crucial for early screening and follow-up. Multi-lead ECG signals， however， are often contaminated by power-line interference， baseline wander and electromyographic artifacts. In addition， marked inter-patient variability， class imbalance and inconsistent acquisition conditions across databases limit the robustness and generalization of existing automated methods. To address these issues， this paper proposes a dual-branch ECG intelligent diagnosis framework. The framework consists of a rhythm detection branch based on the MIT-BIH Arrhythmia Database and a disease classification branch based on the PTB-XL 12-lead database. In the rhythm branch， a CLA model that combines convolutional neural networks (CNN)， long short-term memory (LSTM) networks and an attention mechanism performs beat-level arrhythmia detection and abnormal location marking. In the disease branch， features obtained after preprocessing and principal component analysis (PCA)–based dimensionality reduction are input to an improved MLP+ model with a re-sampling strategy for multi-class cardiovascular disease classification. The two branches are trained and evaluated independently， and their outputs can provide complementary rhythm and disease information in clinical practice. Experimental results show that the proposed framework achieves high accuracy (98.97% ， 89.97%) and weighted F1-scores on both MIT-BIH arrhythmia detection and PTB-XL disease classification tasks， demonstrating good noise robustness and application potential.
[Downloads: 47 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Construction of clinical event prediction models for stroke patients based on multimodal fusion
WANG Muyu;YANG Yang;CHEN Hui;Predicting clinical events in stroke patients has the potential to reduce healthcare burdens and improve patient outcomes. However， existing predictive models typically depend on single-modality data (e.g.， medical images) and have yet to leverage multimodal medical data that could provide richer insights. To address this gap， methods for constructing stroke clinical event prediction models by integrating magnetic resonance imaging (MRI) scans， radiological reports， and laboratory test indicators are proposed and compared. Initially， three-dimensional convolutional neural networks， a Chinese pre-trained language model， and multilayer perceptron were employed to extract features from MRI images， radiological reports， and tabular data， respectively. Both concatenation-based and averaging-based multimodal fusion strategies are designed and evaluated. Subsequently， comparative experiments were conducted on prediction models utilizing single-modality， dual-modality， and tri-modality data to forecast whether patient hospital stays would exceed 7 days or 14 days. The findings indicate that for both prediction tasks， the multimodal model integrating images， text， and tables achieved an area under the receiver operating characteristic curve (AUROC) of 0.747 and 0.813， respectively， and the area under the precision-recall curve (AUPRC) were 0.580 and 0.542， respectively. The findings indicate that the multimodal model， incorporating images， text， and tables， attains an area under the receiver operating characteristic curve (AUROC) of 0.747 and 0.813 for the two prediction tasks， respectively. Additionally， the areas under the precision-recall curve (AUPRC) are recorded at 0.580 and 0.542， respectively， surpassing the performance of both the single-modal and dual-modal models. The multimodal feature fusion method using attention-weighted average demonstrate optimal performance (AUROC and AUPRC of 0.813 and 0.542， respectively)， indicating that multimodal data fusion effectively enhances the performance of clinical event prediction models.
[Downloads: 148 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Research progress on coalescence-based membranes for oil/water emulsion separation
WU Yilin;ZHANG Jichao;FU Shaohai;Traditional oil/water emulsion separation membranes are primarily based on the pore-size sieving mechanism， which makes it difficult to simultaneously achieve good separation efficiency and high permeation flux. To overcome this limitation and realize the concurrent enhancement of flux and efficiency during emulsion separation， coalescence-based oil/water emulsion separation membranes have been developed. By constructing special surface structures or incorporating functional components， these membranes can induce the enrichment and coalescence of emulsified oil droplets on the membrane surface， leading to the formation of larger droplets and thus enabling efficient and rapid demulsification. This paper briefly introduces the concept and fundamental principles of coalescence separation and analyzes the underlying mechanisms of coalescence behavior from the perspectives of surface wettability and structure， emulsion property， and flow environment. Furthermore， the structural characteristics and performance of different types of coalescence-based separation membranes are summarized， and the main challenges and future development directions of this technology are discussed. The research suggests that future efforts should focus on material innovation， interfacial regulation， and large-scale application. In addition， it is essential to further elucidate the intrinsic mechanisms of multiscale interfacial coalescence， thereby providing a solid foundation for the development of high-flux， good-efficiency， and environmentally sustainable membranes for oil/water emulsion separation.
[Downloads: 43 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Optimal Control Analysis of Predation System Based on Invasive Alien Species
WANG Han;ZHANG Yi;In the context of invasive alien species， a predator-prey system model based on dynamic changes in the food chain is studied. By simulating the interactions among three species， a nonlinear biological system with a functional response function is constructed. Firstly， the local and global stability conditions of the system at the positive equilibrium point are discussed in detail. Through introducing appropriate assumptions and parameter settings， sufficient conditions for the stability of the equilibrium point are proved. Secondly， the Pontryagin maximum principle is utilized to solve the optimal control problem of the system， proving the existence of the optimal equilibrium solution. The optimal control strategy and maximum benefit for optimizing the management and protection of the ecosystem are proposed. Finally， numerical simulations are conducted by comparing simulation results under different parameters to observe the dynamic behavior of the system. The optimal control is compared with no control through numerical simulation， obtaining the optimal capture time and maximum benefit， achieving species stability， and verifying the correctness of the theoretical analysis and the feasibility of the model.
[Downloads: 91 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Influence of the conjugacy class lengths of primary real elements upon the structure of groups
YANG Siqiang;LI Xia;LI Xianhua;Consider a finite group G， the relationship between the conjugacy class lengths of primary real elements and the structure of G， and the connection between the conjugacy class lengths of primary real elements and other real elements were discussed. Two conditions related to the conjugacy class lengths of real elements were defined， on this basis， the sufficient conditions for CG(U) to have a normal 2-complement is given， where U is a 2-subgroup of G meeting a specific condition about the conjugacy class lengths of real elements， also， the sufficient conditions for a particular subgroup of G to be a 2-group is given.
[Downloads: 23 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Application of Increasing PID Controlling Method in Temperature Controlling System
YAN Xiao-zhao,ZHANG Xing-guo(School of Mechanical Engineering,Nantong University,Nantong 226007,China)Temperature control is widely applied in scientific experiments and industrial processes.However,the temperature control system has characteristics of being nonlinear,time-varying and has hysteretic complicated large inertial system,and the effect of control is closely related to the algorithms adopted.In this paper,an experimental temperature control system is developed to meet the requirement of innovative ability training for mechatronic undergraduates.An increasing PID control algorithm is designed.The experimental results prove that the effect of the designed algorithm is better than the traditional PID algorithm.
Research and Development in Techniques of Dyeing Wastewater Treatment
JING Xiao-hui 1,YOUKe-fei 2,DING Xin-yu 2,CAI Zai-sheng 1(1.School of Chemistry and Chemical Engineering,Donghua University,Shanghai200051,China; 2.School of Chemistry and Chemical Engineering,Nantong University,Nantong226007,China)Reviewof the progress on treating methods of dyeing wastewater is presented,especially the advanced techniques are introduced,suchas membrane extraction,ultrasonic processes,high-energyphysical processes,advanced electrocatalytic oxidaˉtion processes and advanced photocatalytic oxidation processes.The treating trend for the dyeing wastewater is discussed.
The Application and the Development Foreground of Chitin and Chitosan
ZHANG Wei,LIN Hong,CHEN Yu-yue (School of Material Engineering,Soochow University,Suzhou 215021,China)The structure and the performance of chitin and chitosan are introduced and the application of chitin and chitosan in various fields is analyzed in this paper.The existing problems and the development foreground of chitin and chitosan are summarized in this paper.
Overview of SERS
LAN Yan-na,ZHOULing (Nantong Institute of Technology,Nantong226007,China)The principle of Raman spectrumis expounded first.Then the characteristics of SERS effect in experiment is summaˉrized and the mechanism of SERS is described.It's an accepted viewthat the mechanism of electromagnetic enhancement and the mechanismof chemical enhancement are both in existence.But which one is more important in different experiment depends on specific condition.
Application of ANSYS to Reinforced Concrete Beam
WANG Ya-ping 1,CHEN Jian-ping 2 ,CHEN Wu-zhou 3(1.Nantong Institute of Technology,Nantong226007,China;2.Nantong Architectural Design Institute of Industry,Nantong226001,China;3.Nantong Water Conservancy Construction Company,Nantong226005,China)In this article,taking features of reinforced concrete into account ,FEM software of ANSYS was used to calculate the beam' s deformation and the stress and strain of the normal section.At last,the answers to ANSYS(crack length,stress of reinforc-ing bar and concrete)and theoretical answers were compared in search of reasons and ways or measures that can amend it.
The Application and the Development Foreground of Chitin and Chitosan
ZHANG Wei,LIN Hong,CHEN Yu-yue (School of Material Engineering,Soochow University,Suzhou 215021,China)The structure and the performance of chitin and chitosan are introduced and the application of chitin and chitosan in various fields is analyzed in this paper.The existing problems and the development foreground of chitin and chitosan are summarized in this paper.
Application of Increasing PID Controlling Method in Temperature Controlling System
YAN Xiao-zhao,ZHANG Xing-guo(School of Mechanical Engineering,Nantong University,Nantong 226007,China)Temperature control is widely applied in scientific experiments and industrial processes.However,the temperature control system has characteristics of being nonlinear,time-varying and has hysteretic complicated large inertial system,and the effect of control is closely related to the algorithms adopted.In this paper,an experimental temperature control system is developed to meet the requirement of innovative ability training for mechatronic undergraduates.An increasing PID control algorithm is designed.The experimental results prove that the effect of the designed algorithm is better than the traditional PID algorithm.
Preparation of functionalized carbon nanomaterials and their energy storage applications
LI Qi;QIN Tian;GE Cunwang;Over the past decades, functional carbon nanomaterials(FCMs) have attracted much attention from the materials science community owning to their outstanding physical and chemical properties, such as high electronic conductivity/rapid mass transfer, plentiful active sites, good chemical stability, and robust mechanical stiffness. In view of the anisotropic and synergistic effects stemming from the functionalization as well as small size effect at the nanoscale,these multifunctional FCMs exhibit high potential especially in lithium-ion batteries, sodium-ion batteries, potassiumion batteries, lithium-sulfur batteries, organic solar cells, and supercapacitors. In this review, the functionalization strategies of carbon nanomaterials that have been developed over the last five years are comprehensively summarized and then application of FCMs in energy storage and conversion is introduced exhaustively. Finally, the pressing challenges and research directions are discussed according to the development trend.
Preparation of biochar and its application in environmental pollution management
WANG Jiayue;LING Qian;ZHANG Yunhao;WANG Xinyu;LIN Jiaqi;ZHANG Weitao;LIU Zhixin;WANG Xiangke;Biochar is a kind of environmentally friendly porous material which can be easily synthesized at low cost and in large scale. It has a wide range of applications in environmental pollution control and pollutants′ remediation and immobilization due to its large specific surface area and abundant surface functional groups. In this review, we mainly summarized the preparation of biochar, discussed the effect of preparation conditions on the properties of prepared biochar. The application of biochar in the removal of various pollutants from wastewater and soil improvement,the immobilization and elimination of different pollutants in soils, were reviewed in detail and the interaction mechanism was discussed. The removal of heavy metal ions was mainly attributed to the sorption of metal ions through the formation of surface complexes and part of metal ions could be reduced from high valence to low valence and then immobilized on biochar through adsorption-reduction-solidification strategy. The removal of organic pollutants from solution to biochar was mainly attributed to surface complexation, H bonding and π-π interaction on biochar surfaces.The organic pollutants could also be photocatalytic degraded by biochar or biochar-based materials under visible light irradiation. In conclusion, further research and discussion on the interaction mechanism of pollutant molecular with biochar at molecular level are helpful for the application of biochar in wastewater treatment and soil remediation. This review is of scientific significance for reducing the migration and transformation of pollutants in the environment and reducing the risk of environmental pollutants in the natural environment.
Modeling and simulation of proton exchange membrane electrolyzer system
WANG Huidong;YAO Haiyan;GUO Qiang;XIA Hongjun;Proton exchange membrane(PEM) electrolyzer converts electrical energy into chemical and heat energy,which is a green hydrogen production method, featuring fast response, high current density, compact structure, and other advantages. In the modeling of proton exchange membrane electrolysis water hydrogen production system,existing literature lacks a lumped parameter model that comprehensively describes the voltage and current changes of the electrolysis cell, as well as the temperature dynamics of each component of the system. This study establishes a steady-state voltage model of PEM electrolyzer and the thermal dynamic model of the system based on the basic principles of electrochemistry and the laws of thermodynamics. The simulation analysis was carried out based on MATLAB/Simulink software, and the simulation results were compared with the experimental data. The results showed that the voltage error is less than 0.02 V, and the temperature error is less than 1.6 K, which verifies the validity of the model. The established model can describe and predict the behavior of system parameters and provide support for system design and control. According to the efficiency model of PEM electrolyzer and the simulation results, the influence of different temperature and pressure on the performance of the electrolyzer was analyzed. It is concluded that increasing the temperature and decreasing the pressure can improve the efficiency of the electrolyzer, with temperature being the main factor. Using the simulation model, a feedforward PID controller was employed for temperature control, achieving an overshoot of less than 0.6 K and a settling time within 400 seconds. Comparison with a traditional PID controller demonstrates that the feedforward PID controller has advantages in terms of reduced overshoot and faster response.