Engineering reports : open access最新文献

{"title":"Enhancing Agricultural Management With Internet of Things and Deep Learning","authors":"Chinmaya Prasad Mohanty,&nbsp;Aparna Mohanty,&nbsp;G. Sumathi","doi":"10.1002/eng2.70966","DOIUrl":"https://doi.org/10.1002/eng2.70966","url":null,"abstract":"<p>Modern agriculture faces several critical challenges, including unpredictable weather conditions, inefficient resource utilization, and pest infestations, all of which negatively impact crop productivity and sustainability. Traditional farming practices often lack automation and data-driven decision-making, leading to increased operational costs, resource wastage, and reduced yields. Furthermore, many existing smart farming solutions remain fragmented, focus on isolated tasks, require high computational resources, and are not well-suited for resource-constrained environments, highlighting the need for a more integrated and cost-effective approach. To address these challenges, this work proposes an IoT-enabled smart farming system integrating machine learning, deep learning, and automation for agricultural management. The system automates operations such as plowing, sowing, irrigation, fertilization, pesticide, and weedicide application through an Android-based interface. A Naive Bayes model is employed for crop recommendation based on environmental parameters, enabling data-driven decision-making, achieving an accuracy of 96.5%. The system incorporates real-time soil moisture and temperature monitoring for automated irrigation control, improving irrigation efficiency, resulting in up to 90% improvement in water-use efficiency compared to conventional methods. Pest detection is performed using a hybrid approach combining Passive Infrared (PIR) sensing and sound analysis, achieving an accuracy of approximately 96%. Weed detection is performed using an Inception-based deep learning model, achieving a training accuracy of 99.9% and validation accuracy of 99%. Soil health assessment is performed using an optical transducer-based approach for estimating NPK levels, providing a low-cost, real-time solution. The integrated IoT–AI framework improves productivity, efficiency, and sustainability in resource-constrained agricultural environments.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70966","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148848746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Winter Natural Evaporation Experiment of Deep Brine With Sand-Gravel Pores in Da Lang Tan","authors":"Guo Min,&nbsp;Wang Peiyan,&nbsp;Zhang Jiaojiao,&nbsp;Xue Chaoqun","doi":"10.1002/eng2.70951","DOIUrl":"https://doi.org/10.1002/eng2.70951","url":null,"abstract":"<p>The deep sand gravel-type pore brine in the Qaidam Basin belongs to the unsaturated chloride-type brine with high calcium, high sodium, and low potassium. A total of 600 kg of gravel-type pore brine, located 1000–2000 m underground in the Dalangtan mining area, was selected for the test, evaporated, and crystallized in its natural state during summer. The distribution law of ions in the solid and liquid phases, the salt crystallization behavior, and the crystal morphology were analyzed, with a particular focus on the main precipitation characteristics of potassium. The results show that, when evaporated in the outdoor natural state, the salt precipitation process consists of four stages: a small amount of gypsum + rock salt → rock salt + potassium rock salt → carnallite + rock salt + bischofite → bischofite. The salt evolution law in the halite stage conforms to the phase diagram of Na+, K+, Mg²⁺//Cl− — H₂O. The potassium salt, carnallite, and bischofite stages meet the requirements of the K+, Ca²⁺, Mg²⁺//Cl− — H₂O system. There is a separate precipitation stage of potassium, during which potassium mainly precipitates in the form of carnallite, achieving a recovery rate of 89.62%. The potassium mixed salt contains K+ 8.75 wt.%, Na + 8.60 wt.%, and Mg²⁺ 6.95 wt.%, all existing in the form of chlorides. This makes it a high-quality raw material for the production of potassium chloride using the cold decomposition positive flotation process. When concentrated to a mother liquor density of 1.423 g/cm³ (22.4°C), calcium has not yet precipitated. The experimental results provide a reference for the development and utilization of this type of brine.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70951","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148848744","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Parametric Study of Phosphogypsum-Rock Wool-Phosphogypsum Sandwich Lightweight Partition Wallboards Under Impact Load","authors":"Min Huang,&nbsp;Jun Liu,&nbsp;Wenbin Pan,&nbsp;Xibin Yang,&nbsp;Qinglin Liu,&nbsp;Hao Wang,&nbsp;Jianshuo Wang","doi":"10.1002/eng2.70981","DOIUrl":"https://doi.org/10.1002/eng2.70981","url":null,"abstract":"<p>To investigate the mechanical properties of Phosphogypsum-RockWool-Phosphogypsum (P-RW-P) sandwich lightweight partition wallboards under impact load, a finite element model was built using ANSYS/LS-DYNA. After validating its accuracy, parametric analysis was conducted on five groups of 21 specimens. The effects of rock wool and phosphogypsum board thickness, impact object mass, velocity, and position on failure modes and dynamic responses were examined. Findings reveal that increasing the total board thickness enhances flexural stiffness, raises the maximum impact force and pseudo-static load, and lessens deformation. Thickening the phosphogypsum board is more beneficial for impact resistance. The maximum impact force grows with the impact object's initial kinetic energy. A large—mass impact has a longer duration than a high—velocity one. The failure modes of the partition board are different under various impact parameters. Moving the impact toward the fixed end makes the board more vulnerable. This study provides a reference for modular partition design in building engineering.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70981","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148752657","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research on Architectural Design Method Based on Machine Vision Fusion With Deep Learning","authors":"Chao Wang","doi":"10.1002/eng2.70974","DOIUrl":"https://doi.org/10.1002/eng2.70974","url":null,"abstract":"<p>In order to enhance the intelligence level of architectural design, a design method based on machine vision and deep learning fusion is proposed. Firstly, the overall framework of the method was constructed, and the original building images were processed by machine vision to form color features, texture features, and shape features. Among them, the color features of building images are obtained by extracting color histograms, the texture features of building images are obtained by calculating the energy and entropy of the gray level co-occurrence matrix, and the shape features of building images are obtained by calculating Hu moments. Secondly, a CNN deep learning framework is constructed by replacing VGG16 with VGG19, which has better performance. Deep convolution and pointwise convolution are used instead of conventional convolution to improve the efficiency of deep learning. The color features, texture features, and shape features of building images processed by machine vision are incorporated into deep learning based on VGG19 and fused to generate building design schemes. During the experiment, the V19-CNN method and IV19-CNN method were used as reference methods to design building components, building parts, and building as a whole, respectively. The experimental results show that our method has better performance compared to the V19-CNN method and the IV19-CNN method. This is because our method uses machine vision to obtain richer features, replacing direct input of the original building images. From the quantitative results, our method achieved an accuracy of over 90% in both local and overall design experiments, which is significantly better than the two reference methods.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70974","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148848591","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Computational Investigation of MHD Radiative Flow of Shape-Dependent \u0000 \u0000 \u0000 A\u0000 \u0000 \u0000 l\u0000 \u0000 \u0000 2\u0000 \u0000 \u0000 \u0000 \u0000 O\u0000 \u0000 \u0000 3\u0000 \u0000 \u0000 \u0000 $$ A{l}_2{O}_3 $$\u0000 and \u0000 \u0000 \u0000 F\u0000 \u0000 \u0000 e\u0000 \u0000 \u0000 3\u0000 \u0000 \u0000 \u0000 \u0000 O\u0000 \u0000 \u0000 4\u0000 \u0000 \u0000 \u0000 $$ F{e}_3{O}_4 $$\u0000 Nanofluids in A Semi-Porous Channel","authors":"Mahmmoud M. Syam,&nbsp;Sondos M. Syam,&nbsp;Muhammed I. Syam","doi":"10.1002/eng2.70990","DOIUrl":"https://doi.org/10.1002/eng2.70990","url":null,"abstract":"<p>This study investigates unsteady MHD radiative nanofluid flow in a semi-porous channel with expanding and contracting walls, focusing on <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>A</mi>\u0000 <msub>\u0000 <mrow>\u0000 <mi>l</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>2</mn>\u0000 </mrow>\u0000 </msub>\u0000 <msub>\u0000 <mrow>\u0000 <mi>O</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>3</mn>\u0000 </mrow>\u0000 </msub>\u0000 </mrow>\u0000 <annotation>$$ A{l}_2{O}_3 $$</annotation>\u0000 </semantics></math>-water and <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>F</mi>\u0000 <msub>\u0000 <mrow>\u0000 <mi>e</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>3</mn>\u0000 </mrow>\u0000 </msub>\u0000 <msub>\u0000 <mrow>\u0000 <mi>O</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mn>4</mn>\u0000 </mrow>\u0000 </msub>\u0000 </mrow>\u0000 <annotation>$$ F{e}_3{O}_4 $$</annotation>\u0000 </semantics></math>-water nanofluids. Unlike many earlier studies that treat nanoparticle shape only indirectly or assume spherical particles, this work incorporates the shape factor through the Hamilton–Crosser model to capture morphology-dependent thermal transport. The governing equations are transformed into similarity form and solved numerically using the modified operational matrix method. The results show that magnetic effects suppress the flow, that wall motion strongly alters both the velocity and temperature fields, and that particle shape significantly changes the effective thermal conductivity and heat-transfer rate. The study provides a more realistic description of coupled magnetic, radiative, geometric, and shape-dependent transport mechanisms, which is useful for thermal control and porous-channel applications.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70990","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148752656","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Design and Optimization of Low-Salinity Alkali–Surfactant Flooding for Chemical Enhanced Oil Recovery in Carbonate Reservoirs","authors":"Rasan Sarbast Faisal","doi":"10.1002/eng2.70953","DOIUrl":"https://doi.org/10.1002/eng2.70953","url":null,"abstract":"<p>Oil-wet carbonate reservoirs may also have low recovery efficiency in the Middle East where the reservoir temperature is high and there exists capillary trapping in addition to poor wettability. In this paper, I have estimated the performance of a hybrid low salinity waterflooding (LSWF) and alkaline-surfactant (AS) injection enhanced oil recovery (EOR) method. Laboratory coreflood experiments and numerical simulation were conducted on a representative carbonate reservoir from the Kurdistan Region of Iraq with a reservoir temperature of ˜100°C, and high-salinity formation brine. An anionic surfactant formulation that is thermally stable was mixed with sodium carbonate alkali and the salinity of the injection-brine optimized. The results demonstrate that low-salinity smart water can move the rock towards more water-wet conditions and the alkali-surfactant mixture can reduce the interfacial tension (IFT) by four to five orders of magnitude. The hybrid process was able to reduce the residual oil saturation from 0.32 ± 0.01 to 0.15 ± 0.01 during the coreflood experiments at 100°C. Hybrid LS + AS water flooding incrementals were 20.4% ± 1.0% OOIP in the tertiary mode, compared to 5.5% ± 0.7% OOIP and 9.0% ± 0.8% OOIP for low-salinity water flooding and high-salinity surfactant flooding, respectively. The total oil recovery after baseline waterflooding was 47.3% ± 1.1% OOIP, and 67.7% ± 1.3% OOIP after the treatment. Optimum concentrations of chemicals and water salinity are plotted against predicted recovery gains on the design charts. The study provides a mechanistic insight into the chemical EOR synergy in the carbonate reservoirs and provides useful guidelines for designing hybrid EOR pilots in the high temperature oil fields. Other issues such as scaling, corrosion, chemical cost, produced-water treatment and high pH fluid handling are also considered.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70953","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148752658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Physics Constraint-Guided Neural Network for Solving Partial Differential Equations","authors":"Zhihui Hou,&nbsp;Yaxu Peng","doi":"10.1002/eng2.70987","DOIUrl":"https://doi.org/10.1002/eng2.70987","url":null,"abstract":"<p>Partial differential equations (PDEs) are key tools for modeling continuous physical processes, but traditional solvers are costly for high-dimensional problems with complex boundaries. Existing neural network solvers usually add physical constraints only as loss terms, which limits physical constraint embedding learning and can cause local violations. To address this issue, this paper proposes a Physics Constraint-Guided Network (PCGN) for deep PDE solving. Its main novelty is to introduce physical information at three levels: feature representation, optimization, and output correction. First, governing equations, boundary conditions, and initial conditions are encoded into propagatable constraint features, and neighborhood propagation improves local consistency. Second, adaptive residual balancing adjusts different constraint terms, reducing instability from uneven residual scales. Third, a differentiable constraint projection layer corrects predictions toward feasible solutions. Experiments on Burgers' equation and Darcy flow show that PCGN achieves lower absolute and relative errors than existing deep learning solvers, while improving training stability and physical consistency.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70987","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148752644","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transient Conjugate MHD Flow With Variable Thermal Conductivity: A CFD to ANN Study","authors":"M. A. Forhad,&nbsp;A. K. Azad,&nbsp;M. Sakib Hasan,&nbsp;Md. Ibrahim,&nbsp;Aminur Rahman Khan,&nbsp;M. M. Billah","doi":"10.1002/eng2.70947","DOIUrl":"https://doi.org/10.1002/eng2.70947","url":null,"abstract":"<p>This paper investigates a novel problem of transient magnetohydrodynamic flow and thermal transfer in a hexagonal cavity featuring a star-shaped obstacle located at its center with variable thermal conductivity, a configuration not previously studied. The inclusion of variable thermal conductivity offers new insights for enhancing temperature regulation in intricate engineering systems. A time dependent analysis was conducted for a lid driven cavity applying radiative fluid of Fe₃O₄-CNT-kerosene hybrid nanofluids due to its consistent characteristics at higher temperature. Numerical simulation is conducted using finite element based COMSOL Multiphysics software of version 5.4. The analysis focuses on the effects of magnetic field strength (<i>Ha</i> = 25, 50, 75, and 100) and for four different solid materials such as <i>K</i>_<i>s</i> = 0.17 (PVC), 0.5 (plastic), 0.7 (sandstone), and 1.1 (concrete) for varying time <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mrow>\u0000 <mo>(</mo>\u0000 <mrow>\u0000 <mn>0</mn>\u0000 <mo>≤</mo>\u0000 <mi>τ</mi>\u0000 <mo>≤</mo>\u0000 <mn>10</mn>\u0000 </mrow>\u0000 <mo>)</mo>\u0000 </mrow>\u0000 </mrow>\u0000 <annotation>$$ left(0le tau le 10right) $$</annotation>\u0000 </semantics></math>. Numerical results are presented in terms of flow-thermal field characteristics over time. To support rapid predictions and design optimization, an artificial neural network (ANN) model is trained on the simulation data using the Levenberg Marquardt backpropagation algorithm. The findings show that increasing <i>Ha</i> causes a reduction in the average Nusselt number. Meanwhile, the drag force and pressure gradient around the obstacle increase with higher <i>Ha</i>. On the other hand, higher <i>K</i>_<i>s</i> boosts conduction in the star-shaped obstacle, raising the Nusselt number. ANN accurately predicted the average Nusselt number, drag force, average temperature, and pressure gradient for unseen <i>Ha</i>, <i>K</i>_<i>s</i>, and time values, demonstrating excellent generalization and computational efficiency. The combined numerical and ANN approach offers a robust framework for real-time thermal management, electromagnetic flow control, and smart material design optimization.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70947","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148848590","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Noise and Vibration Analysis of a Tri-Lobe Compressor for Industry Application: A Finite Element and Campbell Diagram Approach","authors":"P. Ashoka Varthanan,&nbsp;C. Rajendran,&nbsp;N. Babu,&nbsp;K. Mallieswaran,&nbsp;A. Johnson Santhosh","doi":"10.1002/eng2.70973","DOIUrl":"https://doi.org/10.1002/eng2.70973","url":null,"abstract":"<p>The tri-lobe compressors have noise and vibration as important issues, which have a lot of impact on the working reliability, energy efficiency, and the adherence to the industrial noise regulation. The paper is numerical research on noise and vibration characteristics in a tri-lobe compressor through finite element analysis through ANSYS. The modal, static structural, and Campbell diagram analyses were conducted to determine the relationship between structural movements and speed of operation at different rotational velocities. The findings reveal that the structural elements of the equipment like the casing, rotor, and the helical gear have characteristic vibrational modes, and a critical speed of about 243 rad/s is established because of the resonance of the backward whirl motion. The static structural analysis was used to verify that there was sufficient stiffness and low stress levels relative to structural integrity when subjected to steady-state loading. On the basis of the simulation results, design-based mitigation of the resonance avoidance, improved structural stiffness, and vibration isolation are discussed. The results are useful in optimization of the design of compressors to improve the attainment of low noise emission, low vibration, and improved operational reliability in industries.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70973","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148616895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Lightweight and Erudite Automated BiLSTM-Based Multi-Cycle GAN for Retinal Blood Vessel and Optic Disc Segmentation in Fundus Images","authors":"Rasmita Kumari Mohanty,&nbsp;Amaravarapu Pramod Kumar,&nbsp;Yuvaraju Macha,&nbsp;Agraharam Siva Kumar,&nbsp;J. Srinivas,&nbsp;G. Shyama Chandra Prasad,&nbsp;Rajesh Kumar Verma,&nbsp;Shashi Kant","doi":"10.1002/eng2.70923","DOIUrl":"https://doi.org/10.1002/eng2.70923","url":null,"abstract":"<p>Accurate segmentation of retinal blood vessels and optic discs is vital for the early diagnosis of ophthalmic conditions such as diabetic retinopathy and glaucoma. However, existing methods often struggle with computational complexity and poor generalization across diverse datasets. To address these challenges, we propose a novel Bidirectional Long Short-Term Memory (BiLSTM)-based Multi-Cycle Generative Adversarial Network (MCGAN). This architecture integrates the temporal modeling strength of BiLSTM with the adversarial learning capacity of a cycle-consistent GAN to improve the spatial coherence and accuracy of thin and thick vessel and optic disc segmentation in fundus images. The BiLSTM module captures spatial-sequential dependencies in retinal structures by encoding image data as ordered sequences, enhancing the continuity of thin and thick vessels and the delineation of anatomical boundaries. These extracted features are passed to MCGAN, which refines and reconstructs the segmentation mask. The discriminator network enforces structural realism by distinguishing between ground truth and generated segmentations. A multi-cycle consistency loss ensures the preservation of anatomical integrity through reconstruction of the original fundus image from the predicted mask, even under weak supervision. An attention-guided refinement module further improves focus on clinically relevant regions. The model is optimized using a composite loss function combining dice loss, focal loss, adversarial loss, and structural similarity measures to handle class imbalance and enforce semantic consistency. Extensive evaluations on publicly available datasets DRIVE, and STARE demonstrate that the proposed model achieves superior segmentation performance with reduced computational complexity. The proposed model achieves an average accuracy of 99.12%, recall of 98.78%, precision of 98.22%, f1-score of 97.13%, and JC of 96.67% on the DRIVE dataset, and 97.80% accuracy, 98.36% precision, f1-score of 98.74%, recall of 97.74%, and JC of 95.80% on the STARE dataset. The proposed framework offers robust, generalizable, and efficient segmentation capabilities, making it highly suitable for real-world ophthalmic diagnostic systems.</p>","PeriodicalId":72922,"journal":{"name":"Engineering reports : open access","volume":"8 8","pages":""},"PeriodicalIF":2.0,"publicationDate":"2026-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/eng2.70923","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148616909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}