To study the impact of land use types surrounding subway stations on passenger satisfaction, this research takes Guangzhou Metro Line 4 as a case. With the land use types around the stations as the independent variable and passenger satisfaction as the dependent variable, while considering the age, gender, education level, and occupation of respondents as covariates, the study employs Propensity Score Matching(PSM) for analysis. A main factor effect model for passenger satisfaction based on a counterfactual framework is established to calculate and match propensity scores, followed by a balance test to ensure the robustness of the findings. The results indicate that after matching, the standard deviations of all covariates are less than 10%, suggesting that the PSM method effectively eliminates the influence of these covariates. It provides a good representation of the impact of land use types on subway passenger satisfaction. Specifically, except for industrial land, which has a negative effect on passenger satisfaction, all other land use types exhibit positive effects. Among them, agricultural land has the largest average treatment effect on satisfaction. To enhance subway passenger satisfaction, measures such as improving the land use functions around stations, strengthening the connection between stations and surrounding lands, and targetedly optimizing environmental facilities can be adopted.

To address the shortcomings in time-series analysis and data imbalance research for operational monitoring data of wind turbine blades, this paper proposes a fault detection method for blade icing based on feature engineering and a HO-CNN-BiLSTM-Transformer framework. Firstly, feature engineering is employed by using the blade icing mechanism model to construct mechanism variables for blade icing. Secondly, a CNN-BiLSTM-Transformer detection model is developed to explore the temporal information in supervisory control and data acquisition (SCADA) system data. Finally, the Hippopotamus optimization (HO) algorithm is utilized to optimize the model’s hyperparameters, enhancing its diagnostic performance and generalization ability. Experimental results demonstrate that this detection method achieves precision, recall, and F₁ scores of 0.983 8, 0.990 2, and 0.987 0, respectively, outperforming other comparative models and optimization algorithms. This method provides valuable insights for optimizing maintenance strategies in wind farms, ensuring safe and efficient operation of wind turbines under cold conditions.

The flow of aqueous humor in human eyes is the key to maintain daily physiological activities. The systematic study of the dynamic behavior of aqueous humor in human eyes is of great practical significance to understand the pathogenesis of eye diseases and improve the therapeutic effect. In this paper, in order to study the influence of personalized geometric features of real human eye tissue on aqueous humor dynamics, three-dimensional geometric reconstruction and numerical simulation of real human eye tissue are carried out based on optical coherence tomography and lattice Boltzmann method, and some meaningful results are obtained. The angle of the anterior chamber is closely related to the flow pattern of aqueous humor. The decrease of the angle of the anterior chamber then leads to the change of flow pattern and the decrease of flow velocity. The maximum flow rate of aqueous humor is reduced from 9.65×10^-5 m/s to 3.10×10^-5 m/s after the current room angle is reduced from 41° to 9°. Corneal depression does not change the flow pattern of aqueous humor, but reduces the flow speed. The maximum flow velocity of aqueous humor decreases from 5.93×10^-5 m/s to 1.45×10^-5 m/s when the depression distance increases from 0 mm to 0.7 mm. Ectopia of pupil results in an asymmetric distribution of aqueous humor flow in the anterior chamber, and the flow pattern becomes more complex with the decrease of the anterior chamber angle.

The understanding of droplet bouncing behavior, which attracts widespread attention, has played a significant role in guiding technologies such as inkjet printing, drug delivery, and microfluidics. The wetting gradient formed by grooved surfaces with varying spacing causes droplets to exhibit complex bouncing behaviors in different directions. Numerical simulations are employed to study this phenomenon. In this work, the Lattice Boltzmann method based on the chemical potential model is used to model and to compare the bouncing behaviors of droplets impacting surfaces with both uniform and gradient-spaced grooves. The droplet’s contact angle, contact line, forces, internal momentum, and velocity vectors during the bouncing process are analyzed to investigate the causes of reverse bouncing. The results indicate that the droplet undergoes a series of asymmetric deformations on the asymmetric groove. When the Weber number of the droplet is 42.4, the droplet completes spreading and tends to bounce along the wettability gradient toward the hydrophilic direction after contraction. However, when the Weber number increases to 117.8, the liquid penetrates deeper into the groove, and the asymmetry becomes more pronounced. This leads to the pinning of the right contact line, and the resulting hysteresis force causes part of the droplet on the interface to contract and shift its center of mass to the right. Consequently, the droplet rebounds against the surface wetting gradient. This study provides insights for the design of microstructured surfaces that manipulate droplet motion.

Boundary treatment is a central issue in fluid flow modelling, especially critical in the lattice Boltzmann method (LBM). While the curved boundary conditions effectively improve the accuracy of complex geometrical boundaries in single-phase flow simulations, the conventional curved boundary conditions often lead to significant mass leakage and computational errors in multiphase flow simulations. This is mainly due to the reason that the nonlinear variation of density in the transition region caused by non-ideal effects is not considered in the traditional format of processing curvilinear boundary conditions. In this paper, non-ideal effects are introduced into the interpolation scheme, and a class of interpolation-based curved boundary algorithms for multiphase flow are proposed, including linear interpolation, quadratic interpolation and cubic interpolation schemes. Validated by a series of static and dynamic multiphase flow simulations with large density ratios, the method effectively improves the computational accuracy of the multiphase flow boundary conditions, with the required mass compensation approaching zero, and the imaginary velocity reducing to below by an order of magnitude compared with the previous method.

Text-to-image person retrieval, which can overcome limitations of traditional image-based methods, emerges as an innovative paradigm in smart city development. However, long-distance imaging and complex backgrounds in surveillance scenarios lead to scale inconsistency and feature contamination, hindering retrieval performance. This paper proposes a cross-modal person retrieval approach based on multi-resolution feature grounding, which effectively addresses detail loss and background interference through integrating multi-scale image feature representations with semantic segmentation boundary information. Two key innovations are introduced: 1) a multi-scale resolution input scheme that processes both low-resolution global features and high-resolution local features, 2) a semantic segmentation-based boundary grounding strategy that precisely segments pedestrian contours to suppress background interference. The Rank-1 accuracies on the CUHK-PEDES, ICFG-PEDES, and RSTPReid datasets are 70.58%, 60.88%, and 55.24%, respectively. Compared with recent methods, the proposed method demonstrates a relatively significant performance advantage in the cross-modal text-to-image person retrieval task.

In order to solve the problems of low detection accuracy, large model parameters and difficult detection of small targets in the detection of bearing appearance defects in the existing deep learning models, a lightweight detection algorithm SBSI-YOLO11 is proposed. Firstly, the SPD-Conv (space-to-depth convolution) module is introduced into the backbone network to reduce the resolution of the feature map, enhance the feature extraction and reduce the number of parameters of the model. Secondly, the Bidirectional Feature Pyramid Network (BiFPN) and the Attention Mechanism (SGE) are introduced in the Neck part to improve the detection performance of the model for small targets. Finally, the Inner-CIoU loss function is introduced to improve the positioning ability of the model. Experimental results show that compared with the basic model, the SBSI-YOLO11 model shows good comprehensive detection performance, with an mAP of 90.4%, an increase of 2.9 percentage points, a decrease of 11.5% in the amount of parameters, and a decrease of 12.7% in the amount of calculations, which can better meet the detection needs of the current industrial site for bearing appearance defects.

Lung cancer is the most deadly cancer in the world, and pulmonary nodules are the main manifestation of early-stage lung cancer. However, existing algorithms have problems with missed detections, false positives and high model complexity when detecting small lung tumors. To solve the above problems, a pulmonary nodules detection algorithm based on channel mask and dynamic channel pruning is proposed. Firstly, the algorithm uses cross-layer connection to fuse shallow features with deep features, and simplifies the path aggregation network, thereby obtaining more abundant features while reducing the number of model parameters. Secondly, the algorithm introduces the SE channel attention mechanism in the residual connection, and adaptively adjusting the weight of each channel to focus on the key information in the lesion in order to improve the detection ability of the algorithm for pulmonary lesions. Finally, the algorithm uses a channel mask for dynamic channel pruning, which completely retains the skip connections in the module, thereby enhancing the feature representation ability of the model. The proposed algorithm is compared with the YOLOv8n algorithm in the performance test on the LUNA16 dataset. The model weight of this algorithm is 0.3 MiB smaller than that of the YOLOv8n algorithm, and the recall rate and mAP@0.5 increase by 2.0 percentage points and 1.7 percentage points, respectively. The proposed algorithm is compared with the YOLOv8n algorithm in the performance test on the Lung-PET-CT-Dx dataset. The model weight of this algorithm is 0.9 MiB smaller than that of YOLOv8n algorithm, and the recall rate and mAP@0.5 increase by 0.8 percentage points and 0.4 percentage points, respectively. The experimental results show that the proposed model has the characteristics of higher accuracy and fewer parameters in lung cancer detection.

The framework of successor features(SFs) and generalized policy improvement(GPI) is recognized as a potential approach for achieving zero-shot transfer in reinforcement learning(RL) among different tasks. This paper investigates the underestimation phenomenon in SFs&GPI: Firstly, it is observed that the estimated Q-value is lower than the true Q-value for the new task during the training process. Then, to shed light on this issue, the expected gap between the estimated and true Q-value is theoretically analyzed, which is proven to be non-positive. Finally, the concepts of distributional RL are integrated into SFs&GPI, leading to the establishment of distributional successor features(DSFs) and distributional generalized policy improvement(DGPI), through which the underestimation gap is effectively narrowed. Experimental results on MuJoCo show that the DSFs&DGPI-based algorithm reduces value estimation bias, enhances transfer potential, and improves transfer stability compared to the SFs&GPI-based approach.

Panel data can explore and explain the dynamic changes and heterogeneity differences behind the data, making it a research hotspot in many fields.This article addresses the issue of variable selection in panel data with individual fixed effects. Firstly, a filter matrix is introduced to eliminate the fixed effects of panel data, and then estimates the regression coefficients and conducts variable selection through the adaptive LASSO-penalized weighted composite quantile regression method. Secondly, the weighted composite quantile regression applies data-driven weights to each different quantile regression. Thirdly, the results of numerical simulations indicate that this method outperforms the adaptive LASSO-penalized composite quantile regression method and the least squares method in both estimation accuracy and variable selection accuracy. Finally, the proposed method is applied to analyze international economic panel data, and the results show that the goodness of fit of the proposed method is higher than that of the other two methods.

In order to study the dynamic response of two kinds of viscoelastic axially moving plates, the governing equations of the axially moving plates are derived by using the variable-fractional model. Firstly, the variable-fractional differential equations are solved by using the shifted Bernstein polynomial as the basis function, and the operator matrix of the shifted Bernstein polynomial is derived. Secondly, the feasibility of the algorithm is verified by numerical examples. Thirdly, the displacement of ceramic plate and PET plate under harmonic load is further studied and compared. Finally the influence of axial moving speed on displacement of axial viscoelastic plate is analyzed, and the displacement changes of different material plates under different thickness changes are compared. The research results break through the variable-fractional order model of viscoelastic moving plate and provide a theoretical basis for the study of viscoelastic structure.

A dynamic Cournot game model with random cost is constructed. Based on the heterogeneous decision-making mechanisms of delay and gradient adjustment, the corresponding nonlinear dynamic system is established. Furfthermore, the equilibrium points of the system are solved and the dynamical stability of the Nash equilibrium point is analyzed.Through numerical simulation, the effects of the delay parameter, the product difference parameter and the random variable on system stability are studied. The conclusions are as follows. Firstly, the effect of the delay parameter on the stability of the system is non-monotonic. With the increase of the delay parameter, the stable region of the system is widened, but when the delay parameter increases to a certain extent, the stability of the system gradually weakens. Secondly, the larger the random variable is, the stronger the stability of the system is. Finally, the sensitivity of two players to the initial value disturbance is different, and the one with delayed rational behavior has stronger ability to resist the initial value disturbance.

To investigate the chemical constituents and biological activities of Curcuma kwangsiensis, the traditional and modern methods such as silica gel and semi-preparative high performance liquid chromatography were used to isolate and purify the ethanol extract of Curcuma kwangsiensis. Identification of the structure of compounds by modern spectroscopic means, and the anti-tumor and anti-inflammatory activities of the compounds were detected by MTS method and Griess method, respectively. A total of eighteen compounds were isolated from this plant and identified as(3R,5S)-3,5-dihydroxy-1-(4,5-hydroxy-3-methoxy-phenyl)-7-(4-hydroxy-2-methoxy-phenyl) heptane (1), 5-((2S,4R,6S)-4-hydroxy-6-(4-hydroxy-3-methoxyphenethyl)-tetrahydro-2H-pyran-2-yl)-3-methoxybenzene-1,2-diol (2), 5-((2S,4S,6S)-4-hydroxy-6-(4-hydroxy-3-methoxyphenethyl)-tetrahydro-2H-pyran-2-yl)-3-methoxybenzene-1,2-diol (3), broussonin B (4), ellisinin A (5), nyasol (6), genkwanin (7), 2,4-di-tert-butylphenol (8), labda-8(17)-,13(14)-dien-15(16)-olide (9),(E)-labda-8(17),12-dien-15,16-olide (10), glutinol (11), limonin (12), β-sitosterol (13), stigmasterol (14), daucosterol (15), berberine (16), sucrose (17) and palmitic acid (18). Compounds 1-12, 16, 17 were isolated from this plant for the first time, and compounds 4-7, 11, 12, 16 were isolated from Zingiberaceae for the first time. The results of antitumor activity showed that the IC₅₀ of compound 5 against leukemia HL-60, lung cancer A549, liver cancer HepG2, breast cancer MDA-MB-231 and colon cancer SW480 cell lines were 5.11, 44.82, 12.27, 28.82 and 15.78 μmol/L, respectively. The IC₅₀ of compound 6 on leukemia HL-60 and liver cancer HepG2 cells were 20.30 and 35.74 μmol/L, respectively. The IC₅₀ of compound 9 on leukemia HL-60 cells was 34.24 μmol/L, respectively. The IC₅₀ of compound 16 on leukemia HL-60, breast cancer MDA-MB-231 and colon cancer SW480 cells were 6.31, 30.57 and 31.39 μmol/L, respectively. The results of anti-inflammatory activity showed that the IC₅₀ values of compounds 9 and 16 against LPS-induced NO production were 45.76 and 44.67 μmol/L, respectively. The results of this study further enrich the chemical composition library of Curcuma kwangsiensis, and lay a foundation for its further research and development.

In this study, the activity and mechanism of a novel anti-tumor compound oxalamide derivative Z21 were studied. Activity screening showed that Z21 could effectively inhibit the growth of various tumor cell lines, especially human hepatoma cell line Hep G2 cells. With Hep G2 cells as experimental subjects, we observed cell morphology by microscope, detected cell viability by MTT assay and ATP assay, detected cell proliferation capacity by EdU incorporation assay and clonal formation assay, detected cell cycle by PI staining, detected cell apoptosis by Annexin V/PI staining, and detected the expressions of apoptosis-related proteins by Western blot. Mechanism studies showed that Z21 suppressed the proliferation of Hep G2 cells in a dose-dependent manner, reduced the expression of Bcl-2, promoted the expression of Bax and significantly induced cell apoptosis by caspase-dependent apoptosis pathway, but no significant cell cycle arrest was observed. These results indicate that Z21 contains great anti-tumor activity, and its ability to suppresses the proliferation of Hep G2 cells may be related to inducing cell apoptosis.

A strategy integrating network pharmacology, molecular docking, and in vitro experimental studies was employed to investigate the inhibitory effects and potential mechanisms of essential oils from Elsholtzia densa Benth. (EOE) and their core active components on the hepatocellular carcinoma (HCC) cell line SMMC-7721. Twenty-two components in EOE were identified by GC-MS, with 2-((3,3-Dimethyloxiran-2-yl)methyl)-3-methylfuran (33.99%), D-limonene (D-LIM, 32.20%), and perillaldehyde (PAH, 14.52%) as the predominant constituents. Network pharmacology revealed 14 active components targeting 123 hepatocellular carcinoma-related proteins. Core components PAH, D-LIM, and linalool (LNL) regulated cancer pathways, PI3K-Akt signaling, and apoptosis by interacting with TNF, IL1B, ESR1, and MAPK3. In vitro experiments demonstrated that EOE, PAH, D-LIM, and LNL concentration-dependently suppressed SMMC-7721 proliferation, showing 24 h IC₅₀ values of 87.4, 89.7, 121.9, and 418.7 μg/mL, respectively. EOE, PAH, and D-LIM further induced apoptosis and inhibited migration by upregulating Bax and Caspase3/9 and downregulating Bcl-2, MMP9, and CD44 expression. The results suggest that EOE and its core components PAH and D-LIM inhibit SMMC-7721 cells through apoptosis induction, proliferation suppression, and migration blockade.

China is the largest producer of edible mushrooms, producing hundreds of millions of tons of spent mushroom dreg every year. However, most of the spent mushroom dreg is directly discarded or incinerated, which not only pollutes the environment but also wastes its rich nutrients, so it is urgent to use resources. In this study, spent mushroom dreg was used as the main material, and the two nitrogen sources, chicken manure and cattle manure, were respectively used for cutting mechanical composting, in order to explore the decomposition effect of bacterial residue and different nitrogen source co-compost (initial C/N was 25) and the characteristics of microbial community structure. The results showed that after 85 days of fermentation, organic matter content ≥30%, total nutrient content ≥4%, pH 5.5-8.5 and germination index ≥70% were obtained from chicken manure residue and cattle manure residue treatment, all of which met the requirements of (NY/T 525-2021 Organic Fertilizer) and (NY/T 3442-2019 Technical Specification for Livestock manure composting). The growth rate of total nitrogen, total phosphorus, total potassium and available potassium content and germination index of chicken manure residue compost were significantly higher than those of cattle manure residue (P < 0.05), and the decline rate of C/N and organic matter were faster than that of cattle manure residue (P<0.05). The pH, conductivity (EC), germination index (GI), organic matter (OM) and other 12 indexes of the two compost treatments were evaluated, and the overall score of the principal component of the chicken manure residue treatment was higher than that of the cattle manure residue treatment. During the co-composting processes, the dominant bacterial phylum changed from Proteobacteria (45.01%) to Actinomycetes (31.84%) and Proteobacteria (23.71%), and the dominant fungous phylum changed from Ascomyces (60.69%) and Basidiomycetes (38.29%) to Ascomyces (91.22%). The dominant bacterial of the treatment changed from the initial Chloroflexi (72.21%) to Proteobacteria (27.38%) and Chloroflexi (23.99%), and the dominant fungous changed from Ascomycetes (82.84%) and unclassified_k_Fungi (13.28%) to Ascomycetes (94.31%). After co-composting, the bacterial diversity index of chicken manure mushroom dreg treatment increased from the initial 3.98 to 5.26, and that of cattle manure mushroom dreg treatment increased from the initial 2.79 to 5.67 (P<0.05). Random Forest model showed that organic matter, water content, carbon to nitrogen ratio, total nitrogen and total phosphorus contents were the key factors affecting microbial community richness and diversity during composting. In conclusion, the difference of nitrogen sources will lead to the difference of microbial communities and physical and chemical indexes in the composting process of spent mushroom dreg, which will affect the composting effect. Among them, the compost with chicken manure as nitrogen source has better rotting effect.

Intestinal bacteria of the Callimenellus fumidus Walker were isolated and purified by the dilution coating method and identified by a combination of morphological and molecular biological methods;at the same time, their antagonistic activity against plant pathogenic fungi was determined by the dual culture method,and the antifungal efficacy of the extracts of the fermentation products of the acvtive intestinal bacteria on plant pathogenic fungi was further evaluated by the mycelial growth rate method. The results showed that 76 culturable bacterial strains were isolated from the intestines of C. fumidus Walker,of which 11 strains were identified through morphological and molecular biological methods as belonging to 9 different genera,including Leucobacter, Microbacterium, Cellulosimicrobium, Arthrobacter, Lysinibacillus, Staphylococcus, Atlantibacter, Klebsiella and Enterobacter. The strain Klebsiella oxytoca BS.X3-A1 was found to have antagonistic activity against all the 9 test pathogens as determined by dual culture method. The inhibitory activity of the fermentation product of the strain against the 9 test pathogens was further determined. At a concentration of 4 g/L,the inhibition rates against Ceratocystis paradoxa,Pestalotiopsis theae,Alternaria citri,Cochliobolus miyabeanus,and Alternaria oleracea all reached 100%. The inhibition rates against anthracnose caused by Colletotrichum capsici,Diaporthe citri and Exserohilum turcicum exceeded 80%, while the inhibition rate against Phytophthora parasitica var. Nicotianae was relatively weaker at 54.68%. The virulence of the strain BS.X3-A1 against 9 phytopathogenic fungi was determined and found to have EC₅₀ values ranging from 0.394 8 to 3.908 9 g/L. Among them,it exhibited the strongest antifungal activity against Ceratocystis paradoxa with EC₅₀ values of 0.394 8 g/L,which was less than 0.5 g/L, followed by Pestalotiopsis theae and Alternaria citri with EC₅₀ values of 1.132 0 and 1.326 8 g/L,respectively both below 1.5 g/L. In addition,it also showed good antifungal activity against Exserohilum turcicum,Cochliobolus miyabeanus,and Diaporthe citri with EC₅₀ values of 2.225 6,2.226 3,and 2.327 2 g/L,respectively,which were all less than 3 g/L;while the antifungal activities against Alternaria oleracea and Phytophthora parasitica var. Nicotianae were relatively weaker,with EC₅₀ values of 3.071 8 and 3.908 9 g/L,respectively. The results showed that the intestine of C. fumidus Walker harbors bacteria with antifungal activity against plant pathogenic fungi,and these intestinal bacteria produce secondary metabolites effective against plant pathogenic fungi. The results of the study provide a reference for further exploration of antifungal compounds derived from the intestinal bacteria of the C. fumidus Walker.

The experiment was to study the strain and the best process of solid-state fermentation of citrus peel residue for protein feed. Using citrus peel residue as raw material, strains were isolated by sodium carboxymethyl cellulose and pectin medium, and the cellulase activity and pectinase activity of the strains were determined, and identified by morphological observation and molecular biology techniques. The results showed that a high cellulase-producing and pectinase-producing strains of Aspergillus niger HCR was screened and identified, with the filter paperase (FPase) , carboxymethyl cellulase (CMCase) and pectinase activities reached 6.548 9 IU/mL, 20.915 2 IU/mL and 12 223.50 U/mL, respectively. Applying three strains of Aspergillus niger HCR, Aspergillus oryzae MQ and Cyberlindnera Jadini 1769 and using the crude protein content of citrus peel pomace fermentation product as the evaluation indexes, the optimal combination of citrus peel pomace fermentation strains and inoculation ratio were screened by a one-way experiment. The results showed that the optimal solid-state fermentation conditions were as follows: the ratio of fermentation substrate citrus peel pomace to bran was 8∶2, the ratio of Aspergillus niger HCR : Aspergillus oryzae MQ : Cyberlindnera jadinii 1769 inoculated at the ratio of 3∶3∶2, the amount of inoculum was 10%, the content of urea was 1.5%, the water content was 60%, and the time of fermentation was 6 d at 30 ℃. Under the conditions, the crude protein of the fermentation product was 33.67%, and 72.75%, higher than those before optimization. It provides a basis for the solid-state fermentation of citrus peel waste to produce protein feed.

Waste leachate is a complex and degradable material that decreases with the increase of landfill time, and most of the non-degradable materials will cause great pollution to soil and groundwater. In this paper, the removal effects of two-stage Fenton oxidation, copper-based Fenton oxidation and copper-based Fenton-trichloroisocyanuric acid (TCCA) combined process on the organic matter in the waste leachate were investigated and the process conditions were optimized. The results showed that the combined copper-based Fenton-trichloroisocyanuric acid process had the highest removal rate of COD_Cr from the waste leachate. The optimal reaction conditions for copper-based Fenton were pH 5.0, CuSO₄·5H₂O dosage of 1.2 g/L, H₂O₂ dosage of 15 mL/L, and the optimal reaction conditions for trichloroisocyanuric acid were pH 5.0, trichloroisocyanuric acid dosage of 7 g/L. The removal rate of COD_Cr was 97.86%, and the residual COD_Cr content in the system was 55.11 mg/L, which was in accordance with the requirements of national emission standards.

To explore the application of algae-bacteria granular sludge in the treatment of cadmium-containing wastewater, a photosequence batch reactor (PSBR) was used to treat simulated cadmium-containing wastewater, and the changes of pollutant treatment effect, sludge physicochemical properties, surface functional groups and microbial community structure were observed. The deep removal mechanism in the process of treating cadmium-containing wastewater was analyzed. The experimental results showed that the algae-bacteria granular sludge had a good removal effect on cadmium-containing wastewater, in which the removal rates of COD, TN and Cd²⁺ reach 88.14%, 76.15% and 93.40%, respectively. The pseudo-second-order kinetic model indicated that the removal of Cd²⁺ was primarily through chemical adsorption, with EPS playing a crucial role. FTIR and XPS characterization revealed that functional groups such as C—O, C—N, and CO on the sludge surface interact with Cd²⁺, forming precipitates or complexes, which facilitated chemical adsorption. Under low Cd²⁺ concentrations, EPS secretion increased from 98.96 mg/g to 144.37 mg/g, demonstrating high resistance to toxicity. However, high Cd²⁺ concentrations inhibited EPS secretion, compromising particle stability. Microbial community analysis showed that under Cd²⁺ stress, the abundance of norank_f_Microscillaceae and norank_f_Saprospiraceae increased. These bacteria were major EPS producers, further enhancing EPS secretion and resisting heavy metal toxicity. In summary, algae-bacteria granular sludge exhibits significant removal capabilities for Cd²⁺, organic matter, and nitrogen and phosphorus, while maintaining particle stability. These findings provide a theoretical foundation and practical guidance for the long-term stable operation of sludge in treating Cd²⁺-contaminated wastewater.