Search Results (1,104)

Vertical-axis wind turbines (VAWTs) have emerged as a promising alternative for wind energy generation, particularly in offshore environments. However, their reliability continues to be limited by a critical component: the main bearing, which constitutes a major bottleneck in operation and maintenance (O&M). Analysis of its loading and operational conditions reveals complex scenarios, notably uneven radial load distributions on bearing elements, aerodynamic vibrations, and particular rotational dynamics, which render the VAWT main bearing highly susceptible to failure. To identify the damage mechanisms influencing its reliability, this review adopts a systematic comparative approach that leverages extensive failure data from various bearings used in horizontal-axis wind turbines (HAWTs). The findings indicate that these bearings are prone to failure during both operational and stationary periods, primarily due to contact fatigue and fretting damage, with stationary failures in offshore environments potentially progressing into tribocorrosion phenomena. By clarifying the domain of applicability of life prediction existing models and mitigation strategies, this review provides a structured framework for interpreting reported bearing failures in VAWTs and for guiding future experimental, modelling, and design efforts aimed at improving bearing reliability. Full article

The Middle Cambrian salt–anhydrite succession in the Tarim Basin has been regarded as an effective regional cap-rock. However, numerous Ordovician hydrocarbon reservoirs have been discovered above the anhydrite, and recent drilling has identified industrial oil and gas flows beneath anhydrite-bearing intervals. These findings call into question the sealing effectiveness of anhydrite rocks in deep subsalt settings. In this study, X-ray diffraction (XRD), petrographic analysis, scanning electron microscopy (SEM), and triaxial compression tests were conducted to investigate the mineral composition, deformation behavior, and failure mechanisms of anhydrite rocks. The results indicate that: (1) dolomite-bearing anhydrite undergoes plastic deformation at depths greater than 4400~4600 m (~70 MPa confining pressure), whereas dolomitic anhydrite enters the plastic deformation regime below 5200~5400 m (~80 MPa confining pressure); (2) the deformation evolution of the cap rocks can be divided into four stages. Stages I–III are dominated by brittle deformation, with plasticity progressively increasing with confining pressure, whereas Stage IV is characterized by pervasive plastic deformation and strong sealing capacity, representing an effective cap rock during the critical period of hydrocarbon accumulation; (3) Middle Cambrian reservoirs in the eastern Tazhong area were destroyed by reverse faults that cut through brittle Middle Cambrian cap rocks. In contrast, Lower Cambrian gas reservoirs were charged during the Himalayan period, when the cap rocks remained intact, and exhibited strong sealing capacity. This study demonstrates the temporal variability in the sealing effectiveness of Middle Cambrian anhydrite cap rocks in the eastern Tazhong area and provides a methodological basis for deep and ultra-deep subsalt hydrocarbon exploration. Full article

The renewed interest in hydrogen peroxide-based space propulsion systems has highlighted the persistent issue of catalyst degradation during long-term operation. Although several studies have investigated the underlying causes of this phenomenon, effective regeneration techniques capable of restoring catalytic activity have not yet been clearly demonstrated. This study investigates the mechanisms responsible for performance degradation and proposes a viable regeneration strategy for palladium-based catalysts. Experimental analyses were conducted on a batch of commercial Al${}_2$O${}_3$/Pd pellets subjected to multiple firing cycles in a 10 N-class hybrid mini-thruster. Monitoring of the propulsive performance revealed a progressive decline in catalytic activity, ultimately preventing ignition of the hybrid rocket engine. To characterize the degradation mechanisms, the pellets were examined through visual inspection, static hydrogen peroxide decomposition tests, and Temperature Programmed Reduction (TPR) analysis. The results indicated significant surface oxidation of palladium, leading to reduced decomposition efficiency. A chemical regeneration procedure based on sodium borohydride (NaBH${}_4$) treatment was subsequently developed to restore catalytic performance. The regenerated pellets were tested under the same experimental conditions that had previously led to ignition failure. Their propulsive performance was then compared with both the degraded pellets and a new batch of equivalent catalysts. The results demonstrate that the regeneration process successfully restored the catalytic activity to levels comparable with the original state, enabling stable and efficient hybrid combustion. These findings confirm the role of surface oxidation in catalyst degradation and demonstrate that targeted chemical treatment can significantly extend catalyst lifetime. The proposed regeneration strategy offers a practical method to reduce costs of ground-based experimental campaigns and support the future deployment of hydrogen peroxide-based propulsion systems in space applications by providing insights into the mechanisms that can degrade the performance of palladium catalysts. Full article

Modern reliability experiments frequently face operational constraints that require balancing test duration, precision, and removal strategies, rendering classical censoring schemes inadequate for contemporary multidisciplinary applications. This study introduces a novel multi-progressive generalized Type-II censoring (MP-GC-T2) framework that unifies and extends existing progressive and generalized censoring structures through the integration of staged failure-proportion controls, dual temporal termination thresholds, and adaptive withdrawal of surviving units. The proposed mechanism provides enhanced flexibility in experiment design while retaining analytical tractability for statistical inference. Assuming Weibull lifetimes, we develop a complete inferential framework including maximum likelihood estimation, asymptotic interval construction, and Bayesian estimation via hybrid Metropolis–Hastings–Gibbs sampling with informative gamma priors, together with multiple interval estimation strategies for reliability characteristics. Extensive Monte Carlo investigations assess estimator bias, precision, coverage behaviour, and interval efficiency across diverse censoring configurations, demonstrating robustness and inferential gains relative to conventional schemes. Furthermore, optimal progressive-removal planning criteria are explored to guide practitioners in selecting censoring patterns that maximize inferential accuracy under practical constraints. The versatility and practical relevance of the MP-GC-T2 design are illustrated through applications to heterogeneous real datasets arising from clinical, chemical, geological, physical, and petroleum sciences, confirming its adaptability to distinct reliability structures and data-generation mechanisms. Collectively, the proposed methodology contributes a unified experimental and inferential platform that advances censoring design, reliability estimation, and cross-disciplinary statistical modelling. Full article

Addressing the engineering challenge of creep instability in weakly cemented fractured sandstones within extremely soft coal-bearing formations under long-term loading in western mining areas, using weakly cemented sandstone from a coal mine in Xinjiang as the study subject. This research employs uniaxial graded loading creep tests combined with full-information acoustic emission technology and DIC high-speed strain field observation to investigate the creep deformation patterns (The full name of “DIC” is the three-dimensional high-speed dynamic and static stress–strain analysis system of the DIC strain field measurement and analysis system. For the convenience of expression, this system will be uniformly referred to as DIC in the following text), damage evolution characteristics, and failure mechanisms of sandstone under intact, pre-fabricated 30° fractures, and pre-fabricated 60° fractures. Results indicate: Fractures significantly weaken rock strength and long-term stability. Unfractured specimens primarily exhibit columnar splitting tensile failure, while pre-fractured specimens show pronounced shear failure. Shear cracks accounted for 83.67% of failures in 30° pre-fractured specimens and decreased to 63.44% in 60° pre-fractured specimens. Intact specimens exhibited acoustic emission ringing responses during accelerated creep stages, whereas fractured specimens showed ringing responses as early as the first loading stage. During graded loading, ringing counts in pre-fractured specimens continuously accumulated, with cumulative counts significantly exceeding those of intact specimens. Pre-fabricated cracks induced significant stress concentration effects at the ends, causing failure cracks to propagate preferentially along the crack direction and forming a non-uniform deformation field bounded by the crack. The study revealed the micro-macro evolution patterns of progressive damage during creep in extremely weak fractured rock, providing theoretical support for early warning and control technologies against creep instability in tunnel rock masses of weakly cemented strata in western regions. Full article

Aluminum conductor cables are exposed to environmental conditions in service, where wind-induced vibrations generate multiaxial stresses and cause partial sliding between the stranded layers. Such dynamic loading can lead to fatigue or wear failure, particularly at the contact zones between wire layers. The influence of heat treatment and cold work on the wear of these aluminum wires remains unstudied. This work aims to evaluate the microabrasive wear of rolled and heat-treated 6201 aluminum alloy wires used in conductor cables. The wear tests were performed using free-ball microabrasive wear equipment and alumina (Al₂O₃) abrasive paste at a concentration of 0.40 g/mL of distilled water. The parameters used were as follows: 100 Cr6 steel balls with a diameter of 25.4 mm, sample inclination of 60°, normal force of 0.3 N, and shaft speed of 0.185 m/s or 280 rpm. The test time was set at 20 min, 30 min, 40 min, 50 min, and 60 min. The wear test data were processed using the Achard equation. The microabrasive wear test results indicate that the wear coefficient decreased by 19.1% after the artificial aging process, compared with the solution-treated alloy (95% CI: 15.5–22.3%), and this reduction was statistically significant (p < 0.001). After the combined treatment of rolling and artificial aging, the alloy had a drop in wear coefficient of 36.1% compared to the same solution-treated alloy (95% CI: 32.6–39.6%), representing the largest statistically significant improvement among the tested conditions (p < 0.001). Cold work (rolling) reduces the mobility of dislocations, requiring greater stress to deform the material, thereby increasing its stiffness and wear resistance. In this 6201 alloy, it is inferred that artificial aging led to the formation of Guinier-Preston zones, which evolved into the formation of metastable β” precipitates in needle-like form, coherent with the matrix. As the aging process progresses, the β’ particles evolve into larger β particles that are no longer coherent with the matrix. The combined processes of rolling and aging decrease the wear coefficient. Statistical analysis demonstrated that microstructural conditions explain approximately half of the total variability in the wear coefficient (η² = 0.495), indicating that the wear performance under the present experimental configuration is primarily governed by intrinsic strengthening mechanisms rather than experimental variability. Full article

Mineral resource extraction and urban expansion in resource-based cities have progressively degraded regional ecosystems, leading to increasingly fragmented ecological patterns. Ecological network resilience plays a critical role in maintaining regional ecological stability. In this study, we integrated landscape ecology and systems science to develop a network model and assess the resilience of ecological networks in the coal mining subsidence area with high groundwater levels. This study employed morphological spatial pattern analysis (MSPA) and circuit theory to construct the ecological network. A cascading failure model was further applied to simulate network dynamics under three attack strategies. Based on a comparative analysis of these strategies, we introduce the concept of “dangerous nodes” to identify priority conservation areas. The research results show that 101 ecological source areas and 255 ecological corridors were identified in the study area. Topologically, its ecological network is characterized by a small number of core nodes and a large number of secondary nodes. When the adjustable parameter is $\alpha <1.2$, targeting low-degree nodes may inflict more severe damage on the network. When $\alpha >1.2$, attacks against nodes with a high-degree or high betweenness centrality may have significant cascading failure implications. Our results show that the network’s critical threshold $T_c$ depends on the number of dangerous nodes in the attack set. The distribution of these nodes differs substantially between low-degree attacks and those targeting high-degree or high betweenness centrality nodes. These findings advance ecological network optimization and provide practical guidance for ecosystem conservation and restoration in resource-based cities. Full article

Background: Obstructive sleep apnea (OSA) is a major modifiable risk factor for atrial fibrillation (AF), promoting arrhythmogenesis through intermittent hypoxia, autonomic activation, and atrial remodeling. Although continuous positive airway pressure (CPAP) effectively treats OSA, real-world evidence linking objectively measured CPAP exposure to clinically relevant AF recurrence remains limited. Aims: We aimed to evaluate the association between CPAP adherence and risk of recurrent paroxysmal AF, and to compare time to first recurrence between patients with mean nightly CPAP use ≥4 h/night versus <4 h/night. Materials and Methods: In this prospective observational cohort (2017–2024), consecutive hospitalized and outpatient adults with severe obstructive sleep apnea (OSA; apnea–hypopnea index > 30 events/h) and documented paroxysmal atrial fibrillation (AF) were enrolled. Persistent and long-standing persistent AF were excluded to ensure a homogeneous population with respect to atrial substrate. OSA was assessed using home sleep apnea testing (ResMed ApneaLink), and all patients initiated continuous positive airway pressure (CPAP) therapy (ResMed AirSense 10). Objective adherence data were obtained via the ResMed AirView telemonitoring platform. Exclusion criteria included permanent AF, prior pulmonary vein isolation, central sleep apnea, left ventricular ejection fraction < 50%, end-stage chronic kidney disease (eGFR < 15 mL/min/1.73 m² or dialysis), or inability to initiate or maintain CPAP therapy. Patients were followed for 12 months. The primary endpoint was time to first documented recurrence of paroxysmal AF (≥30 s on 12-lead electrocardiography or 24-h Holter monitoring). Progression to permanent AF, defined after unsuccessful rhythm control attempts and subsequent transition to a rate control strategy, was assessed as a secondary endpoint. Time-to-event analyses used Kaplan–Meier estimates with log-rank testing, and Cox proportional hazards regression adjusted for age, body mass index, apnea–hypopnea index, heart failure, left atrial volume index, and antiarrhythmic drug therapy. Results: The final analysis included 91 patients (mean age 62.15 ± 8.29 years; 68.13% men). Mean nightly CPAP use was ≥4 h/night in 49 patients and <4 h/night in 42 patients. During follow-up, paroxysmal AF recurrence occurred in 12/49 (24.5%) patients in the ≥4 h/night group and 16/42 (38.1%) in the <4 h/night group. Mean arrhythmia-free survival at 12 months was numerically higher in the ≥4 h/night group (11.25 vs. 10.51 months), without a statistically significant difference in Kaplan–Meier curves (log-rank p = 0.11). In multivariable Cox regression, binary adherence (≥4 h/night) was not independently associated with recurrence (HR 0.52, p = 0.13), whereas mean nightly CPAP use analyzed as a continuous variable remained independently associated with delayed recurrence (per 1-h increase: HR 0.66, 95% CI 0.48–0.91, p = 0.01). Progression to permanent AF occurred in 4/49 (10.0%) versus 9/42 (17.6%) patients, respectively (p = 0.29). Conclusions: In this real-world cohort of patients with severe OSA and paroxysmal AF, higher objectively measured CPAP exposure was independently associated with delayed AF recurrence when analyzed as a continuous variable, suggesting a graded association between objectively measured CPAP exposure and AF recurrence. Larger studies with extended follow-up and continuous rhythm monitoring are warranted to confirm long-term rhythm benefits and effects on AF progression. Full article

Early detection of fire-related risks in lithium-ion batteries (LIBs) remains a critical challenge, as conventional protection mechanisms typically activate only after irreversible degradation or macroscopic failure occurs. In this study, an innovative sensor-based diagnostic framework is proposed for proactive fire prediction in LIBs by simultaneously monitoring low-frequency and high-frequency electrical signatures generated during battery charge–discharge processes. An electromagnetic (EM) antenna sensor and a high-frequency current transformer (HFCT) sensor were employed to capture complementary voltage- and current-based transient signals associated with internal degradation phenomena. Cell-level experiments were conducted under various C-rates and temperature conditions, including high-stress environments, while module-level validation was performed on a 4-series, 1-parallel (4S1P) configuration at a 2C-rate under ambient temperature. Time–frequency characteristics of the measured signals were systematically evaluated using MATLAB-based continuous wavelet transform (CWT) and discrete wavelet transform (DWT) techniques. The results reveal that degradation-induced transient events exhibit non-stationary, impulsive voltage and current signatures with distinct frequency-band localization, which intensify with increasing C-rate, elevated temperature, and aging progression. At the module level, although signal amplitudes were partially attenuated due to current redistribution, characteristic wavelet energy patterns and time–frequency concentrations remained clearly distinguishable, demonstrating the scalability of the proposed approach. The combined EM antenna–HFCT sensing strategy, together with multi-resolution wavelet analysis, enables effective phenomenological differentiation between normal operational noise and incipient internal fault signatures well before conventional thermal or capacity-based indicators become evident. These findings demonstrate feasibility of the proposed method for early-stage fault diagnosis and highlight its potential applicability to advanced battery management systems for proactive fire prevention in large-scale energy storage and electric vehicle applications. Unlike conventional voltage-, temperature-, or gas-based diagnostics, the proposed approach enables the detection of incipient degradation phenomena at the microsecond scale by exploiting complementary low- and high-frequency electrical signatures. This study provides experimental evidence that wavelet-based EM and HFCT sensing can identify MISC-related precursors significantly earlier than conventional battery management indicators. Full article

Background: Systemic inflammation through neutrophil-mediated injury, lymphocyte depletion, and monocyte-driven fibrosis plays a central pathophysiological role in heart failure (HF) progression. We investigated the diagnostic and prognostic utility of contemporary inflammatory indices, particularly the Systemic Inflammatory Response Index (SIRI) and Naples Prognostic Score (NPS). Methods: This retrospective cohort study enrolled 926 participants (500 HF patients, 426 controls). Multiple inflammatory indices (e.g., SIRI, Prognostic Nutritional Index (PNI)) and prognostic scores (e.g., NPS) were calculated from routine hematological and biochemical parameters. Primary outcomes were HF diagnosis discrimination and 3-month and 24-month all-cause mortality. Receiver operating characteristic (ROC) curve analysis, Kaplan–Meier survival curves, and Cox proportional hazards regression were performed. Results: HF patients demonstrated significantly elevated inflammatory burden: SIRI (3.26 vs. 1.06, p < 0.001) and NPS (2.00 vs. 1.43, p < 0.001). For HF diagnosis, SIRI exhibited superior discriminative performance (AUC = 0.893; 95% confidence interval (CI): 0.871–0.912), substantially exceeding all other indices (p < 0.001). For long-term mortality prediction, SIRI maintained the highest accuracy (AUC = 0.677), followed by PNI (AUC = 0.639) and NPS (AUC = 0.613). Kaplan–Meier analysis revealed progressive survival deterioration across NPS categories: 24-month mortality increased from 5.9% (NPS = 0) to 23.0% (NPS = 4), p = 0.002. Multivariable Cox regression confirmed independent prognostic value: SIRI >1.86 (HR = 2.232; 95% CI: 1.280–3.892; p = 0.005) and NPS > 2 (HR = 1.403; 95% CI: 1.180–1.668; p < 0.0001). Conclusions: SIRI and NPS represent powerful, readily accessible prognostic tools capturing distinct but complementary pathophysiological domains in HF. These indices offer substantial clinical utility for risk identification and treatment decisions, particularly in resource-limited settings. Future studies should validate these cut-offs and evaluate biomarker-guided therapeutic strategies. Full article

Background: The early detection of chronic kidney disease (CKD) is critical to preventing progression and reducing associated morbidity. The original SCreening for Occult REnal Disease (SCORED) tool has been widely adopted for CKD screening. However, its length and inclusion of items with limited predictive value affect its practicality in specific settings. This study aimed to validate a modified version of the tool (SCORED-M), which has fewer items and improved predictive performance for the early detection of disease. Methods: A cross-sectional pilot project was conducted and the diagnostic performance of the revised tool (SCORED-M) was evaluated using receiver operating characteristic analysis, sensitivity, specificity, positive predictive value, and negative predictive value (NPV). Items were selected or excluded based on their statistical significance, odds ratios, and clinical relevance to CKD risk. The optimal threshold score for mass screening was determined through a comparative analysis. Results: A total of 116 eligible participants enrolled in this pilot study. SCORED-M, comprising six items, rather than nine, as in the original version, demonstrated superior screening performance. It achieved a higher area under curve (0.89 vs. 0.79), sensitivity (0.97), and NPV (0.97), indicating its improved capability to identify individuals with CKD and rule out those without the condition. The age-related scoring range was recalibrated from 2 to 4 points to a narrower span of 1–3 points, to moderate the influence of age as a standalone risk factor for CKD. Items with limited predictive contribution, such as ‘I am a woman’, ‘I have a history of heart attack or stroke’, and ‘I have circulation disease in my legs’, were removed, while clinically relevant variables like ‘I am diabetic’, ‘I have a history of congestive heart failure or heart failure’, ‘I have protein in my urine’, ‘I have uncontrolled high blood pressure’, and ‘I have a history of renal disease’ were retained. A threshold score of ≥4 was identified as optimal, balancing sensitivity and specificity while supporting resource-efficient screening and ensuring the reproducibility of results. Conclusions: This pilot study provided preliminary evidence that the SCORED-M tool offers a more concise and accurate approach to CKD/diagnosis. While the findings are promising, validation in larger and more diverse populations is necessary to confirm the generalizability of the model and refine it for broader clinical application in mass screening programmes. Full article

This study investigates the complex relationship between fracture patterns and acoustic emission (AE) mechanisms during triaxial deformation experiments on Westerly granite under various confining pressures (5, 10, 20, and 40 MPa). Using numerical simulations with the Particle Flow Code (PFC2D, 6.0, Itasca Consulting Group Inc., Minneapolis, MN, USA), this research emphasizes the significant influence of confining pressure on crack development, AE events, spatiotemporal distribution, energy dissipation, and peak stress in the samples. AE source mechanisms, categorized into T-Type, C-Type, and S-Type, show the dominance of T-Type fractures during post-peak unstable failure and the emergence of C-Type fractures as precursors to critical damage. Additionally, increasing confining pressure is found to correlate with changes in fracture dynamics, evidenced by an increase in big events and a decrease in small events. The analysis of b-values across different pressures reveals fluctuations that indicate change in fracture features. Fractures originate in the model center and propagate towards both ends as loading progresses, ultimately leading to failure. In summary, these findings provide important insights into the fracture patterns of granite and the underlying mechanisms of AE release. Moreover, they carry practical implications for identifying failure precursors and for the potential application of early warning systems in rock engineering. Full article

Cemented paste backfill (CPB) is widely adopted in underground mines, where the shear resistance of the CPB–rock interface critically governs the integrity of backfill–rock systems. This study investigates the effects of polypropylene fiber reinforcement, surface roughness (Joint Roughness Coefficient, JRC = 0 and 1.76), and curing time (1, 3, and 7 days) on the shear strength and deformation characteristics of CPB–rock interfaces. Direct shear tests were performed under normal stresses of 50, 100, and 150 kPa, with synchronous measurements of shear and vertical displacements. Results show that increasing roughness markedly strengthens the interface, with the peak shear stress rising by up to 45% due to enhanced mechanical interlocking and dilation. In contrast, adding 0.5 vol.% PP fibers slightly reduces peak shear capacity but consistently improves post-peak deformability, indicating a transition from brittle interfacial fracture to a more ductile, progressive failure mode. A three-stage mechanical model was established to describe the shear stress–displacement relationship, incorporating elastic, bond degradation, and frictional sliding phases. The model parameters, including the shear stiffness ($K_s$), bond degradation coefficient ($\eta$), and residual strength (${\tau }_r$), were calibrated using the experimental data. Mohr–Coulomb analysis further quantifies the curing-dependent evolution of interfacial strength parameters, highlighting a marked increase in cohesion from 1 to 7 days alongside roughness-governed peak strengthening. This research provides insights into the optimization of the CPB–rock interface design for enhanced geomechanical performance in underground applications. Full article

Progressive collapse is characterized by disproportionate structural failure triggered by localized damage, such as column loss under extreme loading conditions. The objective of this study is to develop a simplified analytical model that is applicable in engineering practice without the need for high-fidelity nonlinear finite element analysis. Although current design guidelines (GSA and DoD) provide analytical procedures and acceptance criteria, they do not explicitly address the tensile resistance of girders after the acceptance criteria are satisfied, particularly under large deformation and connection failure. To address this limitation, this study proposes a simplified theoretical load–displacement model for a fixed-end girder subjected to three concentrated loads, considering the effects of secondary beams and focusing on the local girder response under a column-removal scenario. The proposed model incorporates moment–axial force interactions at plastic sections in the large-deformation range. Based on one-dimensional finite element analysis results, an early-developed axial force of 0.15F_p at the onset of the transition stage and a residual bending moment of 0.3M_p during the catenary action stage are explicitly introduced to better represent actual structural behavior. The girder response is idealized using five characteristic points: yielding (Y), full plasticity (P), transition initiation (T), pure catenary action initiation (C), and collapse governed by connection failure (F_conn). Stress distributions at plastic sections are analyzed using three-dimensional finite element models to establish stress-based formulations and a rational procedure for estimating axial force at collapse. The validity of the proposed model is verified through comparisons with finite element analysis results for girders with different span-to-depth ratios. The results demonstrate reasonable agreement in terms of collapse load and displacement, particularly for slender girders, confirming the applicability of the proposed model for progressive collapse assessment. Full article

Background: Imaging-transcriptomic analysis, through the integration of multimodal magnetic resonance imaging (MRI) and transcriptomic data, provides complementary structural, functional, and molecular information that is crucial for the early detection and mechanistic exploration of Alzheimer’s disease (AD). However, effectively extracting features from heterogeneous multimodal data and capturing the associations between microscopic molecular variations and macroscopic brain alterations remain key challenges. Recent advances in deep learning and multimodal integration have enhanced the ability to model nonlinear cross-modal relationships, enabling more accurate identification of imaging-transcriptomic biomarkers and subtypes. Developing robust multimodal frameworks is therefore essential for early AD detection, subtype identification, and advancing precision medicine in neurodegenerative diseases. Methods: In this study, a two-stage method of multimodal Feature Extraction based on Association Analysis and Graph Convolutional Network with Self-Attention and Self-Expression framework (MFEAA-GCNSASE) for early diagnosis of AD and effective identification of subtypes of MCI with different progression to AD is proposed. In the first stage, the MFEAA model is applied to integrate multiple association analysis methods on sMRI, PET, and transcriptomic data to identify key multimodal biomarkers for AD and mild cognitive impairment (MCI). In the second stage, the GCNSASE model enhances classification accuracy between AD and MCI patients through self-attention and self-expression layers. Additionally, unsupervised clustering was performed on MCI samples using top multimodal biomarkers to explore subtype heterogeneity and conversion risk. Reliable MCI subtypes were also identified through a consensus clustering approach. Results: The proposed algorithm integrates sMRI, PET, and transcriptomic data, identifying robust biomarkers including the Left Hippocampus, Left Angular Gyrus, and key genes such as SLC25A5 and GABARAP. To ensure statistical robustness given the extreme class imbalance, we employed a rigorous repeated stratified cross-validation (RSCV) framework. GCNSASE achieved state-of-the-art discrimination performance with mean AUC values ranging from 0.946 to 0.961 across feature subsets (10–50%), significantly outperforming MOGONET (mean AUC: 0.844–0.875, p < 0.001) and conventional machine learning models with tighter 95% confidence intervals, indicating superior stability despite the limited AD sample size. Clustering analysis revealed two distinct MCI subtypes with divergent molecular landscapes: Subtype A was enriched in energy metabolism and cellular maintenance pathways, whereas Subtype B was enriched in neuroinflammatory and aberrant signaling pathways. Notably, the majority of MCI patients who subsequently converted to AD were concentrated in the immune-inflammatory Subtype B. These findings highlight that neuroinflammation coupled with bioenergetic failure constitutes a critical mechanism driving the conversion from MCI to AD. Conclusions: The proposed methods not only provide the key multimodal biomarkers and enhance the accuracy of the classification model for early AD diagnosis but also identify biologically and clinically meaningful MCI subtypes with distinct molecular signatures and conversion risks. Exploring these associated multimodal biomarkers and MCI subtypes is of great significance, as they help elucidate the heterogeneous mechanisms underlying AD onset and progression, enable the identification of high-risk individuals likely to convert to AD, and provide a foundation for targeted therapeutic strategies and individualized clinical management. These findings have important implications for understanding disease heterogeneity, discovering potential intervention targets, and advancing precision medicine in neurodegenerative diseases. Full article