With the widespread use of the internet, identity authentication has become increasingly important in information systems, while password authentication remains one of the most commonly used authentication methods. In recent years, advances in deep learning have promoted the application of generative adversarial networks (GANs) to password guessing. However, GAN-based password guessing models, such as PassGAN, still suffer from relatively low guessing efficiency. Compared with the traditional password guessing models, these methods generally need to generate more candidate passwords to achieve the same cracking rate and often produce a large number of duplicate candidates. To improve the diversity of generated passwords and improve the guessing rate, a password guessing model named DPAC-Pass is proposed based on reinforcement learning and cross-entropy reward, which adopts the Actor-Critic algorithm in the password guessing task and uses cross-entropy as a single-step reward to guide the model training. The proposed DPAC-Pass model and the PCFG, SeqGAN, and AC-Pass models in the literature are trained on the LinkedIn passphrase set and tested on LinkedIn, RockYou and CSDN set, and compared in terms of the generated password features, guessing efficiency, password diversity, and quality of unique passwords. The experimental results show that the password cracking rate is improved with the proposed DPAC-Pass model by enhancing diversity of generated passwords so that password length and character composition can be more accurately captured.

To explore the sound absorption performance of porous carbon materials, this study prepared three-dimensional ordered porous carbon materials by the template method, and investigated the influence of porous carbon materials prepared at a temperature of 600-1 000 ℃ on the sound absorption performance under a temperature gradient of 100 ℃. Using porous alumina as the template, a mixture of phenolic resin, zinc acetate (Zn(CH₃COO)₂), and tetraethyl orthosilicate (TEOS) was employed to fabricate the 3DOPC sound-absorbing materials. The pore structure, surface morphology, and crystallinity were characterized by nitrogen adsorption-desorption analysis, scanning electron microscopy (SEM), and X-ray diffraction (XRD). The acoustic properties were evaluated using the impedance tube method. The appearance of the characteristic peaks of ZnO in XRD verified the successful thermal decomposition of zinc acetate into ZnO. Acoustic measurements revealed that the 3DOPC material achieved an average sound absorption coefficient of 0.89 in the frequency range of 3 000-4 000 Hz. This work provides an effective strategy for synthesizing porous carbon materials for sound absorption applications.

In the context of the continuous advancement of the "Healthy China" strategy, the public's awareness of the importance of physical and mental health has deepening, leaded to structural changes in pubbcusage patterns and preferences regarding natural parks. This study takes natural exammes in Guangzhou as the research object, based on mobile signaling data from 2019, 2020, and 2022, and systematically analyzes the differentiation characteristics and dynamic trends of natural park use preferences from the dimensions of spatiotemporal patterns, park types, and population groups. The results show that: First, in terms of time, visitor numbers during holidays are much higher than on weekdays, indicating a clear holiday preference. Meanwhile, the proportions of female, underage, and elderly visitors are low and further declined during the pandemic. Sencond, in terms of space, residents show a strong preference for the main urban area. However, after the pandemic, the share of visitors to non-main urban natural parks rose from 21.95% to 30.40%. Some outer suburban parks even exceeded pre-pandemic visitor levels, indicating an increased willingness for long-distance travel. At the same time, young adults tend to travel longer distances, while minors and the elderly mainly visit natural parks in the main urban area and rarely visit those in non-main areas, reflecting intergenerational spatial differentiation. Third, in terms of type, the differences in usage intensity among parks have narrowed. The average daily visitor numbers of wetland parks, forest parks, and scenic areas are roughly equal. This reflects a more balanced use of natural park types, and the public pays more attention to factors such as openness and crowd density. These changes indicate that major public health events have driven a systemic shift in the public's preferences for using natural parks. This study will provide a scientific basis for the spatial optimization of natural parks, the construction of healthy cities, and the response of large urban green spaces to public health emergencies.

Artificial light at night (ALAN) can intensify ecological barriers caused by land development in rapidly urbanizing regions by reducing nocturnal habitat suitability and low-light spatial continuity. To reveal the spatial pattern and connectivity bottlenecks of dark ecological spaces in a highly urbanized bay area, this study develops a dark-conservation-oriented “ecological source-resistance surface-corridor network” framework for the Guangdong-Hong Kong-Macao Greater Bay Area. Ecological sources were identified by integrating MSPA, InVEST habitat quality assessment, and Conefor connectivity analysis, with nighttime light constraints further applied to delineate dark ecological sources. Based on conventional resistance factors, ALAN was incorporated into the composite resistance surface as an independent ecological stressor. Dark ecological corridors were then extracted using circuit theory, and key nodes were identified. Results show that 147 dark ecological sources were identified, mainly distributed in less urbanized areas such as Zhaoqing, Jiangmen, Huizhou, and northern Guangzhou. The composite resistance surface exhibited a “high in the center and low in the periphery” pattern. A total of 330 dark ecological corridors, totaling 3 766.624 km, were extracted, along with 87 ecological pinch points and 115 ecological barrier points. These findings indicate that incorporating ALAN into ecological security pattern construction can effectively reveal connectivity risks in nocturnal ecological processes and provide scientific support for dark ecological network optimization and zoned urban lighting governance in the Greater Bay Area.

To mitigate the grid-connected current distortion and double-frequency power oscillations caused by an unbalanced RL-filter in grid-connected photovoltaic (PV) inverters, this paper proposes an enhanced Single-Degree-of-Freedom Active Disturbance Rejection Control (SDOF-ADRC) strategy. First, a mathematical model of the grid-connected inverter with non-ideal filters is established. Theoretical derivation reveals that filter parameter imbalance introduces DC disturbances. It also generates significant 100 Hz AC disturbances in the synchronous rotating frame, which severely degrades power quality. To address these complex disturbances, a resonant controller is integrated into the extended state observer of the traditional SDOF-ADRC. This allows the improved observer to estimate DC disturbances while effectively observing and compensating for corresponding AC disturbance. Consequently, the system achieves simultaneous estimation and suppression of both DC and AC disturbances. The Bode plots analysis demostrates that the improved strategy maintains the closed-loop bandwidth while achieving high observation gain at 100 Hz. Compared with traditional PI control and standard SDOF-ADRC, this method provides superior disturbance attenuation in specific frequency bands. MATLAB/Simulink simulations verify the effectiveness of the proposed method, which maintains a simple and efficient control architecture. While effectively reducing dq-axis current ripples, this strategy significantly enhances system robustness against filter parameter perturbations and improves dynamic response. Overall, the approach demonstrates significant practical value for engineering applications.

The health-oriented operation and maintenance (O&M) of in-building water supply and drainage systems currently relies predominantly on the "threshold alarm" paradigm. While this approach can detect anomalies, it struggles to reveal the root causes of failures, predict degradation trends, or guide targeted maintenance-a limitation particularly pronounced in aging building systems characterized by complex deterioration mechanisms and dynamic operating conditions. To achieve a fundamental paradigm shift from "passive alarm" to "active mechanism tracing" in health assessment, this paper proposes a novel health assessment theory integrating multi-field coupling mechanisms with lightweight digital twins. At its core lies a physicochemical model of synergistic multi-material deterioration under the coupled effects of water quality, temperature, and stress fields, which addresses the longstanding bottleneck of "unclear mechanisms." A composite assessment model combining the dynamic adaptive entropy weight method with a dynamic Bayesian network is further developed to resolve the challenge of "quantification imbalance" in dynamic uncertainty characterization. On this basis, a lightweight digital twin architecture driven by the aforementioned "mechanism kernel" is designed, which overcomes the limitations of "superficial integration" and "paradigm misalignment" through closed-loop mechanism-data iteration. A simulation case study on the deterioration process of a typical water supply riser verifies the logical feasibility of the proposed theory and method in state identification, adaptive weight adjustment, and trend prediction. The results demonstrate that the proposed theory and method provide a systematic solution for interpretable assessment, predictive maintenance, and precise decision-making regarding the health status of building water supply and drainage systems, advancing the paradigm transformation of infrastructure O&M from experience-driven to mechanism-data fusion-driven.

The engineering investigation industry is currently undergoing a paradigm shift from experience-driven to data- and intelligence-driven practices. This paper constructs an integrated analytical framework incorporating digital transformation theory, human-machine collaboration principles, and a structured analytical model to systematically deconstruct AI integration mechanisms in engineering investigation.Through multi-case empirical studies of three core technologies—intelligent lithological core identification, rapid access to investigation results, and intelligent drawing review—this research reveals the inherent pathways for AI-driven reconstruction of investigation workflows within a data-algorithm-platform architecture. The findings demonstrate that AI applications significantly enhance investigation efficiency, data consistency, and quality control capability. Specifically, drawing review efficiency improved by 72%, layer consistency reached 92%, and the violation rate of mandatory provisions was reduced by 81%.This success relies on an analytical framework following the vertical logic of theory-model-implementation-application and integrating horizontal key elements such as data-algorithm-domain-platform. However, challenges such as insufficient technology generalization, a shortage of interdisciplinary talent, absence of industry standards, and uncertainty regarding return on investment remain major implementation barriers.This paper proposes that the industry should evolve toward intelligent equipment, platform ecosystems, and data-driven services, transitioning from passive investigation to proactive prediction through the development of a "geological brain." This study provides a systematic framework combining theoretical depth with practical operability for the intelligent transformation of the engineering investigation industry.

Steel-precast ultra-high-performance concrete (UHPC) composite beams utilizing high-strength bolts as shear connectors exhibit advantages such as superior mechanical performance and convenient disassembly. Currently, research on the flexural performance of such composite beams remains insufficient. To this end, one such specimen was designed and subjected to a four-point bending test to investigate its flexural behavior. The test results indicate that the composite beam specimen primarily experienced bolt shear fracture as the dominant failure mode, and significant flexural cracks were observed at the bottom of the UHPC slab beneath the loading points. The rigid plastic analysis method and the simplified analysis method recommended by Eurocode 4 were employed to calculate the flexural capacity. The ratios of the calculated values to the experimental values were 0.98 and 0.80, respectively, demonstrating that the former is more applicable for predicting the ultimate flexural capacity of this type of composite beam.

Using the 65-meter Shanghai Tianma Radio Telescope, we carried out mapping observations of molecular H₂CO (2₁₁-2₁₂) and H¹³₂CO (2₁₁-2₁₂) (with rest frequencies of 14.488 GHz and 13.779 GHz, respectively) toward the SgrB2 giant molecular cloud complex in the Galactic center. The H₂CO (2₁₁-2₁₂) line was detected at all 128 observed positions. Follow-up observations of H¹³₂CO (2₁₁-2₁₂) were carried out for 89 sources exhibiting strong H₂CO (2₁₁-2₁₂) emission, and its line emission (signal-to-noise ratio>3) was successfully detected at 18 positions. We analyzed the detected spectral lines to derive parameters including integrated line intensities, and calculated the line intensity ratios of the two molecules at each position. The contour maps show that the spatial distributions of the line intensities of the two molecular gases are similar, both exhibiting a higher intensity in the central region and a gradual decrease toward the periphery. However, the spatial distribution of their intensity ratio, i.e., I (H₂CO (2₁₁-2₁₂))/I (H¹³₂CO (2₁₁-2₁₂)), presents distinct characteristics, with relatively low values in the gas-rich central region. Our analysis indicates that this phenomenon arises from the saturation effect of the H₂CO (2₁₁-2₁₂) line due to its extremely high abundance in the central region. After considering effects such as optical depth, the measured isotopic abundance ratio determined in this work is consistent with the latest results derived from C³⁴S/¹³C³⁴S observations.

Let b and k be positive intrgers with k≥b, and let G be a simple graph. A b-fold k-coloring of G is an assignment of b distinct colors to every vertex, from a set of k colors, such that adjacent vertices do not have any colors in common. The b-fold chromatic number of G, denoted by Χ_b(G), is the minimum number k such that a b-fold k-coloring of G exists, i.e., Χ_b(G)=min{k|there exists a b-fold k-coloring of G}. In this paper, we discuss the relation between the b-fold chromatic number of the join, weak product, and composition of two graphs and the b-fold chromatic number of their original graphs. We also obtain the relation between the b-fold chromatic number of Mycielski's transformation graph of a graph G and the b-fold chromatic number of the original G.

The concept of neighbor sum distinguishing I－total coloring of graphs is defined. By using mathematical induction, the structural coloring method, and combinatorial analytic method, we mainly study the neighbor sum distinguishing I－total coloring of unicyclic graphs, and obtained the neighbor sum distinguishing I－total chromatic number for unicyclic graphs.

This paper investigates the n-fold convolution μ^*n of an in-homogeneous self-similar measure μ on a line, generated by a family of contractive mappings satisfying the equal contraction condition and the open set condition. It is proved that if the in-homogeneous term measure ν of μ also satisfies a certain Fourier integral decay condition, then the Hausdorff dimension, the L^q-dimension, and the entropy dimension of the n-fold convolution μ^*n converge to 1 as n→∞.