Landslide dams are generally characterized by their large volume， sudden occurrence， and wide-ranging impacts， and their breach often leads to catastrophic floods. This study aims to investigate the effects of downstream fold line slope and horizontal particle distribution on the breach process of landslide dams. Eight groups of tests were designed by using the “11.03” Baige landslide dam on the Jinsha River as the prototype for experimental dam dimensions. The results indicate that the downstream curved slope altered the initial starting position of the dam particles， while the steep slope exerted an accelerated collapse effect. The collapse times of downstream convex and concave slopes was shortened by 21.7% and 34.8%， respectively， compared to linear slopes. The erosion caused by water flow on the downstream linear slope is relatively uniform， and its continuous action on the dam body persists over an extended period. Water flow is concentrated at the downstream concave slope， where it easily forms eddies and transports sediment at the spillway’s front end. On the downstream convex slope， the water flow is relatively dispersed， reducing the scouring force in the early collapse stage. The coarse particle layer in a horizontally heterogeneous dam exhibits high resistance to erosion and functions as a protective layer， thereby influencing the dam’s failure strength. The fine particle layer exhibits low erosion resistance， a significant water flow erosion effect， and influences the collapse duration. When the coarse particle layer is positioned upstream of the dam body， particle loss decreases， stability improves， and the peak flow rate is 7.9% to 14.5% lower than that of a homogeneous linear slope dam. When the coarse particle layer is situated downstream of the dam body， collapse duration increases， and the peak flow rate is 1.9% lower than that of a homogeneous linear slope dam. The stability of the horizontal heterogeneous dam structure on the downstream slope is the lowest， with a peak flow rate 2.2% to 8.2% higher than that of a horizontally heterogeneous dam of identical structure. Material distribution’s influence on peak flow dominates， while the erosion effect caused by the downstream slope remains unaffected by the dam material composition. This study deepens the understanding of breach characteristics of landslide dams with downstream fold line slopes and horizontally heterogeneous materials， providing references for the risk prediction and emergency response.

Given that existing research on river confluences is predominantly limited to single tributary-mainstream scenarios， this study investigates the hydraulic characteristics of multi-river continuous confluence scenarios to clarify their behavior. A plane two-dimensional （2D） hydrodynamic model was employed to analyze two distinct confluence patterns—“two tributaries merging before entering the mainstream” and “two tributaries sequentially entering the mainstream”—during the transit of flood peaks under four different flood return periods. The water levels calculated by the numerical model showed good agreement with in-situ data from hydrological stations. The results indicate that due to the large confluence angle （116°）， the low momentum ratio （Mr < 1）， and the buffering effect of the floodplain during overbank floods， typical flow separation zones were absent in both confluence configurations. In the configuration where “two tributaries merging before entering the mainstream”， the backwater effect exerted by the downstream mainstream on the upstream tributary confluence zone intensifies as the flood return period increases from 5-year to 50-year， expanding the area of the stagnation zone within the tributary confluence by 1.94 times. Furthermore， this configuration yields higher peak water levels in the tributaries compared to the “two tributaries sequentially entering the mainstream” configuration. This water level differential is inversely proportional to the flood return period， reaching a maximum of 22 cm at the characteristic tributary cross-section under the 5-year return flood. In the “two tributaries sequentially entering the mainstream” configuration， the backwater effect from the downstream tributary decelerates the flow within the high-velocity zone of the upstream mainstream-tributary confluence. Consequently， the peak water level in the mainstream exceeds that of the “two merging before entering the mainstream” configuration. This differential is directly proportional to the flood return period， culminating in a maximum peak water level difference of 10 cm at the characteristic mainstream cross-section under the 50-year return flood. The research results provide a scientific basis for the design of levee elevations and river regulation in multi-river confluence reaches.

In response to the problems of sedimentation and erosion and damage to the discharge structures in front of the dam caused by low-level open drain sediment in Xixiayuan Reservoir， this study analyzed the spatiotemporal evolution of siltation in the reservoir and the impact of inflow water and sediment and dispatched water level on sedimentation based on actual measured data. By taking into account both the safety of project and sediment discharge effect， an optimized range of the water level in front of the dam during the sediment discharge period is proposed. The results show that severe sediment deposition occurs in Xixiayuan Reservoir. As of the 2023 flood， 49.14% of the storage capacity below the normal water level has been lost， and siltation is mainly concentrated in the wide river section in front of the dam. The erosion and siltation in the reservoir is affected by the synergistic influence of the amount of sand entering the reservoir and the water level in front of the dam. In 2018 and 2020， due to the combined effect of high sand inflow and high water level， the siltation in the reservoir developed rapidly. According to the quantitative relationship between the sediment discharge ratio and the water level in front of the dam， it is recommended to control the water level in front of the dam be controlled at around 130 m during the sediment discharge period （with an allowable fluctuation range of 129.5 to 130.5 m） . The study is expected to provide a certain scientific basis for Xixiayuan Reservoir to carry out precise water and sediment regulation and maintain long-term effective storage capacity.

The water flow pressure distribution of diversion tunnels is crucial for discharge safety during the construction period. To study the pressure distribution law and its influencing factors at the outlet section of diversion tunnels， this paper takes the diversion tunnel of a water control project as the research object. The RNG k-ε three-dimensional turbulent mathematical model is adopted to conduct a simulation analysis on the pressure at the outlet section of the diversion tunnel. The variation law of pressure at the outlet section under different inlet geometries and different cover-slope ratios is studied. A relational expression between the outlet pressure distribution correction coefficient and slope ratio （m）， Froude number （Fr）， contraction coefficient （C_c ）， and discharge coefficient （μ） is proposed， and the prediction accuracy of this expression is verified. The research results show that the existence of negative pressure at the outlet section will lead to the over-discharge phenomenon of the diversion tunnel； Reasonable adjustment of the inlet shape can reduce the negative pressure value at the outlet to a certain extent； Adding a cover-slope section at the outlet can also effectively improve the negative pressure distribution， but the cover-slope may affect the flow capacity. The proposed formula for the outlet pressure distribution correction coefficient reflects the variation law of water flow pressure at the outlet section of the diversion tunnel， which can provide a reference for the design and verification of similar projects.

Effective monitoring of suspended sediment concentration （SSC） is essential for sediment control and hydrological process assessment in rivers， reservoirs， and waterways. Conventional SSC monitoring techniques， however， are often costly， labor-intensive， and limited in their ability to provide continuous and efficient measurements. To address these limitations， this study develops a novel SSC interval classification model based on MobileViT-CBAM. The proposed model exploits high-resolution surface water images captured by cameras to establish the relationship between SSC grades and image features. By integrating the lightweight MobileViT architecture with the Convolutional Block Attention Module （CBAM）， the model enhances feature extraction while effectively reducing background noise. Experimental data were obtained using a mobile sediment–water circulation system， where images corresponding to varying SSC intervals were collected and subsequently divided into training and testing sets at a 7∶3 ratio. The results demonstrate that MobileViT-CBAM significantly outperforms conventional models in terms of accuracy， precision， recall， specificity， and F1-score， achieving an overall classification accuracy of 98.28%. Furthermore， the model requires only 19.2 MB of storage.This study verifies the effectiveness of the lightweight model in identifying surface sediment visual features， and provides technical support for rapid graded evaluation of sediment transport load and water quality early warning.

The middle and lower reaches of the Keriya River lie deep within the Taklamakan Desert， where its channel morphology evolution directly impacts regional water resources patterns and the stability of oasis ecosystems. This study utilizes Landsat series remote sensing imagery from 2000 to 2023， integrating meteorological， population， and land use data. The Google Earth Engine （GEE） platform was employed to extract temporal river water bodies， while comprehensive analysis was conducted using the Digital Shoreline Analysis System （DSAS）， Spearman’s Random Forest model， and Partial Least Squares Structural Equation Modeling （PLS-SEM）. The study reveals： ① MNDWI demonstrates superior identification performance for river channels under complex surface conditions in arid regions， effectively suppressing bare land interference and stably delineating channel boundaries. ② The middle and lower reaches of the Keriya River exhibit an overall evolution from meandering to straightening and stabilization， with decreasing trends in centerline length， water area， and curvature. Channel migration shows spatially active migration in the upper reaches and stable migration in the lower reaches. ③ Driving factor analysis revealed that human activities contributed over 90% to changes in channel centerline length and curvature， while natural factors accounted for approximately 40% of water area variations. PLS-SEM results further indicate that human activities significantly influence channel morphology， with a path coefficient of -0.87， markedly higher than that of natural factors at -0.17. These findings provide scientific basis for water resource regulation and ecological conservation in arid regions within the Keriya River basin.

The pressure difference between the inside and outside of a drop shaft is crucial for its normal operation， and the impingement position of nappe is a key factor in changing the pressure difference. To investigate how the jet impingement position affects the overall air pressure difference of shafts under models with different length scales， three-dimensional numerical simulations with four distinct length scales were conducted to analyze the hydraulic characteristics of the drop shaft under four flow discharge working conditions. The results show that the Z formula can accurately predict the impingement position of nappe in the shaft. Under the same length scale model， the jet impingement position rises rapidly at first and then slowly approaches the horizontal position of the inflow pipe with the increase of flow. Under identical discharge conditions， the jet impingement position rises with the increase of the length scale， while the relative impact position decreases with the increase of the length scale. The absolute pressure difference of the air pressure has a scale effect， while the relative pressure difference and the relative height satisfy a cubic polynomial relationship. The overall pressure difference of the shaft is logarithmically related to the jet impingement position， and the lower the impingement position， the smaller the overall pressure difference. The research results can provide references for hydraulic model tests and the optimal design of shaft structures.

To reveal the evolution law of flow field downstream of curved pipe sections， high-precision spectral element method is adopted to solve the three-dimensional incompressible Navier-Stokes equations. Direct numerical simulations were conducted to investigate turbulent 90° bend pipe flows with varying curvature ratios at a Reynolds number of Re_D =5 300. Self-developed cyclic boundary conditions were implemented to achieve a fully developed turbulent flow at the pipe inlet. Key hydrodynamic parameters downstream of the 90° bend， including the friction coefficient， mean velocity profile， velocity fluctuation， and total turbulence production， were compared and analysed for different curvature ratios to elucidate their variations in the downstream of the bend. The results show that the effect of the bend on the mean velocity， turbulent fluctuation， and total turbulence production gradually diminish as the fluid propagates downstream. The mean velocity profile in the near-wall viscous sublayer， unaffected by the pipe bend， remains consistent with “law of the wall”. The strong shear layer in the pipe core region is found to be the main reason for significant enhancement of turbulence kinetic energy， leading to substantial velocity fluctuations in the same area. A smaller bend curvature ratio prolongs the distance required for flow characteristics downstream of the bend to recover to fully developed flow conditions.

Prolonged droughts can alter rainfall-runoff relationships and induce non-stationary behaviors in watershed hydrological processes. To investigate the characteristics and drivers of such non-stationarity under long-term drought conditions， this study developed a time-varying parameter attribution model based on the Budyko framework， using hydro-meteorological data from 314 catchments across southeastern Australia during 1975-2021. The study period was divided into baseline， drought， and post-drought phases. An 11-year moving window was applied to estimate the temporal variations of parameter ω， and the modified Mann-Kendall test was used to assess the statistical significance of trends. A scenario decomposition method was conducted to quantitatively separate the relative contributions of climate variability and landscape change to runoff variations. Results show significant shifts in rainfall-runoff relationships across the study area. During the drought period， median runoff decreased by 42.1%， far exceeding the concurrent precipitation deficit of -7.2%， indicating a nonlinear amplification effect. Although precipitation recovered in the post-drought phase， 61.8% of catchments exhibited significant lags in runoff recovery. The catchment characteristic parameter exhibited pronounced spatiotemporal variations， with 64% of catchments showing significant increases during drought （median Sen’s slope： +0.027） and slow declines afterward. Changes in water storage could only partially explain parameter variations； most catchments still exhibited non-stationary responses attributed to changes in landscape conditions. Attribution analysis revealed that runoff deficits in most affected catchments were driven by the combined effects of climatic anomalies and landscape changes. Heterogeneity in catchment sensitivity to drought further led to divergent recovery resilience in the post-drought period. This study highlights the driving mechanisms of non-stationary runoff variations under long-term drought and underscores the importance of accounting for landscape dynamics for accurate runoff assessment under changing environmental conditions.

In light of the uneven spatial and temporal distribution of precipitation， the pronounced characteristic of abrupt shifts between droughts and floods， and the escalating drought risks in Hunan Province， this study develops an extreme value analysis model based on a Bayesian Hierarchical Modeling （BHM） framework to scientifically assess regional drought risk. The framework adopts an improved Peaks Over Threshold （POT） model centered on the Meliorated Generalized Pareto distribution （MGP， i.e.， the Burr XII distribution）. It not only strengthens the capability to characterize the heavy-tailed features of extreme drought events， but also integrates provincial and county-scale information through a hierarchical structure， thereby overcoming the limitations of traditional models in addressing spatial heterogeneity and small-sample data. Based on county-level data on agricultural disaster rates and drinking water shortage rates in Hunan Province from 1990 to 2023， the study systematically reveals the statistical patterns of drought risk. The results show that： ① In terms of model performance， the MGP distribution exhibits superior generalization ability on the test set （R ²＞0.93）. Its prediction error for extreme values beyond the 95th percentile can be controlled within 3%， outperforming the 4% achieved by the GP distribution； ② At the provincial scale， extreme drought risks （measured by VaR and CVaR） increase markedly and nonlinearly with rising confidence levels， with agricultural risk displaying higher relative tail intensity and drinking water risk showing greater absolute loss premiums； ③ At the county scale， both agricultural and drinking water risk premiums present distinct spatial heterogeneity. high agricultural risk areas （CVaR-VaR ranging from 15% to 26%） are mainly distributed in central and southern Hunan， whereas high drinking-water risk areas （CVaR-VaR ranging from 25% to 38%） follow an “east-high， west-low” pattern， which profoundly reflects the spatial variability of natural geographical conditions and regional drought-resistance capacity. The findings provide a scientific basis for implementing differentiated drought risk management and adaptive water-resources planning in the Yangtze River Basin.

Addressing the challenges in design coordination arising from the inconsistent standards between drainage and flood control， this study takes Taizhou City， Zhejiang Province as the research area. A coupled return period conversion model based on the binomial probability principle method and the probabilistic connotation method is proposed. The parameters of the long-duration storm intensity formula were optimized using the least squares method， genetic algorithm， and simulated annealing algorithm. The results indicate that the current design standards of pipeline networks in the Jiaojiang， Huangyan， and Luqiao districts are generally low. After return period conversion， the derived flood control standards all fall below the 20-year return period level， whereas the newly planned pipeline system meets this standard. Regarding the storm intensity formula fitting， the least squares method based on the Log-Pearson Type III distribution performed best， with the smallest fitting error. The final optimized parameters were determined as A₁=9.886 7， C=10， b=21.550 7， and n=0.706 2. The research results not only clarify the quantitative correspondence between drainage and flood control standards in Taizhou City but also provide an important theoretical basis and technical support for the scientific planning， standard unification， and parameter revision of urban flood control and drainage systems.

The operational scheduling of cascaded reservoirs during the impoundment period is critical for achieving the synergistic optimization of multiple objectives， including power generation， flood control， and ecological protection. However， a key challenge in practical decision-making is selecting the optimal solution from massive Pareto solution sets while reconciling the conflicting interests of upstream and downstream stakeholders. To address this issue， this study introduces cooperative game theory into the multi-objective decision-making process for cascaded reservoirs. First， a hybrid screening method is employed to extract a set of representative schemes from the Pareto front， thereby reducing the complexity of the decision-making process. Subsequently， based on cooperative game theory， the fairness of benefit distribution among the various operational objectives is quantified using the variance of the Shapley value. Concurrently， the stability of the cooperative alliance between upstream and downstream stakeholders is assessed through the degree of Core constraint violation. Finally， the ideal point method is utilized for the comprehensive optimization and selection of the optimal scheme. A case study on the multi-objective impoundment scheduling for the Nanhe cascaded reservoirs demonstrates that the optimal scheme selected through this decision-making framework yields significant improvements in overall benefits compared to the historical operational plan. Total power generation increased by 18.3%， ecological water deficits and surpluses were reduced by 40.9%， and the flood control risk rate decreased from 23.2% to 3.1%. This decision-making approach provides a scientific method for selecting an operational strategy that not only balances diverse objectives but also ensures the stability of cooperation among stakeholders， offering an effective means to tackle the complex challenge of coordinated scheduling for reservoir groups in complex systems.

Under the combined influence of climate change and human activities， the ecological runoff processes of the Weihe River Basin have undergone significant changes. Clarifying their evolutionary characteristics and driving mechanisms is of great significance for ecological protection and water resource management in the basin. To quantitatively characterize the impact of runoff simulation errors on the attribution of ecological runoff changes， this study takes the Huaxian Hydrological Station in the Weihe River Basin as the research subject. Based on daily runoff and meteorological data from 1960 to 2014， an uncertainty assessment framework for ecological runoff attribution was constructed by coupling probabilistic deep learning with scenario analysis. Runoff mutation points were identified using a sliding t-test， and a Probabilistic Long Short-Term Memory network （Prob-LSTM） model was constructed based on pre-mutation runoff sequences to restore natural runoff and its uncertainty range， thereby quantifying the propagation impact of runoff simulation errors on the attribution results of ecological surplus and ecological deficit. The results indicate that： ① Runoff at the Huaxian Hydrological Station experienced a mutation in 1993， after which the annual average ecological surplus decreased from 0.20 to 0.02， while the ecological deficit increased from 0.07 to 0.30； ② Climate change is the dominant factor in ecological runoff changes， with median contributions to annual-scale changes in ecological surplus and deficit being 90% and 69%， respectively. At the seasonal scale， autumn ecological surplus is primarily controlled by climate change， whereas spring ecological deficit is mainly influenced by human activities； ③ Attribution uncertainty analysis shows that the largest variability occurs in autumn ecological surplus and summer ecological deficit， with uncertainty ranges of 14.3% and 9.2%， respectively， indicating relatively stable attribution results， while uncertainty increases significantly at other temporal scales， reaching up to 73%.

To reveal the spatiotemporal evolution characteristics and major influencing factors of compound drought heatwave events （CDHEs） in Shanxi Province under climate change， this study identified CDHEs using the standardized precipitation evapotranspiration index and daily maximum temperature based on the outputs of three CMIP6 climate models corrected using the Quantile Delta Mapping method. Three indicators， including duration， severity， and magnitude， were constructed to analyze their spatial and temporal variations during the historical period and under the SSP1-2.6， SSP2-4.5， and SSP5-8.5 scenarios. Path analysis was further employed to quantitatively assess the contributions of annual mean daily maximum temperature and annual total water deficit to changes in CDHEs magnitude. The results showed that the characteristics of CDHEs exhibited significant increasing trends under all three future scenarios， and the magnitude of increase became larger with rising emission levels. Compared with the historical period， under the SSP1-2.6， SSP2-4.5， and SSP5-8.5 scenarios， duration increased by 5.22 to 9.55 days， severity increased by 0.63 to 1.07 °C， and magnitude increased by 0.97 to 1.67. The dominant influencing factor of CDHEs in Shanxi Province gradually shifted from water deficit to temperature. During the historical period and under the low-emission and medium-emission scenarios， the contribution of water deficit exceeded 50% in western and northern Shanxi， whereas under the high-emission scenario， the contribution of daily maximum temperature exceeded 60% in most parts of Shanxi Province. In contrast， the synergistic effect of the two factors contributed little to the variation in CDHEs， and their joint coefficient of determination remained at or below 0.2 under all scenarios.

The choice of marginal distribution functions is an important factor in whether the Bayesian Model Averaging （BMA） method can accurately quantify hydrological forecast uncertainty. Inappropriate distribution assumptions can lead to biases in probabilistic forecast intervals， reducing the reliability of forecasts. To evaluate the impact of different marginal distribution functions on the performance of BMA ensemble forecasts， three typical basins with different hydrological and meteorological characteristics—Zijingguan， Yangfangzha， and Tunxi—were selected. BMA ensemble forecast models were constructed based on the Weibull distribution， Gamma distribution， Box-Cox transformation， and log-normal transformation， and the forecast performance of each distribution in different basins was systematically compared. The results show that in basins with large flood peaks， the Weibull and Gamma distributions have stronger fitting ability in deterministic forecasts. For example， in the test set of Zijingguan Basin， the root mean square error（RMSE） are 124.17 and 140.44， respectively， and the Nash-Sutcliffe efficiency （NSE）are 0.74 and 0.67. In contrast， the RMSE of Box-Cox and log-normal transformations are 142.84 and 165.91， with NSE of 0.65 and 0.53. In basins with smaller flood peaks dominated by low flows， Box-Cox and log-normal transformations demonstrate better robustness. However， for the Weibull and Gamma distributions， due to the left-tail characteristics of their probability density functions， the parameter estimation process is easily dominated by low-value samples， causing the 5th percentile in probabilistic forecasts to long-term approach zero and lack fluctuation， failing to reflect the true baseflow variation.

Box-culvert axial flow pump devices are widely used in ultra-low head environments along rivers and lakes， where operating conditions are highly variable and prone to special off-design conditions such as large flow rates， low heads， and stall. These conditions can induce severe pressure fluctuations and flow instability， threatening the operational safety of the unit. To reveal the evolution characteristics and spatial coherence patterns of pressure fluctuations under full operating conditions， an experimental system for a model pump device was established. Eleven pressure fluctuation monitoring points were arranged throughout the entire flow system， from the inlet conduit to the impeller and outlet conduit. Synchronized pressure fluctuation tests were conducted under five typical operating conditions， including near-zero head， large flow rate， best efficiency point （BEP）， high head， and deep saddle zone. Analysis of time-domain statistical features， frequency spectra， and spatial coherence of pressure fluctuation signals under different conditions revealed the following findings： The maximum efficiency of the pump device at BEP was 76.1%， with the dominant pressure fluctuation frequency at the blade-passing frequency （BPF， 77.8 Hz）. Away from the impeller region， BPF intensity rapidly diminished， while the rotational frequency （19.4 Hz） and its harmonics became more prominent. Under ultra-large flow rate （near-zero head） conditions， the rotor-stator interaction effect on the impeller suction side intensified， resulting in pressure fluctuations with complex multi-peak and multi-valley structures. In low-flow conditions， such as high head and deep saddle zone， the dominant frequency of pressure fluctuations decreased， exhibiting chaotic oscillations with significant time delays， along with elevated energy distribution in the ultra-low-frequency range. The intensity of pressure fluctuations was significantly higher in the impeller region than in stationary regions， with the impeller monitoring point （P_imp） exceeding stationary points by more than 10 times. At the deep saddle zone， the pressure fluctuation intensity at the inlet bell mouth （P3） surged， indicating extreme flow instability. Spatial coherence analysis demonstrated that pressure fluctuations at different monitoring points were most strongly correlated at BEP （maximum asynchronous coherence coefficient reached 0.37）. However， under special or unstable conditions， coherence significantly weakened， with near-complete loss of synchronization between impeller and inlet/outlet passage monitoring points. This study provides an important theoretical and experimental foundation for ensuring safe operation， suppressing vibration and noise， and optimizing flow control strategies for box-culvert axial flow pump devices under special operating conditions.

Against the background of high-penetration integration of renewable energy sources such as wind and solar power， hydroelectric units serving as peak-shaving and frequency-regulating power sources face significant stability challenges. To analyze the stability of hydro turbine governor systems in grid-connected mode， this study established a comprehensive mathematical model incorporating the governor， water conveyance system， hydro turbine， generator， and power grid. Building upon this foundation， Hopf bifurcation theory was introduced and applied to the stability analysis of hydro turbine regulation systems. The stable domain of system control parameters in grid-connected operation was determined through theoretical derivation. Time-domain response simulations of system state variables were conducted to investigate their evolution patterns with PID parameter variations. The results validated the accuracy of the obtained stable domain while revealing a superimposed fluctuation phenomenon in state variable responses under grid-connected conditions， which provides theoretical guidance for parameter tuning in hydro turbine control systems.

Under the condition of sediment laden water flow， the jet mechanism of Pelton turbines is subject to long-term erosion by sediment， and the nozzle is an important part of the jet mechanism. The wear and structural deformation of the nozzle seriously affect the efficiency and operational safety of Pelton turbines. To investigate the changes in nozzle wear and nozzle structural deformation， this paper uses the ANSYS Fluent platform to conduct a three-dimensional modeling of the jet mechanism of an Pelton turbine. The SST k- \omega turbulence model， Oka erosion model， and erosion dynamic mesh coupling method are used to numerically simulate the three-phase flow of water， gas， and sand in the jet mechanism， and analyze the effects of different sediment concentrations and water heads on nozzle wear and deformation. The results show that severe sediment erosion occurs at the nozzle throat， and the deformation position is basically consistent with the wear position； The wear rate has a certain regularity and exhibits pulsating characteristics； The deformation is manifested as uneven expansion of the nozzle throat outward， and the degree of deformation is positively correlated with the above variables. This study can provide reference for the safe operation of Pelton turbines under sediment laden water flow conditions.

To investigate the feasibility of using ultrasonic technology to detect cavitation bubbles in water， air bubbles in water are adopted to simulate cavitation bubbles. Theoretical analysis and laboratory experiments are combined to investigate the variation of ultrasonic characteristic parameters after ultrasonic waves pass through water containing bubbles under different bubble content and size. The theoretical model of acoustic speed and attenuation in water containing bubbles are used to calculate the variation of sound velocity and attenuation coefficient with the content and size of bubbles in water. The results indicate that the content and size of cavitation bubble would significantly affect the acoustic speed and attenuation when ultrasonic wave propagate through bubble-containing liquid. With the increase of gas volume fraction， the acoustic attenuation increases and the sound speed decreases gradually. Notably， the acoustic attenuation is more indicative of the bubble content and size in water than the acoustic speed. Additionally， higher ultrasonic frequencies result in greater attenuation. The experiments are designed to use ultrasonic detecting bubbles in water， and the acoustic transmission loss of ultrasonic wave with different frequencies is detected when it passing through the bubble water. The experimental results show that the acoustic transmission loss of ultrasonic waves at various frequencies has a significant response relationship with the bubble content and size， and it increases sharply with the increase of bubble content. The experiments indicate that ultrasonic with frequency exceeding 1.0 MHz can significantly identify the changes of bubble content and size in water. However， the higher the ultrasonic frequency， the greater the acoustic transmission loss， and excessively high frequencies would result in weak signals at the receiving end. Ultrasonic waves within the frequency range of 1.0~2.0 MHz are suitable for detecting bubbles in water.

To support the intelligent transformation of urban water supply pump stations towards proactive control， high-precision， ultra-short-time flow forecasting is essential. Addressing the dual challenges of strong temporal dependencies and complex nonlinear characteristics inherent in five-minute-level flow forecasting， this paper proposes a three-stage hybrid model， termed WD-LSTM-GBRT， based on a “Decomposition-Prediction-Correction” framework. The model first employs Wavelet Decomposition （WD） to decompose the original flow series into multi-scale components， thereby enhancing feature representation. Subsequently， a Long Short-Term Memory （LSTM） network acts as the primary predictor to capture the core temporal dependencies within the series. Finally， a Gradient Boosting Regression Tree （GBRT） is introduced to specifically learn and correct the nonlinear residuals that are highly correlated with external features. Experiments conducted on real-world operational data from a pump station demonstrate that the WD-LSTM-GBRT model significantly outperforms baseline models， including GBRT， LSTM， and CNN-LSTM， across key metrics such as Mean Absolute Error （MAE）， Root Mean Square Error （RMSE）， and the coefficient of determination （R²）. This validates the model’s superiority in ultra-short-term flow forecasting and provides robust technical support for the precise， proactive control of pump stations.

To investigate the influence of variable-load transition regulated jointly by rotational speed and guide-vane opening on the runner of a doubly-fed variable-speed pumped-storage unit， this study performed numerical simulations of the pumping mode under two existing control strategies of a hydropower station. A one-way fluid-structure interaction approach was adopted to compare the dynamic behavior of the runner under different strategies. A blade damage function was then constructed with reference to the Park-Ang two-parameter damage model to assess the degree of blade deterioration. The two control strategies were： ① speed control via the AC excitation system combined with guide-vane opening control， and ② power control via the AC excitation system combined with guide-vane opening control. The results show that， influenced by variations in external characteristics， the maximum equivalent stress and deformation displacement of the runner under the speed-opening control strategy are 167.73 MPa and 0.727 49 mm， respectively， which are significantly lower than those under the power-opening control strategy （279.9 MPa and 1.848 5 mm）. Under both strategies， the main shaft constraint prevented stress release through deformation， causing the maximum stress to occur in the blade region， while deformation decreased gradually from the outer rim toward the main shaft. Based on the constructed blade damage function， it is found that severe damage is more likely to occur at the junction between the blade leading edge and the upper crown， the mid-portion of the blade-lower ring connection， and the junction between the blade trailing edge and the upper crown. Moreover， the maximum damage index under the power-opening control strategy is 0.729 03， which is significantly higher than that under the speed-opening control strategy （0.266 62）， indicating that the runner is more prone to damage under the power-opening control strategy.

The GreenValve integrates fluid regulation and energy recovery into one， and can effectively replace the pipeline turbine to supply power to pipeline equipment. To improve the performance of the GreenValve， computational fluid dynamics software was used to analyze its hydrodynamic performance and compare the blade performance before and after sharp-shaping treatment. According to the structural characteristics of streamlined objects， a turbine blade constructed based on Bézier curve was proposed， and the effects of two parameters including thickness and angle on blade performance were analyzed. The research results show that the blade after sharp-shaping treatment increases the flow coefficient and achieves higher efficiency under the same flow coefficient. Meanwhile， the new impeller based on the Bezier curve further improves efficiency and flow coefficient under the commonly used opening. Therefore， sharp-shaping treatment of blades can effectively enhance the performance of the GreenValve， and the GreenValve impeller based on Bezier curve gains more degrees of freedom in streamlined structure， enabling further improvement of its performance.

Pelton runners are critical load-bearing components in hydropower systems， and the dynamic stresses generated during operation exhibit pronounced nonstationary and multiscale characteristics. Traditional approaches relying on numerical simulations or complete measured sequences struggle to accurately capture transient stress responses and require high data continuity. To address these limitations， this study proposes a modal-decomposition-based and attention-enhanced prediction method tailored to nonstationary multiscale stress features. Firstly， Variational Mode Decomposition （VMD） is employed to extract multi-scale components of stress signals. Subsequently， a Gated Recurrent Unit （GRU） network is used for temporal modeling， while an attention mechanism is introduced to strengthen critical transient responses， enabling the model to maintain stable predictive performance even when only short continuous stress segments are available. Results show that the proposed model achieves a prediction accuracy above 80% across typical operating conditions， with R² values exceeding 0.917 and MAE/RMSE around 1 MPa. Compared with GRU and GRU-Attention models， the proposed method demonstrates notable improvements in both accuracy and stability. These findings provide an efficient and practical approach for dynamic stress prediction and safe operating boundary assessment of hydropower units under constant head conditions and intermittent data availability.

Accurately predicting blasting fragmentation is of great significance for optimizing blasting parameters and controlling costs in mines. Traditional empirical formulas struggle to adapt to nonlinear relationships due to linear assumptions， while existing machine learning models have problems such as relying on experience for feature selection and being prone to falling into local optima during hyperparameter tuning. This study proposes a prediction method that integrates feature selection by Grey Relation Analysis （G）， the Hiking Optimization Algorithm （HOA）， and Least Squares Boosting （LSBoost）. Core features strongly correlated with the average blasting fragmentation （X₅₀） are screened through Grey Relational Analysis， and redundant variables are eliminated. HOA is used to dynamically adjust the search step size to perform global optimization of LSBoost hyperparameters. Finally， LSBoost is driven to iteratively fit residuals to capture nonlinear relationships. Experimental verification based on 97 sets of measured data from 9 mines worldwide shows that the proposed H-HOA-LSBoost model achieves MAE of 0.036 3， MSE of 0.002 7， RMSE of 0.052 3 and R² of 0.907 2 on the test set， outperforming models such as Random Forest （RF）， Extreme Gradient Boosting （XGBoost）， and neural networks （BP）. Ablation experiments verify the synergistic effect of Grey Relational Analysis and HOA. SHAP analysis shows that this model pays more attention to blasting process parameters， providing a scientific tool for engineering parameter regulation.

The bond performance at the rockfill-SCC bond interface formed by rockfill and self-compacting concrete （SCC） in rock-filled concrete （RFC） significantly influences its overall load-bearing capacity. Therefore， it is of great significance to study the influence of interface characteristics on the formation law of its shear strength. Through conducting inclined shear tests on the rockfill-SCC bond interface under different conditions of hydrophilicity/hydrophobicity and roughness， this study analyzed the development pattern of interface shear strength with curing age and the impact of interface characteristics on the failure mode of the shear plane， thereby further examining the influence mechanisms of roughness and hydrophilicity/hydrophobicity on the shear failure of the rockfill-SCC bond interface. The results are as follows： ① The interface shear strength exhibits a power-function increase with curing age， with rapid growth before 14 days and a slower increase thereafter. At 14 days， it can reach approximately 70% of the peak strength. Interface roughness significantly affects the failure mode. Smooth interfaces exhibit brittle debonding failure （a sudden drop in strength after the peak）， while rough interfaces show a mixed failure mode of shear-debonding （gradual softening after the peak）. Hydrophilicity/hydrophobicity has no significant impact on the failure mode. ② At 28 days， under hydrophilic conditions， roughness can enhance the strength by 152.52%， while under hydrophobic conditions， roughness can increase the interface strength by 303.9%. Under rough interfaces conditions， hydrophilicity slightly improves the interface strength by 12.04%， whereas for smooth interfaces， hydrophilicity can enhance the interface strength by 79.22%. ③ Rough interfaces rely on bonding force and interlocking force for shear resistance， resulting in higher strength compared to smooth interfaces. Hydrophilic interfaces promote slurry penetration， leading to higher strength， whereas hydrophobic interfaces have the opposite effect. The research findings provide fundamental support for elucidating the bearing mechanism of RFC.

To achieve accurate prediction of dam displacement and support structural safety assessment， this study focuses on the left bank dam of the Baise Hydropower Project. A finite element model was developed using ABAQUS to simulate displacement responses under various water level conditions， thereby identifying variation patterns and sensitive areas. Subsequently， a prediction model combining the Butterworth Filter （BF） and Gated Recurrent Unit （GRU）， designated as BF-GRU， is proposed to forecast dam displacement trends. The results indicate that the displacement response at the dam crest is the most pronounced， with both horizontal and vertical displacements reaching their maximum values and exhibiting an increasing trend as water levels rise. Consequently， particular attention must be paid to the structural safety of the dam crest under high water level conditions. For the three monitoring points， the BF-GRU model achieved coefficients of determination （R²） of 0.908， 0.869 and 0.801， respectively， with both the Mean Absolute Error （MAE） and Root Mean Square Error （RMSE） remaining below 1.5 mm. Compared to the original GRU model， the BF-GRU model improved the R² values at monitoring points D2-V and D3-V by 18.1% and 26.3%， respectively. Additionally， the maximum reductions in MAE and RMSE exceeded 30%， demonstrating that the filtering process effectively mitigates noise and enhances feature extraction capabilities. In conclusion， by integrating filtering denoising with deep learning， the BF-GRU model achieves high-precision prediction of dam displacement， providing a reliable technical approach for dam deformation monitoring and structural safety assessment.

Traditional sensitivity coefficients are usually used to measure the impact of uncertain factors on the reliability of sluice structures. However， sensitivity coefficients are taken at the design verification points of the influencing factors which are generally obtained through approximate linear expansion of functional functions. In order to overcome the defect of locality， based on the reliability analysis model of sluice structures， a global importance measure index of sluice reliability influence factors is constructed to replace the local sensitivity coefficient， and a global measurement method for the reliability influencing factors of sluice structures is proposed. Then， quantitative analysis of the global importance measure index is achieved by the efficient truncated importance sampling method. Finally， the global importance measure indexes and the sensitivity coefficients of the influencing factors affecting the anti-sliding reliability of a sluice on a clayey foundation are compared. The result shows that the impact effect measurement values calculated by the global importance assessment method are more reasonable. Based on the quantitative results， factors with lower impact effects are treated as fixed values to simplify the reliability analysis process. Meanwhile， factors with significant impact effects can receive more targeted attention in the sluice operation and maintenance plans， which helps to ensure the sluice function reliability.

The hydraulic evolution during the filling process of long-distance water diversion projects is complex. Traditional methods employing the Saint-Venant equations for simulation are time-consuming and fail to meet the real-time requirements of digital twin platforms. To achieve efficient and accurate simulation of the filling process， this paper develops a one-dimensional rapid prediction model based on the physical characteristics of slow initial filling conditions. The filling process is discretized into time steps. Under constant inflow boundary conditions and water level–discharge constraints， quasi-steady hydraulic formulas are used to recursively compute the water levels， flow rates， and flow arrival times at each stage. Validated through a case study， the results show that the deviation between the flow arrival times at key locations predicted by the model and the field measured values is within ±5 minutes， and the simulation time is controllable within 5 seconds. The model offers accurate predictions and high computational efficiency， providing reliable algorithmic support for the initial filling simulation on digital twin platforms for long-distance water diversion projects.

To address the issues of low modeling efficiency and insufficient intelligent reasoning capability in traditional water network planning and decision-making methods for provincial-scale digital water network modeling and water resource allocation， this paper proposes an AI agent-assisted digital water network planning and decision-making method for water networks. The proposed method integrates reasoning chains with external tool calls， alternately iterating planning reasoning and tool execution to achieve the unification of digital modeling and logical reasoning capabilities. Based on an open-source large language model and the LangGraph framework， a specialized AI agent named AquaMind is developed for digital planning and decision-making of the Guizhou water network. Through automatic mapping between front-end/back-end tools and JSON Schema， the agent supports unified invocation of backend application programming interfaces （APIs） as well as frontend geographic information system （GIS） and interface tools. A standardized WebSocket message interaction mechanism is designed to construct an intelligent workflow encompassing semantic parsing， intent recognition， and tool scheduling. By leveraging three working modes—planning， editing， and automatic—the system achieves dynamic collaboration among water network object management， model control， and visual interaction. The results show that the proposed method outperforms traditional expert-knowledge-based reasoning methods in modeling accuracy， planning efficiency， and reasoning reliability. This method provides an accurate， intelligent， and efficient solution for provincial-level digital water network planning and scheduling decision-making， promoting the transformation of water network management systems from digitalization to intelligentization.

The self-compacting concrete （SCC） distributor is the critical equipment for the mechanized construction process of rock-filled concrete （RFC） dams. Accurately identifying the operating status of a SCC distributor is a challenge that must be addressed for remote monitoring of pouring point locations and trajectory. This study proposes an intelligent method for judging the working posture of concrete distributor based on algorithms such as keypoint detection and pose estimation using the YOLOv11-pose deep learning model. Five key feature points of the SCC distributor—including the main shaft， the distribution boom， and the discharge outlet—were selected. An image dataset was constructed and enhanced online using data augmentation strategies. By calculating the displacement and angular changes between feature points and combining them with preset thresholds， the working states of the distributor are categorized into four types： stationary， rotating around the main shaft， rotating around point C， or moving as a whole. Combined with practical engineering applications on a domestic RFC dam， this method has proven effective in tracking the feature points of the distributor and accurately identifying significant motion states， especially a rotation around the main shaft. This method can be used to optimize the concrete pouring path and filling effect， thereby supporting intelligent visual monitoring of mechanical equipment.

Reservoir impoundment exerts a profound influence on the groundwater seepage field in high dam foundations. Identifying the preferential pathways of groundwater flow and the sources of leakage after impoundment is thus crucial for evaluating the performance of seepage control system and ensuring the long-term operational safety of high dams. Grounded in a case study from a high dam project in southwestern China， an integrated approach of groundwater temperature analysis， hydrochemical data clustering， and hydrogen-oxygen isotopes analysis is presented to detect the sources and flow paths of the leakage in the dam site. The results show that the cold groundwater seeped out of the drainage system is primarily originated from the upstream reservoir and the downstream river， while the warm groundwater results from the mixing of the confined water in limestone aquifer and surface water. During the reservoir impoundment， the pH values and the concentrations of Na⁺ and Cl^- of the seeped groundwater gradually increased， while the concentrations of TDS， Ca²⁺， Mg²⁺， and HCO {}_{\mathrm{3}}^{─} concurrently decreased. It reveals that ranked by proportion， the leakage in the plunge pool area mainly originates from three sources： the confined limestone aquifer， the reservoir， and the downstream river， with obvious mixing of seepage water from diverse sources. Seepage water in the fourth grouting and drainage gallery on the right bank powerhouse originates from reservoir water infiltrating along dislocation zones C₇ and C₈， and seepage in the right bank powerhouse is mainly supplied by reservoir water flowing through dislocation zones C₂~C₄. Seepage water in the left bank powerhouse primarily comes from reservoir water infiltrating via dislocation zones C₃~C₅ and is mixed with basalt fissure water. The grout curtain at the dam foundation presents favorable anti-seepage performance， and its seepage water also features multi-source mixing. Overall， reservoir water constitutes the dominant seepage source in the complex area， followed by deep limestone confined groundwater， downstream river water and basalt fissure water on both banks. The primary seepage pathways are interlayer dislocation zones and a small number of unblocked boreholes from geological exploration， which are significantly controlled by topographic conditions and anti-seepage systems. These findings are of merit for the safe construction and efficient operation of similar high dam projects.

After reservoir impoundment， the interaction between fissure water pressure and structural planes significantly influences slope instability and failure. To evaluate the stability of water-affected anti-dip layered rock slopes， this paper uses an engineering case as a geological prototype to analyze the bending and overturning failure process of slopes under water storage conditions， and establishes a mechanical model for the bending and overturning failure of anti-dip rock slopes. Based on the maximum tensile stress theory and Mohr-Coulomb criterion， the mechanical equilibrium equations for arbitrary rock strata under the combined effects of joint water pressure and structural planes are derived. Utilizing the limit equilibrium method and Pan Jiazheng’s principle of minimum value， an analytical approach for the bending and overturning failure of anti-dip rock slopes is proposed. By comparing with the method proposed by Aydan et al.， the reliability of this approach was validated. The research findings provide theoretical significance and reference value for assessing the overall stability of anti-dip layered rock slopes along reservoir banks and predicting their failure modes.

To reveal the characteristics of key meteorological factors and corresponding yield effects in typical humid cotton regions of China， this study constructed an indicator system comprising the Moisture Surplus/Deficit Index （WSD）， Sunshine Duration （SD）， Heat Stress Days （HD）， and High Humidity Days （HH）. Applying methods such as the Mann-Kendall trend test， spatial interpolation， and stepwise regression， we systematically investigated the spatiotemporal evolution characteristics of these indicators and their impacts on yield across different cotton growth stages in Hubei Province from 1961 to 2020. The results indicated that sunshine duration showed a general decreasing trend across Hubei （84.6% of stations were significant）， while heat stress days generally increased （34.6% significant）. High humidity days exhibited a dominant downward trend （42.3% significant）. Regarding decadal variations， WSD increased initially and then decreased， SD continued to decline， heat stress days decreased before rising， and high humidity days stabilized before dropping. Spatially， high-value areas of WSD were located in southwestern and eastern Hubei. Sunshine duration displayed a spatial pattern of “low in the west， high in the east” across all growth stages. High-value areas for heat stress days were found in central-western and eastern Hubei， while high humidity days were concentrated in the southern border regions. The flowering stage was identified as the critical period influencing cotton yield in Hubei. High humidity days during this stage were the primary factor negatively affecting yield， followed by WSD （negative）， sunshine duration during the budding stage （positive）， and heat stress days during the flowering stage （negative）. This study can provide theoretical support for optimizing water and heat management of cotton fields in humid regions.

To address the issues of unfair allocation and water disputes arising from the reliance on rigid historical methods for rural pond water rights allocation， this study constructs a village-household two-tier water rights allocation model， taking Yidu City， Hubei Province as the study area. At the village level， a combined weighting model based on game theory is adopted for allocation； at the household level， a multi-objective optimization model based on the Gini coefficient is developed for multi-household regulation. The results indicate that： at the village level， the combined weighting method based on game theory concentrates the allocation proportion of each village within the range of 12.7%~15.31%， effectively avoiding the limitations of single weighting methods and achieving a synergistic balance of equity， efficiency， and benefit. At the household level， water rights are regulated via the per-unit-area water share of the pond irrigation systems within administrative village groups. This approach enhances the fairness of allocation while respecting the continuity of historical water use， reducing the Gini coefficient from 0.557 7 to 0.246 8. The proposed model effectively narrows the water use gap among households， improves the equity of pond water rights allocation， helps reduce water disputes， and offers reference value for pond management in other regions.