It is the key issues of reasonably and accurately predicting the thrust and torque of tunnel boring machines (TBM) to realize the intelligent control of TBMs. This paper proposes a two-stage prediction method of knowledge-data-driven spatio-temporal stacked convolutional network (KD-NTS-GAT). Firstly, based on expert knowledge and the NTS-NOTEARS method, a new information fusion technique is proposed. The discrete expert experience and the continuous NTS-NOTEARS indicators is mapped and smoothly fused through clustering. The causal relationships among the key operating parameters of the TBM is quantitatively extracted to improve the authenticity of the causal relationships significantly. Then, causality is further combined as a prior knowledge with stacked convolutional network deep learning model for predicting thrust and torque of TBM. Taking the bid Ⅳ of Xinjiang Water Conveyance Tunnel Project as an example, a comparative analysis of the KD-NTS-GAT method and the pure data-driven method shows that the KD-NTS-GAT has better prediction capability on thrust and torque. The conclusions can provide a reference for the intelligent control of TBM construction.

Predicting risks in later diversion stages is essential for the scientific evaluation of initial impoundment schemes for hydropower stations. Considering the gradual development process from dam overtopping to the failure of the diversion system, taking the later diversion systems of the upstream existing and downstream under-construction power stations as the research objects, a later diversion risk model for cascade power stations based on level sets is established, and the risk is expressed in the form of interval numbers. Based on the prospect theory, the initial impoundment scheme of the cascade power stations is evaluated by taking risk loss, relative storage reservoir power generation benefits and storage duration as the indicators. Finally, taking two adjacent under-construction power stations in the upper reaches of the Jinsha River as an example, the results show that when the level sets change from 0 to 1, the equivalent recurrence interval corresponding to the later diversion risk interval of the original impoundment scheme meets the flood design standard with a high safety margin. The staged storage time can be advanced by 40 days, which provides a reference for the safety and economy of advanced impoundment.

Addressing the issues of single model algorithm, low accuracy, and poor generalization in existing shield tunneling speed prediction methods, this study proposes a shield tunneling speed prediction approach to improve prediction accuracy based on Variational Mode Decomposition (VMD), Dung Beetle Optimizer (DBO), and Stacking ensemble learning. Firstly, to obtain more effective data, VMD is applied to decompose and reconstruct the original data to obtain denoised construction parameter data for subsequent model prediction. Secondly, based on the ensemble learning strategy, Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) models are selected as base learners, while Gaussian Process Regression (GPR) is chosen as the meta-learner to construct a Stacking ensemble learning prediction model with higher prediction accuracy and stronger generalization ability. Thirdly, to further enhance prediction accuracy, DBO is employed to optimize the hyperparameters of the ensemble learning model. Finally, this prediction method is applied to the shield tunneling construction of a water diversion tunnel project in Henan Province and compared with other prediction methods. Compared to other single models (SVR, RF, XGBoost), the results indicate that the proposed method achieves higher prediction accuracy, with average accuracy improvements of 7.76%, 6.70%, and 4.97%, respectively, providing a new approach for shield tunneling speed prediction.

Cemented gravel dam construction technique combines the advantages of earth-rock dams and concrete dams, and has broad application prospects. Taking Shaping first-level Hydropower Station as an example, production tests were carried out on C₁₈₀6, C₁₈₀10 and C₁₈₀20 cemented sand and gravel, and core drilling tests were carried out on their compressive strength, splitting tensile strength, permeability and SEM scanning tests at the age of 180 days. The results show that under scientific ratio and reasonable construction technology, the compressive strengths of C₁₈₀6, C₁₈₀10 and C₁₈₀20 have reached 7.1 MPa, 14.8 MPa and 29.4 MPa, respectively. Among them, the impermeability grade of C₁₈₀10 with the largest amount of project consumption is W8. The performance of cemented sand and gravel can achieve the expected results, which verifies the feasibility of the construction plan. Thus, it provides technical support for subsequent construction, and also provides reference examples for subsequent hydropower station construction.

Affected by hydrodynamic excitation and other factors, the opening-closing operation of hydraulic gates exhibits multi-field coupling effects and complex nonlinear dynamic characteristics, leading to difficulties in identifying equipment safety states. Test data of gate operation demonstrate that artificial neural network algorithms can identify hydrodynamic excitation disease features and accurately predict its development trends. To address this, BP and GA-BP neural networks were employed to construct identification and prediction models for hydrodynamic excitation disease. These models were applied to identify and forecast the effective values of reel vibration, with model performance evaluated using metrics including Relative Error (R_RE), Mean Absolute Percentage Error (M_MAPE), and Root Mean Square Error (R_RMSE). Compared to the BP model, the results indicate that the GA-BP model achieves reductions of 20.77% in R_RE, 4.74% in M_MAPE, and 6.27% in R_RMSE, demonstrating superior fitting to measured samples and enhanced stability with extended prediction durations, thus providing critical technical support for engineering risk mitigation and hazard prevention.

Perforated structures are commonly found in engineering applications, where stress concentration around holes significantly affects structural load-bearing capacity and safety. Therefore, accurate mechanical analysis of perforated structures is essential. However, the traditional finite element analysis methods face several challenges, including complex mesh generation, high computational resource requirements, and poor convergence when dealing with perforated structures. To overcome these challenges, a finite element analysis method based on the superposition principle is proposed. This approach replaces the original model with a simplified equivalent model featuring a less complex mesh for simulation. The stress field of sub-models is subsequently employed to correct the stress distribution around holes in the equivalent model. Comparative results indicate that the discrepancy between the equivalent model and the original model is within 5%. The equivalent model achieves a 25.2% reduction in mesh elements and a 75.1% reduction in simulation time, substantially improving simulation efficiency.

To address the issue of reduced detection accuracy under complex working conditions due to the fixed threshold of the isolation forest algorithm, an anomaly detection method for ship lock miter gate monitoring data based on singular spectrum analysis (SSA) and an improved isolation forest (KMIF) is proposed. The SSA is employed to decompose and reconstruct the monitoring data, and separate the trend and noise components. The isolation forest algorithm is improved by incorporating K-Means++ clustering to dynamically set anomaly thresholds for different monitoring datasets. The noise component is then fed into the improved isolation forest algorithm for training and anomaly detection. Taking the stress and vibration data from multiple measuring points of the lower lock miter gate in Jiangsu ship gate project as an example for validation, the results show that the proposed SSA-KMIF method performs excellently in terms of false positive rate, precision, recall ratio, and accuracy. It demonstrates high accuracy and flexibility, which provides a reliable technical support for health monitoring of ship lock miter gates.

Aqueducts are common water conveyance structures in water diversion projects, and accurate prediction of aqueduct deformation is crucial for ensuring the stable operation of water conservancy projects. For this purpose, taking the Liaohe Aqueduct in the South-to-North Water Diversion Project as an example, five different linear additive models, namely elastic net regression, multiple linear regression, stepwise regression, ridge regression and LASSO regression, were established based on the long-term deformation monitoring data of the aqueduct. The prediction results of the aqueduct's deformation behavior by the five different linear additive models were compared. The results indicate that as the prediction time increases, the prediction accuracy of different linear additive models gradually decreases. The LASSO model selects the optimal regularization parameter through cross-validation, achieving variable selection simplification and minimizing model complexity. Additionally, it is verified that the training length affects the prediction performance of multiple linear regression and stepwise regression. The findings of this study provide valuable references for selecting prediction model of aqueduct deformation.

The number of long-distance, high drop, pressurized, and self-flowing water pipeline projects is increasing in the northwest region. Most of the pipelines show undulating shapes, and the hydraulic transition process of the entire pipeline system becomes very complex during operation and regulation. When the water hammer protection setting is unreasonable, it will lead to pipe explosion, seriously threatening the safety of people and property. In order to ensure the safe operation of the entire system, the characteristic line method and the HAMMER V8i water hammer analysis software were used to analyze the hydraulic transition process of a long and high drop inverted siphon in a water transmission project. By setting isolation and maintenance valves, submerged energy dissipation valves, and exhaust valves along the pipeline, and setting regulating valves at the end of the pipeline, the positive pressure of the pipeline system is effectively controlled during normal operation and valve closure. By simulating the hydraulic transition process of the pipeline system under different flow rates after pipe explosion, the installation of water hammer protection equipment minimizes the harm caused by pipe explosion. The flow rate of the pipeline system after complete pipe explosion is not continuous. The action time and operation rules of the water hammer protection equipment for long-distance and high drop inverted siphon lines play a crucial role in the safety of the entire system. The research results can provide reference for the similar projects.

According to the prototype characteristic curve of the pump set, there will be a deviation between the theoretical value and the actual value when guiding the actual scheduling of the pumping station, which will affect the accuracy of daily scheduling and energy consumption evaluation. Therefore, taking Liyuzhou pumping station as an example, this paper proposes a method combining the measured data of the pumping station and the law of similarity to correct the prototype characteristic curve of the pump set. The flow optimization distribution model is established in terms of the actual operating conditions, and the difference of optimal flow distribution scheme of pumping station between the characteristic curve of the pump set before and after the correction is analyzed. The results show that the deviation between the actual operating head and the theoretical head is mainly concentrated in 8%-11%, and the deviation between the actual operating efficiency and the theoretical efficiency is mainly concentrated in 6%-9% when the prototype characteristic curve of the pump set is used. When the flow optimization distribution scheme is selected through the corrected characteristic curve of the pump set, the results are more in line with the actual operation situation, which can effectively strengthen the guiding role of the flow optimization distribution scheme of the pumping station on the actual scheduling.