In this article, the influence of rotor–stator distance on the pump performance and rotor–stator interaction strength within mixed-flow centrifugal pump was investigated based on numerical calculation and test verification. Firstly, the performances of mixed-flow centrifugal pumps with two different rotor–stator distances were obtained and compared with the numerical results, which confirms the high accuracy of the numerical simulation. Next, the performances of mixed-flow centrifugal pumps with five different rotor–stator distances were compared and analyzed. It was found that the hydraulic performance of the mixed-flow centrifugal pump varies slightly as the rotor–stator distance increases. The mean values of the standard deviation of the head and efficiency of the mixed-flow centrifugal pump at each rotor–stator distance under full flow conditions are only 0.16 m and 0.11%, respectively. Then, the strengths of the rotor–stator interaction with different rotor–stator distances were analyzed. It was found that the strengths of the shock interaction, the wake interaction, and the potential interaction were all reduced with increasing rotor–stator distance. Moreover, when the rotor–stator distance is 1.5 mm, the pressure distribution in the circumferential direction of the rotor–stator interference zone shows obvious unstable characteristics: the pressure change amplitude is significantly greater than the other rotor–stator distance of the pressure change amplitude, the maximum and minimum pressure amplitude difference being 56.9 kPa, and with the increase in the rotor–stator distance, the maximum and minimum pressure amplitude difference gradually decreases, with an average value of 32.3 kPa. These findings could provide useful insight into prospects for the improvement of the operational stability of mixed-flow centrifugal pumps, and the results of this study can be extended to all centrifugal pumps using diffusers in the form of vanes as the pressure chamber, which has strong practical application and theoretical value.

Timely and accurate identification of surface damage in hydraulic structures is essential for maintaining structural integrity and ensuring operational safety. Traditional manual inspections are time-consuming, labor-intensive, and prone to subjectivity, especially for large-scale or inaccessible infrastructure. Leveraging advancements in aerial imaging, unmanned aerial vehicles (UAVs) enable efficient acquisition of high-resolution visual data across expansive hydraulic environments. However, existing deep learning (DL) models often lack architectural adaptations for the visual complexities of UAV imagery, including low-texture contrast, noise interference, and irregular crack patterns. To address these challenges, this study proposes a lightweight, robust, and high-precision segmentation framework, called LFPA-EAM-Fast-SCNN, specifically designed for pixel-level damage detection in UAV-captured images of hydraulic concrete surfaces. The developed DL-based model integrates an enhanced Fast-SCNN backbone for efficient feature extraction, a Lightweight Feature Pyramid Attention (LFPA) module for multi-scale context enhancement, and an Edge Attention Module (EAM) for refined boundary localization. The experimental results on a custom UAV-based dataset show that the proposed damage detection method achieves superior performance, with a precision of 0.949, a recall of 0.892, an F1 score of 0.906, and an IoU of 87.92%, outperforming U-Net, Attention U-Net, SegNet, DeepLab v3+, I-ST-UNet, and SegFormer. Additionally, it reaches a real-time inference speed of 56.31 FPS, significantly surpassing other models. The experimental results demonstrate the proposed framework’s strong generalization capability and robustness under varying noise levels and damage scenarios, underscoring its suitability for scalable, automated surface damage assessment in UAV-based remote sensing of civil infrastructure.

A dam deformation prediction method based on interpretable component decomposition and integration is proposed to address the problems of weak interpretability, difficult identification of key factors, and insufficient accuracy in the prediction model of deformation monitoring values of concrete dams due to multiple factors such as environmental loads and time factors. This method first strips the temporal component from the original sequence to obtain the castration sequence. Furthermore, complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is used to decompose and reconstruct it into environmental load components and residual terms. In the process of deformation prediction, based on the characteristics of each deformation component, logarithmic functions, bidirectional long short-term memory (BiLSTM) networks optimized by The Black-Winged Kite Algorithm (BKA), and cloud models are used to fit and predict the temporal components, environmental load components, and residual terms, and the final prediction results are obtained through integration. At the same time, the SHAP (SHapley Additive exPlanations) method is introduced to quantify the contribution of input factors to enhance the interpretability of the model. Case study shows that the model outperforms the comparison model in both prediction accuracy and trend tracking ability, effectively improving the reliability of prediction results and significantly increasing the interpretability of deformation prediction, providing a more reliable analysis technique for dam deformation safety monitoring.

Accurate pixel-level segmentation of small and complex cracks in hydraulic concrete structures is essential for automated inspection and quantitative condition assessment of civil infrastructure. However, reliable deep learning-based crack segmentation remains challenging due to the limited availability of pixel-level annotated hydraulic concrete images and the substantial domain discrepancy between public crack datasets and UAV-acquired hydraulic imagery. To address these challenges, this study constructs a multi-source infrastructure crack database(MICD) by integrating six publicly available crack datasets from heterogeneous infrastructure scenarios. MICD provides diverse prior knowledge of crack texture, boundary characteristics, morphological connectivity, topological patterns, and background interference for transferable representation learning. Based on MICD, a multi-source deep transfer learning framework, termed MSDT-Mask2Former, is developed for fine-grained hydraulic concrete crack segmentation. A Mask2Former architecture with a Swin-Large backbone is pretrained on the multi-source database and subsequently adapted to the hydraulic concrete domain through progressive target-domain fine-tuning. Morphology-sensitive dense point supervision is introduced to enhance the representation of thin, fragmented, weak-contrast, and boundary-ambiguous cracks, while source-centered L2-SP regularization is further investigated for knowledge preservation. On the fixed hydraulic concrete crack test set, the proposed method achieves a Dice score of 81.55%, a crack IoU of 68.85%, an mIoU of 84.12%, a Boundary IoU of 68.30%, and a clDice score of 86.62%. Compared with the architecture-matched target-only Mask2Former, the proposed framework improves Dice and mIoU by 0.31 and 0.22 percentage points, respectively, while increasing recall from 80.16% to 81.70%. Furthermore, it reduces the image-space crack geometry estimation errors of crack area, skeleton length, mean width, and connected-component count by 37.11%, 49.50%, 17.85%, and 41.83%, respectively, compared with U-Net. The proposed framework provides an effective solution for fine-grained crack segmentation and morphology-preserving quantitative assessment of hydraulic concrete structures under UAV-based inspection scenarios.

The inherent uncertainty of concrete dam systems, together with the complex influence of environmental loads and measurement noise, makes it difficult for traditional deterministic point prediction models to provide reliable deformation forecasts. In particular, the prediction performance of conventional models is highly dependent on parameter settings, while the uncertainty and potential deviation of future displacement responses are often not fully quantified. To address these limitations, this study proposes an intelligent data-driven prediction framework for dam displacement based on the integration of convolutional neural networks and long short-term memory networks. In the proposed framework, convolutional neural networks are used to extract local feature information from monitoring data, while long short-term memory networks are employed to capture temporal dependencies in displacement sequences. The Black-winged Kite Algorithm is introduced to optimize the key parameters of the integrated multi-level network, thereby improving the accuracy and robustness of point prediction. Furthermore, quantile regression is embedded into the optimized learning framework to construct an interval prediction model for dam deformation, enabling the conditional predictive uncertainty associated with displacement evolution to be quantitatively characterized. The engineering case study and comparative analyses with other models demonstrate that the proposed model achieves improved prediction performance for the investigated monitoring point. The interval prediction results further show that, for the investigated dam and monitoring point, the proposed framework can effectively characterize conditional predictive uncertainty and provide additional information for deformation interpretation and safety assessment. Further studies involving additional monitoring points and dams are required to assess its broader applicability.