天津海运职业学院

Voronoi structures are widely present in nature, and highly ordered Voronoi structures such as honeycomb structures have gained extensive recognition and in-depth research in the field of sound absorption structure design. However, Voronoi structures in biological tissues are not all highly ordered. Stochastic Voronoi structures are equally prevalent and exhibit excellent multifunctional properties. To further explore the acoustic value of stochastic Voronoi structures, this study proposes a Voronoi sound absorbing porous structure that features both structural stochasticity and performance robustness. First, a theoretical calculation model is established based on microperforated panel theory and Helmholtz resonance theory, enabling the rapid calculation of the structure’s sound absorption coefficient. Then, a systematic analysis is conducted on the effective conditions for absorption performance robustness from four dimensions: unit number, structural randomness, manufacturing errors, and boundary cutting. Results indicate that there exists a unit number threshold associated with absorption bandwidth in the Voronoi structure. When this threshold is exceeded, the structure can exhibit favorable sound absorption robustness against structural stochasticity, manufacturing errors, and boundary cutting. Experimental verification shows that under significant boundary changes, the structure still maintains an average sound absorption coefficient of approximately 0.8 within an absorption bandwidth of approximately 400 Hz. Its favorable low-frequency broadband sound absorption performance and robustness endow it with promising application prospects in engineering fields where cost control, environmental adaptability, and construction efficiency need to be balanced.

The evaluation of underwater small target detection effectiveness is a core challenge for ensuring maritime security and resource exploitation. Traditional static evaluation methods rely on fixed environmental parameters and single metrics, making it difficult to capture the dynamic adaptability of algorithms in complex and time-varying marine environments. To address this bottleneck, this paper proposed a novel dynamic effectiveness evaluation method based on multi-indicator fusion using Python. An environment-coupled dynamic detection probability model was established, extending classical detection theory to time-varying marine environments. The weighted geometric mean was introduced into the modeling of the comprehensive effectiveness index(CEI). An underwater detection dynamic evaluation system(UDDES) was designed and developed, supporting dynamic environment simulation, parallel testing of multiple algorithms, and multi-dimensional effectiveness visualization analysis. Simulation experimental results demonstrate that the detection probability of AI-Det is improved by approximately 99.8% compared with conventional beamforming(CBF). In dynamic environment stress tests involving sudden sea state deterioration and SNR drops, the mean CEI of AI-enhanced algorithms is increased by 36.5% over CBF, along with a 44.8% improvement in robustness coefficient and a 55.3% reduction in tracking stability error. This study shows that the proposed framework and system effectively overcome the inability of traditional static evaluation to quantify dynamic performance evolution, providing systematic theoretical methods and engineering tools for closed-loop testing, optimal selection, and effectiveness prediction of underwater detection algorithms.

The rapid growth of cross-border e-commerce platforms has generated massive volumes of user interaction data, making accurate prediction of customer purchase behavior an important challenge for intelligent online marketing and personalized recommendation systems. Traditional machine learning and sequential deep learning models often fail to effectively capture dynamic behavioral dependencies and influential user interactions during online shopping activities. To address these limitations, this study proposes an Attention-Based Bidirectional Gated Recurrent Unit (ABiGRU) framework for predicting cross-border e-commerce user purchase behavior. Initially, user behavioral data including browsing history, click frequency, cart activity, ratings, session duration, and purchase records are collected from cross-border e-commerce platforms. The collected data are processed using Adaptive Behavioral Noise Filtering (ABNF) and Sequential Interaction Normalization (SIN) techniques to improve behavioral consistency and remove irrelevant interactions. Subsequently, a Feature Embedding Layer transforms sparse behavioral attributes into dense representations for efficient sequential learning. The BiGRU layer is employed to capture bidirectional temporal dependencies in user shopping activities, while the Purchase Attention Mechanism identifies influential interactions that strongly contribute to purchase decisions. Finally, a Sigmoid-based Purchase Prediction Layer performs binary classification to predict purchasing intention. The proposed ABiGRU framework outperforms existing models such accuracy of 98.78%, precision of 98.19%, Recall of 99.30%, F1-score of 99.59% respectively.

Hydrogen serves as a fundamental energy carrier essential for achieving global carbon neutrality, with long-distance pipeline transportation recognized as the most efficient approach for large-scale hydrogen delivery. Nevertheless, hydrogen embrittlement (HE) in pipeline steels constitutes a significant safety challenge for this infrastructure. This review presents a comprehensive and systematic analysis of HE in hydrogen pipeline steels, emphasizing the microstructural fracture mechanisms under hydrogen exposure, the microstructural regulation of hydrogen diffusion and trapping, the interplay of multiple influencing factors, strategies for HE mitigation, and numerical prediction methodologies. Critical factors including microstructural features, hydrogen transport dynamics, and the coupling effects between stress and hydrogen are examined in detail. The review systematically summarizes advanced HE mitigation techniques and strategies, encompassing microstructural engineering, surface modification, alloy composition optimization, finite element modeling, and machine learning applications. Additionally, several principal challenges confronting the field are identified: the ambiguous mechanisms underlying multi-field coupling effects, the limited translation of laboratory-scale findings to practical engineering contexts, especially with regard to multi-scale mitigation of hydrogen-induced fracture, and the incomplete advancement of artificial intelligence-based pipeline integrity assessment tools. To address these issues, the review proposes key research directions and recommendations for future investigations into HE in hydrogen pipeline steels.

With the long-term operation of ships, the performance of marine diesel engines gradually declines due to the wear of internal moving components, increasing the risk of potential failures. Fuel consumption is a critical indicator for assessing engine operating conditions, and accurately predicting baseline fuel consumption under normal operating conditions is essential for evaluating ship energy efficiency and conducting fault diagnosis. To address common issues in marine engine operational data, such as noise pollution, missing values, inconsistent scales, and feature redundancy, a Diesel Engine Data Enhancement and Optimization Framework (DEOF) was developed to systematically improve data quality. Furthermore, to overcome the limitations of existing models, such as insufficient prediction accuracy and poor stability under complex operating conditions, a Meta-learning Diffusion Residual Attention Network (MD-RAN) is proposed. This approach leverages the strengths of diffusion models in nonlinear generative modeling, integrates meta-learning mechanisms to enhance task adaptation speed, employs multi-head attention modules to strengthen feature extraction, and incorporates dynamic residual connections to improve training stability and flexibility. The data used in this study were collected from real-world operations of ocean-going vessels, ensuring high representativeness. This paper systematically benchmarks the proposed model with the traditional learning model. The results are verified to be effective. The MD-RAN algorithm is significantly better than the original model in terms of prediction accuracy, stability, and nonlinear expression ability. The R2 value can reach 0.9853, and the RMSE and MAE are as low as 1.5801 and 1.1879, respectively. Its feasibility will be further evaluated in practical applications in the future. This study not only provides a systematic data-driven modeling framework, offering technical insights for constructing high-quality datasets, but also establishes a novel generative modeling approach for marine diesel engine fuel consumption prediction, providing robust support for intelligent engine maintenance and energy efficiency optimization.