In fuel-cell-powered electric vehicles, the output characteristics of the fuel cell are relatively soft, and the output voltage is unstable. Therefore, a DC-DC converter is required between the fuel cell and the inverter to transform the output voltage of the fuel cell into a suitable voltage for the motor drive. Existing non-isolated DC-DC converters cannot meet the requirements of high voltage gain, high efficiency and a wide range of output voltage simultaneously. To improve these performances, a high-gain DC-DC converter with a wide range of output voltage, based on a switched capacitor structure, is proposed in this paper. The converter supplies power to the load by connecting multiple capacitors with the input source in series in switch-on states, while the input source charges the capacitors through a series connection with an inductor in switch-off states. In comparison to existing converters, the proposed converter maintains high voltage gain at lower duty ratios and offers a wide range of output voltage. The operating principles, key waveforms and parameter design of the topology in Continuous Conduction Mode (CCM) are described and analyzed in detail, and the voltage gain of the proposed converter is compared with some other DC-DC converters. Finally, the results of simulations using Simulink and hardware experiments that were conducted to validate the theoretical analysis are described.

This study develops an interpretable machine learning framework using Shapley additive explanations (SHAP) to analyze gas transmission pipeline failures based on US Pipeline and Hazardous Materials Safety Administration (PHMSA) data (2010–2025). XGBoost achieved optimal performance in classifying six failure modes. SHAP results identified spatial exposure—particularly location type and depth—as dominant risk factors, with right-of-way and shallow-depth segments showing highest vulnerability. Temporal analysis revealed aging effects and seasonal patterns, where natural force damage peaks in winter. Complex nonlinear interactions among operating pressure, pipe diameter, wall thickness, and age underscore the need for integrated safety approaches. These findings support implementing multidimensional integrity management strategies incorporating targeted monitoring and data-driven interventions.

To address issues such as high-dimensionality, small sample size, imbalanced class distribution, and insufficient model interpretability in gas pipeline accident data, we developed an accident severity classification model based on the U.S. PHMSA accident database that integrates data balancing, feature optimization, ensemble learning, and interpretability analysis. After data preprocessing, SMOTENC was employed to oversample underrepresented classes to mitigate class imbalance. Building on this foundation, feature interaction and Relief-F feature selection were used to enhance the representativeness of the accident data. Tree-based models, including LightGBM, XGBoost, CatBoost, and Random Forest, were constructed to establish an ensemble learning model, thereby improving prediction accuracy and stability. Additionally, the SHAP method was introduced to identify key factors influencing accident severity. The results indicate that both feature interaction and feature selection improved model performance to some extent, with the feature interaction method showing a slight overall advantage. Ensemble models outperformed single models in both classification performance and generalization ability, with the best identification results observed for severe accident categories. Key influencing factors were primarily concentrated in the accident-involved objects, accident environment, and accident causes, among which features such as the pipeline structure, pipeline laying environment, and pipeline corrosion made significant contributions to accident severity. These findings provide data support for the identification of gas pipeline accident risks and the implementation of differentiated safety prevention and control measures.

Understanding interactions among risk factors is essential for improving the safety of hazardous liquid pipelines. This study proposes a hybrid framework that integrates association rule mining with Bayesian network modeling to identify key risk factors and quantify their coupling effects. Historical hazardous liquid pipeline accident data are combined with meteorological and soil property data to support coupling-aware risk analysis. A fault tree is constructed and converted into a Bayesian network to evaluate the importance and sensitivity of accident factors across multiple failure scenarios. The results show that neglecting coupling effects leads to systematic bias in risk prioritization. Without coupling, pipeline monitoring system failure and shallow burial depth rank as the most influential factors, whereas incorporating coupling effects elevates internal microbial corrosion to the highest-ranked risk factor and introduces galvanic corrosion among the top contributors. Sensitivity analysis further indicates that corrosion-related and management-related factors exhibit markedly increased influence under coupled conditions, with distinct dominant drivers observed for leak, rupture, and mechanical puncture accidents. These findings demonstrate the critical role of factor coupling in accident propagation and show that explicitly modeling interaction effects improves causal transparency and supports targeted, proactive pipeline safety management.