Predicting small fatigue crack (SFC) propagation remains a persistent challenge, primarily hampered by the scarcity of high-fidelity experimental data and the limited generalization capabilities of conventional models. This study presents a Hybrid Physics-Data Stacking (HPD-Stacking) framework to mitigate these issues. The architecture merges Physics-Informed Neural Networks (PINNs) with adaptive machine learning ensembles, ensuring that rigorous physical constraints guide data-driven flexibility. Bayesian optimization identifies the most effective base learners from a candidate pool, comprising XGBoost, Random Forest, Support Vector Regression, and PINN variants embedding distinct growth laws. A strict "Leave-Group-Out" cross-validation strategy prevents data leakage during validation. In this study, complete experimental curves function as independent test units. Evaluations on titanium alloy datasets indicate that HPD-Stacking outperforms individual models across diverse materials, microstructures, and loading conditions. Crucially, the framework maintains physically meaningful trends while detecting abnormal growth behaviors driven by microstructural interactions. SHAP-based analysis verifies that learned feature attributions adhere to fracture mechanics principles. The proposed approach thus provides a reliable tool for damage tolerance assessment in high-reliability structures.
This study investigates the tensile deformation behavior of TC4-ELI titanium alloys with equiaxed and bimodal microstructures using uniaxial tensile tests and crystal plasticity finite element (CPFE) simulations. The equiaxed structure shows lower tensile strength but higher ductility, with stress concentration mainly at grain boundaries. In contrast, the bimodal structure exhibits higher strength and lower ductility, with more pronounced stress concentration due to grain orientation and phase boundaries. Slip system analysis reveals that equiaxed structures primarily activate basal and prismatic slips, while bimodal structures also engage pyramidal slips, enhancing resistance to deformation. Increasing β-phase content reduces the activation of prismatic and pyramidal slips, while promoting pyramidal and β-<110> slips, leading to lower stress levels during plastic deformation. The tensile strength of different texture models follows the order: basal < transverse < bimodal structure. The basal texture has the highest prismatic slip activity and lowest strength. In contrast, the bimodal structure shows a higher proportion of pyramidal slip activation, contributing to its superior strength. These findings provide insights into the microstructure–property relationships of TC4-ELI alloys, supporting optimized design for high-performance applications.
2,2′‐Azobis(2‐imidazoline) dihydrochloride (AIBI) is a widely used initiator in emulsion polymerization. As a water‐soluble azo initiator, AIBI offers high initiation efficiency and has more environmental protection. However, its heat sensitivity poses a potential risk of fire or explosion. Additionally, the accidental mixing of incompatible substances may increase the risk of thermal runaway accidents during the storage and application of AIBI. In this study, the thermal properties of AIBI, AIBI with toluene, and AIBI with hydrochloric acid (HCl) were examined using differential scanning calorimetry (DSC) and adiabatic accelerated calorimetry (Phi‐Tech II). Kinetic triplet states were determined from experimental data using thermodynamic methods. The risk of AIBI and AIBI‐Toluene was assessed by semi‐quantitative analysis. Finally, the energy changes during the initial pyrolysis of AIBI and its mixtures were investigated by theoretical calculations. The results of the study emphasize the influence of impurities on the thermal hazards associated with AIBI and provide valuable theoretical insights for the protection of personnel and equipment.
As vital infrastructure for offshore energy transportation, submarine pipelines face significant safety challenges from fishing vessel trawl operations. This study analyzes towing force characteristics of trawl gear on submarine pipelines to assess their impact on pipeline integrity and provide engineering recommendations. Using DNV-RP-F111 standards and OrcaFlex software, a numerical model was established to simulate trawl-pipeline interactions across various operational stages. Results demonstrate that trawl gear type, pipeline stiffness, end constraints, span height, and seabed properties significantly affect towing forces. The numerical model achieves excellent validation with 1.0% mean relative error compared to DNV-RP-F111 predictions. Key findings include span height as the most influential parameter, flexible pipelines experiencing 21% lower forces than steel pipes, and axial forces reaching 1800 kN in long systems. The study reveals systematic conservatism in current DNV guidelines for long pipeline systems, suggesting opportunities for design optimization. This research provides theoretical foundation and technical support for submarine pipeline design, operation, and maintenance.
To efficiently degrade printing and dyeing wastewater at low costs, Cu 2 O/Al 2 O 3 catalyst was prepared by electrochemical method. Under its combined action with H 2 O 2 , photocatalytic degradation of methyl orange solution, as simulated wastewater, was carried out. With the aid of X-ray diffractometer, X-ray microanalysis system and scanning electron microscope, the structure and composition of catalyst film were designed by adjusting the electrochemical parameters, and the performance of photocatalyst was evaluated by the degradation rate of methyl orange. It was found that after anodizing aluminum alloy for 2 h in a 2 mol/L H 2 SO 4 solution under a voltage of 15 V, alumina film with reasonable pore size distribution could be prepared on the alloy surface. By electroplating the film in a 0.01 mol/L CuSO 4 solution at 5°C, and then hydrothermally treating it at 200°C, the Cu 2 O/Al 2 O 3 photocatalyst could be obtained. The prepared Cu 2 O/Al 2 O 3 and 10% H 2 O 2 were used jointly as photocatalyst to degrade 10 mg/L methyl orange solution for 480 min, which achieved a methyl orange degradation rate of up to 82.1%.
