Brass，as an important metal alloy in ancient times，has been a focus of academic attention for its corrosion products and corrosion mechanisms in cultural relics. In this study，a Qing Dynasty leaded-brass tripod censer exhibiting three differently colored corrosion products on its body was investigated. X-ray fluorescence spectroscopy (XRF)，X-ray diffraction (XRD)，scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS) were employed to systematically characterize the matrix composition，corrosion morphology，and phase composition of the artifact，and its corrosion mechanisms were further investigated. The results showed that the tripod censer was made of leaded brass containing approximately 2％-3％ Pb. The major corrosion product phases identified on the artifact were (Cu2.81Zn2.19)(CO3)2(OH)6，Cu2(OH)2CO3，K2Cu(C2O4)2·2H2O，CaCO3，and CuO. Among these，(Cu2.81Zn2.19)(CO3)2(OH)6 and Cu2(OH)2CO3 were identified as products of the natural atmospheric corrosion of brass. K2Cu(C2O4)2·2H2O was an uncommon corrosion product and was inferred to be associated with contamination by chemical reagents containing oxalic acid and potassium oxalate. CaCO3 was mainly attributed to reaction and deposition involving lime-containing components in incense during burning，together with exogenous deposition from the surrounding environment. CuO was identified as a high-temperature oxidation product formed as a result of localized heating during incense burning. The results demonstrated that the leaded-brass tripod censer had undergone predominantly long-term atmospheric electrochemical corrosion，accompanied by localized high-temperature oxidation as a secondary corrosion process. They revealed the close relationship between corrosion behavior and the material，function and storage environment of artifacts，offering practical references and a theoretical basis for corrosion mechanism study，conservation and preventive protection of similar brass antiquities.
To conserve ten severely corroded Western Xia iron artifacts in the Ningxia Museum，stereomicroscopy，X-ray fluorescence spectroscopy (XRF)，X-ray diffraction (XRD)，and laser Raman spectroscopy were used to analyze the micromorphology，elemental composition，and phase structure of their corrosion products. The results showed that the corrosion products exhibited multiphase composite characteristics，mainly composed of hematite，magnetite，and goethite. At the same time，highly active goethite and high-chloride akaganeite(β-FeOOH) were detected，and sporadic detection of jarosite was also observed. In areas with high-chlorine corrosion，the electrochemical process driven by chloride ions was the main factor causing corrosion；The mixed composition of multiphase corrosion products indicated that the object was in a complex burial or preservation environment for a long time；The detection of sulfate further indicated that the corrosion cause of this batch of artifacts was the result of multiple coupled factors. Based on the above analysis，following the principle of minimal intervention，measures such as mechanical cleaning，alkali desalination，and corrosion inhibition sealing were adopted for graded protection，which achieved good results and can provide reference for the selection of protection methods for similar iron cultural relics.
Metal artifacts are susceptible to corrosion and deterioration during long-term burial and preservation，placing higher requirements on the compatibility，stability，and reversibility of conservation and restoration materials. Owing to their tailorable structures and tunable properties，polymer materials have considerable potential for applications in the bonding and consolidation，surface protection，corrosion inhibition and protection，controlled cleaning，and temporary protection of metal artifacts. This review summarized the application progress of polymer materials in the conservation of metal artifacts. It particularly emphasized the compositional characteristics，modification methods，current-day applications，and performance characteristics of different polymer material systems. It analyzed their application scopes and existing problems. On this basis，the limitations of existing research concerning the long-term stability，interfacial compatibility，reversibility，and environmental safety of polymer conservation and restoration materials were further analyzed. Future development of polymer materials toward higher precision and functionality in metal artifact conservation was also discussed，with a view to providing a reference for the research and application of related conservation and restoration materials.
To further elucidate the scientific significance of the manufacturing process of Warring States-period openwork bronze square dou vessels in museum collections and to clarify the causes of corrosion-related deterioration of the artifacts，two Warring States-period bronze square dou vessels from the collection of the Gansu Provincial Museum were scientifically investigated using portable X-ray fluorescence spectrometry (p-XRF)，scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDS)，metallographic microscopy，ultra-depth-of-field microscopy，Raman spectroscopy，polarized light microscopy，and ion chromatography. The results showed that all parts of the two bronze dou vessels were composed of ternary Cu-Sn-Pb alloys，with an average matrix composition of82.58％ Cu，10.68％ Sn，and 4.90％ Pb. Both the vessel bodies and lids were each cast as a single piece. The corrosion products on the artifact surfaces included malachite (CuCO3·Cu(OH)2)，cuprite (Cu2O)，anglesite (PbSO4)，atacamite (Cu2(OH)3Cl)，and chalcocite (Cu2S)，with localized deposits of calcium carbonate and silicon dioxide. The corrosion layers exhibited a typical oxidation-type stratified structure:the outer corrosion layer consisted mainly of malachite，whereas the intermediate oxidation layer contained cuprite and cassiterite (SnO2). The energy spectrum data of the rust layer showed a decrease in the Cu/Sn ratio from the substrate to the surface layer，while the lead content increased. However，the soil rust layer exhibited higher copper content，lower lead content，and no detectable tin，indicating that tin and lead were mostly deposited in situ during the corrosion process，with less migration to the soil，while copper element migrated and lost more to the soil. During the corrosion process，chloride and sulfate ions migrated toward the bronze surfaces，resulting in the formation and deposition of secondary corrosion products such as atacamite and lead sulfate. The formation of abundant Cu2S on the bronze surfaces may have been associated with microorganisms，particularly sulfate-reducing bacteria，in the burial environment. In addition，differences in the degree of corrosion and mineralization among different parts of the artifacts were attributed to the combined effects of adhering soil and the openwork structure. These findings provided important scientific evidence for understanding the manufacturing technology of similar artifacts and for their conservation and restoration.
Stress concentration caused by surface roughness is a key factor in metal fatigue failure，and establishing the relationship between surface roughness and fatigue life is important for the safe design of mechanical components. Existing calibration of physical-model parameters through empirical formulas and testing makes it difficult to achieve the required fatigue-life prediction accuracy，whereas the data-driven fatigue life prediction model has shortcomings in expressing the influence of parameters on fatigue life. To address these problems，medium-carbon steel with surface roughnesses of Ra＝0.4，0.8，and 1.6 μm was investigated，and a fatigue-life prediction method combining a physical model with machine learning was proposed. Rotating-bending fatigue lives at the three roughness levels were evaluated. Crack-propagation theory and the Murakami criterion were used to develop a roughness-governed physical fatigue-life model by considering the equivalence between roughness and initial crack size and the quantitative relationship between fatigue limit and roughness. The model parameters included fatiguelimit constants，material constants，and roughness. To overcome the difficulty of determining multiple physical-model parameters simultaneously，a data-driven fatigue-life model was developed using stress amplitude and surface roughness as input features and fatigue life as the output.To eliminate the influence of dimensionality，the input features were normalized. At the same time，taking the logarithm of the original fatigue life to reduce its magnitude difference. During the modeling process，three algorithms were used:Random Forest (RF)，Gradient Boosting Decision Tree (GBDT)，and Gaussian Process Regression (GPR). Samples with Ra＝0.4 and 1.6 μm were randomly divided into training and test sets at a ratio of80％ to 20％. For RF and GBDT，five-fold cross-validation and a root mean square error (RMSE)-based grid search were used to select hyperparameters within the training set；for GPR，the L-BFGS-B algorithm was used to maximize the marginal log-likelihood.The models were then retrained with the optimal hyperparameters，and an iterative validation procedure involving 50 repeated random dataset divisions was used to evaluate model stability and prevent overfitting under small-sample conditions. The GPR model showed the best predictive accuracy and statistical robustness. Its test-set coefficient of determination (R2) was 10.29％ higher than that of RF and 1.30％ higher than that of GBDT，and it had the lowest root mean square error (RMSE). An improved whale optimization algorithm with a global-search strategy(GSWOA) was subsequently introduced to optimize the GPR hyperparameters. After GSWOA tuning，R2 reached 0.965，an increase of 3.54％，whereas RMSE decreased by 29.19％ to 0.114. Validation based on repeated random dataset divisions also confirmed a marked reduction in the dispersion of prediction errors. The GSWOA-GPR model has achieved its precise ability to capture the nonlinear laws of fatigue life and enhanced its generalization stability under different roughness and stress levels. The GSWOA-GPR model was then combined with the physical fatigue-life model，and a differential evolution algorithm was used for global inversion of the physical parameters. The material constants m and C were determined as 2.100 0 and 6.89×10-9，respectively，and the fatigue-limit constants α and β were 348.81 and 10.00，respectively.A roughness-controlled fatigue-life prediction equation for medium-carbon steel was thereby established. Independent validation using fatiguelife data at Ra＝0.8 μm yielded an RMSE of 0.090 0，confirming the accuracy and robustness of the physical model. The model effectively captured the nonlinear fatigue behavior under coupled stress and roughness. It provided a methodological reference for fatigue-reliability assessment and life design of medium-carbon steel components.
