The study was conducted at the Dafeng Atmospheric Corrosion Test Station in Yancheng，Jiangsu Province，which is representative of a subtropical marine climate. Customized automotive corrosion sensors were fabricated from the five test materials，including resistance probes for cumulative corrosion loss and galvanic probes for real-time corrosion rate，and were exposed together with standard corrosion coupons (100 mm×50 mm×3 mm) made of the same materials. During the one-year test period，corrosion rate data and a comprehensive set of environmental parameters，including temperature，relative humidity，rainfall，light intensity，wind speed/ direction，and the concentrations of CO2，SO2，NO2，H2S，PM1.0，PM2.5，and PM10，were synchronously recorded at a frequency of 1 measurement per minute. After exposure，the corrosion coupons were analyzed using scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS). The data were analyzed using corrosion clock diagrams，Pearson correlation analysis，random forest feature importance，SHAP values for model interpretation，and three regression models，namely linear regression，random forest，and gradient boosting regression.
Several key findings were obtained from the integrated analysis. First，significant differences in corrosion resistance were observed. After one year of exposure，2507 stainless steel，304 stainless steel，and T2 pure copper exhibited excellent durability，with only slight surface discoloration.6061 aluminum alloy developed obvious pitting corrosion，and pits were visible after the removal of corrosion products. Galvanized steel suffered the most severe corrosion，manifested as a thick corrosion product layer (mainly ZnO and carbonates) and uniform substrate thinning. EDS analysis confirmed that the corrosion products were mainly oxides of the base metals (Fe，Cr，Al，Zn，and Cu)，while basic copper carbonate was also present on the surface of T2 pure copper. Second，the high-frequency sensor data revealed distinct diurnal corrosion patterns. The corrosion clock diagrams showed that the corrosion activity of 2507 stainless steel，304 stainless steel，T2 pure copper，and galvanized steel was mainly concentrated during the daytime period (08:00－16:00)，which was associated with the temperature rise and dry-wet cycles following nighttime condensation. In contrast，6061 aluminum alloy exhibited continuous，low-level corrosion throughout the day-night cycle. The annual cumulative corrosion loss quantitatively confirmed the durability ranking of the materials. Third，the machine learning analysis of galvanized steel clearly identified relative humidity as the key environmental factor driving its corrosion process. Both the Pearson correlation analysis (the correlation coefficient was about 0.2) and the random forest feature importance analysis ranked humidity above all other variables. The SHAP analysis revealed that high humidity (>80％) had a strong positive effect on the predicted corrosion rate，whereas rainfall events were generally associated with a reduction in corrosion rate，possibly because rainwater washed away corrosive deposits. Temperature and CO2 exhibited secondary or nonlinear effects. Under high-humidity conditions，the dispersity of corrosion rate values increased sharply，indicating a transition toward more aggressive corrosion conditions. Fourth，the comparative evaluation of the predictive models demonstrated the superiority of nonlinear algorithms. The linear regression model，constrained by linear assumptions，showed the poorest performance，especially in predicting higher corrosion rates (>5 μm/ a). Both tree-based ensemble models (random forest and gradient boosting regression) significantly outperformed linear regression. The random forest model exhibited the best overall robustness，and its predicted values were in close agreement with the measured values over the full range from low to high corrosion rates. The gradient boosting regression model showed very high accuracy in the low-corrosion-rate region (<4 μm/ a)，but it was slightly more prone to bias when predicting extremely high-value events.
This study successfully demonstrated a novel big-data-driven paradigm for investigating the atmospheric corrosion of automotive materials.The main conclusions were as follows:(1) In the Yancheng marine atmospheric environment，the five materials exhibited a clear hierarchy of corrosion resistance:the stainless steels and T2 pure copper performed best，the aluminum alloy developed pitting corrosion，and galvanized steel showed the highest corrosion sensitivity.(2) The corrosion behavior exhibited material-specific diurnal patterns，which were associated with temperature-and humidity-driven dry-wet cycles.(3) Relative humidity was confirmed as the most critical single environmental factor controlling the corrosion rate of galvanized steel.(4) Nonlinear machine learning models，especially random forest，were far superior to linear models in predicting corrosion rates from complex environmental data and provided a powerful tool for material selection and service-life prediction. This study provides a scientific basis for developing targeted corrosion protection strategies and improving the environmental adaptability of automotive materials.
The tensile test results showed that the strength of the alloys increased proportionally with increasing Al/ Ti content，which was mainly attributed to the combined effects of grain refinement strengthening and solid-solution strengthening induced by Al/ Ti addition. Among the three tested alloys，Mn30-42-T exhibited the optimal comprehensive properties，with an ultimate tensile strength of 708 MPa，an elongation of 22.3％，and a hydrogen embrittlement susceptibility index of only 2.6％. The SSRT tests were conducted in a hydrogen-saturated environment at room temperature，and the susceptibility index was calculated as the ratio of the tensile strength in the hydrogen-containing environment to that in air. The electrochemical tests conducted in 3.5％ NaCl aqueous solution showed that Mn30-42-T exhibited the best corrosion resistance，with a corrosion potential of－0.312 V and a corrosion current density of 2.75×10－8 A/ cm2. This was attributed to its more stable and denser surface passive film，whose corrosion resistance was significantly better than that of the Mn30-T and Mn30-62-T alloys.
Notably，excessive Al/ Ti addition significantly reduced the hydrogen embrittlement resistance and corrosion resistance of the alloys. This phenomenon was most likely related to local microstructural inhomogeneity caused by excessive Al/ Ti additions，which altered the chemical composition and compactness of the surface passive film and thus reduced the film′s stability. In addition，excessive Al/ Ti may also lead to the formation of a small amount of brittle intermetallic compounds，further deteriorating the ductility and hydrogen embrittlement resistance of the alloys. In summary，an appropriate synergistic addition of Al/ Ti can effectively optimize the microstructure and comprehensive properties of CrFeNiMn-based MEAs，whereas excessive addition will produce adverse effects. This study clarified the synergistic mechanism of Al/ Ti in high-Mn CrFeNiMn MEAs and provided a practical technical route for designing high-performance MEAs with excellent strength-ductility balance，corrosion resistance，and hydrogen embrittlement resistance.
Wear tests were performed on a multi-functional friction and wear tester. The wear scars on the sample surfaces were analyzed using a metallographic microscope，scanning electron microscope (SEM)，and white light interferometer. A ball-on-disc contact configuration was used.By analyzing the friction coefficient (COF)，X-directional force (FX)，and wear morphology，the wear morphology and wear mechanism transitions of 8Cr4Mo4V steel under different experimental conditions were revealed. The results showed that the two ceramic balls caused no abrasive wear on the sample surfaces，only slight wearat 1 800 MPa contact stress under oil lubrication. The dominant wear mechanism was adhesive wear. The domestic ceramic ball pair exhibited the lowest COF and FX values，while the homogeneous steel ball pair exhibited the highest. Under grease lubrication，grooves of varying depths appeared on the sample surfaces for all three pairs. The wear mechanism changed to abrasive wear. However，the ceramic balls showed only slight surface wear. The steel ball pair had the highest COF. The imported ceramic ball pair had the lowest COF and FX values. A large amount of debris caused significant COF fluctuations.
At 2400 MPa contact stress under oil lubrication，the wear mechanisms for the domestic ceramic ball and steel ball pairs were mainly abrasive wear. The imported ceramic ball pair also showed abrasive wear characteristics. Under grease lubrication，the surfaces worn by the steel balls showed wider and deeper grooves，indicating intensified abrasive wear. The two ceramic ball pairs exhibited only slight adhesive wear. The imported ceramic ball pair still exhibited the lowest COF and FX values under both lubrication conditions at high contact stress. The “domestic/imported ceramic ball-steel” heterogeneous pairs were more effective in reducing wear and COF than the “steel ball-steel” homogeneous pair under oil and grease lubrication at low contact stress. However，this gap between different materials gradually narrowed at 2 400 MPa contact stress. At 2 400 MPa high contact stress，lubricating oil was easily squeezed out of the contact area. This resulted in a boundary lubrication state and intensified direct contact and cutting effects between micro-asperities. In contrast，grease maintained its lubricating film under high contact stress due to its excellent adhesion and film-forming ability. This reduced the direct contact area，leading to more stable tribological properties and lower FX values.
