| 131 | 0 | 118 |
| 下载次数 | 被引频次 | 阅读次数 |
铸态双相不锈钢因其高强高韧及优异耐点蚀特性，已广泛应用于海洋平台、石化管道及核电等苛刻环境。然而，传统以目标性能为导向的试错法成本高、周期长，难以满足材料快速设计的需求。本研究通过整合文献数据，收集铸态双相不锈钢的冲击韧性相关数据，构建涵盖“成分—工艺—组织—性能”的多尺度特征空间，提出“机器学习—SHAP—最优子集”建模框架。采用八种主流机器学习算法构建预测模型，十折交叉验证结果表明，XGBoost算法表现最优，其决定系数R2=0.767。结合SHAP分析与最优子集选择的解释性方法，将特征降维至6个关键指标，分别为Mn、N、铁素体含量、抗拉强度、时效温度及时效时间。通过逐级递进引入并量化多尺度特征的边际增益，模型R2提升至0.875。该研究不仅为双相不锈钢冲击韧性的低成本快速预测与优化提供了数据驱动支撑，也为其他钢种的相关研究奠定了基础。
Abstract：Cast duplex stainless steel has been widely applied in harsh environments such as off shore platforms, petrochemical pipelines and nuclear power due to its high strength, high toughness, and excellent pitting corrosion resistance. However, the traditional trial-and-error method oriented to target performance is characterized by high cost and long cycle, which makes it difficult to meet the demand for rapid material design. In this work, by integrating literature data, relevant impact toughness data of as-cast duplex stainless steel were collected, a multiscale feature space covering “composition-processing-microstructure-performance” was constructed, and a modeling framework of “machine learning-SHAP-optimal subset” was proposed. Eight mainstream machine learning algorithms were adopted to build prediction models. Results of ten-fold cross-validation showed that the XGBoost algorithm performed the best with a coefficient of determination R2=0.767. Combined with the interpretive methods of SHAP analysis and optimal subset selection, the features were reduced to 6 key indicators, namely Mn, N, ferrite content, tensile strength, aging temperature, and aging time. By gradually introducing and quantifying the marginal gain of multi-scale features, the R2 of the model was improved to 0.875. This work not only provides data-driven support for low-cost and rapid prediction and optimization of the impact toughness of duplex stainless steel, but also lays a foundation for related research on other steel grades.
[1]YAO Y, WEI J, WANG Z. Effect of long-term thermal aging on the mechanical properties of casting duplex stainless steels[J].Materials Science and Engineering:A, 2012, 551:116-121.
[2]FAN Y, LIU T, XIN L, et al. Thermal aging behaviors of duplex stainless steels used in nuclear power plant:A review[J]. Journal of Nuclear Materials, 2021, 544:152693.
[3]ELSABBAGH F M, EL-SABBAGH A, HAMOUDA R M, et al. Analysis of super duplex stainless steel properties as an austeniteferrite composite[J]. Materials Sciences and Applications, 2015, 6(12):1121.
[4]DOGGETT R D. Welding of FERRALIUM alloy SD40 super duplex stainless steel[J]. Anti-Corrosion Methods and Materials, 1996,43(4):4-7.
[5]LIU C, WANG A, WANG H, et al. Effect of Solution Treatment Temperature on Microstructure, Mechanical Properties, and Corrosion Resistance of 5A Super Duplex Stainless Steel[J]. Journal of Materials Engineering and Performance, 2023, 32(15):6622-6629.
[6]DAS Y, LIU J, EHTESHAMI H, et al. Quantitative nanostructure and hardness evolution in Duplex stainless steels:Under real lowtemperature service conditions[J]. Metallurgical and Materials Transactions A, 2022, 53(2):723-735.
[7]KOKAWA H, TOMITA M, KUWANA T. Effect of nitrogen on impact toughness of duplex stainless steel weld metal[J]. Welding international, 1995, 9(3):217-724.
[8]JIANG Y, TAN H, WANG Z, et al. Influence of Creq/Nieq on pitting corrosion resistance and mechanical properties of UNS S32304duplex stainless steel welded joints[J]. Corrosion Science, 2013, 70:252-259.
[9]WANG G, WANG Q, CHE H, et al. Effects of silicon on the microstructure and propertiesof cast duplex stainless steel with ultrahighchromium and high carbon[J]. Acta Metall Sin, 2019, 56(3):278-290.
[10]YAMADA T, OKANO S, KUWANO H. Mechanical property and microstructural change by thermal aging of SCS14A cast duplex stainless steel[J]. Journal of Nuclear Materials, 2006, 350(1):47-55.
[11]Pineau A, Besson J. Thermal embrittlement of cast duplex stainless steels:observations and modeling[C]//Duplex stainless steels.Hoboken, NJ, USA:John Wiley&Sons, Inc., 2013:161-208.
[12]ZHU Z-H, ZHANG W-D, TU X-H, et al. Effect of sigma phase precipitation on microstructure and properties of cast ZG0Cr26Ni5Mo3Cu3 duplex stainless steel under different heat treatments[J]. China Foundry, 2018, 15(3):182-188.
[13]HASSEL T A, MARKEN L A, ARBO S M, et al. Microstructure development in duplex stainless steels from additive manufacturing with coaxial directed energy deposition and heat treatment[J]. Metallurgical and Materials Transactions A, 2025, 56(2):474-505.
[14]IBRAHIM O H, IBRAHIM I S, KHALIFA T A F. Effect of aging on the toughness of austenitic and duplex stainless steel weldments[J].Journal of Materials Science&Technology, 2010, 26(9):810-816.
[15]Kim T K, Han C H, Kim S H, et al. Effects of ferrite content on the tensile strength and impact toughness of 2.25 Cr-1Mo-0.25 V steels[C]//Materials Science Forum. Trans Tech Publications Ltd, 2010, 654:520-523.
[16]GOTO T, NAITO T, YAMAOKA T. A study on NDE method of thermal aging of cast duplex stainless steels[J]. Nuclear engineering and design, 1998, 182(2):181-192.
[17]商春磊,王传军,刘文月,等.数据驱动的文献辅助管线钢产线落锤撕裂韧性内禀特征关联[J].工程科学学报, 2023, 45(8):1390-1399.
[18]李丰范,匡健隆,季佳浩,等.机器学习在金属材料服役性能预测中的应用[J].工程科学学报, 2024, 46(1):120-136.
[19]SHEN C, MU W, WANG C, et al. Uncovering the generic and alloy-specific governing parameters of deformation-induced martensitic transformation in austenitic steel[J]. Journal of Materials Science, 2024, 59(7):3087-3100.
[20]MU W, RAHAMAN M, RIOS F L, et al. Predicting strain-induced martensite in austenitic steels by combining physical modelling and machine learning[J]. Materials&Design, 2021, 197:109199.
[21]Marcílio W E, Eler D M. From explanations to feature selection:assessing SHAP values as feature selection mechanism[C]//2020 33rd SIBGRAPI conference on Graphics, Patterns and Images(SIBGRAPI). Ieee, 2020:340-347.
[22]SAI N J, RATHORE P, CHAUHAN A. Machine learning-based predictions of fatigue life for multi-principal element alloys[J].Scripta Materialia, 2023, 226:115214.
[23]ZHU D, WU H-H, HOU F, et al. A transfer learning strategy for tensile strength prediction in austenitic stainless steel across temperatures[J]. Scripta Materialia, 2024, 251:116210.
[24]SHANG C, ZHU D, WU H-H, et al. A quantitative relation for the ductile-brittle transition temperature in pipeline steel[J]. Scripta Materialia, 2024, 244:116023.
[25]CHOPRA O, SATHER A. Initial assessment of the mechanisms and significance of low-temperature embrittlement of cast stainless steels in LWR systems[R]. Nuclear Regulatory Commission, Washington, DC(USA). Div. of Engineering; Argonne National Lab., IL(USA),1990.
[26]CHOPRA O K. Estimation of fracture toughness of cast stainless steels during thermal aging in LWR systems[R]. US Nuclear Regulatory Commission(NRC), Washington, DC(United States). Div. of Engineering; Argonne National Lab.(ANL), Argonne, IL(United States),1991.
[27]MICHAUD W, TOBEN P, SOPPET W, et al. Tensile-property characterization of thermally aged cast stainless steels[R]. US Nuclear Regulatory Commission(NRC), Washington, DC(United States). Div. of Engineering; Argonne National Lab.(ANL), Argonne, IL(United States), 1994.
[28]CHOPRA O K, SHACK W J. Mechanical properties of thermally aged cast stainless steels from Shippingport reactor components[R]. US Nuclear Regulatory Commission(NRC), Washington, DC(United States). Div. of Engineering Technology; Argonne National Lab.(ANL),Argonne, IL(United States), 1995.
[29]CHEN Y, WANG S, XIONG J, et al. Identifying facile material descriptors for Charpy impact toughness in low-alloy steel via machine learning[J]. Journal of Materials Science&Technology, 2023, 132:213-222.
[30]JIANG X, JIA B, ZHANG G, et al. A strategy combining machine learning and multiscale calculation to predict tensile strength for pearlitic steel wires with industrial data[J]. Scripta Materialia, 2020, 186:272-277.
[31]TAKANO Y, MIYASHIRO R. Best subset selection via cross-validation criterion[J]. Top, 2020, 28(2):475-488.
[32]BHATTACHARYA A, BISWAL S, BARIK R K, et al. Comparative interplay of C and Mn on austenite stabilization and low temperature impact toughness of low C medium Mn steels[J]. Materials Characterization, 2024, 208:113658.
[33]李国平,裴海祥,李建春,等.σ相对S32760超级双相不锈钢组织和性能的影响[J].工程科学学报, 2016, 38(3):364-370.
[34]YAN F, SONG K, LIU Y, et al. Predictions and mechanism analyses of the fatigue strength of steel based on machine learning[J].Journal of Materials Science, 2020, 55(31):15334-15349.
[35]XIONG J, PENG Y, SAMIUDDIN M, et al. Common mechanical properties of diffusion bonded joints and their corresponding microstructure features[J]. Journal of Materials Engineering and Performance, 2020, 29(5):3277-3286.
基本信息:
中图分类号:TP181;TG142.71
引用信息:
[1]姜峒伯,朱德鑫,商春磊,等.多尺度特征融合与可解释机器学习在双相不锈钢冲击韧性预测中的应用[J].华北理工大学学报(自然科学版),2026,48(01):29-38.
基金信息:
国家重点实验室自主课题(2022Z-09)
2026-01-12
2026-01-12