基于綜合智能模型的碳鋼大氣腐蝕重要變量提取和依賴關系挖掘

張明; 付冬梅; 張達威; 馬菱薇; 邵立珍

doi:10.13374/j.issn2095-9389.2022.01.10.007

基于綜合智能模型的碳鋼大氣腐蝕重要變量提取和依賴關系挖掘

doi: 10.13374/j.issn2095-9389.2022.01.10.007

張明^{1, 2,},
付冬梅^{1, 3, ,},
張達威^{4, 5, ,},
馬菱薇^{4, 5},
邵立珍^{1, 2}

1.
北京科技大學順德研究生院，佛山 528300
2.
北京科技大學自動化學院，北京 100083
3.
北京市工業波譜成像工程技術研究中心，北京 100083
4.
北京科技大學新材料技術研究院，北京 100083
5.
國家材料腐蝕與防護科學數據中心，北京 100083

基金項目: 科技部科技基礎資源調查專項資助項目（2019FY101404）；北京科技大學順德研究生院科技創新專項資助項目（BK20AE004）

詳細信息

通訊作者:
付冬梅，E-mail: fdm_ustb@ustb.edu.cn

張達威，E-mail: dzhang@ustb.edu.cn

中圖分類號: TG172.3
計量
- 文章訪問數: 482
- HTML全文瀏覽量: 147
- PDF下載量: 68
- 被引次數: 0
出版歷程
- 收稿日期: 2022-01-10
- 網絡出版日期: 2022-02-21
- 刊出日期: 2023-03-01

Extraction of important variables and mining of dependencies of atmospheric corrosion of carbon steel based on a comprehensive intelligent model

ZHANG Ming^{1, 2
,},
FU Dong-mei^{1, 3
, ,},
ZHANG Da-wei^{4, 5
, ,},
MA Ling-wei^{4, 5},
SHAO Li-zhen^{1, 2}

1.
Shunde Graduate School, University of Science and Technology Beijing, Foshan 528300, China
2.
School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing 100083, China
3.
Beijing Engineering Research Center of Industrial Spectrum Imaging, Beijing 100083, China
4.
Institution for Advanced Materials and Technology, University of Science and Technology Beijing, Beijing 100083, China
5.
National Materials Corrosion and Protection Data Center, Beijing 100083, China

More Information

Corresponding author: FU Dong-mei, E-mail: fdm_ustb@ustb.edu.cn; ZHANG Da-wei, E-mail: dzhang@ustb.edu.cn

摘要

摘要: 針對碳鋼在大氣腐蝕過程中影響變量多且作用機制復雜的問題，提出一種基于綜合智能模型的重要變量挖掘框架，利用該框架可以挖掘影響碳鋼早期大氣腐蝕的重要環境變量及其對腐蝕電偶電流產生的影響。本文通過大氣腐蝕監測儀（ACM）收集了我國5個試驗站點的大氣腐蝕數據，首先，構建了隨機森林（RF）、梯度提升回歸樹（GBRT）和BP神經網絡（BPNN）三種機器學習模型；其次，利用多模型集成重要變量選擇算法（MEIVS）量化環境變量的重要性并提取影響碳鋼早期大氣腐蝕的重要環境變量；最后，繪制了環境變量與腐蝕電偶電流的局部依賴曲線（PDP）。仿真結果顯示，MEIVS算法挖掘出的重要環境變量更符合大氣腐蝕的先驗規律；PDP與MEIVS算法的結論具有很好的一致性，重要環境變量對應的PDP的變化幅度大，且PDP的變化趨勢能夠反映環境變量對腐蝕電偶電流的影響。
- 大氣腐蝕 /
- 碳鋼 /
- 模型集成 /
- 重要變量提取 /
- 局部依賴曲線
Abstract: Machine learning algorithms are widely used to predict the corrosion rate of materials in a specific environment. However, the interpretability of such black-box models is poor, which hinders their application in the field of material corrosion. Therefore, to increase algorithm transparency in practical applications, the causal relationship in the material corrosion phenomenon based on machine learning models needs to be further explored. To solve the aforementioned problems, this study analyzed the corrosion process of carbon steel in the atmosphere with many variables and complex mechanisms and proposed an important variable mining framework based on the comprehensive intelligent model. This framework can mine the important environmental variables that affect the early atmospheric corrosion of carbon steel and their influence on the corrosion galvanic current. This study collected the hour-level atmospheric corrosion data, including relative humidity, temperature, rainfall, and O₃, SO₂, NO₂, PM_2.5, and PM₁₀ concentrations, of 45# carbon steel from five test sites in China using the atmospheric corrosion monitor of the China Meteorological Administration. To ensure the stability of the results, three machine learning models with different fitting strategies, namely, random forest, gradient boosted regression trees, and backpropagation neural network, are constructed. Then, the multimodel ensemble important variable selection (MEIVS) algorithm is used to quantify the importance of environmental variables and extract important environmental variables that severely affect the early atmospheric corrosion of carbon steel. Eventually, the partial dependence plot (PDP) of the environmental variables and corrosion galvanic current is drawn. Based on the simulation results, three significant conclusions are obtained: (1) Compared with Pearson’s and Spearman’s correlation coefficients, the important environmental variables mined using the MEIVS algorithm are more consistent with the prior law of early atmospheric corrosion of carbon steel. Relative humidity, temperature, and rainfall have the most significant impact on the early atmospheric corrosion of carbon steel, and O₃ has a considerable influence on the early atmospheric corrosion of carbon steel in Sanya. Moreover, other pollutants in various regions have a weak impact on the early atmospheric corrosion of carbon steel. (2) PDP shows that, in most cases, the corrosion galvanic current of 45# carbon steel is negatively correlated with temperature and positively correlated with relative humidity. (3) PDP and MEIVS are well consistent. The simulation reveals that PDP corresponding to important environmental variables has a greater range of change, and the changing trend of PDP can reflect the influence of environmental variables on the corrosion galvanic current.
- atmospheric corrosion /
- carbon steel /
- model integration /
- important variable extraction /
- partial dependence plot

HTML全文

圖 1 MEIVS算法流程圖. (a) MEIVS算法主流程; (b) 排列算法流程

Figure 1. Flowchart of the multimodel ensemble important variable selection (MEIVS) algorithm: (a) main process steps of the MEIVS algorithm; (b) process steps of the permutation algorithm

下載: 全尺寸圖片幻燈片

圖 2 各地區環境變量重要性得分. (a) 北京; (b) 杭州; (c) 武漢; (d) 青島; (e) 三亞

Figure 2. Importance score of the environmental variables in each region: (a) Beijing; (b) Hangzhou; (c) Wuhan; (d) Qingdao; (e) Sanya

下載: 全尺寸圖片幻燈片

圖 3 各地區相對濕度的局部依賴圖.(a)北京;(b)杭州;(c)武漢;(d)青島;(e)三亞

Figure 3. PDP of the relative humidity in each region: (a) Beijing; (b) Hangzhou; (c) Wuhan; (d) Qingdao; (e) Sanya

下載: 全尺寸圖片幻燈片

圖 4 各地區溫度的局部依賴圖. (a) 北京; (b) 杭州; (c) 武漢; (d) 青島; (e) 三亞

Figure 4. PDP of the temperature in each region: (a) Beijing; (b) Hangzhou; (c) Wuhan; (d) Qingdao; (e) Sanya

下載: 全尺寸圖片幻燈片

圖 5 各地區特定環境下溫度、相對濕度及降雨對對數腐蝕電偶電流的影響.(a)北京;(b)杭州;(c)武漢;(d)青島;(e)三亞

Figure 5. Influence of temperature, relative humidity, and rainfall on the logarithmic corrosion galvanic current in a specific environment of each region: (a) Beijing; (b) Hangzhou; (c) Wuhan; (d) Qingdao; (e) Sanya

下載: 全尺寸圖片幻燈片

圖 6 各地區污染物的局部依賴圖.(a)北京;(b)杭州;(c)武漢;(d)青島;(e)三亞

Figure 6. PDP of pollutants in each region: (a) Beijing; (b) Hangzhou; (c) Wuhan; (d) Qingdao; (e) Sanya

下載: 全尺寸圖片幻燈片

表 1 大氣腐蝕試驗場地理和氣候信息

Table 1. Geographic and climatic information of the atmospheric corrosion test sites

Region	Longitude and latitude	Climate type	External environment	Corrosion grade
Beijing	116°21′E, 39°59′N	Temperate monsoon climate	Country	C3–C4
Hangzhou	120°30′E, 30°22′N	Subtropical monsoon humid climate	Industrial zone	C4
Wuhan	114°15′E, 30°34′N	Subtropical monsoon climate	Country	C3–C4
Qingdao	120°26′E, 36°04′N	Temperate monsoon climate	Coastal industrial zone	C5–CX
Sanya	109°21′E, 18°17′N	Tropical marine monsoon climate	Coastal industrial zone	C4

下載: 導出CSV

表 2 三種模型在不同地區的擬合表現

Table 2. Fitting performance of the three models in different regions

Region	Model	Training set		Testing set
Region	Model	R²	RMSE	R²	RMSE
Beijing	RF	0.956	0.408	0.815	0.743
	GBRT	0.878	0.679	0.788	0.795
	BPNN	0.767	0.938	0.642	1.034
Hangzhou	RF	0.951	0.472	0.795	1.053
	GBRT	0.822	0.903	0.765	1.125
	BPNN	0.687	1.197	0.696	1.280
Wuhan	RF	0.946	0.438	0.751	1.085
	GBRT	0.895	0.611	0.761	1.064
	BPNN	0.727	0.986	0.700	1.191
Qingdao	RF	0.952	0.620	0.756	1.381
	GBRT	0.892	0.926	0.708	1.511
	BPNN	0.785	1.304	0.661	1.628
Sanya	RF	0.931	0.587	0.688	1.501
	GBRT	0.786	1.036	0.654	1.581
	BPNN	0.666	1.292	0.598	1.704

下載: 導出CSV

表 3 不同地區PCC和SCC分析結果

Table 3. Pearson’s and Spearman’s correlation coefficient results in different regions

Region	Method	Result
Region	Method	T	RH	Rainfall	O₃	PM2.5	PM10	SO₂	NO₂
Beijing	PCC	?0.32	0.68	0.69	?0.08	0.24	0.21	0.11	0.11
Beijing	SCC	?0.17	0.56	0.43	?0.03	0.28	0.20	0.08	0.12
Hangzhou	PCC	?0.65	0.70	0.67	?0.25	?0.04	?0.06	?0.16	0.23
Hangzhou	SCC	?0.74	0.78	0.63	?0.34	0.03	0.02	?0.14	0.42
Wuhan	PCC	?0.63	0.65	0.50	?0.32	?0.10	?0.24	?0.25	?0.02
Wuhan	SCC	?0.73	0.73	0.36	?0.41	?0.16	?0.27	?0.36	0.01
Qingdao	PCC	?0.60	0.46	0.52	?0.15	?0.20	0.03	?0.24	?0.19
Qingdao	SCC	?0.62	0.48	0.49	?0.11	?0.16	0.06	?0.45	?0.23
Sanya	PCC	?0.53	0.59	0.51	0.07	0.03	?0.06	?0.02	0.10
Sanya	SCC	?0.54	0.64	0.43	0.01	0.06	?0.02	?0.11	0.07

下載: 導出CSV

www.77susu.com

參考文獻(33)

[1]	Tsuru T, Nishikata A, Wang J. Electrochemical studies on corrosion under a water film. Mater Sci Eng A, 1995, 198(1-2): 161 doi: 10.1016/0921-5093(95)80071-2
[2]	Wan S, Hou J, Zhang Z F, et al. Monitoring of atmospheric corrosion and dewing process by interlacing copper electrode sensor. Corros Sci, 2019, 150: 246
[3]	Palsson N S, Wongpinkaew K, Khamsuk P, et al. Outdoor atmospheric corrosion of carbon steel and weathering steel exposed to the tropical-coastal climate of Thailand. Mater Corros, 2020, 71(6): 1019 doi: 10.1002/maco.201911340
[4]	Li Z L, Fu D M, Li Y, et al. Application of an electrical resistance sensor-based automated corrosion monitor in the study of atmospheric corrosion. Materials, 2019, 12(7): 1065 doi: 10.3390/ma12071065
[5]	Wang J R, Bai Z H, Xiao K, et al. Influence of atmospheric particulates on initial corrosion behavior of printed circuit board in pollution environments. Appl Surf Sci, 2019, 467-468: 889 doi: 10.1016/j.apsusc.2018.10.244
[6]	Meng J T, Zhang H, Wang X, et al. Data mining to atmospheric corrosion process based on evidence fusion. Materials, 2021, 14(22): 6954
[7]	Oesch S. The effect of SO₂, NO₂, NO and O₃ on the corrosion of unalloyed carbon steel and weathering steel—The results of laboratory exposures. Corros Sci, 1996, 38(8): 1357 doi: 10.1016/0010-938X(96)00025-X
[8]	Cai Y K, Zhao Y, Ma X B, et al. Influence of environmental factors on atmospheric corrosion in dynamic environment. Corros Sci, 2018, 137: 163 doi: 10.1016/j.corsci.2018.03.042
[9]	International Organization for Standardization. ISO 9226—2012 Corrosion of Metals and Alloys—Corrosivity of Atmospheres—Determination of Corrosion Rate of Standard Specimens for the Evaluation of Corrosivity. Geneve: International Organization for Standardization Press, 2012
[10]	International Organization for Standardization. ISO 9223—2012 Corrosion of Metals and Alloys— Corrosivity of Atmospheres—Classification, Determination and Estimation. Geneve: International Organization for Standardization Press, 2012
[11]	Zhi Y J, Jin Z H, Lu L, et al. Improving atmospheric corrosion prediction through key environmental factor identification by random forest-based model. Corros Sci, 2021, 178: 109084
[12]	Yan L C, Diao Y P, Lang Z Y, et al. Corrosion rate prediction and influencing factors evaluation of low-alloy steels in marine atmosphere using machine learning approach. Sci Technol Adv Mater, 2020, 21(1): 359 doi: 10.1080/14686996.2020.1746196
[13]	Mansfeld F, Kenkel J V. Electrochemical monitoring of atmospheric corrosion phenomena. Corros Sci, 1976, 16(3): 111
[14]	Yang D M, Mei H W, Wang L M. Corrosion measurement of the atmospheric environment using galvanic cell sensors. Sensors, 2019, 19(2): 331 doi: 10.3390/s19020331
[15]	Mizuno D, Suzuki S, Fujita S, et al. Corrosion monitoring and materials selection for automotive environments by using Atmospheric Corrosion Monitor (ACM) sensor. Corros Sci, 2014, 83: 217
[16]	Shi Y N, Fu D M, Zhou X Y, et al. Data mining to online galvanic current of zinc/copper Internet atmospheric corrosion monitor. Corros Sci, 2018, 133: 443 doi: 10.1016/j.corsci.2018.02.005
[17]	Pei Z B, Cheng X Q, Yang X J, et al. Understanding environmental impacts on initial atmospheric corrosion based on corrosion monitoring sensors. J Mater Sci Technol, 2021, 64: 214 doi: 10.1016/j.jmst.2020.01.023
[18]	Lipton Z C. The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery. Queue, 2018, 16(3): 31
[19]	Pei Z B, Xiao K, Chen L H, et al. Investigation of corrosion behaviors on an Fe/Cu-type ACM sensor under various environments. Metals, 2020, 10(7): 905
[20]	Pei Z B, Zhang D W, Zhi Y J, et al. Towards understanding and prediction of atmospheric corrosion of an Fe/Cu corrosion sensor via machine learning. Corros Sci, 2020, 170: 108697
[21]	Zhang M, Fu D M, Cheng X Q, et al. A two-step method for constructing a cusp catastrophe model based on the selection of important variables. Chin J Eng,http://www.jiang8.net/article/doi/10.13374/j.issn2095-9389.2021.07.19.006 張明, 付冬梅, 程學群, 等. 基于變量選擇的尖點突變模型的兩步構建方法. 工程科學學報,http://www.jiang8.net/article/doi/10.13374/j.issn2095-9389.2021.07.19.006
[22]	Breiman L. Random forests. Machine Learning, 2001, 45(1): 5 doi: 10.1023/A:1010933404324
[23]	Fisher A, Rudin C, Dominici F. All models are wrong, but many are useful: Learning a variable's importance by studying an entire class of prediction models simultaneously [J/OL]. arXiv preprint (2018-06-23) [2022-01-10].https://arxiv.org/abs/1801.01489
[24]	Biecek P. DALEX: explainers for complex predictive models [J/OL]. arXiv preprint (2018-06-23) [2022-01-10].https://arxiv.org/abs/1806.08915
[25]	Zhang Z H, Beck M W, Winkler D A, et al. Opening the black box of neural networks: Methods for interpreting neural network models in clinical applications. Ann Transl Med, 2018, 6(11): 216 doi: 10.21037/atm.2018.05.32
[26]	Nishikata A, Ichihara Y, Hayashi Y, et al. Influence of electrolyte layer thickness and pH on the initial stage of the atmospheric corrosion of iron. J Electrochem Soc, 1997, 144(4): 1244 doi: 10.1149/1.1837578
[27]	Xia Y, Liu P, Zhang J Q, et al. Corrosion behavior of weathering steel under thin electrolyte layer at different relative humidity. J Mater Eng Perform, 2018, 27(1): 202 doi: 10.1007/s11665-017-3101-0
[28]	Xiao K, Gao X, Yan L D, et al. Atmospheric corrosion factors of printed circuit boards in a dry-heat desert environment: Salty dust and diurnal temperature difference. Chem Eng J, 2018, 336: 92 doi: 10.1016/j.cej.2017.11.017
[29]	Koushik B G, van den Steen N, Mamme M H, et al. Review on modelling of corrosion under droplet electrolyte for predicting atmospheric corrosion rate. J Mater Sci Technol, 2021, 62: 254
[30]	Zhi Y J, Yang T, Fu D M. An improved deep forest model for forecast the outdoor atmospheric corrosion rate of low-alloy steels. J Mater Sci Technol, 2020, 49: 202
[31]	Ha M Y, Jeon S H, Jeong Y S, et al. Corrosion environment monitoring of local structural members of a steel truss bridge under a marine environment. Int J Steel Struct, 2021, 21(1): 167 doi: 10.1007/s13296-020-00424-3
[32]	Rajendran K, Sumi A, Bhattachariya M K, et al. Influence of relative humidity in Vibrio cholerae infection: A time series model. Indian J Med Res, 2011, 133: 138
[33]	Yoon Y, Angel J D, Hansen D C. Atmospheric corrosion of silver in outdoor environments and modified accelerated corrosion chambers. Corrosion, 2016, 72(11): 1424