我国不同地区钢材大气腐蚀预测算法评估与筛选

doi:10.11902/1005.4537.2023.257

中国腐蚀与防护学报

2024, Vol. 44

Issue (4): 939-948 CSTR: 32134.14.1005.4537.2023.257 DOI: 10.11902/1005.4537.2023.257

研究报告

本期目录 | 过刊浏览

我国不同地区钢材大气腐蚀预测算法评估与筛选

沈坚¹^,², 吴柯娴¹^,²^,³(

), 何晓宇¹^,², 方兴龙⁴

1.浙江数智交院科技股份有限公司杭州 310006
2.综合交通运输理论交通运输行业重点实验室杭州 310006
3.浙江大学结构工程研究所杭州 310058
4.浙江海港内河港口发展有限公司杭州 310005

Evaluation and Screening of Atmospheric Corrosion Prediction Algorithms of Steels in Different Regions of China

SHEN Jian¹^,², WU Kexian¹^,²^,³(

), HE Xiaoyu¹^,², FANG Xinglong⁴

1. Zhejiang Institute of Communications Co., Ltd., Hangzhou 310006, China
2. Key Laboratory of Integrated Transportation Theory and Transportation Industry, Hangzhou 310006, China
3. Institute of Structural Engineering, Zhejiang University, Hangzhou 310058, China
4. Zhejiang Seaport River Port Development Co., Ltd., Hangzhou 310005, China

引用本文:

沈坚, 吴柯娴, 何晓宇, 方兴龙. 我国不同地区钢材大气腐蚀预测算法评估与筛选[J]. 中国腐蚀与防护学报, 2024, 44(4): 939-948.
Jian SHEN, Kexian WU, Xiaoyu HE, Xinglong FANG. Evaluation and Screening of Atmospheric Corrosion Prediction Algorithms of Steels in Different Regions of China[J]. Journal of Chinese Society for Corrosion and protection, 2024, 44(4): 939-948.

全文: PDF(2940 KB) HTML

摘要:

针对我国不同地区的各类钢材，提出了一种基于机器学习算法的钢材大气腐蚀深度预测方法，并对不同算法的适用程度进行评估。首先，收集了我国10个大气暴露站的腐蚀检测数据、环境特征和材料特征，采用规范公式与6种机器学习算法预测钢材腐蚀深度，分析预测误差，对比环境腐蚀性等级评估的准确率，筛选适用于我国钢材大气腐蚀的预测模型。进一步分析材料与环境特征敏感性，揭示影响钢材大气腐蚀的主要材料与环境因素。结果表明，相比于规范公式，应用随机森林(RF)和长短期记忆循环神经网络(LSTM)算法的预测模型精度大幅提升；除了规范公式中的温湿度、硫酸盐和氯盐沉积率外，有关雨水酸碱性和雨水腐蚀性离子浓度的特征对钢材腐蚀行为有较大影响，应予以考虑。

关键词 ：钢材, 大气暴露站, 机器学习, 特征敏感性, 环境腐蚀性等级

Abstract：

Atmospheric corrosion of steels is a universal problem. Improving the prediction accuracy of atmospheric corrosion rate of steels in China is of great significance for setting corrosion margin, preventing corrosion failure and reducing the corrosion induced economic loss. For different type of steels in different regions of China, a prediction method of corrosion depth for steels based on machine learning algorithm was proposed, including data acquisition and processing, model training and testing, model evaluation and screening, feature ranking and other steps. The applicability of different algorithms was evaluated and the optimal algorithm of the corrosion prediction for steels was selected. Firstly, the corrosion data, environmental- and materials-features of 10 atmospheric exposure stations in China were collected. The corrosion depth of steels was predicted by using standard formulas and 6 machine learning algorithms. The grades of environmental corrosivity of atmospheric exposure stations were evaluated. Then, the prediction errors were analyzed, the accuracy of environmental corrosivity grade assessment was compared, and the prediction model suitable for steel corrosion was screened. Moreover, the sensitivity of materials- and environmental-features were analyzed, revealing the main factors of environments and materials affecting the atmospheric corrosion of steels. The results show that compared with the standard formula, the accuracy of the prediction models is greatly improved by RF and LSTM algorithms. In addition to the terms such as temperature, humidity, sulfate- and chloride-deposition rates mentioned in standard formulas, the acidity, alkalinity and rainwater corrosive ion concentration of rainwaters have a great impact on the corrosion of steels, which should be considered.

Key words： steel atmospheric exposure stations machine learning feature sensitivity environmental corrosive grade

收稿日期: 2023-08-17 32134.14.1005.4537.2023.257

ZTFLH:

TG172.3

基金资助:浙江省交通运输厅科技计划项目(2020003);浙江省交通运输厅科技计划项目(2023007);交通运输行业重点科技项目(2020-GT-010)

通讯作者: 吴柯娴，E-mail: wukexianzju@163.com，研究方向为结构性能评估、全寿命设计及优化

Corresponding author: WU Kexian, E-mail: wukexianzju@163.com

作者简介: 沈坚，男，1969年生，正高级工程师

	服务

	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	沈坚
	吴柯娴
	何晓宇
	方兴龙
44.8K

链接本文:

https://www.jcscp.org/CN/10.11902/1005.4537.2023.257 或 https://www.jcscp.org/CN/Y2024/V44/I4/939

图1 大气站腐蚀数据箱型图

图2 腐蚀深度与特征的相关系数矩阵热力图

表1 化学元素的影响系数

图3 不同方法对腐蚀深度预测结果

表2 不同机器学习模型预测结果的误差指标

图4 大气腐蚀深度预测结果的RMSE

Features	GB	SVM	RF	RBFNN	BPNN	LSTM	CNN	RF importance
Experiment time	16.1%	6.8%	18.1%	5.5%	7.4%	17.6%	31.2%	15.0%
R_pH	-	18.1%	27.8%	22.2%	52.1%	14.5%	0.1%	9.6%
Cr	0.1%	6.5%	0.3%	2.5%	3.2%	8.1%	4.5%	8.1%
R $S O 4 2 -$	-	3.9%	1.4%	5.5%	1.4%	1.5%	20.1%	5.6%
t_sun	-	2.7%	5.5%	3.2%	3.9%	0.5%	0.5%	5.5%
D_dust	-	1.5%	2.4%	0.5%	0.6%	1.8%	1.0%	5.2%
D_sulfation	10.3%	3.1%	0.1%	2.1%	0.4%	2.0%	0.2%	4.6%
T	4.7%	3.3%	2.5%	3.8%	3.9%	4.2%	9.8%	4.5%
D $N O 2$	-	2.2%	0.9%	3.1%	0.5%	3.0%	0%	4.4%
C	0%	5.6%	6.7%	1.5%	0.5%	1.1%	0.1%	4.3%
Cu	0.1%	3.0%	0.6%	10.7%	1.9%	2.6%	0%	3.9%
Si	0.5%	1.0%	7.9%	4.5%	1.9%	5.5%	0.4%	3.3%
D_seasalt	7.1%	0.1%	4.9%	1.5%	2.0%	3.8%	0.1%	3.2%
D $N H 3$	-	1.5%	1.2%	2.3%	2.9%	1.0%	0%	3.2%
RH	59.5%	13.8%	3.3%	6.5%	5.6%	13.6%	7.1%	2.9%
H	-	0.1%	0.5%	0.6%	0.4%	1.3%	13.2%	2.9%
Mn	-	2.5%	2.8%	6.7%	0.6%	1.4%	0.7%	2.7%
S	1.0%	0.6%	3.5%	6.9%	2.3%	2.3%	0%	2.3%
D $H 2 S$		1.3%	1.0%	0.8%	0.2%	2.5%	0%	2.3%
P	0.4%	7.1%	6.3%	1.5%	3.6%	5.4%	0%	2.1%
R_Cl⁻	-	0.6%	1.6%	4.7%	2.6%	2.5%	10.8%	1.7%
V_wind	-	1.3%	0.3%	1.6%	0.8%	1.8%	0%	1.7%
Ni	0.1%	2.8%	0%	0.1%	0%	0.1%	0.1%	1.5%
Mo	-	0.1%	0.2%	0.1%	1.1%	0.4%	0%	0.7%
Ti	-	5.5%	0%	0.2%	0.2%	0.1%	0%	0%
V	-	3.1%	0%	1.0%	0.1%	1.0%	0%	0%
Nb	-	1.2%	0%	0.2%	0%	0.6%	0%	0%
Al	-	0.8%	0.1%	0.2%	0.1%	0%	0%	0%

表3 不同机器学习模型的特征敏感性指标

图5 不同机器学习模型3类特征敏感性百分比

图6 模型预测准确率雷达图

图7 大气环境腐蚀性等级评估准确率的材料类型差异

图8 大气环境腐蚀性等级评估准确率的地区差异

[1]	Zhi Y J, Fu D M, Zhang D W, et al. Prediction and knowledge mining of outdoor atmospheric corrosion rates of low alloy steels based on the random forests approach [J]. Metals, 2019, 9(3): 383
[2]	Wang S S, Ma S J, Che K, et al. Application status of machine learning in field of natural environment corrosion assessment and prediction [J]. J. Chin. Soc. Corros. Prot., 2023, 43: 441
[2]	王莎莎, 马帅杰, 车琨等. 机器学习在自然环境腐蚀评估与预测领域的应用现状 [J]. 中国腐蚀与防护学报, 2023, 43: 441 doi: 10.11902/1005.4537.2022.147
[3]	Ministry of Science and Technology of the People’s Republic of China. National environmental corrosion platform of China [EB/OL]. (2023-07-20). https://most.gov.cn/ztzl/kjzykfgx/kjzygjjctjpt/kjzyptml/201407/t20140716_114280.html
[3]	中华人民共和国科学技术部. 国家材料环境腐蚀野外科学观测研究平台 [EB/OL]. (2023-07-20). https://most.gov.cn/ztzl/kjzykfgx/kjzygjjctjpt/kjzyptml/201407/t20140716_114280.html
[4]	Fang S F, Wang M P, Qi W H, et al. Hybrid genetic algorithms and support vector regression in forecasting atmospheric corrosion of metallic materials [J]. Comput. Mater. Sci., 2008, 44(2): 647
[5]	Jiménez-Come M J, Turias I J, Ruiz-Aguilar J J. A two-stage model based on artificial neural networks to determine pitting corrosion status of 316L stainless steel [J]. Corros. Rev., 2016, 34(1-2): 113
[6]	Pintos S, Queipo N V, de Rincón O T, et al. Artificial neural network modeling of atmospheric corrosion in the MICAT project [J]. Corros. Sci., 2000, 42(1): 35
[7]	Zhi Y J. Data-driven prediction model for small sample data of metal materials atmospheric corrosion [D]. Beijing: University of Science and Technology Beijing, 2019
[7]	支元杰. 大气环境下小样本金属材料腐蚀的数据驱动预测模型 [D]. 北京: 北京科技大学, 2019
[8]	Zhi Y J, Jin Z H, Lu L, et al. Improving atmospheric corrosion prediction through key environmental factor identification by random forest-based model [J]. Corros. Sci., 2021, 178: 109084
[9]	Yang X J. Research on the control of corrosion resistance of Cr-containing low-alloy steel based on corrosion big data technology [D]. Beijing: University of Science and Technology Beijing, 2021
[9]	杨小佳. 基于腐蚀大数据技术的含Cr低合金钢耐蚀性能调控研究 [D]. 北京: 北京科技大学, 2021
[10]	Li Q, Xia X J, Pei Z B, et al. Long-term corrosion monitoring of carbon steels and environmental correlation analysis via the random forest method [J]. npj Mater. Degrad., 2022, 6: 1
[11]	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 [J]. Corros. Sci., 2020, 170: 108697
[12]	Zhang M, Fu D M, Zhang D W, et al. Extraction of important variables and mining of dependencies of atmospheric corrosion of carbon steel based on a comprehensive intelligent model [J]. Chin. J. Eng., 2023, 45(3): 407
[12]	张明, 付冬梅, 张达威等. 基于综合智能模型的碳钢大气腐蚀重要变量提取和依赖关系挖掘 [J]. 工程科学学报, 2023, 45(3): 407
[13]	Feng D C, Wu G. Interpretable machine learning-based modeling approach for fundamental properties of concrete structures [J]. J. Build. Struct., 2022, 43(4): 228
[13]	冯德成, 吴刚. 混凝土结构基本性能的可解释机器学习建模方法 [J]. 建筑结构学报, 2022, 43(4): 228
[14]	China Gateway to Corrosion and Protection. Data management system [EB/OL]. (2023-07-20). http://data.ecorr.org.cn/
[14]	中国腐蚀与防护网. 数据管理系统 [EB/OL]. (2023-07-20). http://data.ecorr.org.cn/
[15]	Evans U R. The Corrosion and Oxidation of Metals: Scientific Principles and Practical Applications [M]. London: Edward Arnold, 1960
[16]	State Administration for Market Regulation, Standardization Administration of the People's Republic of China. Corrosion of metals and alloys—Corrosivity of atmospheres—Part 1: Classification, determination and estimation [S]. Beijing: Standards Press of China, 2018
[16]	国家市场监督管理总局, 中国国家标准化管理委员会. 金属和合金的腐蚀大气腐蚀性第1部分: 分类、测定和评估 [S]. 北京: 中国标准出版社, 2018
[17]	State Administration for Market Regulation, Standardization Administration of the People's Republic of China. Corrosion of metals and alloys—Corrosivity of atmospheres—Part 2: Guiding values for the corrosivity categories [S]. Beijing: Standards Press of China, 2018
[17]	国家市场监督管理总局, 中国国家标准化管理委员会. 金属和合金的腐蚀大气腐蚀性第2部分: 腐蚀等级的指导值 [S]. 北京: 中国标准出版社, 2018
[18]	Cortes C, Vapnik V. Support-vector networks [J]. Mach. Learn., 1995, 20(3): 273
[19]	Ho T K. The random subspace method for constructing decision forests [J]. IEEE Trans. Pattern Analy. Mach. Intell., 1998, 20(8): 832
[20]	Broomhead D S, Lowe D. Multivariable functional interpolation and adaptive networks [J]. Complex Syst., 1988, 2: 321
[21]	Goodfellow I, Bengio Y, Courville A. 6.5 Back-Propagation and Other Differentiation Algorithms [M]. MIT Press, 2016
[22]	Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural Comput., 1997, 9(8): 1735 doi: 10.1162/neco.1997.9.8.1735 pmid: 9377276
[23]	Valueva M V, Nagornov N N, Lyakhov P A, et al. Application of the residue number system to reduce hardware costs of the convolutional neural network implementation [J]. Math. Comput. Simul., 2020, 177: 232

[1]	李乐民, 张洁, 卞亚飞, 缪春辉, 陈国宏, 汤文明. 安徽省内电网设备常用Q235和40Cr钢大气腐蚀特性及其规律[J]. 中国腐蚀与防护学报, 2023, 43(3): 535-543.
[2]	王莎莎, 马帅杰, 车琨, 杜艳霞. 机器学习在自然环境腐蚀评估与预测领域的应用现状[J]. 中国腐蚀与防护学报, 2023, 43(3): 441-451.
[3]	陈梦成, 温清清. 钢材腐蚀损伤过程的元胞自动机模拟[J]. 中国腐蚀与防护学报, 2018, 38(1): 68-73.
[4]	王佳; 孟洁; 唐晓; 张伟 . 深海环境钢材腐蚀行为评价技术[J]. 中国腐蚀与防护学报, 2007, 27(1): 1-7 .

Viewed

Full text

Abstract

Cited

Shared

Discussed