Effective Utilization for Data of Natural Environment Corrosion of Materials
Haitao Wang,Leijiang Yao,Xiaofeng Li,Enhou Han,Xiaoyan Tong
DOI: https://doi.org/10.5334/dsj-2015-018
2015-01-01
Data Science Journal
Abstract:An analysis architecture is built to make full use of natural environment corrosion of materials data. All environmental factors, such as average temperature, average relative humidity, rainfall hours, sunshine duration, chloride ion settlement, etc., in nine atmospheric test stations, including Beijing, Wuhan, Jiangjin, and Wanning, and seawater temperature, salinity, dissolved oxygen, pH values, etc. in four marine test stations, including Qingdao, Zhoushan, Xiamen, and Yulin, are modelled monthly using a statistical analysis technique. Normal distribution, logarithmic normal distribution, Weibull distribution, and uniform distribution of the samples are checked by the Shapiro-Wilk and Kolmogorov-Smirnov methods. Subsequently, the probability distribution function, the expected population value and variance, and the confidence interval are estimated. Corrosion data for twenty-two kinds of carbon steels, low alloy steels, and stainless steels, including A3, 3C, 16Mn, 10CrMoAl, 1Cr18Ni9Ti, in atmospheric and marine test stations are analyzed. The main environmental factors for atmospheric and marine corrosion of each type of steel are determined by grey relational analysis. Subsequently, a predictive model incorporating the main corrosion environmental factors, exposure time, and corrosion rate is built by the BP artificial neural network. After the evolution of corrosion pit depth is predicted by the artificial neural network, a fracture mechanics calculation is used to evaluate the residual life of a structure with corrosion pit defects.