Materials Performance

AUG 2018

Materials Performance is the world's most widely circulated magazine dedicated to corrosion prevention and control. MP provides information about the latest corrosion control technologies and practical applications for every industry and environment.

Issue link: https://mp.epubxp.com/i/1007306

Contents of this Issue

Navigation

Page 60 of 92

58 AUGUST 2018 W W W.MATERIALSPERFORMANCE.COM CM CORROSION MANAGEMENT T The main objective of this study was to develop a forecasting model of car- bon steel corrosion rates in atmo- spheric exposures. The corrosion rates in the industrial and coastal zones were all higher than the corro- sion rates in the rural zone. The corro- sion rates of samples in the industrial zone all approached 200 μm/y. For industrial exposures, the artificial neu- ral network (ANN) and principal com- ponent analysis (PCA)-ANN models showed good forecasting capabili- ties. The multilayer perceptron (MLP)- ANN and PCA-ANN models were suitable for rural environments. Taiwan is an island surrounded by seas on its four sides. In recent years, an evalua- tion of the atmospheric corrosion of carbon steel (CS) in Taiwan reported that its usage was mainly observed in the coastal areas, and most of the data trends showed that the corrosion rate gradually decreased with increasing distance from the seashore. In this study, the research test stations were located in different environments— industrial, coastal, and rural—as shown in Figure 1. Atmospheric corrosion is a very complicated nonlinear problem; therefore, an accurate forecasting model usually can- not be set up through regression analysis. However, the artificial neural network (ANN) approach has better nonlinear pro- cessing capability. Some studies have used an ANN to predict metal corrosion. For example, Kenny 1 and Jan ˇ cíková 2 used an ANN to predict the corrosion rate of low- carbon steel (CS), copper, and aluminum. Their results showed that an ANN can be used to evaluate local atmospheric corro- sion rates. The neural network method was used to set up a forecasting model for metallic corrosion, and it is anticipated that the forecasting results can improve the accu- racy of the environmental classification of atmospheric corrosion, and can be used as a reference by engineering personnel in the management, planning, designing, and maintenance stages. Consequently, engi- neering, safety, and the service life of public structures can be enhanced. Research Method Artificial Neural Network An ANN is a highly connected array of elementary processors called neurons. The multilayered perceptron (MLP)-ANN is the most widely used model. The root mean square of error (RMSE) and determination coefficient (R 2 ) were employed to evaluate the prediction performance of the ANN models. Multiple Linear Regression Model Structure Multiple linear regression (MLR) was used to determine the correlation formula among metallic corrosion rate, pollutant, and meteorological factors. IBM's SPSS † software was used for the model construc- tion. 3 If the correlation among simulation variables is too high, the parameters of the regression model cannot be fully estimated. In this study, the setup model had to pass Atmospheric Corrosion Prediction of Carbon Steel C.M. l o, Department of Environmental Engineering, National Chung Hsing University, Taiwan, Republic of China; and Harbor & Marine Technology Center, Institute of Transportation, Ministry of Transportation and Communications, Taiwan, Republic of China l . h . t sai, Harbor & Marine Technology Center, Institute of Transportation, Ministry of Transportation and Communications, Taiwan, Republic of China h . h . l ai and W. t . Wu, Department of Materials Science and Engineering, National Chung Hsing University, Taiwan, Republic of China M. d . l in, Department of Environmental Engineering, National Chung Hsing University, Taiwan, Republic of China † Trade name.

Articles in this issue

Links on this page

Archives of this issue

view archives of Materials Performance - AUG 2018