Keywords: 
• low alloy steels /
• seawater corrosion /
• dual-rate data /
• comprehensive index value /
• relevance vector regression

Abstract: With the rapid development of Internet of Things technology, the use of front-end sensors makes the corrosion potential online detection of low alloy steels in marine environment a reality, thereby obtaining oceans of corrosion data. Concerning the problems of data information loss and modeling accuracy reduction caused by traditional mean value method when processing dual-rate corrosion data, a new dual-rate data processing and modeling algorithm combining Comprehensive Index Value (CIV) and Improved Relevance Vector Regression (IRVR) was proposed. Firstly, the CIV was constructed to characterize the comprehensive influence of the input data and Beetle Antennae Search (BAS) algorithm was applied to optimize its parameters. Then, linear regression models between the best CIV sequence and the output data were established to convert the dual-rate corrosion data into single-rate data for modeling, which retained more information of the original corrosion data. Finally, IRVR method based on BAS optimization of compounding kernels was given to establish the prediction model for dual-rate seawater corrosion data of low alloy steels. The results show that, compared to mean value method, the proposed model CIV-IRVR increases the number of modeling samples from 196 to 1834. Besides, compared with the commonly used methods like Artificial Neural Network (ANN) and Support Vector Regression (SVR), the Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Coefficient of Determination (CD) of the CIV-IRVR model are 1.1914, 1.5729 and 0.9963, respectively, which outperforms the comparison algorithms. It can be concluded that the proposed model not only reduces the information loss and improves the modeling accuracy but also has practical significance for modeling dual-rate seawater corrosion data.