Data - Driven Prediction of Corrosion Rate of 3C Steel in Marine Environment

Abstract

For predicting the corrosion rate of 3C steel quickly and accurately,a support vector regression (RBF-SVR) model based on Gaussian kernel function and an ensemble model based on three models were developed using the data obtained from literatures.The RBF-SVR model was established after dimension reduction of the data set realized respectively through the SVR-based genetic algorithm,forward algorithm and backward algorithm,and after optimizing super-parameter using leave-one-out cross validation method and grid search.Finally,the obtained model and two other models from published literatures (GA-BPNN model and BPNN model with four layers) were integrated to form an ensemble model.The research results indicated that the ensemble model had higher prediction accuracy and generalization ability.

Keywords

• 3c steel|corrosion rate|support vector regression|prediction model|marine environment