Corrosion rate prediction model of carbon steel in regional soil based on BP artificial neural network

A short-term prediction model for soil corrosion of carbon steel in the regional soil environment of Daqing area was established by measuring the physical and chemical properties of soil in this area, the short-term corrosion data of carbon steel and analyzing the logical relationship among mass transfer processes. The reasonableness of the corrosion model was verified by using BP artificial neural network to learn, train, simulate and compare to the corrosion test results of buried carbon steel samples. The results show that water content, air content, pH, Cl^- content, SO₄^2- content and total dissolved salts are the six key factors on soil corrosion of carbon steel in the local soil environment. It is indicated that a stable forecasting model with good generalization ability can be built based on BP artificial neural network through Matlab platform software, by continuous accumulation of soil corrosion information and after adequate training. The model predicts the corrosion rates of carbon steel in Daqing soil accurately.