The mathematical relationship between corrosion degree and time is referred to as a corrosion model. Existing corrosion models can only be used to predict the corrosion wastage of a certain material based on its available historical corrosion data, but the corrosion wastage of newer steel grades cannot be obtained if the data are not available. To solve this problem, two advanced algorithms, i.e., generalized regression neural network (GRNN) and optimizing grey model (OGM (1, N)), are introduced, based on which corrosion models can be obtained for steel classes even in the absence of historical corrosion data, as long as the chemical compositions of the material are known. First, the theoretical basis and operational procedures of GRNN and OGM (1, N) are introduced. Grey relational analysis of corrosion wastage influencing factors is subsequently conducted. Last, the time-dependent atmospheric corrosion wastages of Q345 and Q460 steels, two typical structural steel grades but their corrosion models have not been well established, are predicted based on their chemical compositions by these two advanced algorithms. The results show that the main chemical compositions that influence the atmospheric corrosion wastage of steels are C and S. Both GRNN and OGM (1, N) can accurately predict the corrosion wastage of the steels, and the predicted results can be fitted by quadratic function or power function, where the goodness of fit is greater than 0.95, which indicates a high fitting accuracy.
Advanced Algorithms to Predict Time-Dependent Atmospheric Corrosion Wastage of Low-Alloy and High-Strength Steels Based on Chemical Compositions
Yuelin Zhang, Ruyan Zheng; Advanced Algorithms to Predict Time-Dependent Atmospheric Corrosion Wastage of Low-Alloy and High-Strength Steels Based on Chemical Compositions. CORROSION 1 October 2023; 79 (10): 1122–1134. doi: https://doi.org/10.5006/4363
Download citation file: