| Popis: |
In corrosion monitoring, the prediction of material constitutive and environment relationship can improve the optimization design process of pipeline. Recently, the artificial neural network (ANN) models are considered as a powerful tool to describe the electrochemical corrosion behavior of materials. Based on the experimental data from the potentiodynamic polarization of high-strength low-alloy (HSLA) steels, an ANN was trained with standard back-propagation learning algorithm to predict the passivation current density and potential of microalloyed steels. The inputs of the model were chemical compositions comprising of “carbon”, “magnesium”, “niobium”, “titanium”, “nitrogen”, “molybdenum”, “nickel”, “aluminum”, “copper”, “chromium”, “vanadium” and “carbon equivalent” (weight percent) and microstructure consist of diffusion (ferrite/pearlite) and shear (bainite/martensite) transformations, corrosion cell characteristics such as “reference electrode”, “scan rate”, “temperature”, relative pressure of oxygen”, “pressure of purged CO 2 ”, “chloride ion” and “bicarbonate concentration” whereas “passivation current density” and “passivation potential” were the outputs. According to the predicted and experimental results, it was indicated that the developed model showed a good capacity of modeling complex corrosion behavior and could accurately tracks the experimental data in a wide steel chemical compositions, microstructures, temperature ranges and corrosion cell characteristics. Scatter plots and statistical criteria of “absolute fraction of variance ( R 2 )”, and “mean relative error (MRE)” were used to evaluate the prediction performance and universality of the developed models. Based on analyses, the proposed models could be further used in practical applications and corrosion monitoring of microalloyed steels. |
Full Text from ScienceDirect