Combined EIS and BAS-BP neural network analysis of electrochemical corrosion on pipeline steel in silty soil in a Salt–Temperature coupling environment

Abstract

Soil corrosion negatively affects long-distance pipeline transportation. Studies on this phenomenon have revealed the potential safety risks associated with metal durability. In previous studies on the soil corrosion of pipeline steel, corrosion behaviour, which is influenced by the salt–temperature coupling environment, was seldom considered. In this study, a laboratory-simulated corrosion environment for electrochemical impedance spectroscopy (EIS) measurements was designed to analyse the corrosion behaviour of X80 steel immersed in silty soil mixed with different concentrations of NaCl and Na2SO4 over a temperature range of 20 to −20 °C. Nyquist plots displayed three capacitive reactance arcs in various soil environments in the range of 100 kHz to 0.1 Hz. Furthermore, an equivalent circuit model was utilised to fit the Nyquist and Bode plots. The effects of the Cl− concentration, SO42− concentration, and temperature on the EIS characteristics were analysed. The change in the temperature considerably affected the degree of corrosion. Cl− reacted with the metal and catalysed the corrosion process, whereas SO42− enhanced the chemical inertness of the metal surface. The mechanism of the metal corrosion reaction was revealed by performing scanning electron microscopy and X-ray photoelectron spectroscopy. Based on the experimental results and beetle antennae search (BAS) algorithm, an optimised back propagation (BP) neural network model was established to predict the Nyquist plot of the soil corrosion process. The predicted Nyquist plots exhibited the same electrochemical characteristics as those shown by the experimental curves. A comparison of the evaluation parameters revealed that the equivalent circuit of the corrosion system can be applied to the prediction results obtained from the BAS-BP neural network model. Moreover, it demonstrated the practicality of the neural network prediction method in the field of corrosion science.