Study On First-principles Calculation Of Micro-Galvanic Effect Of Compounds And Prediction Of Corrosion Behavior Of Aluminum Alloy

This paper focuses on the micro-galvanic effects of the compunds and matrix at the initial stage of atmospheric corrosion and its influence mechanism on the corrosion behavior of aluminum alloy.14 typical intermetallic compounds(IMCs)are selected for the calculations and experiments,and the methods of modeling and prediction using machine learning for corrosion data assisted by first-principles calculation based on density functional theory(DFT)are explored.The atomic models of IMCs and aluminum matrix are built to study the work functions,Volta potentials,adsorption(O2,H2O)and permeation(Cl-)behavior by the first-principles calculation.The micro-galvanic corrosion between IMCs and aluminum matrix,the effect of IMCs on the properties of oxide film and electrochemical characteristic are studied by scanning Kelvin probe force microscopy(SKPFM)and electrochemical atomic force microscopy(ECAFM)in this paper.The random forest(RF)algorithm is used to establish the prediction model of atmospheric corrosion rate of aluminum alloy.Combining the calculated Volta potential differences(VPDs)data between the IMCs and aluminum matrix,the research for prediction of corrosion rate is carried out,and the outdoor exposure test data in atmospheric environment of Southeast Asian is used to verify the generalization ability and accuracy of the model.The result shows that:The theoretical Volta potential differences calculated by first-principles calculation between IMCs and aluminum matrix is consistent with the experimental Volta potential difference.It can be seen that first-principles calculation can be used to predict the tendency of galvanic corrosion of IMCs in aluminum alloy.The micro-galvanic effect of typical intermetallic compounds from weak to strong is,the anode phase:Mg2Si<Al2CuMg<Al3Zr<Al3Mg2<Al3Sc<MgZn2.The cathodic phase:Al3Ti、Al6Mn<Al7Cu2Fe<Al2Cu<Al3Fe<Al23Fe4Cu<Al13Cr4Si4、Al12Fe3Si.In elastic and plastic deformation stages,the strain of AA7075-T6 aluminum alloy is mainly concentrated within the Mg2Si particles and at the aluminum matrix around the Al23Fe4Cu particles.The local strain affects the micro-galvanic effects of IMCs and aluminum matrix.The experimental and theoretical results show that the Volta potential differences between IMCs and aluminum matrix decrease under elastic deformation stage,while the Volta potential differences between IMCs and aluminum matrix increase significantly after yielding.Therefore,the micro-galvanic interaction between IMCs and aluminum matrix in the elastic deformation stage is weakened and the pitting resistance is improved,while the micro-galvanic interaction between IMCs and aluminum matrix in the plastic deformation stage is enhanced under plastic deformation and the pitting corrosion resistance is reduced.The oxide film on Al2CuMg surface is loose,thin and defected compared to the oxide film covered on aluminum surface.It’s found that O2 and H2O prefer to dissociate and absorb on the aluminum surface rather than the surface of Al2CuMg by first-principles calculation.Thus,the thickness of the oxide film on the aluminum surface is larger than that on the Al2CuMg surface.Moreover,the Cu-rich layer at the Al/Al2O3 interface promotes the formation of defects and decreases the energy barrier of Cl-penetrating from surface to Al/Al2O3 interface,which causes the rupture of the oxide film.It can be seen that the protective performance of the oxide film covered on the Al2CuMg surface is worse than the oxide film on the Al surface,so the localized corrosion preferentially occurs on the Al2CuMg particles.The random forest method was used to establish the prediction model of outdoor atmospheric corrosion rate of aluminum alloy.Combined with first-principles calculation,the Volta potential differences data expand the dimension of machine learning features.Through importance analysis,it is found that the Volta potential differences between the IMCs and aluminum matrix is the key characteristic variable of the random forest.The cross-validation results show that the random forest algorithm combined with first-principles calculation can effectively improve the goodness of fit(R2)and prediction accuracy of the models.The outdoor exposure experiments are carried out on 5083,7N01 and 6N01 aluminum alloys in Singapore,Jakarta and Bangkok.It is found that the average error between the experimental corrosion rates and the predicted corrosion rates is less than 15%,which verifies the generalization ability of the prediction model for the atmospheric corrosion rate of aluminum alloy.The goodness of fit of the prediction model reaches 0.85,which achieves the prediction and verification of the atmospheric corrosion rate of aluminum alloy in Southeast Asia.