材料期刊网

运用人工神经网络研究苯胺及其衍生物在酸性介质中的缓蚀性能与分子结构参数的关系。研究对象包括苯胺在邻、间、对位的氟、氯、溴以及甲基取代物。建立了可用于预测苯胺类衍生物缓蚀效率的预测模型，并对苯胺邻、间、对位的碘、氰、羟基、甲氧基以及胺基取代物的缓蚀效率进行预测。探讨了结构参数对缓蚀剂的影响规律。

The approach of neural networks was made on its applications to the correlation between molecular electronic structrue and corrosion inhibition properties of aniline and its halogenated and methylated derivatives with substituents at different positions.It was found by reference [2] that there exists the relation between the inhibitive efficiency of the above compounds in acid media for mild steel and their electronic structures;and a tentative interpretation was given for the phenomenon that when the inhibitiors are substituted with the same kind of substituents at different positions in the molecule,or with different substituents at same position,they exhibit different inhibitive effciency.The mechanism of corrosion inhibition was also wxplained from the point of view of electron transfer and the information provided in that paper could be probably useful for developing new more effective inhibitors.The present authors believe that the molecules of the inhibitors were possibly adsorbed on the surface of iron electrode in horizontal state.The inputs of neural networks included the molecular structure parameters,such as hydrophobic index (π,π-),molar refraction (MR),steric parameter (Es),electronic index (σ,σ*),as well as net charge and πcharge of N atom (qN),and low-unoccupied molecular orbital(LUMO),which could be calculated by means of semi-empirical methods of quantum chemisty,including HMO and CNDO/2 methods.The neural network's outputs were the corrosion-inhibiting efficiencies of aniline and its halogenated and methylated derivatives for mild steel in acid media.The corresponding relationship and modelling prediction were established by using neural networks with 8-15-1 topological structures trained by a modified backpropagation (MBP) algorithm,respectively.The learning precision was high and the predicting performance was excellent.The corrosion inhibiting properties or electrode parameters of some new aniline's derivatives were predicted quantitatively and the results were quite good.