In research of the low temperature parts of atmospheric pressure device, by using BP neural network, the connection of PH value, Cl-, H2S and Fe+2 was setup which can predict Fe+2 content accurately, and obtain the requirement accuracy, hence more accurate corrosion can be predicted and providing more suggests for corrosion protection.

The relation model established by artificial neural network that includes 4 corrosive material parameters that are Cl-, H2S, NH3 and pH value and corrosion testing parameter that is Fe2+ of the fractionator overhead recycle system, studied the corrosion sensitivity of the corrosion parameters, has got the sensitive areas, put forward the suitable range of the corrosive material parameters of corrosion control in the production process.