Abstract: Artificial neural network (ANN), which simulates frame, the character and the cognizing ability of nerve element of cerebrum, is a new system that can handle many kinds of information and semaphore. By use of the obtained-by-test internal corrosion velocity values caused by H2S, CO2 and corrosion-control agent concentration in gas phase or liquid phase, the predicting model of artificial neural network to inter- nal corrosion velocity in gas pipeline with Levenberg-Marquardt algorithm is established in the article. The predicting model of neural network with the data samples of relationship between H2S, CO2 and corrosion-control agent and the internal corrosion velocity is trained and applied to forecast the internal corrosion velocity in gas pipeline, the results is content.