Abstract: Based on the data of low temperature corrosion of atmospheric and vacuum distillation unit, a model was established for predicting corrosion rate of oil and gas system at top of atmospheric tower by artificial neural network. The corrosion factors of pH value, the concentrations of chloride and iron ion as well as sulfide were used as input data and the average corrosion rates as output data. The experimental results showed that the model has good prediction accuracy with a relative error of 10% and average relative error of 7.5%, indicating that the model can reflect the relationship between corrosion factors and corrosion rate predicted.

Key words: top of atmospheric tower, corrosion, artificial neural network, corrosion rate, prediction