Aiming at the problem that corrosion is easy to occur

in oil pipelines

a genetic algorithm（ GA）optimized back propagation neural network（ BPNN） is used to predict the corrosion rate of pipelines in oil pipelines. The specific optimization process is given. Using GA to optimize the starting weight and threshold of BPNN model

it effectively avoids the problem that a single BPNN model falls into local optimum

which improves the accuracy of prediction. Taking an oil pipeline as an example

the improved GA-BPNN model is verified and analyzed. The results show that the highest relative error predicted by BPNN model is as high as 24. 49 %

and the average relative error is 11. 13 %. Compared with the BPNN model

the prediction accuracy of the GA-BPNN model has been greatly improved

the maximum relative error is only 8. 16 %

and the average relative error is3. 10 %. Therefore

using GA-BPNN model to predict pipeline corrosion can provide a reliable theoretical basis for pipeline inspection and maintenance.