Abstract: Aiming at the problems of poor adaptability and low accuracy of traditional prediction methods for corroded pipelines, a model combining random forest algorithm (RF), evolution of mind (MEA) and elman (RF-MEA-Elman model)was proposed: First, RF was used for pipelines data preprocessing, using MEA to optimize the weight and threshold parameters of the Elman neural network, so as to establish a combined prediction model for the remaining life of the corroded pipeline.Taking a certain pipe section as an example, the simulation training and prediction were carried out with the help of MATLAB.The results show that compared with the other two traditional single models, this model has smaller error and higher prediction accuracy and generalization ability, which provides a useful tool for the study of the remaining life of the pipeline.The new thinking also provides a reference for the risk prevention and maintenance management of the oil and gas transportation system.