Research on Corrosion Rate Prediction of Buried Pipeline Based on KPCA-Improved PSO-BP Neural Network Model

Corrosion will damage the structure of buried pipeline and lead to its failure. Corrosion rate varies with different soil properties. In order to predict the corrosion rate of buried pipeline accurately and reliably, KPCA- improved PSO-BPNN prediction model is proposed. In this paper, a buried oil pipeline section in Shaanxi Province is selected for research. Firstly, an 8-dimensional external corrosion index system is constructed. Secondly, in order to improve the fitting effect and remove redundant information, KPCA is introduced for preprocessing to obtain a 6-dimensional index system. Thirdly, BPNN is optimized by PSO algorithm with improved weight parameters, and corrosion rate is predicted through training. Finally, by comparing the prediction accuracy of single BPNN, KPCA-BPNN and KPCA-improved PSO-BPNN model, it is found that the accuracy of KPCA-improved PSO-BPNN model can reach 94.73%, which is 13.26% and 8.04% higher than the previous two models respectively. Therefore, the proposed new model can better meet the actual engineering requirements.