Internal Corrosion Risk Prediction of Submarine Pipeline Based on Kernel Principal Component Analysis

The kernel principal component analysis (KPCA) was used to optimize the algorithms such as gaussian kernel support vector machine (SVM-rbf), poly kernel support vector machine (SVM-poly), linear kernel support vector machine (SVM-linear), artificial neural network (ANN) and random forest (RFR). The maximum corrosion rate of 20 submarine pipelines was predicted by grid search method. The root mean square error (RMSE), mean absolute error (MAE) and squared correlation coefficient (R²) were used as evaluation indexes to evaluate the prediction effect of the model before and after optimization. The results showed that the R² value of each model was significantly increased after optimization, and the maximum value was 0.987 7. KPCA could reduce both the feature dimension and the noise interference and improve the prediction performance of the model. The optimized support vector machine algorithm had high accuracy in predicting the corrosion rate of submarine pipelines, which could provide a reference for early warning and protection of submarine oil and gas pipeline corrosion.