Abstract
The development of a precise model for predicting pipeline corrosion rates is essential for ensuring the safe operation of pipelines. To address the issues of inadequate stability and prolonged execution time associated with traditional models, the KPCA algorithm is used here to reduce the dimensionality of corrosion rate data for subsea pipelines, and the primary factors that influence the corrosion rate are identified. Based on the data characteristics, four algorithms (BP, LSSVM, SVM, and RF) were compared. Ultimately, the LSSVM algorithm was selected as the final prediction model. Then the LSSVM prediction model is subsequently developed, and the NGO algorithm is utilized to optimize the weights and thresholds of the LSSVM model, thereby increasing the accuracy of the prediction model and effectively reducing prediction instability. A combined KPCA-NGO-LSSVM model is developed to predict the corrosion rates of subsea pipelines and is compared with three other models: KPCA-PSO-LSSVM, PSO-LSSVM, and NGO-LSSVM. The mean absolute percentage error (MAPE), root mean square error (RMSE) and determination coefficient (R(2)) of the integrated KPCA-NGO-LSSVM model are 1.791%, 0.06105 and 0.9922, respectively, these metrics are significantly lower than those of benchmark models, a finding consistently validated across multiple experimental datasets. This demonstrates the KPCA-NGO-LSSVM framework's enhanced prediction accuracy and stability. The model demonstrates effective performance in predicting the corrosion rates of subsea pipelines and offers novel methodologies and concepts for future research in this area.