Objective To improve the accuracy of internal corrosion rate predictions for natural gas gathering and transportation pipelines in marine environments, evaluate their residual strength, develop anti-corrosion measures, and ensure their safe operation, a corrosion rate prediction model based on Kernel Principal Component Analysis (KPCA), Improved Hunt-Prey Optimizer (IHPO), and Kernel Extreme Learning Machine (KELM) was proposed.

Methods Using the internal corrosion data from a natural gas gathering and transportation pipeline in the South China Sea, KPCA was first employed to extract features from corrosion-influencing factors, eliminate the impact of redundant data on the prediction results, and determine the input variables. Second, population initialization was performed using Circle mapping. Cauchy mutation was applied to enhance the local development capability of the Hunter-Prey Optimization (HPO) algorithm, while opposition-based learning was utilized to improve its global search capability. The Improved Hunt-Prey Optimizer (IHPO) was then employed to optimize the regularization coefficient (C) and the kernel function parameter (γ) of the KELM. Finally, corrosion rate predictions were made using Matlab, and the results of the KPCA-IHPO-KELM model were compared with those of the KELM, KPCA-KELM, and KPCA-HPO-KELM models.

Results The results indicated that numerous initial influencing factors were present in the case. A total of three principal components were extracted using the KPCA algorithm, which effectively eliminated the influence of redundant data while preserving the main features of the original dataset, thus reducing prediction error. The optimal regularization coefficient C and kernel function parameter γ for the KELM model, determined by IHPO, were 3.83 and 0.01, respectively, at which the model achieved the best prediction performance. After feature extraction and algorithm improvement, the predictions of the KPCA-IHPO-KELM model closely aligned with the actual corrosion rate, demonstrating superior performance. Its root mean square error, mean absolute error, and determination coefficient were 0.0245, 0.0204, and 0.9976, respectively. Compared to the other three models, it achieved the highest prediction accuracy and the lowest mean error.

Conclusion The proposed KPCA-IHPO-KELM combined prediction model for corrosion rate exhibits excellent prediction performance. It offers a new approach for predicting the internal corrosion rate of natural gas gathering and transportation pipelines in marine environments, providing valuable insights for their operation and maintenance management, and risk early-warning systems.