Intelligent Prediction of the Internal Corrosion Rate in Submarine Oil and Gas Pipelines Based on KPCA-ISSA-MKSVR Combined Strategy

Given the increasingly prominent problem of pipeline internal corrosion leakage in long-term service submarine oil and gas pipelines, this paper proposes a KPCA-ISSA-MKSVR combined prediction model for accurate internal corrosion rate prediction. The model uses kernel principal component analysis (KPCA) to reduce the dimensionality of influencing factors, improving computational efficiency. An improved sparrow search algorithm (ISSA) optimizes model parameters, employing a tent chaotic map and periodic nonlinear adaptive convergence factor to enhance exploration and search capabilities. Polynomial mutation disturbance is incorporated to prevent local optima traps. By leveraging mixed kernel support vector regression (MKSVR), which integrates multiple kernel functions, the model successfully captures the diversity and complexity of the corrosion data, solving the kernel function selection issue faced by traditional SVR models. In Example 1, the model reduced the mean absolute error (MAE) to 0.015, the root mean square error (RMSE) to 0.018, and the mean absolute percentage error (MAPE) to 0.546, and achieved an R2 of 0.991. In Example 2, the MAE decreased to 0.013, the RMSE to 0.017, the MAPE to 0.493, and R2 reached 0.994. The KPCA-ISSA-MKSVR strategy demonstrates superior accuracy in predicting internal corrosion rates, providing strong technical support for submarine oil and gas pipeline maintenance and safety management.