Corrosion rate prediction for long-distance submarine pipelines based on MWIWOA-SVM

ObjectiveTo ensure the safe operation of long-distance submarine oil and gas pipelines, it is essential to enhance the prediction accuracy of their internal corrosion rates. Most existing models rely on Support Vector Machine (SVM), which has limitations including low convergence accuracy, unbalanced optimization, and a tendency to get stuck in local optima. MethodsTo address these issues, Multi-Way Improved Whale Optimization Algorithm (MWIWOA) was proposed to optimize the SVM-based prediction model for the internal corrosion rate of long-distance submarine pipelines. The population initialization utilized Tent chaotic mapping in conjunction with an opposition-based learning mechanism. Global optimization and local search functions were balanced through the introduction of adaptive weights and nonlinear convergence factors. The extended search mode was refined by integrating the simplex method, while the optimization capability of the Whale Optimization Algorithm (WOA) was enhanced with Levy flight to improve step length. Consequently, the improved WOA facilitated the optimization of kernel function parameters and penalty factors in the SVM-based model, thereby increasing the rigor of parameter selection. ResultsUsing the corrosion data from the SP74-FPSO pipeline segment as a case study, various algorithm model improvement strategies were applied to construct internal corrosion rate prediction models for long-distance submarine pipelines, including MWIWOA-SVM, WOA-SVM, PSO-SVM, and SVM. These models were trained to predict internal corrosion rates and subsequently compared against one another. The internal corrosion rate prediction model based on MWIWOA-SVM for long-distance submarine pipelines achieved an average absolute percentage error and root mean square error of less than 2％, indicating a very low error level. Both the determination coefficient and fitting degree exceeded 98％, while the relative error between the predicted and actual internal corrosion rates did not exceed 0.99％. These performance metrics significantly outperformed other prediction models, demonstrating higher prediction accuracy. ConclusionThe successful implementation of the MWIWOA enhances prediction accuracy and outperforms comparative models, demonstrating the feasibility of the algorithm improvement. It effectively addresses issues such as low convergence accuracy, tendency to get stuck in local optima, and unbalanced computational power. The experimental results indicate that the MWIWOA-SVM-based prediction model for the internal corrosion rate of long-distance submarine pipelines demonstrates better prediction performance and effectiveness. This model can serve as a valuable reference for future research on risk assessment and maintenance recommendations for submarine pipelines.