A data-driven prediction model for maximum pitting corrosion depth of subsea oil pipelines using SSA-LSTM approach

Abstract

Pitting corrosion is considered to be one of the most dangerous failure forms of offshore steel structures, and corrosion depth is treated as an important indicator of corrosion condition. This paper presents a data-driven model to predict maximum pitting corrosion depth of subsea oil pipelines using the integrated SSA and LSTM approach. LSTM is utilized to learn the relationship between pipeline corrosion depth and its influencing factors. SSA with the strong global search ability and the fast convergence speed is used to optimize hyperparameters of LSTM model to improve its prediction accuracy. A total of 300 samples of maximum pitting corrosion depth of subsea oil pipelines are used to develop the data-driven model. These data are divided into training set and testing set to train and verify the model, respectively. The developed model is compared with LSTM alone and SSA-BP model. The results indicate that SSA-LSTM model performed superior in the prediction accuracy and robustness which evaluation parameters are the smallest values in these models (MAE = 8.84%; RMSE = 0.0607; MSE = 0.36%; MAPE = 9.58%). The developed model can serve as a useful online tool to support the digitalized safety of subsea process systems.