Research on Corrosion Rate Prediction Model for Long-Distance Pipelines Based on Gravimetric-CART Fusion

Xin Wen; Huiwen Chen; Xiaodi Zhang; Xinyi Xu; Jingyi Liu; Jiarun Xie

doi:10.63313/AERpc.9015

Authors

Xin Wen College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author
Huiwen Chen College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author
Xiaodi Zhang College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author
Xinyi Xu College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author
Jingyi Liu College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author
Jiarun Xie College of Petroleum and Natural Gas Engineering, Southwest Petroleum University, Chengdu 610500, China Author

DOI:

https://doi.org/10.63313/AERpc.9015

Keywords:

Corrosion Rate Prediction, CART Algorithm, Gravimetric Method, Machine Learning, Long-Distance Pipelines

Abstract

As a classical method for corrosion rate determination, the gravimetric method provides fundamental data for corrosion analysis by measuring material mass loss. However, its reliance on laboratory environments and inability to perform real-time prediction limit its engineering applications. This paper proposes a corrosion rate prediction model based on the CART (Classification and Regression Tree) algorithm, which integrates historical gravimetric data with multi-source environmental parameters to achieve dynamic corrosion rate prediction under complex working conditions. Experiments show that the CART model significantly outperforms the traditional gravimetric method in prediction accuracy (MAE=0.05 mm/a) and variable interaction analysis capability (MAE=0.12 mm/a). The study reveals how data-driven methods inherit and extend the core value of the traditional gravimetric method, providing an efficient and low-cost solution for corrosion management in long-distance pipelines.

References

[1] Wen Yaxing, Jia Fusheng, Liu Pengcheng, et al. Study on the Probability Distribution Law of Pipeline Internal Corrosion Detection Data and Its Application[J]. Oil-Gas Field Surface En-gineering, 2024, 43(7): 70–76.

[2] Wang Shasha, Ma Shuaijie, Che Kun, et al. Application Status of Machine Learning in Natural Environment Corrosion Assessment and Prediction[J]. Journal of Chinese Society for Cor-rosion and Protection, 2023, 43(3): 441–451.

[3] Xu Qingyan. Calculation Method of Corrosion Rate for Long-Distance Pipelines Based on CART [J]. Scientific and Technological Innovation, 2023, (12): 217-220.

[4] Hu Jun, Hao Lin, Zhang Shaoyang. Application of Machine Learning Algorithms in Internal Corrosion Prediction of Mixed-Transport Pipelines[J]. Total Corrosion Control, 2023, 37(4): 7–12. DOI:10.13726/j.cnki.11-2706/tq.2023.04.007.06.

[5] Wen Yaxing, Han Bin, Wang Feng, et al. A Corrosion Rate Calculation Method for Long-Distance Pipelines Based on CART[J]. Corrosion and Protection, 2022, 43(12): 78–82+108.

[6] Wang Hongying, Jiang Tao, Qin Xiaoxia, et al. Soil Corrosion Coupon Test for the Si-chuan-to-East Gas Pipeline[J]. Oil & Gas Storage and Transportation, 2010, 29(10): 769–771+718.