The reliability prediction of corroded pipelines is an important part of ensuring the safe operation of pipeline systems. Aiming at the deficiencies of the existing reliability evaluation methods for corroded pipelines in dealing with the correlation of multiple defects and computational efficiency, a method for predicting the reliability of corroded pipelines based on the model of LSTM(long short-term memory) integrated with the attention mechanism was proposed. First, this method constructed the limit state equation of the pipeline section failure and calculated the failure probability of a single corrosion defect. Second, the Gaussian Copula function was used to characterize the correlation of multiple defects, and a joint failure density function was established. Then, failure samples were generated through Monte Carlo sampling, and the dynamic prediction of the failure probability was realized by the model of LSTM integrated with the attention mechanism. Combined with the pipeline internal inspection data of the China-Russia Eastern Gas Pipeline, the influences of the number of defects and the corrosion growth model on the reliability of the pipeline section were analyzed. The results show that, compared with the traditional sampling algorithm, the average absolute error of the proposed method for predicting the reliability of corroded pipelines is only 0.039, and the calculation time is only 1% of that of the Monte Carlo method. The research achievements provide a new technical idea for the reliability evaluation of corroded oil and gas pipelines with multiple defects.
LI Rong-guang, CHEN Si-xun, SUN Ling.
Reliability Prediction Method for Corroded Pipelines Based on LSTM Model Integrated with Attention Mechanism[J].
Science Technology and Engineering, 2026, 26(2): 589-596 DOI:10.12404/j.issn.1671-1815.2500635
目前相关研究主要聚焦于管道失效的建模方法和求解方法两方面,常采用的独立变量假设的方法对腐蚀管道进行建模,并借助蒙特卡洛模拟(Monte Carlo simulation, MCS)方法进行腐蚀管道可靠性评估[2-4]。上述方法虽然可以初步解决对单一腐蚀缺陷管道的可靠性评价,但并未考虑多个腐蚀缺陷生长的相关性,且基于MCS的抽样效率较低。为解决这一问题,中外学者针对腐蚀管道可靠性问题进行了大量研究工作。李长俊等[5]重点考虑失效模式竞争性和变量相关性,提出了腐蚀管道时变失效概率预测方法,所提方法可适用于极小概率事件的预测。Wen等[6]提出了一种基于ANN的建模方法,通过优化训练样本的输入顺序、神经网络初始权重和偏差进而提高了模型的计算效率,减少了训练集随机行为对计算结果的影响,但该研究仅考虑了单个缺陷。张鹏等[7]研究了4种随机变量相关性对腐蚀管道失效概率的影响,明确了不同变量相关性对不同管道失效模式的影响,但该研究仅考虑了随机变量相关,且没有对计算效率进行深入讨论。Li等[8]开发了一种基于多种故障模式的腐蚀管道可靠性动态评价方法。该方法考虑了失效发生时间对构建失效模式的影响,研究了腐蚀缺陷几何分布对腐蚀管道可靠性的影响等,虽然取得较好的定性结论,但该方法增加了计算时间。王赤宇等[9]提出了基于两层相关性的腐蚀管道失效概率计算方法,该方法改善了采用传统独立假设计算管道失效概率的弊端,但缺乏对所提方法求解效率的分析,且其是否可以应用于多层缺陷相关性还有待验证。
MaGang. Prediction of residual strength of single defect pipeline based on correlation analysis and SVM algorithm[D]. Xi’an: Xi’an Shiyou University, 2020.
[3]
AbyaniM, BahaariM R. A comparative reliability study of corroded pipelines based on Monte Carlo simulation and Latin hypercube sampling methods[J]. International Journal of Pressure Vessels and Piping, 2020, 181.DOI: 10.1016/j.ijpvp.2020.104079.
[4]
SongC, KawaiR. Monte Carlo and variance reduction methods for structural reliability analysis: a comprehensive review[J]. Probabilistic Engineering Mechanics, 2023, 73.DOI: 10.1016/j.probengmech.2023.103479.
[5]
TeixeiraA P, Guedes SoaresC, NettoT A, et al. Reliability of pipelines with corrosion defects[J]. International Journal of Pressure Vessels and Piping, 2008, 85(4): 228-237.
LiChangjun, ZengJianghao, YangFan. Prediction of pipeline failure probability with multiple failure modes and variable correlation[J]. China Safety Science Journal, 2024, 34(5): 36-43.
[8]
WenK, HeL, LiuJ, et al. An optimization of artificial neural network modeling methodology for the reliability assessment of corroding natural gas pipelines[J]. Journal of Loss Prevention in the Process Industries, 2019, 60: 1-8.
ZhangPeng, PengYang. Failure probability of corroded pipeline considering random variables correlation[J]. Acta Petrolei Sinica, 2016, 37(10): 1293-1301.
[11]
LiC, YangF, JiaW, et al. Pipelines reliability assessment conside-ring corrosion-related failure modes and probability distributions characteristic using subset simulation[J]. Process Safety and Environmental Protection, 2023, 178: 226-239.
WangChiyu, YuWeichao, ChenLei, et al. A method for calculating the failure probability of corroded oil and gas pipelines considering the two layers of correlation[J]. Petroleum Science Bulletin, 2024, 9(5): 819-830.
LiuLang, LiuHuoyi, MaChanglong. Effect of corrosion correlation among components on reliability of steel-concrete composite bridge system[J]. Science Technology and Engineering, 2024, 24(10): 4247-4254.
PengLifeng, SuGuoshao, WangZhicheng, et al. Reliability analysis method for geotechnical engineering structure using gaus-sian process classification based monte carlo simulation[J]. Science Technology and Engineering, 2013, 13(21): 6150-6157.
ZhaoShuai, YanXiangzhen, HanLiling. Reliability analysis of natural gas storage roof based on Monte Carlo method[J]. Science Technology and Engineering, 2013, 13(36): 10908-10912.
ZhengXiaoliang, YangXiaoliang, LaiWenhao. Classification and research progress of new energy power generation forecasting methods[J]. Science Technology and Engineering, 2024, 24(34): 14503-14521.
WuDonghui, XuJing, ChenJibin, et al. Human activity recognition algorithm based on CNN-LSTM with attention mechanism[J]. Science Technology and Engineering, 2023, 23(2): 681-689.