A tubing life prediction method based on multi-factor time sequence coupling and mechanism fusion

By using differentiated preprocessing and mechanistic feature extraction of multi-source time series data, combined with an improved TA-RLSTM model, the accuracy and stability issues of tubing life prediction under complex downhole conditions were solved, achieving high-precision tubing life prediction.

CN122412960APending Publication Date: 2026-07-17CHONGQING UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING UNIVERSITY OF SCIENCE AND TECHNOLOGY
Filing Date
2026-05-25
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing tubing life prediction technologies cannot meet the requirements for high accuracy and high stability under complex downhole conditions. They also cannot effectively integrate multi-source data, dynamic time-series evolution, and engineering mechanisms, resulting in limited prediction accuracy and stability.

Method used

Multi-source time-series data of the entire life cycle of tubing were collected, and after differential preprocessing, multi-source time-series factors were constructed and combined with the time-series coupling matrix. The improved TA-RLSTM model was used to predict the life of tubing. The mechanism characteristics of electrochemical corrosion, fatigue damage and structural strength were integrated. The LSTM model was optimized by time-series attention and residual connection mechanism, and trained by combining the improved AdamW optimizer and mechanism-data dual-driven loss function.

Benefits of technology

It achieves high-precision and stable prediction of tubing life under complex working conditions, improves the accuracy of damage quantification and identification and the reliability of prediction results, and adapts to the dynamic changes of downhole working conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122412960A_ABST
    Figure CN122412960A_ABST
Patent Text Reader

Abstract

本发明公开了一种多因子时序耦合与机理融合的油管寿命预测方法，涉及油气田井下设备状态监测技术领域。本发明采集井下多源时序数据，构建标准化时序数据集；结合腐蚀、疲劳、强度三类机理模型提取量化机理特征因子，基于互信息与动态时间规整算法融合动静关联指标，构建时序耦合矩阵；通过改进型CNN损伤判断模型输出油管腐蚀深度与疲劳损伤参数；结合时序注意力机制、残差连接改进LSTM，构建TA‑RLSTM预测模型，通过改进型AdamW优化器与机理‑数据双驱动融合损失函数进行模型训练；该模型融合损伤参数与耦合矩阵强度特征，实现油管剩余寿命定量预测。本发明结合工程机理约束与数据驱动建模，提升了复杂井下工况下油管寿命预测的准确性与稳定性。
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Coiled tubing fatigue life early warning method based on digital twinning

    CN116415479A

  • Oil and gas pipeline life prediction system based on deep belief network learning

    CN118964976A

  • Method for predicting fatigue damage degree and residual life of coiled tubing

    CN120217599A