A comprehensive prediction framework for offshore downhole collapse pressure based on machine learning and multi-attribute decision analysis: Insights from the East China Sea,

Huayang Li, Quanyou Liu, Jiaao Chen, Dantong Liu, Zehui Shi, Fuzhi Chen,
A comprehensive prediction framework for offshore downhole collapse pressure based on machine learning and multi-attribute decision analysis: Insights from the East China Sea,
Engineering Analysis with Boundary Elements,
Volume 180,
2025,
106510,
ISSN 0955-7997,
https://doi.org/10.1016/j.enganabound.2025.106510.
(https://www.sciencedirect.com/science/article/pii/S0955799725003972)
Abstract: Wellbore collapse during offshore drilling poses serious safety and economic risks. Using the Xihu Sag (East China Sea) as a case study, this paper proposes an integrated framework for predicting offshore downhole collapse pressure. Specifically, grey relational analysis is used to select pore pressure, formation density, interval transit time, and borehole diameter as input layer variables. Ten machine learning models were then developed to predict collapse pressure, and an objective multi-attribute decision method (CRITIC–TOPSIS) was used to identify the most robust performer. The LightGBM model achieved the best results, with a relative closeness to the ideal solution of 0.9081 and prediction errors within −1.35% to 1.35%. The predicted collapse pressure distribution closely matches field observations of wellbore instability. Finally, this study clarifies the wellbore collapse mechanisms under different lithological conditions and puts forward corresponding drilling optimization measures. The proposed framework streamlines the prediction process, enhances drilling safety, and provides a transferable solution for collapse pressure prediction in complex offshore environments.
Keywords: Collapse pressure; Prediction method; Wellbore stability; Machine learning; Drilling safety; TOPSIS