Intelligent Stuck Pipe Risk Prediction Based on Dynamic Condition Recognition and Multi-Source Data Fusion

Sticking is one of the critical factors affecting drilling efficiency and safety. To effectively prevent sticking incidents and improve the accuracy and timeliness of risk prediction, this study proposes a multi-source data fusion-based early warning method driven by dynamic drilling condition recognition. By establishing a novel early warning framework that integrates drilling condition recognition, multi-parameter synergy analysis, and hybrid window-based decision-making, this method enhances the accuracy and efficiency of stuck-pipe risk prediction. It achieves real-time identification of drilling conditions using a dynamic domain adaptation neural network model. It constructs a stuck-pipe risk quantification model through multi-parameter coupling analysis. Additionally, it applies a hybrid sliding window strategy and state continuity detection to enable continuous risk assessment and early warning, ensuring stable and adaptive performance under varying operational conditions. The framework improves the precision of sticking risk detection, enhances model interpretability, and provides a robust technical foundation for safe and intelligent drilling operations. Experimental results show that the proposed method can accurately identify various drilling conditions and effectively monitor abnormal states in axial movement, rotational friction, and the circulation system during drilling operations. It is capable of detecting potential stuck-pipe risks up to 15 minutes in advance and issuing tiered early warnings. The prediction accuracy reaches 93.3%. The algorithm demonstrates high prediction capability and real-time responsiveness, significantly reducing the likelihood of sticking incidents, minimizing non-productive time and economic losses, and offering decision-making support for field operations, thereby enhancing the safety of drilling activities.