Improvement of Rate of Penetration in Drilling Process Based on TCN-Vibration Recognition

Rate of penetration (ROP) optimization is critical for improving efficiency in the drilling process. Yet most optimization strategies overlook the effect of drill-string vibration, a significant ROP inhibitor. Moreover, effective methods for recognizing vibration using only surface drilling data are scarce. To address these issues, an intelligent ROP optimization strategy taking vibration mitigation into account is proposed. First, a stacked temporal convolutional network (TCN) is developed to extract the temporal features in multisensor drilling data for vibration recognition. Then, a modified ROP optimization model is developed, integrating a process evaluation index to assess vibration severity. Based on the recognition and evaluation result, an adaptive strategy is devised to constrain the optimization space of operational parameters for mitigating excessive vibration. Finally, sliding time window technique and hybrid bat algorithm (HBA) are applied to optimize the operational parameters in real-time. Experiments on industrial data from an actual drilling field demonstrate the efficiency of the proposed strategy. The vibration model achieves recognition accuracy over 90% and surpasses existing methods. Furthermore, the recommended operational parameters effectively mitigate severe vibrations induced by high weight on bit and low rotational speed, leading to a 27% improvement in ROP compared to manual operation.