While drilling operators always want to move as fast as possible, drill bit failure will inevitably lead to increased costs and added delays. When such a failure occurred for Hawthorne, the team at Nsight Analytics took on the task of diagnosing the issue. We began by examining the drilling data, quickly identifying notable spikes in differential pressure and torque at the time of the break. This initial observation was crucial in understanding the underlying factors contributing to the failure.
Further investigation into the drilling data revealed damage to a cutter on the drill bit's shoulder, predominantly on its rear side. This wear pattern suggested the occurrence of “backward whirl,” a destabilizing force causing the bit to rotate counter to the drilling direction. Such motion can lead to unintended contact with the formation, damaging the cutter. We identified “stick slip,” a cyclical event of stalling and sudden rotation in the drill string, as a key factor contributing to this phenomenon.
The analysis further showed that the drill bits were suffering from torsional fatigue. This type of fatigue results from cyclic twisting or rotational loads, leading to the development of microcracks in the material. Over time, these stresses can cause significant damage, ultimately leading to the breakage of the MWD tool, as observed in this case.
To prevent such failures in the future, our analysts recommended consistent analysis of the rate of penetration (ROP) against mechanical specific energy (MSE). We employed a rigorous process involving specific data selection, normalization of data using StandardScaler, and elbow analysis to identify the optimal number of clusters for data analysis. This method ensured a focus on relevant conditions and balanced the risk of overfitting or underfitting the data.
The final step involved making the analysis actionable. We used K-means clustering on normalized data to segment it into distinct groups, aiding in the identification of inefficiencies in drilling operations. Visualization techniques, such as scatter plots, were utilized to analyze patterns and relationships within the drilling data. This approach highlighted certain clusters as being particularly inefficient, providing valuable insights for optimizing ROP and reducing unnecessary wear on the drill bit during future operation.