Real-Time Recognition of Drill Bit Wear State Using Clustering Algorithm and GRU Neural Network

Drill bits are a critical factor influencing drilling efficiency. Due to the complexity of downhole conditions, it is challenging to achieve real-time and quantitative assessment of drill bit wear during the drilling process. Traditional approaches often rely on rate of penetration to assess downhole drill bit wear, which has inherent limitations and insufficient accuracy. To address the issues, this study presents a novel method for real-time identification of drill bit wear based on rock-breaking efficiency and mechanical specific energy, utilizing clustering algorithms and GRU neural networks. Initially, the working conditions of the drill bit, including bit sticking, vibration, bit bouncing, and bit balling, are annotated. Subsequently, an unsupervised clustering approach is employed to segment the drill bit wear process based on logging data, and the GRU neural network is used to establish a mapping relationship between surface drilling parameters and drill bit wear. This model was applied in Tianjin, China, where it monitored 18 drill bit trips. Among these, there were 3 instances with deviations, leading to an overall accuracy rate of 91.3%.