ABSTRACT The conglomerate formation in China's Mahu oilfield is characterized by complex lithology, gravel development, and strong non‐homogeneity, resulting in difficult bit selection, low mechanical drilling speed, long drilling cycle, and difficulty in evaluating its formation drillability by conventional methods. In this paper, the relationship models of gravel volume content, median particle size, uniaxial compressive strength, and conglomerate drillability value of PDC bit are discussed; a conglomerate drillability logging evaluation model and a depth neural network prediction model are established based on logging data and depth learning method. The results show that the drillability value of PDC bit conglomerate is linearly related to the volume content of conglomerate, logarithmically related to the median size of conglomerate, and power function related to the uniaxial compressive strength of conglomerate, and the prediction accuracy of the model ranges from 80.8% to 96.8%, with an average accuracy of 91.2%. The deep learning method can predict the drillability of PDC bits in conglomerate formations more accurately and provide a basis for bit selection and drilling parameter optimization in conglomerate formations, thus improving drilling efficiency in conglomerate formations.
Xia et al. (Mon,) studied this question.