Intelligent Identification of Formation Lithology While Drilling

Abstract During drilling engineering, the rate of penetration (ROP) is the basic index to measure the drill ability of various rocks, and drilling parameters are the main control factors that affect the ROP. Identifying formation lithology while drilling can promptly adjust drilling parameters and effectively improve drilling efficiency. In this paper, drilling parameters and intelligent models are combined to realize formation lithology identification while drilling. Different from previous research, this method uses K-means, Fuzzy C-means (FCM), and Mean Shift algorithms to cluster the data set after dimension reduction, uses support vector machine (SVM), random forests (RF), and extremely randomized trees (ET) algorithms to train multiple models to identify formation lithology according to the clustering results, and analyzes the identification accuracy of the models under different combinations of dimension reduction and clustering methods. The results indicate that: 1) without clustering, the accuracy of the model in identifying formation lithology is poor, and the final identification result is biased towards one or two types of lithology; 2) the linear kernel function is the best among the three kernel functions, and the classification results lead to high accuracy in identifying lithology. The classification results of the Gaussian kernel function and polynomial kernel function are biased towards a certain kind of lithology; 3) in the multi-model of linear kernel function + K-Means + SVM, the identification accuracies of sandstone, mudstone, limestone, and shale are 60%, 80%, 70%, and 90% respectively, with an average identification accuracy of 82.5%. This paper puts forward a method for identifying formation lithology while drilling by combining drilling parameters and intelligent models, and takes the YX block as an example to carry out application test. The results show that this method can effectively improve the accuracy of lithology identification while drilling.