Automatic detection of formations using images of oil well drilling cuttings

Abstract

Abstract In oil well drilling process, a perennial issue is formations detection particularly in passing through high and low pressure formations. However, automatic classification of keybeds in the Gachsaran and Asmari formations by applying drill cutting images can help in decision-making, especially in oil wells of Iran, about mud weight and casing design for oil well drilling process. First, this study focuses on color analysis and fuzzy c-mean clustering to extract relevant features from images of the drill cuttings. Furthermore, a support vector machine and different kernel functions are utilized to classify the samples into different keybeds. Second, due to changing color of drilling cutting in each well, this study proposes texture analysis for keybeds classification. In this method, a co-occurrence matrix and features of energy, homogeneity, entropy and brightness are applied as feature vectors and classification is done by using the support vector machine too. This study, moreover, introduces the accuracy and response speed of the above techniques. To sum up, the results show that this method can be used to detect different formations (particularly between Gachsaran and Asmari) by approximately 95% accuracy.