Unsupervised learning of well drilling operations: Fuzzy rule-based approach

The paper proposes a method for operation identification in the context of well drilling. This task is usually trusted to domain experts, however, we introduce a fuzzy rule-based classifier for the automatic detection of ongoing operations at a drilling site. The operations of the drilling rig are usually monitored and sensory data is stored. The proposed classifier is identified from real data from an already drilled well via unsupervised learning. The results of our experiment are encouraging, since we manage to separate a number of distinct drilling operations and the classifier is transparent to interpretation thus its decisions are understandable to domain experts.