Wellbore failure is the most headache and common problem in drilling engineering. Wireline logging like caliper log and image log is usually using to identify wellbore failure. However, it is impossible to discern the failure types and take steps in real-time using the post-drilling methods. Various cavings will circulate to the ground with drilling fluid when wellbore instability. These cavings are the foremost indicator of wellbore deterioration and have a potential link to the mechanism of borehole failure. Unfortunately, the current approaches for classification and mechanism of wellbore instability using cavings are inclined to qualitative description and fragmentary. In this paper, we propose a new method for quantitative characterization of caving and a caving interpretation model based on image processing and machine learning. Firstly, caving contour is recognized from the image automatically and be quantitatively characterized by 3 features. Then classification is performed in the hyperdimensional space composed of these features. Finally, the mechanism of borehole failure that generates this caving can be inferred based on the classification results. We demonstrate this model using various classification algorithms including LogisticRegression, DecisionTree, RandomForest, GBDT and Bayes in 11wells in the K and T oilfield. The results show that the average accuracies of these algorithms are 82.1%, 79.4%, 82.8%, 83.5%, and 77.2% respectively, which reflects that the model has an excellent performance in caving interpretation as well as that the features have high representativeness and stability for caving characterization. This method opens up a new path to early warning of borehole instability and diagnosis of instability mechanism, thus corresponding remedies can be taken in time to avoid serious downhole accidents.
The problem of borehole instability often occurs and leads to the frequent occurrence of complex situations like diameter shrinkage, sticking, borehole collapse during drilling, which reduces the drilling efficiency greatly. Although considerable efforts have been made from the communities of drilling, engineering geology and geo-mechanics, many oil wells are still troubled by the problem of wellbore instability. More critically, much uncertainty still exists concerning exactly where, when, and why the instability occurred in most cases (Edwards, 2003).