Stuck Pipe Prediction Using Adaptive Nero-Fuzzy Inference System Based on Feature Analysis

In the process of underground drilling in coal mine, there are some cases of drilling resistance but not stuck pipe. The method of data training model to predict stuck drilling will often lead to a high false alarm rate and reduce the prediction accuracy. Therefore, this paper proposes an improved adaptive nero-fuzzy inference system prediction method based on feature analysis (ANFIS-FA). The rotational speed with the strongest correlation was selected as the feature analysis object through principal component analysis, and then the sliding window was used to extract the amplitude change to construct the penalty factor, so as to improve the prediction results of ANFIS. In this paper, the actual drilling data are used for experiments. The results show that ANFIS-FA reduces the false alarm rate and improves the prediction accuracy, which has certain guiding significance for practical drilling.