Model and algorithm for predicting technological tool sticking by wellbore depth based on the four-module neural network

Sticking the technological tools is considered one of the most capital-intensive types of accidents in oil and gas well drilling, which have a significant impact on the cost of the well. The prediction of stuck tools at the design stage and in the process of drilling allows to minimize the risks of its occurrence, and also allows to choose the optimal method of prevention for specific geological and technical conditions. The work is devoted to the model and algorithm of predicting technological tool sticking by the wellbore depth based on a 4-module neural network. In the work presented classification of a group of methods for predicting sticking and the main disadvantages of existing methods. In the paper is presented the method of transformation the input data elements, on the example of textual and categorical data types. Due to this, it is possible to include into the list of input data elements such geological parameters as rock types, which are one of the most important factors influencing the process of sticking tools, and previously not perceived by models. In order to form a list of significant input data elements, the calculation of correlation coefficients between input data elements and target variables is presented.The type and architecture, as well as the hyperparameters of the modular neural network are chosen experimentally. Based on the trained and tested 4-module prediction model, we propose an algorithm for conducting a sticking tool prediction procedure at the design and drilling stage of the well. A specific feature of the proposed method is that the prediction model uses a wide diapason of universal factors as input data elements, such as geological, technological and rheological parameters of drilling mud as well as technical and technological parameters of drilling influencing the process of all types of stuck tools. Model in perspective has the ability to retrain and adapt to new data, which often happens when drilling wells in new fields.