A kick monitoring method for deepwater open-circuit drilling based on convolutional neural network

Abstract

In the case of shallow drilling in deepwater without installation of riser and subsea BOP, open-circuit drilling is usually used. In the process of open-circuit drilling, drilling fluid is directly returned to the seabed. Therefore, observing the return of seabed drilling fluid through Remotely Operated Vehicle (ROV) has become one of the main methods for monitoring the kick in open-circuit drilling. However, the current ROV monitoring method only relies on the professional technical personnel to observe the video screen, which is difficult to identify the kick quickly and accurately in the early stage. In view of this problem, an intelligent kick monitoring method based on deep learning image recognition technology is proposed in this paper. By collecting the actual video image data shot by ROV and simulating the drilling fluid return morphology under different gas influx using Ansys Fluent software, a typical sample database is constructed. Then, the pretrained GoogLeNet neural network model is used for transfer learning, and the video images shot by ROV on site are used to test the trained model. The results show that this method can identify the kick more quickly and accurately, and the model has good robustness. Using this method can greatly reduce the economic losses in the process of deepwater open-circuit drilling, and provides technical support for safe and efficient drilling of deepwater open-circuit drilling.