A novel PIG and an intelligent pigging scheme based on Deep-learning technology

Abstract

In a previous paper, a novel PIG is reported, and in this paper, more elements are given and a Deep-learning technology-based pigging scheme is proposed. The PIG mainly consists of a speed-control unit and a wax-removal unit which are connected by a couple of Hook joints. It differentiates itself from the previous proposal by outstanding features including measuring arms, turbine generator, etc. The realization of jetting, cutting, and bypass valve-based motion control mechanism is elaborated. To overcome the disadvantages of traditional wax-removal methods and improve the processing efficiency, a new intelligent wax-removal algorithm based on Convolution Neural Network is proposed to realize a real-time identification of different wax layer thickness. In addition, the training results of the Convolution Neural Network method are compared with some other traditional machine-learning algorithms to prove its superiority, and it lays a foundation for the practical pigging application in the future.