Research and prediction of pipeline inspection gauges velocity based on simulation and neural network

Abstract

Pipeline transportation is a highly efficient transportation mode, and pipeline inspection gauge (PIG) is a widely used pipeline cleaning, detection and maintenance device. The PIG's ability to fulfill its intended purpose hinges heavily on its motion velocity. And it is vital to predict and control the velocity of PIG to ensure optimal functioning. This paper establishes several models with different rotary valve orifice numbers, orifice diameters, valve openings, and sealing disc compression amounts. The effects of the factors on PIG force state and velocity are analyzed by finite element method (FEM). With these simulation outcomes, the paper establishes velocity prediction models based on polynomial fitting, support vector machine (SVM) and neural network methods. In addition, by combining neural networks with genetic algorithms, the PIG size is optimized. The results demonstrate that compared with polynomial fitting and SVM, neural networks are more convenient and accurate and have a broader application prospect.