Method of Temporal Interpolation of the Corroding Gas Pipeline Wall Thickness Values Coordinated with a Physical Model

The analysis of processes evolving over time plays an increasingly important role in the modern world with the development of computing power. In this paper, the process of corrosive wear of the gas pipeline wall is considered, namely, the problem of regression of the pipe wall thickness value. A new method of temporal interpolation of the values of the wall thickness produced in accordance with the physical parameters of the transported gas condensate is proposed. Experiments on machine learning of regression models using the RANSAC algorithm are carried out, definitions of two metrics of correspondence of the trained models to physical reality are introduced. The experiments results showed that the use of the proposed interpolation method instead of spline interpolation allows for the increase of the first metric value by an average of 2 times and of the second metric value by 3 times.