FEATURE EXTRACTION OF INDICATOR CARD DATA USING MACHINE LEARNING

Abstract

In this article, three feature extraction techniques for sucker-rod pump indication card data based on Fourier Descriptors (FD), Geometric Moment Vector (GMV), and Grey Level Matrix Statistics (GLMX) have been investigated, simulated, and compared. Due to the non-optimal amount of Fourier Descriptors used in the technique, the Fourier Descriptors algorithm may result in information loss in numerical tests. The Geometric Moment. While the Grey Level Matrix Statistics approach produces low-dimension feature vectors with greater time consumption and memory space, it also consumes more time and resources. Additionally, the Fourier Descriptors approach and the Geometric Moment Vector algorithm's rotational invariance property may lead to incorrect pattern identification of indicator card data when utilised for sucker-rod pump operating condition diagnostics