Enhancement of the accuracy of ultrasonic flowmeters by applying the PCA algorithm in predicting flow patterns

Abstract

In the current study, a novel arrangement of acoustic paths with an on-off technique upon ultrasonic waves is proposed to enhance the accuracy and reduce the frequency of calibration of the ultrasonic flowmeter (USFM). In addition, a web of chord probes on the velocity profile with a novel rotating method is also proposed. Twenty velocity profiles, including 19 selected Salami asymmetric velocity profiles and a centrosymmetric developed velocity profile, are combined with the proposed chord-probe mesh to build a novel velocity database. The built database includes 2241 900 sets of 12 mean velocities, ranging from 0.1 m s−1 to 25 m s−1. Furthermore, the principal component analysis algorithm is employed and trained by the developed database to predict the flow patterns in a practically flowing measurement. The predictive ability of the developed algorithm is examined in 10 different scenarios. In all cases in this study, it is found that the proposed USFM can not only measure the volumetric flow rate of the asymmetric flows with extremely high accuracy, but can also eliminate the influences of flow regimes and the effects of the installed location.