An Experimental Study of Different Signal Processing Methods on Ultrasonic Velocity Profiles in a Single Phase Flow

Abstract

Ultrasonic velocity profile (UVP) measurement methods have been continuously developed in the field of engineering. A UVP can visualize a fluid flow along a benchmark line. This provides a significant advantage over other conventional methods such as differential pressure, turbine, and vortex. This paper presents an experimental study of using different signal processing methods including autocorrelation (AC), fast Fourier transform (FFT), maximum likelihood estimation (MLE), multiple signal classification (MUSIC), and Estimation of signal parameter via rotational invariance technique (ESPRIT) under diverse situations as the number of pulse repetitions (N_prf), frequency of repetitions (f_prf), velocity profiles, computation – time requirements and flowrates. Experimental results express that there is an optimal number and frequency of pulse repetitions for each signal processing method that depended on f_prf, N_prf, and flowrate. Moreover, computation-time and statistical tests were verified from experimental results. From the comparisons, MLE was experimentally the best algorithm even though the trade-off of moderate computation-time requirements was realized. However, considering the optimization of both accuracy and computation-time consumption, MLE was determined as the preferred signal processing method based on UVP for estimating flowrate in existing water reactors.