Ultrasonic flowmetering with reflected pulses

Abstract

A transit time type ultrasonic flowmeter was tested with two different reflected pulse trajectories in flowing air at ambient conditions against an orifice meter. The flowmeter was designed to be highly accurate, to require minimal excavation for installation (both transducers to be placed on the upper surface of the pipe), and to require no service shutdown for installation or calibration. The two trajectories were two successive tilted diameters with a single reflection, and three successive tilted midradius chords with two reflections. High frequency (100 kHz) narrowband pulses were used. Both ultrasonic flowmetering configurations were tested in 12 inch pipe in fully developed turbulent flow, and in the abnormal flow downstream of a 90 degree elbow. The velocity range was 5.5 fps - 25 fps. The triple midradius chord configuration performed extremely well, with maximum errors of 1.3, and 2.0 percent of reading, in the normal and abnormal flows, respectively. The double tilted diameter configuration gave maximum errors of 7.2, and 9.3 percent of reading in the normal and abnormal flows, respectively. Recommendations for field testing of the two ultrasonic configurations are made.

A numerical simulation of ultrasonic flowmetering in an abnormal flow using single, double, and triple midradius chords, and a double tilted diameter was conducted prior to the experimental tests. The simulation showed that the triple midradius chord and double tilted diameter were, respectively, the most accurate and second most accurate of the four trajectories.

An amplitude difference between the acoustic signals received at the upstream and downstream transducers, in flowing air, was measured. This amplitude difference is believed to be caused by flow effects. A two-dimensional model was developed to explain the amplitude difference in terms of focusing of the downstream ultrasonic beam and defocusing of the upstream beam, due to velocity gradients. The focusing and defocusing predicted by the model was found to be too small to explain the amplitude difference, however.