This thesis has two objectives. The first is to investigate the relative importance of the installation related setup parameter errors on uncertainties of a fully defined reference configuration of conventional clamp-on ultrasonic flow meters (UFMs). The second objective is to improve the overall accuracy of the UFMs by using a different ultrasonic transmission method to address the major contributors found in the first objective.

The first objective relates to carrying out uncertainty analysis on conventional UFMs. There is very little work published reporting on the relative importance between these installation parameters and this is addressed in this thesis. First, a reference configuration of a conventional direct-path clamp-on UFM is presented. The upstream and downstream signals of this reference UFM were simulated using 2D finite element analysis including some simplifying assumptions to simulate the effect of flow. The uncertainties were quantified by comparing simulations that contain deliberately induced parameter errors in relation to the reference configuration. The uncertainty analysis revealed that the pipe roughness was the biggest contributing factor, approximately 2% for a moderately corroded pipe (RMS 0.2mm, correlation length 5mm). This is followed by the effect of wedge angle (>1%), pipe thickness (1%), wedge properties(1%) and pipe properties (<1%). To verify the simulation results, experiments were carried out by studying the effect of horizontal separation distance between transducers and internal pipe wall roughness on flow measurement error. The experimental result for separation distance matches well with the simulated result. The experimentally measured roughness related uncertainties also have the same magnitude levels as the simulation.

The second objective is to address the important contributors found in the uncertainty analysis, internal pipe wall roughness. To reduce the roughness related uncertainties, leaky Lamb wave that uses longer wavelength and lower frequency (200kHz, 5 times lower than conventional UFMs) than the conventional UFMs was chosen. The ultrasonic signals of this leaky Lamb wave UFM were simulated using 2-D finite element (FE) analysis by unwrapping the pipe and only considering the cross section as a plate. The simulation results show that the uncertainties related to pipe wall roughness of leaky Lamb wave UFMs is approximately half of that of conventional UFMs for corroded pipe walls with RMS values larger than 0.1 mm (0.2, 0.35 and 0.5 mm). This method also improves the physical limit of the velocity sensitivity (60% improvement) and interrogates the average velocity of the whole volume of the fluid that passes the inspection area. Initial experiments were carried out. The experiments verified that a pure A0 mode can be transmitted using an EMAT transducer. The experiments also successfully detected the leaky A0 mode and prove the validity of using leaky Lamb wave in flow metering.