In the oil and gas industry, it is crucial to employ appropriate drilling fluids to maintain equilibrium of formation pressure throughout the various stages of drilling operations. During the recycling process, the drilling fluid may precipitate gas and as a result, exhibit non-full pipe flow upon return to the surface. Accurate measurement of the volume flow rate of it is imperative in obtaining valuable information from the bottom of the well. Commonly, on-site drilling operations use a multiphase target flowmeter in conjunction with an empirical model to rectify calculation results. Therefore, the theoretical potential of utilizing non-contact ultrasonic sensors for measuring multiphase volume flow rate of non-full pipe flow is significant. In this research, an apparent flow velocity calculation model was established by integrating the ultrasonic Doppler shift model and pipeline fluid mechanics utilizing a four-channel ultrasonic array. Subsequently, the invariant scattering convolution -long short-term memory network was trained on the data-fused ultrasonic signal to identify the liquid level. The velocity-area method was also employed to establish a new multiphase volume flow calculation model. To evaluate the validity of the proposed model, comparison experiments of liquid single-phase flow and liquid-solid two-phase flow were conducted. The experimental results show that, compared with the comparative flow measurement system (CFMS), the accuracy of the ultrasonic flow measurement system (UFMS) is reduced by 0.965%, the nonlinear error (NLE) by 2.293 %, the average relative error (ARE) by 2.570 %, the standard deviation (SD) by 1.395, and the root mean square error (RMSE) by 14.394.

No abstract