| Liquid-liquid dispersed flows are widely used in petrochemical,food processing and other industrial processes.Accurate measurement of phase fraction is important for parameter monitoring,flow modelling and process control.Ultrasonic attenuation is a non-invasive approach for measuring the phase fraction of fluids because of its high penetration capability,no restrictions on the transparency of the medium,and safety from radiation.However,ultrasonic attenuation is influenced by the size and spatial distribution of the dispersed phase,and there is a non-linear relationship between the attenuation rate and the phase fraction.Because the liquid-liquid dispersed flow is a typical nonlinear dynamic system,where the droplet diameter and spatial distribution of dispersed phase are diverse and time-varying,ultrasonic propagation characteristics are complex.Although the multi-frequency ultrasonic attenuation can provide observation information support for the inversion of droplet size characteristics,it is difficult to develop the ultrasonic attenuation model and the inversion of phase fraction.To address the above problems,this work starts from the ultrasonic propagation mechanism in liquid-liquid dispersed flow,constructs the multi-frequency ultrasonic attenuation model based on the theoretical relationship between multi-frequency ultrasonic attenuation and the droplet diameter distribution of dispersed phase,solves the non-linear problem between single-frequency acoustic attenuation and phase fraction;then,the objective function between multi-frequency attenuation and the phase fraction is constructed.The optimization algorithm is used to determine droplet diameter characteristics of liquid-liquid dispersed flow and achieve the inversion of the phase fraction.In this paper,the ultrasonic attenuation model and the phase fraction inversion algorithm are investigated as follows:(1)Aiming at the time variability of liquid-liquid dispersed flow,an ultrasonic swept-frequency excitation strategy is proposed and a demodulation method is designed to ensure the consistency of multi-frequency attenuation measured.The ultrasonic propagation of different excitation modes is modeled by multi-physical field simulation,and the consistency of the results measured by the excitation modes of swept-frequency and single frequency is verified.The absolute error of ultrasonic attenuation at different frequencies is less than 1 Np/m under the two excitation modes,and the relative error of the attenuation obtained from the swept-frequency measurenment is less than 5% compared with the single frequency measurement.(2)Aiming at the diversity of droplet diameter distribution for liquid-liquid dispersed flow,an ultrasonic swept-frequency attenuation method based on droplet diameter distribution is proposed for the fraction measurement of oil-water dispersed flow.According to the ratio of ultrasonic wavelength to droplet diameter and the physical parameters of oil-water two phases,an ultrasonic attenuation model based on scattering attenuation theory and absorption attenuation theory is constructed to explain the relationship between the droplet diameter distribution of dispersed phase and ultrasonic attenuation.An improved ultrasonic attenuation model based on equivalent fluid is proposed by referring to the Core-Shell model of sound energy dissipation to solve the problem of ultrasonic multiple scattering and predict ultrasonic attenuation at high phase fraction.The maximum measurement error is within 3.67% by the hybrid optimization algorithm.(3)Aiming at the influence of the inhomogeneous distribution of dispersed phase on ultrasonic attenuation in pipeline at low flow rate,the ultrasonic attenuation model based on cluster fractal is proposed by taking the inhomogeneous dispersed flow as fractal medium to characterize the influence of the inhomogeneous distribution of dispersed phase on ultrasonic attenuation.The validity of the fractal-modified ultrasonic attenuation model is verified by the experiments of liquid-solid dispersed flow with calibrated particle size distribution,which is used to simulate dispersed phase agglomeration with non-uniform spatial distribution.The experimental results show that the correlation coefficient between measured particle size distribution and calibration results is higher than 98%,and the cumulative error is less than 0.15.(4)Aiming at the computational complexity of the hybrid optimization inversion algorithm,Covariance Matrix Adaptive Evolutionary Strategy(CMA-ES)is introduced to reduce the computational complexity and avoid complex parameter adjustment.The Restart Strategy(RS)and Local Search Strategy(LS)are combined to improve CMA-ES,and the RL-CMA-ES algorithm is proposed to solve the problem that the sample sampling of CMA-ES distribution model leads to the lack of population diversity at the later stage of search.Thus,the search performance of the algorithm is enhanced and the inversion accuracy of phase fraction is improved.The maximum error and average error of the measured phase fraction for oil-water dispersed flow are 2.45% and 0.93% respectively at different flow rates.(5)In order to ensure the real-time performance of the measurement method,a proxy model is introduced based on the proposed RL-CMA-ES algorithm to solve the time-consuming problem of calculating a large number of objective functions in the iterative process.Meanwhile,an ultrasonic attenuation measurement system based on CPCI(Compact PCI)industrial bus standard is designed to meet the requirements of multi-mode compatibility and integration in the field of multi-phase flow parameter measurement.The maximum error of phase fraction is 3.26%,and the confidence probability of phase fraction error less than 2% is 75%. |
