Void fraction measurement of gas-liquid two-phase flow based on a multi-frequency CCEIT system

This work proposes a novel void fraction measurement method of the gas-liquid two-phase flow based on a multifrequency capacitively coupled electrical impedance tomography (MFCCEIT) system. The proposed method fully leverages the multi-frequency data of the fluid impedance by integrating the statistical and frequency features of the impedance spectroscopy. An innovative feature extraction approach is presented. It combines statistical feature analysis and impedance spectroscopy analysis to develop a hybrid feature vector consisting of both the statistical features and frequency features of the multi-frequency impedance data that are closely correlated with the void fraction. Firstly, data preprocessing steps including electrode pair grouping and frequency range segmentation are employed. Subsequently, the statistical and frequency characteristics of the multi-frequency impedance data are investigated. Correspondingly, a hybrid feature vector consisting of the average standard deviations of the impedance across multiple frequencies and the impedance spectroscopy fitting parameters is established for each flow pattern. Comparison study is conducted to seek the best fit data mining algorithm for the hybrid features, and the original void fraction measurement method combining the proposed feature extraction approach and the Gaussian Process Regression (GPR) algorithm is obtained to develop accurate void fraction measurement models. Experimental results show the effectiveness and potential of the proposed method. Under bubble flow, annular flow and stratified flow, the maximum absolute errors of the void fraction measurement method are 1.79 %, 1.91 % and 1.61 %, respectively, and the root mean squared errors (RMSEs) are 0.0046, 0.0064 and 0.0053, respectively. The research results also show that the proposed method outperforms the traditional single-frequency methods, highlighting the contributions of the multi-frequency measurement acquisition and the feature extraction approach in fully mining and utilizing the fluid impedance information.