ECT Flow Pattern Recognition Of Two-Phase Flow Based On Feature Extraction

Two-phase flow systems are widely used in petroleum,chemical,aerospace and other fields.Flow pattern is the most basic characteristic parameter of two-phase flow.Accurate measurement of other parameters of two-phase flow usually depends on the understanding of the convection pattern.Therefore,it is of great significance to accurately identify the two-phase flow pattern.Electrical capacitance tomography(ECT)is a modality of electrical tomography(ET).It has the characteristics of safety,no radiation,non-contact,simple structure,low cost,and good real-time.This technology determines the medium distribution by measuring the dielectric constant distribution,and is mainly used for real-time imaging and flow pattern identification of the two-phase flow inside the pipeline.The use of ECT system to realize two-phase flow pattern recognition has received more and more attention.This paper focuses on the research of flow pattern identification based on ECT.The main work and achievements can be summarized as follows.1.Researched the automatic modeling method of two-phase flow pattern based on COMSOL WITH MATLAB.Most flow pattern identification methods require a large-scale data set to obtain higher identification accuracy and better model generalization performance.Aiming at the practical problems of low modeling efficiency,strong subjectivity of flow patterns,and difficulty in obtaining large-scale data sets,a automatic flow pattern modeling method based on COMSOL WITH MATLAB was proposed,which realized 8 typical two-phase flow patterns The automatic modeling design laid a foundation for the identification of flow patterns.2.Two normalized capacitance value preprocessing methods based on equal width sub-bins and equal frequency sub-bins are proposed.On this basis,an ECT flow pattern identification method based on data binning and support vector machine(SVM)is proposed.The basic idea is to perform data binning preprocessing on the normalized capacitance value obtained by the ECT system,replace the data in the box with the box number,and reduce redundant information that is not related to classification;Then input to the SVM classifier optimized by the improved particle swarm optimization algorithm to realize the classification of 8 typical flow patterns.The simulation and experimental results show that the proposed data preprocessing method can effectively improve the sensitivity of flowpattern features,reduce redundant information not related to classification,and improve the accuracy of flow pattern identification.Compared with the equal width binning method,the equal frequency binning method is more resistant to noise.3.The test conditions of different data preprocessing methods and different flow pattern identification classifiers are compared.Using different data preprocessing inputs in the literature and the data binning-box number smoothing preprocessing method of this paper,they are input to different flow type identification classifiers.The results show that after the data preprocessing of this paper,different classifiers have achieved good classification The results verify the effectiveness of the data binning-box number smoothing data preprocessing.Among them,the equal frequency binning-box number smoothing preprocessing method has the best accuracy rate of 99%,and the equal-width binning-box number smoothing preprocessing The accuracy rate of the method reached 98%.This study provides a new way for ECT flow pattern identification.