Research On Flow Regime Identification Based On Support Vector Machine For Electrical Resistance Tomography System

Two-phase fluid has complex flow characteristic, accurate identification of flow regime is the foundation of measuring two-phase flow's parameter accurately. As a result, the on-line intelligent identification of flow regime is an important role of two-phase flow research. Electrical resistance tomography (ERT) technology is process tomography technology based on electrical resistance sensing mechanism, because it has the advantages of visualization and non-intrusive detection, it has been applied on all kinds of measuring on-line and identification of two-phase flow regime. It is the fastest development process tomography technology in recent years. Support vector machine (SVM) is the latest part of statistical learning theory. It is a kind of pattern recognition method based on structural risk minimization (SRM) principle. It has been applied well on classification, regression estimation and density estimation, etc. It has the more generalization ability than neural network.The research in this paper is based on electrical resistance tomography system and flow regime of oil-water two-phase flow. First, principal component analysis is adopted to extract the feature of the border measurement voltage data of the electrical resistance tomography system, then the extracted feature data is taken as input information of the support vector machine multi-class classifier which is based on one to all strategy, so the four kinds of two-phase flow regime can be identified. The main study content of this paper is listed as follows:1. The research status of common used flow regime identification method presently is summarized. It is pointed out the problem of the flow regime identification. 2. The composition and technical features of electrical resistance tomography is analyzed in theory. It is pointed out the problem which is must be resolved in the development of electrical resistance tomography.3. The necessity of using principal component analysis to extract feature is analyzed. The principle of principal component analysis method which was used in the paper—that is to compress optimally the discrimination information which is included in average vector of class is expounded.4. Aiming at the characteristics of oil-water two-phase flow, and according to basic principle, basic algorithm and common classification method of support vector machine, support vector machine multi-class classifier which is based on one to all strategy has been designed.5. Through MATLAB simulation experiment, the sample's feature has been extracted. And through the comparison of the experiment, the optimal kernel function and related parameters has been selected. The four kinds of oil-water two-phase flow regime can be identified, and the ideal result has been obtained.