Flow Pattern Identification Based On Neural Network In Electrical Resistance Tomography System

Process tomography technique which is from medical CT technique is mainly used to the research of multiphase flow regime, and also can be used to real-time detection in multiphase flow parameter measurement. With process tomography technique, multiphase flow two-dimensional or three-dimensional space-time partial and microscopic distribution information can be obtained in multiphase flow parameter measurement, which provides an effective approach in solving the difficulty of multiphase flow parameter measurement. Process tomography technique has made a great progress in the last ten years. According to the way of acquiring information and principle of sensor, process tomography technique can be divided into more than ten kinds such as electrical capacitance tomography, electrical resistance tomography, electromagnetic tomography and so on. The subject investigated in this dissertation is electrical resistance tomography system, which is on the basis of resistance sensor principle. First, fuzzy clustering is used to fuzzy the measurement voltage data of the ERT system, then the fuzzy data is taken as input information for the BP network, the fuzzy data of measure voltage are trained repeatedly in BP network, so the four kinds of two-phase flow regime can be identified.The important meaning of multiphase flow measurement to the science research and industry production is summarized, at the same time main parameter, current status and developing trend of multiphase flow measurement technique are analyzed. In addition, the structure and technique characteristic of electrical resistance tomography system is clarified. The current status and broad industrial application prospect of electrical resistance tomography are discussed. The key problems that must be solved in the development of electrical resistance tomography are also pointed out. It is emphasized that the soft field characteristic and the low resolution of image reconstruction are the main problems that result in the difficulty of its practical application. Compared with the other flow regime identification methods, fuzzy neural network method is proposed for regime identification, and the necessity of which is discussed. The model of fuzzy neural network is built, and it is taken as the base of analizing the number of flow regime identification. The experiments show that fuzzy neural network model is feasible, which provides the basis for related image reconstruction algorithms. The four kinds of two-phase flow regime is researched, and the four kinds of two-phase flow regime is experimented by fuzzy neural network model. Finally the simulation software of flow pattern identification based on neural network in ERT system is developed, whose interface is simple and utility. It is easier to flow pattern. Also, the software promotes the other following research to study.