Study On Identification Of Flow Regimes Based On Neural Networks For Electrical Capacitance Tomography System

The flow regime is one of the important process parameters in the industrialsystem of production two-phase flow, and the measurement of other parametersare often dependent on the flow regime, thus the identification of flow regime isvery important for industrial production. The technology of ElectricalCapacitance Tomography (ECT) is a multiphase flow measurement technology,its working principle is to compute the distribution of the dielectric constant bymeasuring the capacitance value between the object surface electrode, its hasmany advantages such as simple structure, low cost, non intrusive, securityfeatures. Currently, there are still many technical difficulties to break through, soit has important significance and application value to carry out the relevant theoryand technology research.In this thesis, the research object is the12-electrode of ECT system, basedon the analysis of flow regime identification theory of electrical capacitancetomography, the thesis studies problem solving model and flow patternidentification issues in electrical capacitance tomography system. The mainresearch contents in this thesis are as follows:Firstly, based on the research background, the significance of the researchtopic is discussed, and the research status of electrical capacitance tomographyand identification measure of flow regimes are also studied in this thesis.Secondly, this thesis analyses the composition of electrical capacitancetomography system, the capacitance value between the plates is obtained throughestablishing mathematical model of electrical capacitance tomography systemand simulating with Matlab. The images are reconstructed by different reconstruction algorithms, then analyse and compare different reconstructionalgorithms.Thirdly, through the research of Elman neural network, this thesis proposeda identification of flow regime method for electrical capacitance tomographysystem. According to the characteristics of Elman neural network and ECTsystem, the new feature extraction method of capacitance value is proposed. AfterMatlab simulation, the results are compared with other algorithms.Finally, the application of stacked-self coding technique and Softmaxregression technology is to design and realize the deep learning network system.The deep learning network system is used to identify flow regime in the electricalcapacitance tomography, and the validity of the algorithm is verified byexperiment. Compared with the traditional neural network algorithm, deeplearning shows better effect.