Research On Electrical Capacitance Tomography To Multiphase Pipeline Detection

Multiphase flow is a common flow exists in the industry and nature. With the development of science and technology, people began to study its internal characteristics. The detection of the phase distribution and flow characteristics is of great significance to the industry development and human progress. complex nonlinear system, because of the relative velocity and interface effect, the measurement is difficult because of the complex components. Process tomography is a hot technology in recent years, which has great potential in industrial multi-phase detection. Electrical capacitance tomography is one of the Process Tomography. It has the advantages of non-contact, low cost, simple structure, safety, fast response speed, wide adaptation range and so on, which can be applied in the visual monitoring of multiphase flow. However, in order to be used in the actual industry, further research and exploration is still needed.This paper is for the purpose of multiphase flow visualization monitoring in coking wastewater treatment. By studying the finite element simulation and image reconstruction problems in electrical capacitance tomography system, proving the feasibility of ECT technology. The main works and contributions of this dissertation are as follows:1. Through the literature research, it introduces the composition, characteristics, development and classification of ECT, analyzing the structure and working principle of it, and using the mathematical description for the forward and inverse problem.2. By the means of APDL in large finite element analysis software ANS YS, we realize the simulation of ECT sensor field. By calculating the node potential values, the capacitances can be obtained which reflects the distribution, and the sensitivity is analysed which confirming the existence of "soft field".3. According to the necessity of image reconstruction algorithm, we compare several common image reconstruction algorithms at present. Based on the strong associative ability and nonlinear mapping characteristics of BP neural network, we analyse the feasibility of the application of BP neural network to ECT. Because the BP training process is easy to fall into local minimum. So we improve it by using the genetic algorithm to search the initial weights and thresholds globally. Then the BP neural network is utlized to the local exact search. The simulation results show that this improved algorithm can effectively reduce the training times and improve the reconstruction quality obviously.