Research On The Application Of Particle Filter And Compressed Sensing In ECT

Electrical capacitance tomography (ECT) is a kind of process image technologywith advantages of rapid response, no radiation hazard and low cost, and it is oftenused for parameter detection of multiphase flow. The basic idea of ECT is that thedielectric constant distribution is calculated by the capacitance values between plateelectrodes around the measured material. In ECT system, ECT image reconstructiontechnology reconstruct the image of the dielectric constant distribution of two phaseflow, which provides a new means for on-line detection system and optimizationdesign; ECT flow pattern identification technology identify and project the two-phaseflow pattern to real-time control the transportation equipment of pipeline, which isimportant to prevent the plugging pipe phenomenon and improve the system stabilityand efficiency.On the base of reading mass home and abroad references, thetheory and mathematical model of ECT image reconstruction and flow pattern havebeen analyzed and summarized. And then some improvement ECT imagereconstruction and flow pattern identification algorithms based on the basic ideas ofparticle filter and compressed sensing have been put forward in this paper. The mainworks of this paper are as follows:1. To overcome the shortcoming of easily immerging in partial minimum infunction optimization problem, a new optimization algorithm based on particle filteris put forward. Firstly, the basic principle of particle filter is introduced; Secondly, thebasic frame of converting the optimization process of optimization problem into thestate estimation process of dynamic time-varying system is established, and then theoptimization algorithm based on particle filter is put forward. Finally, this algorithmis applied to solve the actual problems such as nonlinear constrained optimizationproblem, nonlinear equation problem and traveling salesman problem, and thesimulation experiments have shown that this algorithm is effective and of highprecision. 2. To improve image reconstruction quality of ECT, on the basis of particle filteroptimization algorithm, a new image reconstruction algorithm of ECT based onparticle filter is put forward. Firstly, the principle and commonly used algorithms ofimage reconstruction of ECT are analyzed. Secondly, the image reconstructionprocess of ECT system is described as the state estimation process of dynamictime-varying system, and then the image reconstruction algorithm of ECT based onparticle filter is put forward. Finally, the experiment results have shown that theimage error and correlation coefficient are much better than LBP algorithm,IMNSNOF algorithm and Landweber algorithm. So it is a kind of ECT imagereconstruction algorithm with low error and high accuracy.3. The reconstruction results of particle filter algorithm have still a certain gapswith the original images, and it requires a lot of samples which lead to a longeroperation time. In order to overcome this defect, a new ECT image reconstructionmethod based on compressed sensing is put forward. Firstly, the basic principle ofcompressed sensing and its decoding process are analyzed. Secondly, the datecollection process of ECT system is regarded as sampling process of compressedsensing, and the measurement matrix is designed by zero expansion of the sensitivitymatrix and its row vectors random restruction. Finally, the dielectric constantdistribution signal is recovered by signal reconstruction method of compressedsensing. The experiment results have shown that the image error and the correlationcoefficient of reconstruction results obtained by this algorithm are much better thanparticle filter algorithm. So it is a kind of ECT image reconstruction algorithm withlow error and high accuracy.4. In view of lower recognition rate of flow pattern identification of ECT, a flowpattern identification algorithm of ECT based on compressed sensing is put forward.Firstly, the principle and commonly used algorithms of flow pattern identification ofECT are analyzed. Secondly, the normalized measurement capacitance vectorsobtained by ECT system are represented as a sparse linear combination of trainingsample bases, and the measurement matrix is constructed using random Gaussianmatrix to make sampling from the test and standard samples respectively. Finally, thesparse solutions of measured and standard samples are obtained by signal reconstruction method of compressed sensing, and the linear correlation degrees areused for flow pattern identification. The experiment results have shown that it is akind of flow pattern identification algorithm of ECT with high accuracy, and has goodnoise immunity.