Research On Key Technologies Of Electrical Capacitance Tomography System

Electrical capacitance tomography (ECT) technique is a kind of process tomography technique based on the sensitive principle of capacitance. It has been the most popular research direction and the main development technique of process tomography with a wide application prospect in the field of industry, due to its many distinct advantages such as no radiation, no invasion, high speed of response, simple structure, low cost, wide application range, fine safety and so on. This paper explores the profound study focused on problems such as sensitive field distribution of transducers, optimized design of transducers’structure parameter, image reconstruction algorithm, hardware implement of the algorithm based on FPGA and design of the digital system. The study we have completed is as follows.First of all, profound study is executed on the technical characteristics and the system composition of ECT system. The mathematical model of capacitance sensitivity field is presented according to the principle analysis of ECT system. Computer simulation data of different structure parameters is received by using Matlab and ANSYS, the sensitivity matrix is calculated. The affection of structure parameter on sensor performance is studied as well. On this basis, an improved chaos search strategy is proposed, which is introduced into particle swarm optimization algorithm (PSO). An optimal group of transducers structure parameter is obtained which improved sensitivity and evenness of the sensitive field. The sensor performance is improved significantly.Secondly, more profound study is focused on the several kinds of typical reconstruction algorithms. According to the problem of soft field characteristics and poor imaging stability, a new image reconstruction method based on Chebyshev polynomial was proposed. In the algorithm, improved k-means algorithm is firstly adopted on clustering the large scale training sample set, characteristic samples are selected and added into the training sample set which greatly reduced the scales of the training sample. Then we adjust sample input interval by unipolar S function, use Chebyshev polynomial as the activation function of hidden layer and Delta rule is used to adjust weight. The result shows that the algorithm has many advantages, such as simple algorithm, high speed and great imaging stability.To the problem of ill-posed characteristic, a new set of reconstruction algorithm based on neural network types supported algorithm (NSSN) was proposed. Network type supported algorithm was proposed, which was used as the activation function of hidden layer. The output matrix of hidden layer was a strictly diagonally dominant one, so it can enhance anti-interference ability of the network. We make the linear equations which included the unknown weight from hidden layer to output layer keep excellent state by setting the same number of the hidden layer and input samples, and it has unique solution. According to sensitivity distribution, we divided the whole NSSN network into six sub-systems, reduced the scale of network. The solution shows that the improved NSSN not only enhances training speeds, but also improves the quality of imaging, especially on the flow regime identification.Once again, in view of the non-linear non-separating problem, a kind of image reconstruction algorithm based on C-SVM was studied and analysed, which was improved by using the method of prefetching boundary vector, it can reduce the scales of the training sample. A kind of non-stationary biasing hardening algorithm based on C-SVM is proposed. It is implented on FPGA by using serial calculation-parallel transmission mode. The speed of image reconstruction is highly improved.Finally, we design a kind of ECT digital system. On that basis of analyzing the influence on the response time of each part of digital management, in order to improve the ability of real time and long-distance quick transmission, we used FPGA hardware implement mage reconstruction algorithm, task scheduling algorithm, interrupt stack management and network communication control. Meanwhile, a kind of using software call the interfaces of hardware function was given, so we reduced the execution time of system calls and improved the execution efficiency of processor. In addition, ANSYS programming language and C language are used to implement the method and algorithm which was proposed in this paper which has functions of transducer modeling, automatic subdivision, automatic calculates of capacity and sensitivity distribution, structural parameter optimization and image reconstruction.