### Finite Element Approximation and Numerical Analysis of Three-dimensional Electrical Impedance Tomography

•  Yirang Yuan
•  Jiuping Li
•  Changfeng Li
•  Tongjun Sun

#### Abstract

Electrical impedance tomography is solved by solving an inverse problem of elliptic equation, and a new numerical method or a new technique is argued to consider finite element (such as normal element and mixed element) in this paper on three dimensional region. Introducing different perturbations to boundary restrictions and using different spacial steps, the authors obtain numerical solutions and give comparison with exact solutions. Numerical data show that numerical solution can approximate exact solution well as spacial step taken small and the approximation of Neumann boundary condition is more stable than that of Dirichlet case.
For Newton iterations on finite element method, a large-scaled system of massive linear equations is solved in each iteration, thus the computation is quite expensive. So two techniques are argued in the first half of this paper. Firstly, the invariance property of quasi-element stiffness matrix is used in the  iterations and a type of special current model is introduced. Then the minimum number of direct problems solved is considered. Later a local conservative numerical approximation, low order mixed element (block-centered method) is presented in the latter part and the positive semi-definiteness and the existence of its solution are proved. Computational formula of error functional Jacobi matrix is derived and the least direct problems in each iteration are solved by using the symmetry of algorithm and a special current basis. This method has been applied successfully in actual numerical simulation of three-dimensional electrical impedance tomography.

• Full Text: PDF
• DOI:10.5539/jmr.v8n4p99 This work is licensed under a Creative Commons Attribution 4.0 License.
• ISSN(Print): 1916-9795
• ISSN(Online): 1916-9809
• Started: 2009
• Frequency: bimonthly

### Journal Metrics

• h-index (December 2021): 22
• i10-index (December 2021): 78
• h5-index (December 2021): N/A
• h5-median (December 2021): N/A