Optimization Of Finite Element Model And Research On Image Reconstruction Algorithms For Electrical Resistance Tomography

As a new non-invasive and visible measurement technology, electrical resistancetomography has important applications in industrial and medical fields for theadvantages of low cost and rapid response, etc, and becomes one well-establishedrelatively mature electrical tomography technique. In order to improve the speed andaccuracy of image reconstruction, finite element model and image reconstructionalgorithms for electrical resistance tomography are deeply studied in this paper:1. An improved genetic algorithm is used to optimize the node numbering inelectrical resistance tomography off-line, which determines the bandwidth of theglobal stiffness matrix, and the bandwidth is effectively reduced under the sameexperimental condition, improving the efficiency of the forward problem.2. By taking the enhancement of the accuracy of solving the forward problemand the improvement of the ill-posedness of the sensitivity matrix with homogeneoussensitivity field as two respective targets, different kinds of improved geneticalgorithms are applied to optimize the topology of the finite element model inelectrical resistance tomography off-line. Experimental results demonstrate that theoptimization measures can improve the accuracy of solving the forward problemeffectively, reduce the condition number of the sensitivity matrix, improving the spaceresolution effectively.3. The decreasing strategy of inertia weight applied in the improved particleswarm optimization is adopted to determine the regularization factor of the modifiedNewton-Raphson algorithm, and an improved Newton-Raphson algorithm is proposedby updating the sensitivity matrix automatically, which can improve the real-timeperformance without reducing precision, on the basis, the resistivity is corrected byusing a threshold, and the space resolution is improved.4. By optimizing the sensitivity matrix and choosing the gain factor off-line, animproved pre-iteration Landweber image reconstruction algorithm is proposed.Experimental results demonstrate that compared to the pre-iteration Landweberalgorithm with choosing the gain factor empirically and the off-line iteration on-linereconstruction algorithm, the novel algorithm can improve the space resolution of thereconstructed images effectively.