Nuclear Science and Engineering / Volume 176 / Number 2 / February 2014 / Pages 240-253
Technical Paper / dx.doi.org/10.13182/NSE12-26
Iterative algorithms for computerized tomography reconstruction employ a variety of grids, interpolation techniques, and solution procedures. A new projection-intersection (PI) grid is presented in this work. It comprises all the intersection points between the projection rays passing through the object. A few advantages include (a) a user-independent discretization process and (b) a reduction in reconstruction error caused by nonparticipating nodes. Computerized tomography reconstruction results by PI are compared with existing conventional grids. The multiplicative algebraic reconstruction technique (MART) and entropy maximization are used as solution techniques. We note that for simulated data, the PI grid gives better results when compared with the square-pixel grid. Two different sets of experimental data (obtained previously for a mercury-nitrogen flow loop and one with a known specimen with a static known profile) are processed with the above-mentioned options. A basic theoretical model (but experimentally correlated) is also used to verify the void reference level. Computerized tomography results for experimental projection data indicate a trend similar to the previous MART results, but a major difference is visible in the void-fraction distributions. This fact is important, as heat transfer coefficients are strongly dependent on the distribution of voids.