A novel neutron computed tomography partial volume voxel water quantification technique.

摘要

Since neutrons are attenuated by small amounts of water, but readily penetrate most metals, neutron imaging is ideal for the observation and quantification of water mass in operating hydrogen-powered polymer electrolyte fuel cells (PEFCs). PEFC's have a special need for non-destructive analysis techniques for the imaging of liquid water because the liquid water stored in the porous media can be a source of reduced performance, degradation and cause more lethargic start-up from freezing conditions. Traditional two-dimensional (2D) neutron radioscopy has been successfully applied to the quantification of liquid water in PEFC's, but it suffers from the drawback of making it difficult to determine the exact "depth" within a PEFC the liquid water mass exists. Neutron computed tomography (NCT) produces a three-dimensional (3D) volumetric reconstruction that offers the ability to determine the exact spatial location of a liquid water mass within a PEFC. This makes possible the isolation of liquid water slugs that can block the channels of either the anode or cathode reactant flow fields. Water mass quantification of these slugs would provide insight into improving PEFC design.;Thus, a method was developed for the precise quantification of water mass in neutron computed tomography (NCT) reconstructions. A three-dimensional (3D) volumetric reconstruction is comprised of individual volume elements, or voxels. The gray level value of a voxel represents the total macroscopic cross section, Sigmat, of the material present at the voxel's spatial location. For voxels along interfaces, the gray level represents a combination of Sigmats for the various materials present. The fractional amount of water, also known as a partial volume, represented by such a voxel must be quantified for an accurate result. This calculation requires removing or compensating for the influence of other materials on the voxel's gray level. This is accomplished by background normalizing the raw data used to produce the volumetric reconstruction. The resulting volumetric reconstruction contains voxels that represent only water. Normalizing to the gray level value of a voxel of known water mass produces a matrix of voxels with gray levels that now represent fractional amounts of water. These fractional amounts are tallied and multiplied by the known water mass of the normalizing voxel to determine the total.;The NCT water quantification technique was tested using MCNP simulations of samples containing liquid phase water and ice phase water. Quantification of the MCNP simulations yielded results within 0.2% of the theoretical. For liquid phase and ice phase water samples at ∼30mm from the detector, results were within 2% of the theoretical. The ability to quantify an ice water mixture to within 2% of the theoretical was also demonstrated. For liquid phase water samples at 140mm from the detector, significant error in the quantified water mass, as large as 47%, was observed and determined to be the result of geometric un-sharpness effects and cupping artifacts.;Deconvolution of the imaging system's blurring function was performed to correct for the geometric un-sharpness. Results of the devoncolution showed a reduction in the geometric un-sharpness by ∼14.4% yielding an average increase in quantified water mass of 6.7%. The effects of magnification, cupping artifacts, and geometric unsharpness on the final quantification results were also investigated. Magnification was determined to have no effect while cupping artifacts accounted for 1.4% of the error. Geometric un-sharpness accounted for 45% of the error, making it the dominant source of error.