Estimating the composition of multicomponent particle slurries using focused beam reflectance measurement
Kernick, Tristan James
MetadataShow full item record
This thesis uses focused beam reflectance measurement (FBRM) in conjunction with a linear empirical model to measure the solid concentration in a nuclear waste simulant in situ. Nuclear waste represents a significant environmental hazard at the Hanford site in Washington State, and the research in this thesis works towards developing a means of monitoring the solid particles present in nuclear waste. The waste simulant, which consisted of six individual components mixed with water, was placed in a stirred vessel and monitored with FBRM at a range of concentrations. The assumptions inherent to the linear model, which incorporates experimentally obtained FBRM histograms from each of the simulant components, were assessed, and the model was applied to two- and three-component mixtures before being used to estimate the composition of the complete simulant. The linear model was capable of estimating the composition of the two-component system containing glass beads and tungsten shavings, and it was able to track changes in composition over time for this simplified system. For the other four components, the model yielded less accurate results. All three large components (alumina, silica, and glass beads from a larger size range) produced much fewer FBRM counts per unit mass of simulant than the smallest three components, making them difficult to detect when the small components were included in the monitored mixture. Silicon carbide, the smallest component in the waste simulant, saturated the FBRM probe at the concentration specified in the simulant, thereby impeding detection of the other five components. Ultimately, the model presented in this thesis can produce accurate composition estimates if the materials being used behave linearly under FBRM observation. For a simple two-component system with particles that backscatter a large percentage of incident light, the linear model has yielded accurate composition estimates with computation times on the order of a few seconds. Materials that generate few chord counts or materials with nonlinear chord count behavior require more advanced models and signal processing to predict composition from their chord length histograms. For the full six-component simulant, the model used in this thesis lacked sufficient complexity to describe the behavior of the system. Further work is necessary to enhance the model’s ability to interpret data produced by multicomponent mixtures.