(70b) Generic Simulant Design Using Statistical Methods and Development

Savannah River National Laboratory is investigating generic "base" simulants to blend for upcoming sludge batches of waste product at the Defense Waste Processing Facility (DWPF). A statistical approach has been implemented using two data clustering algorithms: K Means and Normal Mixtures. These data clustering algorithms were applied to a parameter space formed from related past batches processed at the DWPF. Statistical methodology was used to identify the number of general simulants needed to make this task viable. Afterwards the simulants were computationally evaluated against projections to determine goodness of fit and model adequacy. The evaluation was initiated using restricted regression analysis that determined the proportions of simulants needed to match various test cases. The results were corroborated by matching Python and SAS output. Moving forward, concentration will be on optimizing and expanding the design space and model confirmation using regression of the new simulants to target expanded case projections.