Practical identifiability and uncertainty analysis of the one-dimensional hindered-compression continuous settling model

Abstract

The practical application of the one-dimension hindered-compression settling models remains a challenge, since the model calibration strongly depends on experimental observations with limited information. In this study, the identifiability of parameter subsets of the hindered-compression models is evaluated for various experimental layouts. Global sensitivity analysis is used to preliminarily select the influential parameters which can be reasonably estimated, while the identifiability analysis of parameter subsets is conducted based on the local sensitivity functions and collinearity measures. The batch settling curve observations are informative for calibrating hindered parameters, and to determine the compression parameters, the concentration profile observations may need to be collected. For different experimental layouts, at least three parameters are identifiable, and the number of identifiable parameters can potentially increase to five, if both batch settling curve and concentration observations are available. The parameter subset identifiability is sensitive to the choice of initial parameter values, and determining the initial values of hindered parameters and gel concentration by measuring the hindered settling velocities and the top concentration of the static sediment respectively allows efficient reduction of the sensitivity. Parameter subset estimates are sensitive to the values of fixed parameters, and reliable estimation of identifiable parameter subsets is possible to significantly decrease model prediction uncertainties.