Application of time domain curve-fitting to parameter estimation in RTD models

Abstract

Parameters in residence time distribution (RTD) models are commonly estimated by least-squares curve-fit in the time domain. Two procedures suggested in the literature for this purpose involve (i) fixing τ (the mean residence time) as equal to μ (first moment of the experimental response) prior to curve-fitting and (ii) estimating τ along with other model parameters by curve-fitting. In this work these two procedures, i.e., curve-fitting with and without the constraint τ = μ, are compared. Four models in conjunction with eight sets of RTD data are considered. The results indicate that consistently better results are obtained when the constraint τ = μ is relaxed prior to curve-fitting. This is particularly so for models with fewer parameters like the dispersion model.