Estimation of empirical coefficients of a granular filtration model

In this paper a methodology comprising a non-linear least-squares algorithm for determining the empirical parameters of a filtration model is presented and is tested against literature data. Through the case study it is found that incorporation of a multiplication factor for the sum of the square of errors associated with effluent quality could lead to a more balanced estimation of parameters, while achieving a better fit for both effluent quality and headloss. In the case study it is found that the shape of the fitted effluent quality curve has significant impact on the values of the model parameters. The proposed algorithm appears to be well suited for obtaining the best fit for both effluent quality and headloss, simultaneously. The value of the multiplication factor should be fixed based on the accuracy required under both effluent quality and headloss predictions. The model parameters predicted under each of the case studies, correlated well with the experimental observations.