Computerised interpolation algorithms as well as the empirical model for analysing the flocculent settling data were developed. A mechanistic semi-empirical model developed from fundamental physical principles of a falling particle in a viscous fluid was tested against actual flocculation column data. The accuracy of the mechanistic model was evaluated using the sum of the squared errors between the interpolated values (real values) and the model predictions. Its fitting capabilities were compared with Özer's model using nine flocculent data sets of which four were obtained from literature and the rest was actual data from the performed experiments. The developed model consistently simulated the flocculation behaviour of particles in settling columns better than Özer's model in eight of the nine data sets considered. It is recommended that the model's performance be further compared with other models like the Rule based and San's model. The errors due to the use of interpolated values when determining the performance of the empirical models need to be investigated. Furthermore, a three-way rather than two-way interpolation should now be achievable using the interpolation algorithm developed in this study thereby reducing the effects of interpolation bias. The above work opens the way to full automation of design of flocculation sedimentation basins and other gravitational particle separation systems which at present are designed manually and are susceptible to a wide range of human and random errors. Keywords: sedimentation, flocculation, total suspended solids, iso-percentile profiles, empirical modelling, automated design.
INTRODUCTIONDuring the design of flocculent sedimentation tanks, data from settling column tests is interpreted by a graphical technique. Samples at different times and depths are analysed for suspended solids concentration and removal rate profiles are manually plotted as a function of time. From the graphs, one can predict or calculate the removal efficiency, overflow rate, and settling velocities. The results from the graphs are then used to determine the desired sedimentation tank parameters. The obvious shortcoming of this procedure is that no two individuals can end up with the exact same design and the process is in itself tedious. In addition to the problem of irreproducibility, the graphical method is sometimes misleading in the region of less variability typically at the beginning of the settling process (San, 1989).