Uncertainty propagation in a model of dead-end bacterial microfiltration using fuzzy interval analysis
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2017 Elsevier B.V. Uncertainty is inherent in experimentation, modeling, and analysis. Variations and errors in parameter estimates or physical processes are unavoidable and can affect the reliability of model predictions. Therefore understanding the role of uncertainty is embedded in the process of modeling and approximating the real world. In this manuscript we consider uncertainty propagation in a theoretical model of water/wastewater treatment. In dead-end microfiltration contaminated water is fed through a membrane that filters out colloids, bacteria, and protozoa. However, these particles foul the membrane reducing the filter productivity, which is alleviated by periodically reversing the flow, i.e. backwashing. We investigate how uncertainty in sensitive parameter estimates propagates to the estimates of the optimal amount of volume of water that is filtered in a fixed time period and the associated backwashing timing and duration. We find that the model provides conservative estimates for the total volume since the uncertainty is not propagated symmetrically with respect to over and underestimating specific measurable quantities. The uncertainty in the timing is more symmetric implying that there is essentially an equal amount of uncertainty for increasing or decreasing the frequency and duration of backwashing. We identified biofilm production as propagating the most uncertainty in the volume estimate. The fouling rate has the most effect on the timing estimates. Additionally we explored the affect of asymmetric parameter distributions and find that, for most parameters, asymmetry does not lead to increased asymmetry in predicted optimal regimes, implying that uncertainty in the skewness is likely not an issue.
