A general approach, called robust optimal design, is developed in a companion paper and applied here to a comprehensive, nonlinear, activated sludge treatment system. The treatment system robustness is related to the sensitivity of steady-state effluent biochemical oxygen demand and total suspended solids concentrations to changes in 55 uncertain model parameter values. Treatment system designs with lower sensitivity of the effluent water quality measures are likely to be more robust than designs with higher sensitivity. The robust optimal design approach is used to generate alternative designs that represent the trade-off between total system cost and robustness. The robust design trends are generally more consistent with recommended design practice than are typical minimum-cost design trends. In particular, improvements in the robustness of effluent biochemical oxygen demand concentration are associated with higher activated sludge mean solids-residence times, whereas typical minimum-cost designs are associated with low solids-residence times. The robustness of the alternative designs is evaluated further using a nonlinear worst-case analysis. The robust optimal design framework should be applicable to many design problems where uncertainty in parameter values is significant.
Read full abstract