A generalized model is formulated to predict the time-dependent permeate flux by extending previous models to include the particle transport mechanisms of Brownian diffusion, shear-induced diffusion, inertial lift and concentrated flowing layers. A new model for estimating the capital costs of membrane plants is developed which incorporates individual cost correlations for different categories of manufactured equipment. The effects of particle size, design, and operating variables on permeate flux and treatment costs are investigated numerically. Optimization problems are formulated and solved to investigate (a) optimal membrane design and system operation, (b) optimal backflushing frequency, and (c) optimal selection of hybrid filtration configurations, over variable raw water quality. The combined theory predicts an unfavorable particle size, on the order of 10\sp−1 μm, where net back-transport is at a minimum. This implies minimum permeate fluxes in the size range of 10\sp−2 μm - 10\sp−1 μm, depending on the operating time. These results support experimental observations of minima in back-transport (Chellam and Wiesner, 1996) and permeate flux (Fane, 1984). Inside-out hollow fiber geometry is predicted to be favorable for feed suspensions with small particles and/or low concentrations. The constant pressure mode of operation is predicted to yield higher specific permeate fluxes compared to the constant flux mode, particularly for particles which demonstrate mass-transfer limited behavior. Comparisons and parameter estimations made with available experimental data on polydisperse suspensions give solidosity estimates ranging from 0.70 to 0.77. Membrane design is predicted to be optimized at values of fiber radius (narrow) and length (short) where the permeate fluxes are maximized. Particles affected by mass-transport limitations demonstrate comparatively lower optimal transmembrane pressures. For unfavorable particles, treatment costs are predicted to be minimized at intermediate recoveries and backflushing frequencies. At small capacities, the hybrid hollow fiber ultrafiltration and spiral wound nanofiltration system with higher non-membrane capital costs is predicted to be largely non-optimal compared to hollow fiber nanofiltration. Membrane costs are expected to play a significant role in determining the optimal configuration at large capacities, where the hybrid configuration is predicted to become largely optimal.