Microfiltration is an important unit operation in downstream processing. However, due to the influence of membrane fouling, prediction of the filtration performance for biological suspensions is difficult. This paper describes a modeling approach that allows a comprehensive description of filtration performance. On the basis of experimental data and linguistic information, a specific artificial neural network was developed that predicts the process behavior within a certain range of parameters. This approach allows us to analyze influences of fermentation on filtration. By using extensive simulations, the interactions of 17 parameters were examined and the fouling causes determined. The model was developed for cell harvesting of Escherichia coli through a shear-enhanced module. The method can be applied to any cross-flow filtration process.