Modeling of permeability of membrane bioreactor system using artificial neural network

Modeling for complex systems such as membrane bioreactor due to run virtual tests a lot in a short time is a powerful tool, however, requires experimental validation and conversion the process to mathematical model. In this study, the modeling of filtering process by using neural network software MATLAB 8.1 (2013) with experimental data from a submerged MBR system that is equipped with Kubota membranes for municipal wastewater treatment with high Mixed Liquor Suspended Solids (MLSS) concentration was investigated.
2/3 of empirical data were used for build, training and assessment the network, then the designed network was used to estimate the permeability of 1/3 of the data as well as other similar membrane bioreactor system. trainlm algorithm is applied for training. The value of Coefficient of determination (R^2) for predicting the permeability of 1/3 of datas of the first system is 0/93 and 0/92 for the same system.