Prediction of permeate flux decline in crossflow membrane filtration of colloidal suspension: a radial basis function neural network approach

The capability of a radial basis function neural network (RBFNN) to predict long-term permeate flux decline in crossflow membrane filtration was investigated. Operating conditions of transmembrane pressure and filtration time along with feed water parameters such as particle radius, solution pH, and ionic strength were used as inputs to predict the permeate flux. Simulation results indicated that a single RBFNN accurately predicted the permeate flux decline under various experimental conditions of colloidal membrane filtrations and eventually produced better predictability than those of the regular multi-layer feed-forward backpropagation neural network (BPNN) and the multiple regression (MR) method. We believe further development of the artificial neural network approach will enable us to design and analyze full-scale processes from results of laboratory and/or pilot-scale experiments.

Due to the complex physicochemical microphenomena occurring during membrane filtration, conventional theoretical models have been able to predict the filtration procedure only under limited conditions with specific and/or inevitable assumptions. An artificial neural network method provides a unified approach for analysis and prediction of the membrane performance under various combinations of operational conditions. In this study, the RBFNN is applied to predict the permeate flux on crossflow membrane filtration as a function of transmembrane pressure, ionic strength, solution pH, particle size, and elapsed filtration time. Results show that the transient profiles of the permeate flux during crossflow membrane filtration can be predicted by a single RBFNN with acute accuracy, given a limited number of training points. Comparison of the performances of RBFNN, BPNN, and MR confirms the superiority of RBFNN in terms of the R-square, RMSE, N10 and simulation time required of higher accuracies. As a consequence, it may not be necessary to carry out an entire series of expensive pilot or fullscale tests to collect and verify filtration data. The RBFNN can interpolate the performance of membrane filtration under other conditions of interest by using widely ranged and sparse data points to reduce the time and costs required. Although pilotscale experimental tests restrict the number of operating parameters that generally are of importance in larger-scale applications, they still can closely simulate full-scale operations. In the same manner, RBFNN can also serve as potential prediction tools for full-scale operations, possibly combining several trained networks of different scales into one smart neural network.