Dead-end filtration of yeast suspensions: Correlating specific resistance and flux data using artificial neural networks

Abstract

The specific cake resistance in dead-end filtration is a complex function of suspension properties and operating conditions. In this study, the specific resistance of resuspended dried bakers yeast suspensions was measured in a series of 150 experiments covering a range of pressures, cell concentrations, pHs, ionic strengths and membrane resistances. The specific resistance was found to increase linearly with pressure and exhibited a complex dependence on pH and ionic strength. The specific resistance data were correlated using an artificial neural network containing a single hidden layer with nine neurons employing the sigmoidal activation function. The network was trained with 104 training points, 13 validation points and 33 test points. Excellent agreement was obtained between the neural network and the test data with average errors of less than 10%. In addition, a network was trained for prediction of the filtrate flux directly from the system inputs and this approach is easily extended to crossflow filtration by adding inputs such as the crossflow velocity and channel height. An attempt was made to interpret the network weights for both the specific resistance and flux networks. The effective contribution of each input to the system output was computed in each case and showed trends that were as expected. Although network weights, and consequently the computed effect of each parameter, is different each time a network is changed (depending on the initial weights used in the training process), the variation was low enough for information contained in the network to be interpreted in a meaningful way.Keywords: Dead-end filtration; Specific cake resistance; Artificial neural network; Bakers yeast

Artificial neural network model for desalination by sweeping gas membrane distillation

“…Several studies have been considered for the application of ANN in modeling of various processes in membrane technology [1][2][3][4][5][6][7][8][9][10]. A feed-forward ANN was developed by Abbas and Al-Bastaki [1] for the prediction of a reverse osmosis (RO) performance using a FilmTec SW30 membrane for desalination of various salt solutions ranging between brackish water and seawater salinities.…”

“…Sahoo and Ray [9] worked on the prediction of permeate flux decline in crossflow membranes using ANN and genetic algorithms. Mhurchú and Foley [10] employed the dead-end filtration of yeast suspensions by correlating specific resistance and permeate flux data using artificial neural networks.…”

The use of neural network to estimate mass transfer coefficient from the bottom of agitated vessel

“…The number of neurons in the hidden layer was varied from 1 to 12 neurons. As previously reported, the network performance was found to be sensitive to the randomly generated initial guess [32,60]. Therefore, each test was performed around 50 times and the averages of the K values were used in the subsequent calculations.…”

Effects of particle size and its distribution on specific cake resistance during rice wine microfiltration

“…The cake is incompressible for n = 0 and is more compressible for higher values of n. Specific cake resistance is dependent on several factors including particle shape, size distribution, porosity, and particle density (Endo and Alonso, 2001), as well as suspended solids concentration (Lee et al, 2003b;Mhurchu and Foley, 2006;Tanaka et al, 2001).…”