抄録
The specific filtration resistance of a slurry was predicted from several fundamental properties of the particles in the slurry using neural network (NN) modeling. Polyethylene beads, polymethyl methacrylate beads etc. of several to several hundred micrometers in volume-based median diameter were dispersed in water or methanol to obtain the compression-permeability data. The porosity of the sediment obtained from the slurry in gravitational and centrifugal fields was newly added as one of the input parameters of the NN. The other input parameters were the particle size distribution and the porosity of the compressed cake. The product of the specific filtration resistance and the particle true density of the cake was chosen as the output parameter of the network. The NN trained with these data was found to be fairly effective in predicting the filtration property of another slurry containing particles used to obtain the training data. The porosity of the compressed cake could also be predicted successfully by another NN trained with the sedimentation results, the particle size distribution and the loaded pressure in compression-permeability tests. The predicted cake porosity could be utilized as the input value to predict the filtration property. With the method, after the database of filtration resistance and the properties of various slurries is developed, the filtration resistance of a slurry can be predicted with much less amount of the test slurry compared with conventional filtration experiments.
