Optimizing Preparation of Modified Granular activated carbon Using Orthogonal Design and BP Neural Network

Abstract:

Granular activated carbon (GAC) was modified with FeCl2•4H2O and KMnO4. The scheme was optimized by orthogonal design combining with BP neural network to get effective removal efficiency of arsenite. The four factors five levels orthogonal design table, such as overall mole concentration, the mole ratio of FeCl2•4H2O and KMnO4, water bath temperature and drying temperature as factors and arsenite removal rate as the goal factor, was established. The optimal modified scheme was found via predicting and selecting the BP network model through four factors as inputs and arsenite removal rate as output, which was combined with the effective fitting function. The overall mole concentration and the mole ratio of FeCl2•4H2O and KMnO4 were 0.12mol•L-1and 3:1, water bath temperate and drying temperate were 45℃ and 190℃,respectively. The arsenite removal rate was 0.765, so there was 3.00% error discrepancy compared with the model value 0.788. XRD and SEM were applied to gain the optimal modified GAC surface property. At the same time, the iron and manganese amount was measured. These supplied fundamental information for the further research.