Yield prediction in activated carbon production process using artificial neural network and genetic algorithm

Activated earrons arc produced from a varicty of carbonaceous raw materials. 'be yield of production proccss and charactcristics of activatcd carbons dcpcnd on the physical and chemical properties of the starting materials and the activation methods used. The mechanism of activatedcarron production is only partially known; thcrcfore the fonnu lat ion of a correct model for this process is a difficult task. In this study, llctivated carbon yield was predicated with artificial illlelligence. Artificial neural networks (ANN) arc inlonnation processing systems that have been used lor making predictions based upon their perceived inllucncing lac tors. A radial basis neuralnetwork was first designed for yield of process eval uation, then the results were applied for optimization with genetic algorithm. Raw matcrialtypc, ash content, activation agent, impregnation ratio of activation agent to raw material, activation temperature and activation time were selected asinput variables and yield as the desired output variable_ The results indicated that artificial neural networks arc able to predict yield of activated carron production. Mean squared error lor train data and test data in RBF neural network , were 0.0007 and 0.0012 respectively. The best trained network was eOlllieeted to genetic algorithm (GA) to find the optimum value for input parameters. For di lTerelll raw materials and dilTcrent chemical agents, optimization was done with GA Optimization results showed that there is a good agreemcnt between uetual va lue for yield of activated carron production and predicted value applying this network