Design method for reasonable operation of industrial crystallizer using neural network model

The use of a neural network model to determine the operational conditions for the production of crystals of specific size in a continuous industrial crystallizer with an actual heat exchange area of 400m(2) is discussed. The product crystal size is well determined by the neural network model consisting of three explanatory variables: steam flow rate, suspension density of crystals in a crystallizer and frequency of the circulation pump. The most suitable learning number of iterations for the neural network model obtained by the leave-one-out cross-validation method is 50,000, and the mean estimated error of the product crystal size is about 0.03mm. From these results, it is believed that the neural network model is sufficiently accurate for practical use, and is effective for the design of the operational conditions required for manufacturing products with a specific crystal size in industrial crystallization. A practical method for constructing the neural network model is proposed.