In Czochralski method, the shape of crystal-melt interface and its position play a major role on the quality of single crystals. In the Czochralski crystal growth process having a nearly flat interface, a single crystal with less structural defect, uniform physical properties and homogenous chemical composition is obtained. In the present study, firstly a 2-D fluid flow and solidification model was developed to simulate the YAG single crystal growth process using a finite volume method. The fluid flow and solidification heat transfer model was further tested by available experimental data. The verified fluid flow and solidification heat transfer model was used to build an artificial neural network and trained to optimize the parameters affecting the shape and position of the interface. Finally, the trained neural network was employed to optimize the operating parameters such as pulling rate, rotation speed of the crystal, ambient gas and temperature of crucible wall to obtain a closely flat crystal-melt interface. The optimized variables were eventually used in fluid flow model to evaluate the performance of the optimization model. (C) 2008 Elsevier B.V. All rights reserved.