Grain design is the most important part of a solid rocket motor. The aim of this study is finocyl grain design based on predetermined objective function with respect to ballistic curves in order to satisfy various thrust performance requirements through an innovative design approach using a genetic algorithm optimization method. The classical sampling method has been used for design space-filling. The level set method has been used for simulating the evolution of the burning surface in the propellant grain. An algorithm has been developed beside the level set code that prepares the initial grain configuration using Pro/Engineer software to export generated models to level set code. The lumped method has been used to perform internal ballistic analysis. Two meta-models are used to surrogate the level set method in the optimization design loop. The first method is based on adaptive basis function construction and the second method is based on the artificial neural network. In order to validate the proposed algorithm, a grain finosyl sample has been investigated. The results show that both grain design method reduced the design time significantly and this algorithm can be used in designing of any grain configuration. In addition, data have more accuracy in grain design based on the artificial neural network, so this method is the more effective and practical method to grain burn-back training.