Manufacturing of high-quality metal plates for various industries such as the automotive, aerospace, appliance and food packaging, both in terms of mechanical and physical properties has always been a very important issue. From the consumer point of view, flatness is one of the most important parameter that has a great impact on final product quality. Improving the quality of the plate is costly and therefore from the technical point of view, finding the optimal setting rolling parameters to minimize the number of operation and hence the manufacturing cost, is an important issue. In this thesis, the impact of three most effective parameters on flatness of rolled plate namely; reduction, work roll bending and work roll crown were evaluated. For this purpose, the cost function was created using rolling mill data, slit beam method (SMB) and artificial neural network (ANN). The required data for SBM is obtained from the computer logs of Mobarakeh Steel Company (MSC). The optimum structure of ANN is determined by setting the number of neurons in the hidden layer and learning algorithm. Cost function is optimized by different metaheuristic algorithms such as genetic algorithm (GA), particle swarm optimization (PSO), differential evolution (DE) and co-evolutionary particle swarm optimization (CPSO). Firstly, reductions and weighted sum of flatness in sequential stands were considered as input variables and output of cost function. The required data for neural network training provided by SBM code. The cost function derived from the neural network is optimized by the mentioned metaheuristic algorithms. Finally the appropriate optimization algorithm is chosen by comparison between the results obtained from different metaheuristic algorithms and the cost function is optimized for different strip widths. Since the desired flatness was not achieved for wide strip by considering the reductions as only input variables of cost function, work roll crown and work roll bending were also considered as input variables of cost function to achieve appropriate flatness. Finally a significant improvement in flatness is resulted by optimizing the reduction, work roll crown and work roll bending in each stand Keywords: Strip flatness, Artificial neural network (ANN), Genetic algorithm (GA), Particle swarm optimization, Differential Evolution (DE), Hot rolling