The buffer allocation problem, i.e. how to distribute a certain amount of buffers among the buffer locations of a production line, is an NP-hard combinatorial optimization problem. In this study, a hybrid approach combining simulated annealing and genetic algorithms is presented to find optimal buffer sizes for unreliable production lines. The objective is to improve the capacity of production line by implementing proposed hybrid approach. To evaluate the capacity of the line, a detailed simulation model is developed. Numerical examples are given for evaluating the performance of the proposed heuristic, traditional genetic algorithm, and simulated annealing procedures. The results show that the proposed hybrid approach gives good solutions in reasonably short computation times.