The optimal design of gear parameters in the movements of clocks and watches is a combinatorial optimization problem, which is too difficult to be solved optimally using traditional optimization methods. The GA(genetic algorithm) is a natural evolution-based heuristic searching method, which was recently introduced and has been successfully applied to a variety of problems. The difficulties in using this algorithm are how a particular problem is to be modeled to fit into the genetic algorithm framework, and how the operators (selection, crossover, mutation) work due to the code strings. The indexes of various gear parameters of transmission is presented in this paper as a string, and the optimal design of gear parameters is successfully solved. This method can be also applied to general mechanical designs. Through comparison, it can be seen that the parameters calculated in the GA are better than those in the expert system.