Taking the double clutch transmission gear system as the research object,the virtual prototype model of the gear transmission system is established by using Romax software. The bearing capacity,contact spot and transmission error of the unmodified first gear are analyzed. Based on the theoretical basis of helical gear axial modification and tooth profile modification,the optimum three-dimensional comprehensive modification parameters of the first gear are designed by applying the optimal Latin sampling principle. Then,the optimal three-dimensional modification parameters are obtained with the help of the neural network and genetic algorithm. The simulating calculation results indicated that the three-dimensional modification of the gear can greatly reduce the gear contact stress,optimize the gear contact spots and reduce the meshing transmission error.The comparison of the results before and after optimization also proved the effectiveness of this method and it can be used as a reference for optimum design of transmission gears.