Ceramic/Metal FGM Armor is a new type armor, and it can provide more strong protection than conventional Ceramic/Metal armor for combat vehicles, because there is no significant interface between ceramic and metal, which make the FGM armor transfer the energy to larger region than conventional ceramic armor and reduce the reflected stress wave's strength, especially tensile stress wave strength, so that the hard face of ceramic can be used sufficiently. In this paper, a new kind of optimize method is advanced to research the design of FGM armor. This method is based on the neural net theory and conjugate gradient method. The process of this method is as following: (1)The impact process of the armors with different FGM layer is analyzed by numerical method or test method, in this process, two aspects of factors are taken into account, one is the design parameters of FGM armor: it includes FGM's layers number, the percent of ceramic in each FGM layer, the thickness of each FGM layer, the thickness of ceramic layer and the metal layer; another is the object parameter of optimization: the penetration depth is only one parameter. (2)The neural net theory is used to build the relation between the design parameters and object parameter. Due to the high non-linear characteristic of this relation, it can't be expressed by ordinary function. The results of last step are as the training sample of neural net, which is made of 5 input parameters (the design parameters) and 1 output parameter (the object parameter). (3)The relation expressed by neural net is selected as the object function of optimization, then the conjugate gradient method is used to build optimization algorithm. (4)The optimized results are attained by optimization algorithm. At the end of paper, the ceramic/metal FGM armor is optimized by this method, and then validated by numerical method.