In order to reduce the hazards of earthquake disasters on human, basic principle of reducing vibration is discussed by establishing passive vibration isolation model. The calculation results show vibration can be reduced by isolation materials whose deformation modulus (stiffness) should be smaller and damping should be appropriate. Light reinforced soil is a good man-made material, its all physical mechanical parameters can be changed, and anti-pull capacity of soil can be increased. According to the basic principle of reducing vibration, reducing vibration idea using light reinforced soil is brought forward, and the idea has a very important significance for research on reducing earthquake disasters from the angle of ground and foundation.

Based on the RCC（roller compacted concrete） gravity dam, considering the multiphase, heterogeneity and microscopic structure characteristics of RCC, we use the RFPA analysis system to analyze the whole destruction process of the RCC weak level under the action of the pure water pressure in this paper. By simulation, we obtained the distribution of the stress, elastic modulus, water pressure on the whole failure process of RCC weak level and several parameters, such as the up crack load, instability load and critical crack propagation length. By calculation, we got the double K fracture toughness of RCC under the different of water pressure, and the results show that the up crack toughness and instability toughness change regularly with the increasing of water pressure initial value in the mass.

This paper introduces the neural network PID control method, in which the parameters of PID controller is adjusted by the use of the self-study ability. And the PID controller can adapt itself actively. The dynamic BP algorithm of the three-layered network realizes the online real-time control, which displays the robustness of the PID control, and the capability of BP neural network to deal with nonlinear and uncertain system. A simulation is made by using of this method. The result of it shows that the neural network PID controller is better than the conventional one, and has higher accuracy and stronger adaptability, which can get the satisfied control result.