In order to cope with excessive displacement requirements of the isolation layer in the base isolated structure，a hybrid isolation system is proposed in this paper through the base isolated system （BIS） + tuned tandem mass dampers-inerters （TTMDI）. The Bouc-Wen hysteretic model is used to simulate the nonlinear force-deformation behaviors of the isolation layer. Employing the stochastic equivalent linearization，pattern search optimization algorithm，as well as the earthquake ground motion model simultaneously，the optimum design framework of the BIS+TTMDI system is established in the frequency domain. The performance of the BIS+TTMDI system is systematically evaluated in terms of its robustness，effectiveness，stiffness and damping coefficient，stroke and sensitivity to seismic frequency contents，and compared with BIS + the tuned mass damper （TMD），tandem tuned mass damper （TTMD），and tuned mass damper-inerter （TMDI），respectively. The dynamic elasto-plastic analysis of a seven story hybrid base-isolated system，respectively including the BIS+TTMDI and BIS+TMDI systems，is carried out under the near-field earthquake ground motions. The results show that the BIS+TTMDI system has the best seismic performance and strong robustness. Furthermore，the total damping requirement of the TTMDI in the BIS+TTMDI system is less than half of the TMDI in the BIS+TMDI system，which is more economical and practical.

In order to improve the application of the damping force of isolation device for free-standing objects，a new damping mechanism is developed. Its structure and working mechanism are expounded，and its mechanical characteristic is deduced and verified. Combined with the target performance of an isolation device for cultural relics，the parameters of the damping mechanism are designed by numerical method，and contrasted with the shaking table test. The results indicate that the damping mechanism has a simple structure and stable performance. The isolation device optimized by damping parameters achieves the set isolation target under the premise of meeting the displacement limit. The test results are highly consistent with the numerical analysis results，which not only prove the correctness of the damping design，but also verifies the installation quality and operation effect of the isolation device. In addition，the coincidence degree between two results can provide a new method for design comparison and quality evaluation of the isolation device，especially the coincidence degree of relative displacement curves.

The one-dimensional and two-dimensional time-domain inversions of the bedrock incident waves cannot reflect the spatial arbitrariness of the incident azimuth and oblique incident angles，and the free field constructed based on these two methods might deviate is far from reality. It is necessary to carry out a three-dimensional time-domain inversion of the bedrock incident waves. Based on the wave function combination method，the free field of the ground surface control point is expressed by the time histories of the incident waves of P，SV and SH waves. A three-dimensional time-domain inversion method of bedrock incident waves based on design ground motions is established. Then，the spatially non-uniform free field under the oblique incidence of multi-wave combination is constructed. Finally，the free field deviations constructed based on one-dimensional and two-dimensional inversion methods are analyzed，and the influence mechanisms of different inversion and free field construction methods on the seismic response of the asphalt concrete core dam are revealed. The results show that the free field constructed based on the three-dimensional inversion method of ground motion has spatial inconsistency and can reflect the site ground motion field more comprehensively. Compared with the ground motion field constructed based on three-dimensional inversion，when the incident azimuth angle is 0°~60° and the oblique incident angle is 40°~90°，the feature point free field in the x-direction constructed by the one-dimensional inversion method has a large deviation with the maximum deviation55.8%； When the incident direction is parallel to a certain horizontal coordinate axis，the other horizontal free field deviation of the feature point constructed by the two-dimensional inversion is larger，and the maximum deviation is 100%； When the Poisson's ratio of the medium increases，the deviation of the x-direction free field decreases under the one-dimensional inversion，and the free-field deviation of the x-direction increases under the two-dimensional inversion，and the maximum deviations of the feature points are 46% and 36%，respectively. Compared with the maximum tensile stress of the core wall based on the three-dimensional inversion method of ground motion，the maximum tensile stresses under the two-dimensional and one-dimensional inversion methods are reduced by 83.3% and 20.0%，respectively. Therefore，it is necessary to perform a three-dimensional inversion of bedrock incident waves and construct a spatially non-uniform free field based on the design ground motion in the seismic design of dams.

Based on the actual project in Dalian，this paper studies the dynamic interaction （SSSI） of the double tunnel sand bridge pile system under earthquake through shaking table test，obtains the dynamic response law of structure and site，and compares it with ABAQUS numerical simulation. The Kelvin constitutive model subroutine is introduced into the numerical model，and the equivalent linear method is used to deal with the nonlinear problem of sand in the calculation process. The experimental results are compared with the numerical model to verify the reliability of the numerical simulation. On this basis，eight working conditions are designed，and the interaction law between structures in the system is studied through comparative analysis. The results show that the tunnel will amplify the peak acceleration of the bridge pile and adjacent tunnel，but the bridge pile will weaken the peak acceleration of the side tunnel； the existence of the tunnel and bridge pile will increase each other's section shear force and bending moment. The main affected areas are concentrated in the upper and lower arches of the tunnel and the interface between the pile bottom of the bridge pile and the pile-soil.

With the advent of the era of big data，the mechanical equipment fault diagnosis method based on deep learning has attracted more attention. However，the traditional deep network model seriously limits its application in practical engineering due to the excessive amount of parameters and calculations. Based on this，a GhostConv lightweight network model is proposed and used for fault diagnosis. GhostConv generates a small part of the feature maps through conventional convolution，and performs multiple feature extraction on the generated feature maps to generate the remaining feature maps. Contact the feature maps of the two parts to obtain a complete feature map. GhostConv structure saves the cost of generating redundant feature maps in conventional convolution to the maximum extent，and reduces the model parameters to ensure the performance of the model. In the experiment，the continuous wavelet transform is used to transform the vibration signal to generate a two-dimensional time-frequency diagram，and then the designed GhostConv is used to establish a lightweight fault diagnosis network model. The original dataset and noisy dataset of Case Western Reserve University are used for experimental verification，and compared with the conventional convolution structure network model and depth separable convolution structure model in terms of parameters，calculation and recognition rate. The experimental results show that the GhostConv lightweight network model still has high recognition accuracy and strong anti-noise ability under the condition of fewer parameters and calculations with good robustness and generalization ability. The parameters of the model are only 6% of the conventional convolution model and 56% of the deep separable convolution model. Under the condition of strong noise interference，the fault diagnosis and recognition rate is still higher than that of the conventional convolution model，which confirms its engineering application value.