Zhenning Ba

Conventional gradient-based elastic full-waveform inversion methods encounter challenges associated with local minima and trade-offs in multiparameter inversion. To address these issues, a new method was developed, called the elastic direct waveform inversion (eDWI) scheme. By explicitly enforcing the time-space causality of the wavefield, eDWI converted, without loss of information, the global iterative nonlinear inversion problem into a sequence of localized recursive inversions performed from shallow to deep depths. The method was tested on two numerical examples for stratified layered elastic media with a plane-wave source and a point source. The stiffness matrix method was used to decompose and extrapolate wavefields into up/down-going P and S wavefields, and specific P-to-P and P-to-S reflection events were used for the localized recursive inversion. Numerical examples showed that the new approach did not need a global initial model, avoided local minima, and addressed parameter coupling issues. It automatically used multiples and maintained efficacy even without low-frequency content in the recorded seismic data. The new inversion approach may find a wide range of imaging and sensing applications using seismic and other waves.

Spatiotemporal prediction of seismic ground motion represents a critical scientific challenge in earthquake engineering and seismology, with important implications for seismic design, rapid post‐earthquake damage assessment, and emergency response. Traditional methods, which primarily rely on physics‐based simulations and ground motion prediction equations, face inherent limitations in balancing prediction accuracy and computational efficiency. In recent years, the continuous expansion of strong motion records and numerical simulation databases has enabled machine‐learning (ML) approaches to demonstrate significant advantages in ground motion prediction. This study systematically reviews advances in ML–based spatiotemporal ground motion prediction, focusing on two key aspects: spatial ground motion parameter prediction and time series prediction. The reviewed studies are further categorized according to input data type, including feature parameters, sparse spatial fields, and station time series data, while recent developments in physics‐informed learning are also summarized. Existing research indicates that ML approaches show strong potential for modeling complex nonlinear spatiotemporal characteristics of seismic motions and for improving computational efficiency under specific data conditions. However, important challenges remain regarding data imbalance, cross‐region generalization, physical consistency, interpretability, and uncertainty quantification. Current evidence suggests that ML‐based approaches should presently be regarded as complementary tools rather than replacements for traditional ground motion prediction frameworks in engineering‐critical applications. This review aims to provide a critical reference for future development of ML‐based spatiotemporal ground motion prediction methods.

When large liquefied natural gas (LNG) storage tanks are constructed on deep non-rock sites, soil–foundation–structure interaction (SFSI) alters the dynamic characteristics of the system and affects the actual control effectiveness of the isolation layer. However, the current understanding of the coupling mechanism between the two remains insufficient. This paper takes a 220,000 m³ full-containment LNG storage tank as the study object and establishes a three-dimensional finite element model of the tank-pile group-site system in ABAQUS. Through comparative analyses of three model configurations, namely a rigid foundation model, a non-isolated model considering SFSI, and a lead-rubber bearing (LRB) isolated model considering SFSI, the SFSI effects and LRB isolation effectiveness are systematically separated. For the SFSI effects, the deep site attenuates medium- and high-frequency content while amplifying the response around approximately 1.6 Hz through site–foundation flexibility, transforming the heightwise acceleration amplification profile from an approximately linear pattern to a curvilinear one that bulges at mid-height, with peak pile-cap accelerations increasing by 25.1–76.9% relative to the rigid-base values. For the LRB isolation performance, the introduction of LRBs shifts the dominant system frequency below 1.0 Hz and reduces the maximum tank-wall acceleration amplification factor from 2.64 to 0.81. The resulting attenuation of superstructural inertial forces leads to reductions of 49.6–82.0% in pile-head shear and 57.4–78.0% in near-head bending moment, while the outer-to-inner pile-head moment ratio decreases from 2.94 to 1.13, indicating substantially improved pile-group force uniformity. Nevertheless, the beneficial effect of isolation diminishes with depth, and internal forces at abrupt soil-stiffness interfaces remain governed by kinematic interaction that the isolation layer cannot mitigate. The findings of this study can provide references for the seismic isolation design and pile foundation seismic optimization of super-large LNG storage tanks on deep overburden sites.

Dynamic structure–soil–structure interaction (SSSI) for periodically arranged structure system under incident plane SH wave is investigated using the indirect boundary element method. The solution is presented through response of an infinite number of equally spaced, identical structures (periodically arranged structure system) supported by rigid foundations embedded in a soil layer overlying elastic bedrock. The presented method has the advantages of high accuracy, low computational cost, and low memory requirement. In parametric analysis, whether only immediately adjacent structures contribute significantly to the structural response in dense areas is analyzed first. Numerical results showed that the structural response for periodically arranged structure system may be evidently different from that of a finite array of structures model. While the immediately adjacent structures have the most significant effect, the effect of farther structures is still non‐negligible. Then, the effects of separation distance and soil layer are investigated. It is shown that the separation distance may reduce or amplify the structural response compared with that for single structure, which could be as large as 67.5% and 59.7%, respectively, for building relative displacement. The soil layer also has prominent influence. With decreasing soil layer stiffness or soil layer thickness, the building's relative displacement amplitude increases. The conclusions may be useful for prediction of seismic response for design and interpretation of recorded response of full‐scale structures in densely built areas.

The large shaking table at Tianjin University's National Facility for Earthquake Engineering Simulation (NFEES) is the largest in the world, featuring a rigid‐flexible foundation differing from the conventional rigid block type. This design meets the structural requirements for large‐scale shaking tables and is suitable for sites without ideal soil conditions. However, the vibration response characteristics and spatial distribution patterns of this foundation under dynamic loading remain insufficiently studied. This paper presents a test investigation into this newly developed foundation's response, employing white noise sweep and sinusoidal excitation tests to obtain actual acceleration response data and its spatial distribution. Furthermore, the key factors influencing the acceleration response distribution are analyzed by considering both the structural properties and the excitation characteristics. The results show that at excitation frequencies below 15 Hz, the foundation exhibits good overall integrity. In contrast, high‐frequency horizontal excitations tend to localize the response in the excitation region, while high‐frequency vertical excitations amplify the response in the flexible area, leading to a clear deviation from rigid‐body motion. This study provides a reliable and authentic data source for future theoretical and numerical research, and the findings offer valuable engineering insights and practical guidance for the design and application of similar large‐scale shaking table foundations.