Electrical cabinets are critical for the reliable operation of nuclear power plants (NPPs), yet their seismic fragility under high-frequency ground motions remains insufficiently addressed. This study develops a computationally efficient three-dimensional finite element model for an anchored NPP electrical cabinet that explicitly accounts for excitation frequency relative to cabinet dynamics. The cabinet is idealized with beam and shell elements, while omitted panel masses are represented by an equivalent lumped mass whose location is optimized using Response Surface Methodology. The optimized model reproduces shaking-table-identified fundamental frequencies with a maximum error of 1.99% and accurately matches measured top-of-cabinet acceleration histories. Seismic fragility is evaluated through incremental dynamic analysis using two controlled suites of motions: a low-frequency set dominated below the fundamental mode and a high-frequency set concentrated near the first-mode frequency. High-frequency excitation produces a consistent leftward shift of fragility curves, indicating earlier exceedance of a functional acceleration limit, and reduces the HCLPF capacity by about 22% compared with low-frequency excitation. These findings demonstrate that excitation frequency relative to cabinet dynamics critically influences seismic vulnerability, and neglecting high-frequency effects can yield unconservative capacity estimates. The proposed framework provides an efficient basis for frequency-sensitive seismic fragility evaluation of NPP cabinets.

This study investigates the influence of inclined bedrock on seismic site response. Conventional one-dimensional seismic analyses in earthquake engineering assume flat bedrock. However, subsurface bedrock layers are rarely horizontal, and assuming a flat bedrock surface may lead to inaccurate predictions. Additionally, the numerical approaches to capture the resulting 2D wave interactions for inclined bedrock also remain limited. To address these limitations, two-dimensional (2D) site response analyses were conducted for soil-bedrock models with bedrock slopes of 0°, 5°, 10°, 20°, and 30° to assess how bedrock slope affects the surface response. Additionally, a parametric study was carried out to explore how different earthquake excitations, soil properties, and depth of inclined bedrock interface affect seismic response. Numerical analyses were performed using ABAQUS, and the models were validated against analytical solutions. The results show that the one-dimensional assumption becomes unreliable beyond 10° inclination, while the effect weakens with increasing interface depth and varies with observation location. The slope side amplifies high-frequency components, while the flat side responds more strongly at lower frequencies, resulting in unequal demand at support. The lateral influence of the slope extends to about six times the soil-column height. Increasing bedrock inclination shifts the dominant peaks to shorter periods and reduces long-period response along the sloping interface, and the stiffer site experiences higher amplification, contrary to one-dimensional site response predictions, due to complex wave interference associated with faster wave propagation. These results suggest the need for spatially varying design spectra and micro-zonation-based design to capture true seismic demand on sloping bedrock sites.

This paper evaluates the soil-foundation-structure interaction (SFSI) of structures supported by shallow foundations and deep basements using a dynamic centrifuge test at 60 g in dry sand. Two testing series were investigated including shallow foundation and deep basement systems. Each series comprises one free-field ground motion case and two model structure cases using single-degree-of-freedom (SDoF) and two-degree-of-freedom (2-DoF) on various types of foundations. To address the effects of the SFSI, the spectral accelerations imposed on the model structures supported by respective foundation systems were measured using different ground input motions. All dynamic tests were performed on one directional shaking table mounted with an equivalent shear beam containing dry dense sand with a relative density of 77 to 80

Control cabinets in nuclear power plants (NPPs) and Small Modular Reactors (SMRs) must maintain functional integrity under enhanced seismic conditions. Standard cabinets are highly susceptible to resonance near their fundamental frequency (similar to 12 Hz), causing excessive top-level acceleration amplification. To address this, a Tuned Mass Damper (TMD) prototype was designed and experimentally validated. A TMD with a 16.5 kg tuning mass, tuned to 13 Hz, was installed on a 939 kg standard cabinet and subjected to seismic tests using an enhanced 0.42 g Required Response Spectrum (RRS). The TMD successfully suppressed resonance, reducing cabinet top peak spectral acceleration (PSA) by over 50%, from approximately 91 g to below 36 g. This study confirms the robust efficacy of TMDs as a seismic mitigation solution for NPP or SMR control cabinets. The results validate their potential for widespread application in enhancing the seismic resilience of critical nuclear infrastructure.

Artificial intelligence is increasingly being applied across engineering disciplines, including the study of seismic soil-structure interaction (SSI). While deep learning models offer strong potential for predicting structural time-history responses in SSI systems, their practical adoption remains limited by the high computational cost of generating extensive high-fidelity (HF) training data. To address this challenge, this study introduces a one-step multi-fidelity (MF) deep learning framework that simultaneously integrates low-fidelity (LF) and HF numerical simulation data. The key innovation lies in leveraging informative yet computationally efficient LF models, such as free-field soil column or fixed-base structure, to complement limited, computationally expensive HF data that fully capture nonlinear material behavior and soil-structure contact. The results demonstrate that models trained with the proposed MF approach achieve superior predictive accuracy and stability compared with HF-only methods, even when using a small dataset. This strategy substantially reduces computational demands and provides a practical, scalable pathway for the broader application of deep learning in complex seismic SSI analysis.

Deep learning-based structural health monitoring (SHM) of large-scale structures requires extensive damage labels. Data augmentation is commonly performed with numerical simulations based on finite element (FE) models. However, a significant distributional discrepancy or domain gap, exists between the measured signals and simulation data, rendering them heterogeneous. This disparity arises because measured signals are sensitive to external conditions, while simulation results are predicated on idealized numerical analysis. This study proposes an end-to-end framework, the contrastive domain adaptation-autoencoder (CDA-AE), to learn noise-robust features and mitigate the domain gap with an unsupervised model. The proposed method generates synthetic data by extracting noise from measured signals and synthesizing it onto the clean simulation data using signal processing techniques. Since both the measured data and the synthetic data pertain to the same structure, their embedded features share key dynamic properties, i.e., eigenfrequencies. The CDA-AE architecture is designed to minimize the impact of noise and map the invariant physical features onto a common latent space. By employing contrastive learning, the framework aligns source features, enhancing sensitivity to subtle eigenfrequency changes that indicate the structure's health state. The proposed approach was validated through its application to anomaly detection in bridge behavior using ambient vibration data. The framework successfully captures variations in the acceleration response of a real bridge, and, even with limited measured data, anomaly detection accuracy of 99.75% was achieved. These findings suggest that the proposed framework holds significant potential for the development of future real-time, digital twin-based SHM systems.

This study evaluates the seismic performance of a nuclear power plant (NPP) electrical cabinet with emphasis on frequency-dependent in-cabinet response amplification. An experimentally validated finite element model is developed to capture global cabinet behavior and local in-cabinet responses under seismic excitation. Shaking table test data are used to verify the model based on natural frequencies and acceleration responses measured at multiple cabinet locations. Seismic performance is assessed using spectrally matched artificial floor response motions representing varying frequency content and intensity levels. Results indicate that low-frequency excitation resulted in nearly rigid-body behavior with minimal amplification. Under intermediate-frequency input, noticeable amplification develops at the cabinet top and within in-cabinet shelves, indicating increased dynamic participation of cabinet components. The most critical response occurs under high-frequency excitation, where resonance near the cabinet’s fundamental mode leads to peak spectral accelerations (SAs) of about 29.4 g at the cabinet top and up to 26.1 g at the upper in-cabinet shelf. Findings further indicate that in-cabinet response amplification is governed primarily by frequency content rather than a constant amplification factor recommended by the EPRI NP-7146-SL guidelines. This suggests that frequency-dependent analysis provides a more realistic and less conservative approach to the seismic qualification of electrical cabinets.

Accurate nonlinear seismic site response analysis requires site-specific damping. In current finite element (FE) practice, Rayleigh damping is widely used as an equivalent viscous model, but the mass- and stiffnessproportional coefficients alpha, beta are typically chosen from empirical rules or trial-and-error calibration, making results user-dependent and difficult to reproduce. This study proposes a physics-informed neural network (PINN) framework that inversely identifies Rayleigh damping from paired bedrock-surface acceleration records. Synthetic training data are generated by nonlinear site response analyses in ABAQUS using a Mohr-Coulomb constitutive model with Rayleigh damping, and centrifuge shaking table tests on a dense sand deposit provide independent validation data. The model takes the magnitude and phase of the frequency response function (FRF) as input features and employs a multi-task architecture with a damping ratio classification head and a physics head that predicts Rayleigh coefficients and modal weights for FRF reconstruction. A hybrid loss function combines cross-entropy on damping ratio classes, FRF-matching, time-domain reconstruction error, and a physics-consistency term linking alpha and beta to the equivalent damping ratio. In the initial simulations, an equivalent damping ratio of about 10% is used, whereas half-power bandwidth analysis of centrifuge FRFs indicates an effective damping of about 5%. Applied to these experimental records, the trained PINN consistently infers damping-ratio classes and Rayleigh coefficients corresponding to approximately 5% damping, thereby quantifying the discrepancy between empirical damping assumptions and the actual site response. To further validate the framework beyond laboratory conditions, a site-specific free-field model was developed for the KiK-net station FKSH11 in Fukushima, Japan. Using the same workflow as for the centrifuge-based dataset, additional synthetic bedrock-surface response pairs were generated for the FKSH11 profile and used to fine-tune the pretrained PINN; the adapted model was then applied to the FKSH11 field record to infer Rayleigh damping. Numerical simulations using the PINN-inferred Rayleigh damping (corresponding to 5%) reproduce the FKSH11 measured surface response spectrum within 5 -15% error at the predominant spectral periods, whereas the conventional 10% damping assumption underestimates peak spectral accelerations by approximately 30 -40%, consistent with the centrifuge validation results. The framework achieves 98.5% classification accuracy for unseen motions and provides a reproducible inverse-identification workflow for site-specific calibration of Rayleigh damping in nonlinear seismic site response analysis when paired bedrock-surface records are available.

This study investigates the flexural behavior of steel-concrete composite beams under bending load conditions. Unlike the conventional research studies on H-shaped sections, this study introduces a novel built-up channel (BC) section, comprised of two channel sections connected to outer plates through bolted connections, which provides adaptability for site conditions and compares its flexural performance with the H-shaped section. Experimental tests were conducted on several members with a Linear Variable Differential Transformer (LVDT) and strain gauge attached to measure force-displacement and force-strain responses. Numerical analysis was performed using Abaqus, where the models were validated against the experimental data to closely reflect the real test conditions. A parametric study is conducted to analyze the impact of bolted and welded connections on the structural performance of the BC-specimen, along with damage assessment to evaluate failure patterns in compression and tension zones. The result shows that the addition of concrete significantly improved the stiffness and load capacity of each composite member. The results also indicate that the BC-specimen demonstrates better performance in terms of initial stiffness, key aspect of flexural loading. It is also observed that the welded connections are more efficient numerically and offer a slight improvement in post-yield performance compared to bolt connections. The study concluded that the BC-specimen, as an adaptive structural design, shows higher resistance to initial deformation, and ductile failure compared to the H-shaped section, making it a feasible alternative to conventional beams for structural applications such as temporary and permanent underground diaphragm walls.

This study aims to accurately identify periods of ground and lengthened period of multi-degree-of-freedom structures installed on shallow foundations in dynamic centrifuge model tests using the frequency domain decomposition (FDD) method. A series of dynamic centrifuge model tests were conducted with single-and multi-degree-of-freedom structures installed on shallow foundations and sandy ground. The frequency domain decomposition method was utilized to identify the ground period and the lengthened structure period. The results obtained from the FDD method for the soil-structure system, including periods and mode shapes, are compared with those calculated using conventional methods such as response spectrum (RS), ratio of response spectrum (RRS), fast Fourier transforms (FFT), and ratio of fast Fourier transforms (RFFT). The comparison demonstrates that FDD is effective in terms of both simplicity and accuracy for the modal analysis of complex soil-structure systems based on centrifuge test data.

The residual strength evaluation under a quasi-static loading system after a fire event in an electrical cabinet of an NPP is the main concern of this study. The capacity reduction, ductility displacement change, local joint behavior estimation, maximum displacement changes with corresponding base shear show the residual performance. The full-scale three-dimensional finite element model (FEM) is generated by considering the constitutive material model to adopt the fire event. Modal parameter estimation is taken into account to capture the dynamic property of the cabinet by the shake table test for comparing the fundamental frequencies, which leads to the calibration of the FEM. The quasi-s residual strength; fire-damaged FEM; capacity degradation; displacement ductility; local joint behavior tatic load has been applied under the fire event along with the gravity load to get the reduction pattern of the structural performance. The results analysis shows that the elevated temperature due to fire has a significant impact on the cabinet, on which the capacity reduction expresses an abrupt change after a specific temperature. This is also notified in the displacement ductility, joint behavior as well as maximum displacement change.

Rapid population growth is driving construction onto increasingly uneven terrains, where topography critically alters seismic site effects and soil-structure interaction (SSI). Despite evidence of slope-induced amplification and damage during past earthquakes, current seismic design codes lack clear provisions for buildings on irregular topography, and the role of earthquake frequency content remains unquantified. This study investigates how different seismic frequency characteristics influence the dynamic response of steel moment resisting frame (MRF) structures located on slope crests. A combined experimental-numerical approach was adopted: two geotechnical centrifuge tests A combined experimental-numerical approach was adopted: two geotechnical centrifuge tests were used to validate the numerical framework. Two prototype building models were then placed on the validated flat and sloped ground, and subjected to ground motions with varying frequency components. Responses were evaluated in terms of free-field response spectra, structural accelerations, displacements, base shear, and foundation rocking. Results demonstrate the correlation between the topographic frequency, SSI and earthquake frequency content, and their impacts on the seismic characteristics of buildings on sloped terrain that can aid the improvement of current seismic design provisions.

This study pioneers the application of deep learning-based approaches to predict nonlinear seismic time history responses of above-ground structures with shallow foundations considering soil-structure interaction (SSI). Long short-term memory (LSTM) and gated recurrent unit (GRU) models are selected for training. The training data are generated through numerical simulations validated by a centrifuge test. Nonlinear materials for both soil and structure, as well as nonlinear contact, are fully considered. The seismic response of the structure is calculated using two types of models: (i) an advanced SSI system, integrating soil and structure in one model, and (ii) a fixed-base structure, a simplified model commonly used in practice. A limited number of real earthquakes with a wide range of properties are chosen to demonstrate the predictive ability of the deep learning models. Hyper-parameters of the models are carefully investigated in terms of validation loss and training time. The effectiveness of the models is demonstrated through representations of time history responses and response spectra. The results are discussed and compared between the LSTM and GRU models, as well as between the SSI system and the fixed-base structure.

Evaluation of seismic risk by capturing the influences of strong motion duration and frequency contents of ground motion through probabilistic approaches is the main element of this study. Unlike most existing studies that mainly focus on intensity measures such as peak ground acceleration or spectral acceleration, this work highlights how duration and frequency characteristics critically influence dam response. To achieve this, a total of 45 ground motion records, categorized by strong motion duration (long, medium, and short) and frequency content (low, medium, and high), were selected from the PEER database. Nonlinear numerical dynamic analysis was performed by scaling each ground motion from 0.05 g to 0.5 g, with the drift ratio at the dam crest used as the Engineering Demand Parameter. It is revealed that long-duration and low-frequency ground motions induced significantly higher drift demands. The fragility analysis was conducted using a lognormal distribution considering extensive damage threshold drift ratio. Finally, the probabilistic seismic risk was carried out by integrating the site-specific hazard curve and fragility curves which yield the height risk for long durations and low frequencies. The outcomes emphasize the importance of ground motion strong duration and frequency in seismic performance and these findings can be utilized in the dam safety evaluation.

This study investigates the effects of adjacent deep excavation on the seismic performance of buildings. For that purpose, the numerical models are constructed for different buildings (i.e., 5-Story building and 15-Story building) considering the deep excavation-soil-structure interaction (ESSI) and soil-structure interaction (SSI). The results achieved from the ESSI and SSI systems are discussed and compared. Fully nonlinear numerical models with material, geometric, and contact nonlinearities are developed. Eleven earthquakes with different intensities, epicentral distances, significant durations, and frequency contents are applied to the models; and, the numerical results are given in terms of average records. The buildings are carefully designed and verified based on common design codes. The numerical modelling procedure of the deep excavation-soil system is validated using centrifuge test data. The comparisons between the ESSI and SSI systems are carried out in terms of accelerations, lateral displacements, inter-story drifts, story shear forces, and the nonlinear behavior of the soil medium under the buildings. The results show that it is necessary to consider the ESSI effect, and it might significantly change the seismic behavior of buildings adjacent to the deep excavations. The findings from this study can provide valuable recommendations for engineers to design buildings close to deep excavations under earthquakes.

The current study investigated the boundary effects induced by an Equivalent Shear Beam (ESB) container in a dynamic centrifuge test of a silica sand deposit. A fully nonlinear two-dimensional (2D) numerical model is established for the ESB container with soil by adopting nonlinear sand and hyperplastic rubber materials. A nonlinear frictional connection is adopted between the ESB and the soil. This study also proposes a numerical simulation procedure for ESB containers as boundaries for seismic analysis. A relationship between the fundamental frequency of the ESB container and rubber material properties is also proposed and provides a guide for the simulation of centrifuge tests with ESB containers. The established numerical model is validated via a dynamic centrifuge test. Investigation revealed that the boundary effects are minimal and are less than 20

Rapid urbanization has led to limited land availability, prompting the construction of residential and underground utility facilities in close proximity, forming an interconnected system with the surrounding soil. However, studies on the seismic behavior of such integrated systems are scarce due to their complex soil-structure interaction (SSI) mechanism. Furthermore, the dynamic behavior of these coupled structures depends heavily on earthquake characteristics like frequency content, an aspect that remains largely unexplored. This study aims to investigate the effects of earthquake frequency on the dynamic behavior of soil-underground structure-aboveground structure coupled systems. Using a novel experimental-numerical approach, two centrifuge tests were used to validate the numerical modeling framework. Three cases (soil-building, soil-tunnel, and soil-tunnel-building coupled systems) are then analyzed under earthquakes with varying frequency content. Results demonstrate how seismic response and SSI indicators of the structures evolve with changing seismic frequencies, offering valuable insights for enhancing safety measures in urban megacities.

The mutual behavior between the response of structure and soil, named soil-structure interaction (SSI) may become significantly important to be considered when structures are founded on deformable soil. In particular, critical infrastructures like nuclear power plants (NPPs) are particularly vulnerable to the consequences associated with the failure of any structural components because of the mutual interaction between the structure and the soil. In addition, soil mechanisms under strong earthquakes are highly non-linear and this reflects the discordance among the researchers for the selection of the most representative analysis model of nuclear structures. In this article, the role of SSI is assessed by presenting a critical discussion of the contemporary methods performed for the seismic analyses of NPPs. In particular, the paper proposes a discussion of the key issues that might facilitate the overall understanding of SSI phenomenon in case of NPPs.