Prediction Of Seismic Response Research Articles

Machine learning-based prediction of seismic response of elevated steel tanks

Seismic response analysis assesses structural safety and integrity during earthquakes and guides design for optimal performance, compliance with regulations, and risk mitigation. This study proposes a novel approach utilizing machine learning (ML) techniques to predict the seismic response of elevated steel tanks accurately. ML is a powerful tool that can handle complexity, improve accuracy, provide data-driven insights, and optimize designs for predicting seismic responses of structures. Moreover, this study evaluates various ML models containing Random Forest, Support Vector Regression, Adaptive Boosting, LightGBM, Histogram-based Gradient Boosting, XGBoost, CatBoost, Bagging Regression, and Decision Trees to identify the most effective predictive model. In general, the developed ML-based predictive models provide a promising tool for engineers involved in the seismic design and retrofitting of elevated steel tanks, aiding in optimizing designs and enhancing structural safety and resilience against seismic hazards. As a result, the HGB model obtained the most suitable performance compared to other developed models with higher R2 and lower RMSE. In addition, a sensitivity analysis was performed based on SHAP values, and it was found that the height of the liquid and the intensity measure were the most influential variables in the seismic response of the tank.

Pulse-type seismic response prediction for uncertain structures based on the attention mechanism and wavelet decomposition augmented decomposition learning

Accurate prediction of structural responses under seismic action is important for the performance assessment of structures. However, depending on the variability of the structure (e.g., the design information) or ground motion characteristics (e.g., the presence of pulses), the seismic response characteristics of the may be significantly different. To overcome this problem, this study proposes a novel structural feature attention mechanism and wavelet decomposition augmented decomposition learning method (SFA-WD-DL) for predicting the time-series of the structural response under pulse-type ground motions. Unlike previous studies, the proposed method can be applied to structures with different shapes. Innovatively, a structural feature attention mechanism is embedded within a neural network model, enabling the consideration of different structural parameters. In addition, a wavelet decomposition-based velocity pulse identification method is combined with decomposition learning, using decomposed pulses and high-frequency features as inputs to the neural network model. This process simplifies pulse feature recognition task of the model. To avoid an unbalanced distribution of ground motion features affecting model performance, a balanced sampling strategy for ground motion datasets combining ground motion clustering and category reorganization methods is proposed, which effectively improves the model robustness. The accuracy and applicability of the proposed method are validated using a dataset containing reinforced concrete frame structures of different shapes and a pulse-type ground motion dataset containing 148 records. In addition, the importance levels of the structural features are obtained through the weight distribution of the structural feature attention mechanism. This indicates that the proposed SFA-WD-DL method is easily interpretable and can extract key features affecting the structural response.

Seismic response prediction of composite plate shear walls- concrete filled (C-PSW/CF) using machine learning methods

In the seismic design of composite plate shear walls filled with concrete, referred to as C-PSW/CF, cyclic backbone curves provide invaluable information for effectively withstanding seismic disturbances. The curves evaluate lateral strength, ductility, and energy dissipation capabilities, and guide performance-based design strategies. This study harnesses machine learning (ML) techniques to predict critical parameters of the cyclic backbone curve of C-PSWs/CF. Various types of existing C-PSWs/CF from the literature were categorized for this analysis. ML models, comprising Random Forests (RF), Support Vector Regression (SVR), eXtreme Gradient Boosting (XGB), and Gaussian Process Regression (GPR), were developed to facilitate the predictive tasks. The findings indicate that all models demonstrate robust performance in predicting maximum strength and ultimate strength, yielding R2 scores close to 1.00 for training and testing, and exceeding 0.90 for cross-validation. Additionally, RF, SVR, and XGB exhibit commendable accuracy in predicting yield strength and its corresponding drift ratio.

Using image-based inspection data to improve response predictions of earthquake-damaged unreinforced masonry buildings

AbstractExplicit representation of uncertainties is essential to improve the reliability of seismic assessments of earthquake-damaged buildings, particularly when dealing with unreinforced masonry buildings. Modern inspection techniques use images for detecting and quantifying the damage to a structure. Based on the principle of falsification, this paper evaluates how the use of information of damage that is obtained from images taken on earthquake-damaged buildings reduces the uncertainty when predicting the seismic response under a future earthquake. New model falsification criteria use information on the residual state of a building, such as shear cracks, residual roof displacements, and observation of out-of-plane failure. To demonstrate the effectiveness of these criteria in reducing the uncertainty in response predictions, results from a four-story unreinforced masonry building stiffened with reinforced concrete walls, which was experimentally tested under a sequence of ground motions, are assessed. Three commonly used modeling approaches (single degree of freedom (DOF) systems, multi DOF systems with four DOFs, and equivalent frame models) are used, where uncertainties in model parameters and model bias are included and propagated through the analysis. Out of the models used, and in the absence of any additional source of information, the proposed falsification criteria are most effective in connection with the equivalent frame model because this model can simulate the response at the element-level, while the simpler models can only represent the global response or the response at the storey-level. The results show that when using only the information on the presence of shear cracks, which might be the first and only source of information after an earthquake, the effectiveness of model falsification is increased, thus reducing the uncertainty in model parameter values and seismic response predictions through the use of image-based inspection.

Seismic response prediction for buildings with sliding live loads using neural networks

The seismic response of primary structures (PS) can be significantly influenced by live loads, particularly when these loads consist of stacked bodies that are capable of sliding during seismic events. Conventional design methods often fail to account for the energy dissipation resulting from such sliding, leading to overly conservative structural estimates. This study examines the effects of sliding live loads on the seismic response of a PS, where two rigid bodies, referred to as secondary bodies (SBs), are stacked as live loads on a PS. The interaction between the lower SB and the PS, as well as between the SBs, is modelled using Coulomb's friction model. The governing equations are solved using the fourth-order Runge-Kutta method. Seismic Zones III and V from IS 1893:2016 are considered as hazard levels. A parametric investigation explores the impact of varying dynamic parameters of both the PS and SBs on the seismic response. The results demonstrate significant energy dissipation within the stack due to sliding, which, if not considered, results in conservative displacement estimates. The study also evaluates how friction coefficients, mass ratios, and excitation levels affect the stack's energy dissipation capacity. A novel method to compute the modified primary structural period (Tnew) for design purposes is proposed, and an artificial neural network (ANN) model is developed, achieving a high prediction accuracy (R2 = 99 %). Sensitivity analysis is performed to assess the influence of input parameters on the output. This research highlights the need to include sliding loads in seismic design.

Seismic response prediction method of train-bridge coupled system based on convolutional neural network-bidirectional long short-term memory-attention modeling

Seismic response prediction is crucial for the safety analysis of train-bridge coupled systems. However, due to the complexity, suddenness, and high-risk nature of earthquakes, there are strong nonlinear relationships among different parts of bridges, making it challenging to express their spatial correlations using analytical models and traditional neural networks. To address this, this paper establishes a ballast track shaker scaling model and employs the grating monitoring measurement method to construct a spatial quasi-distributed monitoring system for the ballast track, thereby collecting seismic strain responses of the train-bridge coupled system under various seismic conditions. A hybrid neural network method is proposed for predicting the seismic responses of the train-bridge coupled system. This hybrid neural network integrates the features of a Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory Neural Network (BiLSTM), and the attention mechanism, thereby termed the CNN-BiLSTM-attention hybrid neural network. The model was validated using strain responses from 54 seismic scenarios. The results indicate that the model has a Mean Absolute Error (MAE) of 0.2349 and a coefficient of determination (R2) of 0.9446. Comparing the prediction results with those from RNN and LSTM models, it was found that the CNN effectively extracts features under various seismic parameters, while the BiLSTM better captures the temporal information of the strain responses, ensuring effective prediction regardless of the magnitude of strain responses. Therefore, the CNN-BiLSTM-attention hybrid neural network model is recommended for predicting seismic response.

An efficient method based on shear models for structural seismic response prediction considering hysteretic characteristics

An efficient method based on shear models for structural seismic response prediction considering hysteretic characteristics

A nonlinear structural pulse-like seismic response prediction method based on pulse-like identification and decomposition learning

Abstract Accurate prediction of structural responses under seismic action is important for the performance evaluation of structures. However, depending on the differences in the characteristics of ground motions, particularly whether they contain impulses, the seismic-response characteristics of structures may differ significantly. Hence, a new method, i.e., WD–AttDL, is proposed in this study to predict structural response under pulse-like ground motions. This method innovatively combines a wavelet decomposition-based velocity pulse-identification method with decomposition learning, where decomposed pulses and high-frequency features are used as inputs to the neural-network model, thus simplifying the identification of pulse features for the model. The decomposition learning module integrates different types of submodules, such as CNN feature-extraction, LSTM temporal-learning, and self-attention mechanism submodules. To reduce the nonlinear time-range analysis calculations required to generate training data, a screening method for impulsive ground motions based on time-series feature clustering is proposed. The accuracy and applicability of the proposed method are verified by numerically simulating three different cycles of reinforced-concrete frame structures and a three-dimensional masonry structure. Compared with existing structural seismic-response models, WD–AttDL synergistically integrates different modules and thus offers a higher prediction accuracy. In particular, it reduces the peak error of the predicted response, which is important for the evaluation of structural performance. Additionally, based on the response predicted by WD–AttDL, a vulnerability analysis of the structure under pulse-like seismic effects can be performed rapidly and accurately.

Probabilistic Prediction on Seismic Responses of a Long-Span High-Speed Railway Bridge Using BPINN

Efficient probabilistic prediction of seismic responses is crucial for assessing the seismic-resistant capability of long-span continuous girder high-speed railway bridges. Thus, a Bayesian physics-informed neural network (BPINN) is adopted to rapidly and effectively predict the probabilistic seismic responses of such bridges. The BPINN model combines deep learning and physics to improve the accuracy and consistency of predictions, while also quantifying the uncertainties using Bayesian inference methods. Various seismic excitations, including pulse, far-field and near-field types, are employed to probabilistically predict the seismic responses of the top of bridge pier. Evaluation metrics, including mean squared error and prediction interval coverage probability, are used to assess the deterministic and probabilistic estimates of BPINN. Results demonstrate that BPINN performs better in deterministic results for far-/near-field ground motions compared to pulse-like earthquakes, with most cases exhibiting a close approximation to the 95% confidence interval. The flexibility and adaptability of BPINN in handling different types of ground motions can provide valuable insights for assessing the seismic performance of such structures.

Practice-oriented numerical model for seismic response of cold-formed steel-framed mid-rise buildings

Cold-formed steel (CFS) framed buildings are becoming increasingly common, notably in the high seismic regions of the United States (US). However, research to understand the seismic performance of CFS-framed buildings specifically using large-scale (building-level) specimens has begun only recently. Therefore, efforts to validate and improve numerical modeling of whole building systems have been limited due to this lack of experimental data. A recently completed experimental program, conducted at the Large High-Performance Outdoor Shake Table at the University of California, San Diego, involved earthquake testing of a full-scale 6-story CFS-framed building. The test building adopted an assembly of continuous tie-rod and built-up compression stud pack at the ends of individual shear walls to resist seismic overturning moment and uplift forces within the building. The objective of this paper is to present a practice-oriented phenomenological finite element modeling approach intended for design engineers and compare the predictions from this model against the measured seismic response of the 6-story building. The model aims to provide a robust, yet computationally efficient approach, with minimal degrees-of-freedom, while also capturing the salient features of seismic response of mid-rise CFS-framed buildings suitable for nonlinear time history analyses-based design. The building model, developed within the OpenSeesPy finite element platform, includes the shear walls and gravity walls as well as the lateral strength and stiffness contribution from exterior and interior gypsum sheathing. The continuous tie-rod system is also modeled to capture the cumulative floor displacements caused by the axial elongation in the steel rods. The effect of built-up compression stud packs on the lateral behavior of a shear wall was estimated using available experimental data and a detailed numerical model of an isolated shear wall, then subsequently incorporated in the nonlinear hysteretic material model. The numerical model captured the natural frequencies of the translation modes of the building in its undamaged state with <3% error. It also predicted the peak roof drift ratio within an error of 0.03% drift ratio under the design earthquake event. These results provide confidence in the proposed numerical modeling approach as a tool for seismic response prediction of mid-rise CFS-framed buildings with similar structural detailing features.

Deep Learning Frequency Loss Functions for Predicting the Seismic Responses of Structures

Novel loss functions for seismic response prediction, such as physics-informed loss functions, have attracted considerable attention. However, existing loss functions are based on the time domain and do not consider the frequency domain differences contained in structure response signals. This study accordingly designed three loss functions based on frequency domain information — one pure frequency domain loss function and two combined frequency- and time-domain loss functions — using Fourier transforms. The proposed loss functions were applied using a trusted publicly available dataset for a frame structure, and their performances were compared with those obtained using the conventional mean squared error (MSE) loss function. The proposed frequency domain loss function exhibited superior performance, accelerating the convergence of learning curves for long-period signals while simultaneously enhancing their frequency domain accuracy and facilitating the prediction of high-sampling-rate response signals. In addition, the designed loss functions exhibited more stable performances and superior accuracy than the MSE when random dataset partitioning was employed. Finally, the combined time- and frequency-domain loss functions were shown to predict the seismic displacement time-history of tall, long-span bridges more accurately than the MSE by training the model to predict structural acceleration responses, then integrating them to predict displacement responses.

Time history seismic response prediction of multiple homogeneous building structures using only one deep learning‐based Structure Temporal Fusion Network

AbstractStructural response prediction under earthquakes is crucial for evaluating the structural performance and subsequent functional restoration. Deep learning provides the potential to rapidly obtain the responses by skipping the time‐consuming nonlinear finite element analysis. However, a single deep learning network may only predict the time history responses of one specific structure, resulting in redundancy and resource waste when building multiple networks for modeling different structures. Thus, this study proposes a Structure Temporal Fusion Network (STFN) that can predict responses of various homogeneous structures using a single network. The key concept is that the seismic waves and the structural characteristics, such as story numbers, are fused together to predict diverse time history responses. Two numeric experiments are conducted, including predicting responses of ideal single‐degree‐of‐freedom (SDOF) structures and regular multistory reinforced concrete frames. Furthermore, a series of ablation analyses are carried out to validate the network architecture. The results indicate that STFN can predict nonlinear time history responses of different structures with mean square errors in the magnitude of and for two experiments, respectively. The solutions also highlight the importance of fusing static characteristics for the modeling of various structures with only one network. The STFN presents a promising solution for time history response prediction across multiple structures in regions.

Structural nonlinear seismic time-history response prediction of urban-scale reinforced concrete frames based on deep learning

Efficiently predicting the seismic response of urban building clusters is essential for preemptively identifying potential seismic hazards prior to an earthquake and optimizing resource allocation post-event. However, complete information of buildings at a city scale is generally un-accessible or non-existent. Existing methods struggle to reconcile low information demands, high computational accuracy, and computational efficiency. This paper proposes a fast prediction method for structural seismic time-history responses that combines deep learning methods with easy-getting structural parameters at an urban scale. An end-to-end network with adaptive multilevel fusion output is designed, which incorporates the autoencoder concept for predicting the structural seismic time-history responses based on ground motions records and five easy-getting structural parameters. The models are compared and optimized considering the training hyperparameters and network architecture, resulting in an optimized model with low complexity that provides valuable reference values for structural seismic response. Besides, the proposed model is applied to four actual buildings with different construction time, occupancy types, and floor sizes, demonstrating its good prediction performance and significant computational advantages comparing to the universally used MDOF method.

Adaptive GN block-based model for seismic response prediction of train-bridge coupled systems

Deep learning (DL) methods has found extensive application in predicting structural responses. However, conventional neural networks often lack the capacity to effectively capture inter-data relationships, relying solely on training data for end-to-end prediction. Consequently, these methods may struggle with generalization, particularly when faced with new structural configurations absent from the training dataset. To overcome this limitation, this study adopts a graph-based framework, representing structures as graphs facilitates the transfer of features between different components, mimicking the transfer of forces in the real world. Subsequently, a Graph Neural Network Blocks (GN Blocks) model is introduced. Leveraging the holistic inference structure inherent in graph representations, the GN Block-based model as an adaptive model exhibits superior generalization performance in response analysis tasks involving unknown structural forms. Illustrated through the application to train-bridge coupled (TBC) systems, the GN Block-based model demonstrates high prediction accuracy and robust generalization performance, underscoring its suitability for tackling complex structural response prediction.

Seismic Response Prediction of Porcelain Transformer Bushing Using Hybrid Metaheuristic and Machine Learning Techniques: A Comparative Study

Although seismic response predictions are widely used for engineering structures, their applications in electrical equipment are rare. Overstressing at the bottom of the porcelain insulators during seismic events has made power transformer bushings in substations prone to failure. Thus, this paper proposed and compared six integrated machine learning (ML) models for seismic stress response predictions for porcelain transformer bushings using easily monitored acceleration responses. Metaheuristic algorithms such as particle swarm optimization were employed for architecture tuning. Prediction accuracies for stress response values and classifications were evaluated. Finally, shaking table tests and simulation analyses for a 1100 kV bushing were implemented to validate the accuracy of the six ML models. The results indicated that the proposed ML models can quickly forecast the maximum stress experienced by a porcelain bushing during earthquakes. Swarm intelligence evolutionary technologies could quickly and automatically aid in the retrofitting of architecture for the ML models. The K-nearest neighbor regression model had the best level of prediction accuracy among the six selected ML models for experimental and simulation validations. ML prediction models have clear benefits over frequently used seismic analytical techniques in terms of speed and accuracy for post-earthquake emergency relief in substations.

Seismic response prediction and fragility assessment of high-speed railway bridges using machine learning technology

Nowadays, reliable and accurate prediction of the seismic response of high-speed railway bridges and then rapid assessment of their seismic fragility is crucial for the risk assessment of regional high-speed railway bridges. However, the traditional seismic fragility assessment based on nonlinear time history analysis (NTHA) has a characteristic of high computational cost. Therefore, this study explores different machine learning (ML) methods to develop a reliable rapid seismic fragility assessment framework for high-speed railway bridges. Taking a typical high-speed railway continuous (HRC) bridge as a case, a large number of numerical model and ground motion pairs considering various structural and ground motion features are established. Then, the two types of models including six popular ML methods are investigated, in which Lasso regression, artificial neural network (ANN) and support vector machine (SVM) are single models, and random forest (RF), extreme gradient boosting (XGBoost) and light gradient boosting machine (Light GBM) are ensemble models. A database containing a large number of seismic responses of different components is established through the NTHA to develop prediction models, and the optimal hyperparameters of different ML models are determined by ten-fold cross-validation and grid search. The study results show that the ensemble model can reliably predict the seismic response of the HRC bridge compared with the single model, and the XGBoost model is recommended for seismic fragility assessment. On this basis, the optimal ML model obtained by the performance evaluation is applied to develop seismic fragility curves of the HRC bridge, which are compared with the fragility curves based on NTHA. The results indicate that the XGBoost model can effectively replace the traditional NTHA for rapid seismic fragility assessment of the HRC bridge, which can achieve a sufficient assessment level while saving calculation costs.

Seismic fragility analysis of RC frame structures based on IDA analysis and machine learning

This study introduces a machine learning-based methodology designed to rapidly predict the seismic responses of reinforced concrete (RC) frame structures. The research focuses on three types of RC frame structures: low-rise, multi-story, and small high-rise buildings. Ground shaking records are selected according to the conditional mean spectrum (CMS). A sample database, constructed via Incremental Dynamic Analysis (IDA), facilitates the prediction of structural responses using ground shaking intensity and structural details as inputs. Concurrently, the study performs a feature importance analysis of the model. Machine learning algorithms, including integrated learning and neural networks, are utilized to predict the seismic responses of the RC frame structures. This methodology also assists in evaluating the seismic fragility of these structures. The results show that the discrepancy between the neural network-based seismic fragility assessments and the IDA results is minimal, indicating a high degree of accuracy in the proposed methodology. Among the characteristic parameters, Average Spectral Acceleration (AvgSa) is identified as the most significant. This methodology serves as a valuable tool for the rapid prediction of seismic responses in RC frame structures, demonstrating substantial practical value.

Development of Machine Learning Based Seismic Response Prediction Model for Shear Wall Structure considering Aging Deteriorations

Development of Machine Learning Based Seismic Response Prediction Model for Shear Wall Structure considering Aging Deteriorations

Machine Learning based Seismic Response Prediction Methods for Steel Frame Structures

Machine Learning based Seismic Response Prediction Methods for Steel Frame Structures

Surrogate models for seismic and pushover response prediction of steel special moment resisting frames

For structural engineers, existing surrogate models of buildings present challenges due to inadequate datasets, exclusion of significant input variables impacting nonlinear building response, and failure to consider uncertainties associated with input parameters. Moreover, there are no surrogate models for the prediction of both pushover and nonlinear time history analysis (NLTHA) outputs. To overcome these challenges, the present study proposes a novel framework for surrogate modelling of steel structures, considering crucial structural factors impacting engineering demand parameters (EDPs). The first phase involves the development of a process by which 30,000 random steel special moment resisting frames (SMRFs) for low to high-rise buildings are generated, considering the material and geometrical uncertainties embedded in the design of structures. In the second phase, a surrogate model is developed to predict the seismic EDPs of SMRFs when exposed to various earthquake levels. This is accomplished by leveraging the results obtained from phase one. Moreover, separate surrogate models are developed for the prediction of SMRFs’ essential pushover parameters. Various machine learning (ML) methods are examined, and the outcomes are presented as user-friendly GUI tools. The findings highlighted the substantial influence of pushover parameters as well as beams and columns’ plastic hinges properties on the prediction of NLTHA, factors that have been overlooked in prior studies. Moreover, CatBoost has been acknowledged as the superior ML technique for predicting both pushover and NLTHA parameters for all buildings. This framework offers engineers the ability to estimate building responses without the necessity of conducting NLTHA, pushover, or even modal analysis which is computationally intensive.