Deep learning techniques for predicting nonlinear multi-component seismic responses of structural buildings

Torky, Ahmed A.; Ohno, Susumu

doi:10.1016/j.compstruc.2021.106570

Cited by 45 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other approaches have constructed a KD tree organized by nearest neighbors (NNs) which reduces the search time in a large data set during ongoing earthquake records [ 70 ]. Deep learning techniques (ConvLSTM) have also been used to indirectly predict the behavior of seismic events with time–frequency analysis of the acceleration response and filtering of time series data with discrete wavelet transform [ 66 ]. All of these efforts can reduce the warning delivery time during a seismic event.…”

Section: Resultsmentioning

confidence: 99%

“…In this literature review, we mentioned that several studies applied new methodologies such as machine learning [ 31 , 41 ], deep learning [ 49 , 66 , 72 ], convolutional neural networks [ 36 , 61 ], among others. Mainly, they have focused on the processing and post-processing of feature data and the interpretation of seismic signals.…”

Section: Discussionmentioning

confidence: 99%

“…The database was focused on values of peak ground acceleration, velocity, and displacement (PGA, PGV, and PGD), instead of common parameters such as hypocenter distance. Torky et al [ 66 ] used hybrid convolutional neural network (ConvLSTM) techniques to indirectly predict seismic response of mid-rise and high-rise buildings in earthquake-prone areas, and to assist earthquake early warning systems. They used accelerometer mesh with a sampling frequency of 10-20-100 Hz for torsional vibration and waveform detection.…”

Section: Methodology Validationmentioning

confidence: 99%

See 2 more Smart Citations

Analysis of Information Availability for Seismic and Volcanic Monitoring Systems: A Review

Arrais

Urquiza-Aguiar

Tripp-Barba

2022

Sensors

View full text Add to dashboard Cite

Organizations responsible for seismic and volcanic monitoring worldwide mainly gather information from instrumental networks composed of specialized sensors, data-loggers, and transmission equipment. This information must be available in seismological data centers to improve early warning diffusion. Furthermore, this information is necessary for research purposes to improve the understanding of the phenomena. However, the acquisition data systems could have some information gaps due to unstable connections with instrumental networks and repeater nodes or exceeded waiting times in data acquisition processes. In this work, we performed a systematic review around information availability issues and solutions in data acquisition systems, instrumental networks, and their interplay with transmission media for seismic and volcanic monitoring. Based on the SLR methodology proposed by Kitchenham, B., a search string strategy was considered where 1938 articles were found until December 2021. Subsequently, through selection processes, 282 articles were obtained and 51 relevant articles were extracted using filters based on the content of articles mainly referring to seismic–volcanic data acquisition, data formats, monitoring networks, and early warnings. As a result, we identified two independent partial solutions that could complement each other. One focused on extracting information in the acquisition systems corresponding to continuous data generated by the monitoring points through the development of mechanisms for identifying sequential files. The other solution focused on the detection and assessment of the alternative transmission media capabilities available in the seismic–volcanic monitoring network. Moreover, we point out the advantage of a unified solution by identifying data files/plots corresponding to information gaps. These could be recovered through alternate/backup transmission channels to the monitoring points to improve the availability of the information that contributes to real-time access to information from seismic–volcanic monitoring networks, which speeds up data recovery processes.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Methodology Validationmentioning

confidence: 99%

See 1 more Smart Citation

Analysis of Information Availability for Seismic and Volcanic Monitoring Systems: A Review

Arrais

Urquiza-Aguiar

Tripp-Barba

2022

Sensors

View full text Add to dashboard Cite

show abstract

“…[8][9][10] The direct relevance to real-time data synchronization between physical spaces and dynamic models in DL based algorithms offers the benefit of improved decision supports and automatic planning and maintenance, such as system control, prediction, diagnose/assessment and so forth. Some pioneered previous work includes constructing nonlinear structural response model through long short-term memory (LSTM) for building structures using numerical and experimental data 11 as well as real sensed seismic data, 12 approximating governing equations by means of Neural Ordinary Differential Equations (Neural ODEs), 13 modeling structural complex hysteretic behaviors based on Transformer network architecture, 14 and employing generalized linear regression approach to uncover the full parametric form of damping behaviors. 15 Specifically, considerable progress has been made in physics-informed learning strategy for modeling and prediction of dynamics.…”

Section: Introductionmentioning

confidence: 99%

Combination of physics‐based and data‐driven modeling for nonlinear structural seismic response prediction through deep residual learning

Guo

Enokida

et al. 2023

Earthq Engng Struct Dyn

View full text Add to dashboard Cite

Despite great progress in seeking accurate numerical approximator to nonlinear structural seismic response prediction using deep learning approaches, tedious training process and large volume of structural response data under earthquakes for training and validation are often prohibitively accessible. In our methodology, the main innovation can be seen in the incorporation of deep neural networks (DNNs) into a classical numerical integration method by using a hybridized integration time-stepper. In this way, the linear physics information of the structure and the obscure nonlinear dynamics are smoothly combined. We propose to use residual network (ResNet) to learn time-stepping schemes specifically for the nonlinear state variables of the system. Our Physics-DNN hybridized integration (PDHI) time-stepping scheme provides important advantages over current pure data-driven approaches, including (i) a flexible framework incorporating known time-invariant physics information, (ii) requirement of structural seismic response data being circumvented by simple short bursts of trajectories collected from underlying nonlinear components, and (iii) efficiency in training and validation process. Besides, our results indicate that a simple feedforward or convolutional architecture outperforms recurrent networks to fulfill the requirement of prediction accuracy as well as long-range memory in structural dynamic analysis. Several numerical and experimental examples are presented to demonstrate the performance of the method.

show abstract

“…Over the last few years, artificial intelligence has gotten a lot of attention in the field of civil engineering. Regarding structural and earthquake engineering, research on machine learning can be divided into three areas including structural health monitoring [1,2], performance assessment of buildings [3][4][5][6][7], and prediction models of the mechanical behavior [8][9][10][11][12]. Estimating the response and predicting the behavior of structural members is one of the applications of machine learning techniques in civil engineering.…”

Section: Introductionmentioning

confidence: 99%

Efficiency of Hybrid Algorithms for Estimating the Shear Strength of Deep Reinforced Concrete Beams

Barkhordari

Feng

Tehranizadeh

2022

Period. Polytech. Civil Eng.

View full text Add to dashboard Cite

Earthquakes occurred in recent years have highlighted the need to examine the strength of reinforced concrete (RC) members. RC beams are one of the elements of reinforced concrete structures. Due to the dramatic increase in the population and the number of medium/high-rise buildings, in recent years, the beams of buildings have been mainly designed and executed in the type of deep beams. In this study, the artificial neural network (ANN) with optimization algorithms, including particle swarm optimization (PSO), Archimedes optimization algorithm (AOA), and sparrow search algorithm (SSA), are used to determine the shear strength of reinforced concrete deep (RCD) beams. 271 samples from experimental tests are employed to develop algorithms. The results of this study, design codes equations, and previous research are compared. Comparison between the results shows that the PSO-ANN algorithm is more accurate than previous methods. Finally, SHApley Additive exPlanations (SHAP) method is utilized to explain the predictions. SHAP reveals that the beam span and the ratio of the beam span to beam depth have the highest impact in predicting shear strength.

show abstract

Deep learning techniques for predicting nonlinear multi-component seismic responses of structural buildings

Cited by 45 publications

References 17 publications

Analysis of Information Availability for Seismic and Volcanic Monitoring Systems: A Review

Analysis of Information Availability for Seismic and Volcanic Monitoring Systems: A Review

Combination of physics‐based and data‐driven modeling for nonlinear structural seismic response prediction through deep residual learning

Efficiency of Hybrid Algorithms for Estimating the Shear Strength of Deep Reinforced Concrete Beams

Contact Info

Product

Resources

About