Application of pool-based active learning in reducing the number of required response history analyses

Kiani, Jalal; Camp, Charles V.; Pezeshk, Shahram; Khoshnevis, Naeem

doi:10.1016/j.compstruc.2020.106355

Cited by 25 publications

(13 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In mid-and farfuture, it will be reasonable to try for improving the SEB THAAT, especially to overcome the third challenge addressed in Section 3. This seems possible, either directly with attention to the very details of the technique [1], the integration methods, or even by combining the SEB THAAT with other techniques, such as those proposed in [5][6][7]14]. Furthermore, and in view of the mathematical basis of the SEB THAAT [1,13], this technique can be tested in non-seismic problems, where the excitation is not originated in earthquakes.…”

Section: A Look At the Futurementioning

confidence: 99%

“…Some main approaches to lessen the computational effort are: (1) reducing the structural systems by replacing the structures finite element models with models with less degrees of freedom [3,4], (2) reducing the number of essential earthquake records, e.g. see [5][6][7], (3) reducing the number of oscillatory modes [8,9], and (4) using higher order time integration methods [10][11][12]. Meanwhile, in the last two decades, approaches are developed to reduce the computational effort by reducing the earthquake records' data [13][14][15].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Technique for Time Integration With Steps Larger Than the Excitation Steps: Review of the Past and Addressing the Existing Challenges and a Perspective of the Future

Soroushian¹

2021

Proceedings of the 8th International Conference on Computational Methods in Structural Dynamics and Earthquake Engineering (COM

View full text Add to dashboard Cite

Time history analysis using a time integration method is a powerful broadly accepted tool for structural dynamic analysis. In practical areas, such as earthquake engineering, computations based on time history analysis might be computationally very expensive. In 2008, a technique was proposed to considerably reduce the computational effort by reducing the data in the earthquake record. This technique, which is recently named as a SEB THAAT, is reviewed in this paper. After a brief review on the formulation and implementation, the positive points, the limitations, and the challenges facing versatile implementation of the SEB THAAT are addressed, and a future perspective is presented.

show abstract

Section: A Look At the Futurementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Technique for Time Integration With Steps Larger Than the Excitation Steps: Review of the Past and Addressing the Existing Challenges and a Perspective of the Future

Soroushian¹

2021

Proceedings of the 8th International Conference on Computational Methods in Structural Dynamics and Earthquake Engineering (COM

View full text Add to dashboard Cite

show abstract

“…Active learning has been implemented in bridge archetype selection for surrogate model development in the context of single‐scenario regional seismic risk assessment (Mangalathu & Jeon, 2020). Active learning has also been used for ground motion record selection, also within the context of surrogate models (Kiani, Camp, Pezeshk, & Khoshnevis, 2020). Online machine learning is another learning procedure where the goal is to update the features and model as new data become available.…”

Section: Introductionmentioning

confidence: 99%

Active learning method for risk assessment of distributed infrastructure systems

Tomar

Burton

2021

Computer aided Civil Eng

View full text Add to dashboard Cite

Event‐based methods are commonly used to assess the risk to distributed infrastructure systems. Stochastic event‐based methods consider all hazard scenarios that could adversely impact the infrastructure and their associated rates of occurrence. However, in many cases, such a comprehensive consideration of the spectrum of possible events requires high computational effort. This study presents an active learning method for selecting a subset of hazard scenarios for infrastructure risk assessment. Active learning enables the efficient training of a Gaussian process predictive model by choosing the data from which it learns. The method is illustrated with a case study of the Napa water distribution system where a risk‐based assessment of the post‐earthquake functional loss and recovery is performed. A subset of earthquake scenarios is sequentially selected using a variance reduction stopping criterion. The full probability distribution and annual exceedance curves of the network performance metrics are shown to be reasonably estimated.

show abstract

“…In [22], the authors used nonlinear regression analysis techniques for accelerating seismic analysis. In [23], the authors developed a pool-based active learning algorithm to choose partial data from ground motion records to meet informativeness, representativeness, and diversity criteria. They applied this method to reduce the computation burden in finite element analyses.…”

Section: Introductionmentioning

confidence: 99%

Neural Networks and Imbalanced Learning for Data-Driven Scientific Computing With Uncertainties

Pourkamali-Anaraki

Hariri-Ardebili

2021

IEEE Access

View full text Add to dashboard Cite

Uncertainty quantification in complex engineering problems is challenging because of necessitating large numbers of expensive model evaluations. This paper proposes a two-stage framework for developing accurate machine learning-based surrogate models in structural engineering. The studied numerical model considers aleatory and epistemic uncertainties, i.e., ground motion features and material properties. Our framework's first step trains classification algorithms on the collected data from our numerical model with a disproportionate ratio of observations from two categories, i.e., failed and safe simulations. We investigate the performance of imbalanced learning strategies along with artificial neural networks to achieve high classification accuracy. The second step of our framework aims to estimate three quantities of interest using the same network architecture, comparing our approach with regularized linear regression models. Moreover, we present a new approach to reducing the number of numerical simulations for developing machine learning-based surrogate models with limited training data. This approach employs Gaussian processes as a powerful probabilistic technique, providing an inherent uncertainty measure to determine the quality of estimated response values. Extensive numerical experiments demonstrate the superior performance of neural networks with three hidden layers compared to traditional machine learning algorithms for both classification and regression tasks. Also, empirical investigations corroborate that Gaussian processes enable us to predict the values of missing simulations for reducing the computational cost associated with numerical models. To conclude this work, we present several applications and future research directions. INDEX TERMS Uncertainty, Gaussian processes, neural networks, imbalanced classification, regression.

show abstract

Application of pool-based active learning in reducing the number of required response history analyses

Cited by 25 publications

References 29 publications

A Technique for Time Integration With Steps Larger Than the Excitation Steps: Review of the Past and Addressing the Existing Challenges and a Perspective of the Future

A Technique for Time Integration With Steps Larger Than the Excitation Steps: Review of the Past and Addressing the Existing Challenges and a Perspective of the Future

Active learning method for risk assessment of distributed infrastructure systems

Neural Networks and Imbalanced Learning for Data-Driven Scientific Computing With Uncertainties

Contact Info

Product

Resources

About