A novel machine learning framework for efficient calibration of complex DEM model: A case study of a conglomerate sample

Shentu, Junjie; Lin, Bor-Luh

doi:10.1016/j.engfracmech.2023.109044

Cited by 10 publications

(5 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sun 61 and Zhai 62 , for example, leveraged macroscopic and microscopic parameters obtained through experimental design to train their neural network, which was found to exhibit low verification error when compared against experimental data. Meanwhile, Chen 47 and Shentu 63 employed algorithms such as random forests and support vector machines to train models for macroscopic and microscopic parameters, and compared the accuracy of each method. Additionally, Chen and others [64][65][66][67] utilized a filtering framework for calibration which accounted for the influence of load history on the elastic-plastic behavior of granular materials, resulting in a more precise model.…”

Section: Machine Learning Methodsmentioning

confidence: 99%

A Review of Calibration Strategies for Microscopic Parameters in Discrete Element Method

Liu

Wang

et al. 2023

Preprint

View full text Add to dashboard Cite

Discrete element method (DEM) offers a distinct advantage in addressing fracture problems due to its exemption from partial differential equations. However, the governing equations of DEM incorporate microscopic parameters that cannot be accessed via experimental methods. Parameter calibration represents the most prevalent technique currently utilized to obtain the required microscopic parameters for DEM. This article initially introduces the development history of DEM calibration, then elaborates on the governing equations of DEM and the microscopic parameters required for calibration. Next, the principles of classical DEM calibration methods are explicated in detail, alongside the validation and discussion of their properties. Lastly, this study investigates the model dependence of DEM and presents guidance for the calibration application of large-scale simulations.

show abstract

Section: Machine Learning Methodsmentioning

confidence: 99%

A Review of Calibration Strategies for Microscopic Parameters in Discrete Element Method

Liu

Wang

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Compared to the design of experiments method, machine learning methods can more efficiently handle high-dimensional and nonlinear issues. Various machine learning methods have been utilized by researchers to address parameter calibration issues, including random forest method 44 , 56 , support vector machine method 44 , 56 , Bayesian filtering method 110 – 113 , and neural network method 30 , 42 , 61 , 95 , 114 .…”

Section: Review Of Calibration Strategiesmentioning

confidence: 99%

“…Shentu employed the random forest algorithm for the calibration of the discrete element method and observed promising precision. The calibration process involved the utilization of 500 data sets 44 . This method has the possibility of overfitting issues when a small amount of data is available.…”

Section: Review Of Calibration Strategiesmentioning

confidence: 99%

“…The notable advantage of this method lies in its robustness when it encounters limited samples. Shentu et al has applied the support vector machine algorithm for the calibration of the discrete element method with 500 data, observing favorable precision in calibration results 44 . However, this method is susceptible to noise and overlapping data.…”

Section: Review Of Calibration Strategiesmentioning

confidence: 99%

“…In 2011, the application of neural network training was utilized in DEM calibration 41 . Now various machine learning methods have been explored for parameter calibration 42 – 44 . In 2019, Qu made a contribution by presenting a theoretical derivation for DEM calibration 45 .…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Review of calibration strategies for discrete element model in quasi-static elastic deformation

Liu

Wang

et al. 2023

Sci Rep

View full text Add to dashboard Cite

This study first reviewed theories of the mechanical response of structures under loading, and the discrete element method provides a route for studying mechanical response including elastic deformation and structure failure. However, the direct acquisition of the microscopic parameters from the governing equations of the discrete element method via experiments encounters challenges. One possible strategy to obtain these microscopic parameters is parameter calibration that are widely used by researchers. Secondly, the governing equations and failure criterion of the discrete element method are summarized, and the microscopic parameters that would be calibrated are pinpointed. Next, the principles of classical calibration methods of discrete element method are explicated in detail, alongside the validation and discussion of their properties. Lastly, this study examined the applicability of calibrated parameters and points out that the size ratio, porosity, maximum radius, and minimum radius of particles should be identical in both the geometric calibration model and that for applications.

show abstract

Investigation of hydraulic fracture propagation in conglomerate rock using discrete element method and explainable machine learning framework

Shentu,

Lin,

Jin

et al. 2024

Acta Geotech.

View full text Add to dashboard Cite

A novel machine learning framework for efficient calibration of complex DEM model: A case study of a conglomerate sample

Cited by 10 publications

References 40 publications

A Review of Calibration Strategies for Microscopic Parameters in Discrete Element Method

A Review of Calibration Strategies for Microscopic Parameters in Discrete Element Method

Review of calibration strategies for discrete element model in quasi-static elastic deformation

Investigation of hydraulic fracture propagation in conglomerate rock using discrete element method and explainable machine learning framework

Contact Info

Product

Resources

About