Predictive modeling of dynamic fracture growth in brittle materials with machine learning

Moore, Bryan A.; Rougier, Esteban; O’Malley, Daniel; Srinivasan, G.; Hunter, Abigail; Viswanathan, Hari

doi:10.1016/j.commatsci.2018.01.056

Cited by 77 publications

(43 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the concept can be traced back to the works done by Fourier (1768-1830), there exist many recent monographs [232]- [239] and review articles [240]- [253]. In practice, ROMs (i.e., emulators) have great promise for applications especially when multiple forward full-order numerical simulations are required (e.g., data assimilation [254]- [259], parameter identification [260]- [263], uncertainty quantification [264]- [270], optimization and control [271]- [278]). These models can quickly capture the essential features of the phenomena taking place, and we often might not need to calculate all full order modeling details to meet real-time constraints.…”

Section: Nonintrusive Data-driven Modelingmentioning

confidence: 99%

Digital Twin: Values, Challenges and Enablers From a Modeling Perspective

2020

View full text Add to dashboard Cite

A digital twin can be defined as an adaptive model of a complex physical system. Recent advances in computational pipelines, multiphysics solvers, artificial intelligence, big data cybernetics, data processing and management tools bring the promise of digital twins and their impact on society closer to reality. Digital twinning is now an important and emerging trend in many applications. Also referred to as a computational megamodel, device shadow, mirrored system, avatar or a synchronized virtual prototype, there can be no doubt that a digital twin plays a transformative role not only in how we design and operate cyber-physical intelligent systems, but also in how we advance the modularity of multi-disciplinary systems to tackle fundamental barriers not addressed by the current, evolutionary modeling practices. In this work, we review the recent status of methodologies and techniques related to the construction of digital twins. Our aim is to provide a detailed coverage of the current challenges and enabling technologies along with recommendations and reflections for various stakeholders.

show abstract

Section: Nonintrusive Data-driven Modelingmentioning

confidence: 99%

Digital Twin: Values, Challenges and Enablers From a Modeling Perspective

2020

View full text Add to dashboard Cite

show abstract

“…The mean absolute error (MAE) in the prediction is 15.4% of the average time to failure, and the correlation coefficient is r = 0.46. This result is competitive with the most recent ML approaches [21] based on similar HOSS data, which have an error of approximately 16% and a correlation coefficient varying from r = 0.42 to r = 0.68, even though those methods (unlike ours) are specifically oriented towards predicting failure paths. The MAE in our results reflects the non-negligible scatter in predicted val-ues for a given actual time to failure, which may arise in part from our features imperfectly characterizing the geometry of fractures.…”

Section: Materials Failurementioning

confidence: 67%

“…For the problem of dynamic graph representations of fractures, Miller et al [20] has employed an image-based approach, extracting video representations of HOSS simulations and using graph convolutional networks to learn features from these. Finally, Moore et al [21] and Srinivasan et al [22] have used a variety of ML methods to predict fracture coalescence and time to failure. These studies suggest that important aspects of fracture dynamics can be learned from a modest sample of simulation training data, and then predicted through modern algorithmic techniques.…”

Section: Introductionmentioning

confidence: 99%

Learning to fail: Predicting fracture evolution in brittle material models using recurrent graph convolutional neural networks

Schwarzer

Rogan

Ruan

et al. 2019

Computational Materials Science

Self Cite

View full text Add to dashboard Cite

We propose a machine learning approach to address a key challenge in materials science: predicting how fractures propagate in brittle materials under stress, and how these materials ultimately fail. Our methods use deep learning and train on simulation data from high-fidelity models, emulating the results of these models while avoiding the overwhelming computational demands associated with running a statistically significant sample of simulations. We employ a graph convolutional network that recognizes features of the fracturing material and a recurrent neural network that models the evolution of these features, along with a novel form of data augmentation that compensates for the modest size of our training data. We simultaneously generate predictions for qualitatively distinct material properties. Results on fracture damage and length are within 3% of their simulated values, and results on time to material failure, which is notoriously difficult to predict even with highfidelity models, are within approximately 15% of simulated values. Once trained, our neural networks generate predictions within seconds, rather than the hours needed

show abstract

“…We note that this method has also been applied using RFs and DTs in place of the ANNs used here. Further details on these methods and the comparison across ML algorithms can be found in [38].…”

Section: Micro-crack Pair Informed Coalescence (Mcpic)mentioning

confidence: 99%

“…(2) Feature vectors for each crack pair includes information such as the length of each crack, their respective orientations, the distance between the two micro-cracks, the minimum distance to the boundary from either crack, and the stress intensity factors at the crack tips. (3) Uses ANNs for the ML algorithm, but can also utilize RF or DT approaches [38].…”

Section: Mcpic Model Summarymentioning

confidence: 99%