Unsupervised discovery of interpretable hyperelastic constitutive laws

Flaschel, Moritz; Kumar, Siddhant; Lorenzis, Laura De

doi:10.1016/j.cma.2021.113852

Cited by 138 publications

(99 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Apart from being related alternatives to model ageing at the mesoscale, the modular paradigm and multiscale techniques in [ 7 ] can be further combined with the models discussed here in order to upscale degraded mechanical properties from the mesoscale up to even higher scales of interest (e.g., component scale, structure scale). Furthermore, the current trend in computational mechanics to employ machine learning for surrogate modelling [ 172 ], model selection [ 173 ], and data assimilation [ 174 ], which was briefly described in [ 7 ], is expected to also profoundly impact the class of models treated in the present discussion.…”

Section: Discussionmentioning

confidence: 99%

Modelling of Environmental Ageing of Polymers and Polymer Composites—Durability Prediction Methods

et al. 2022

View full text Add to dashboard Cite

Polymers and polymer composites are negatively impacted by environmental ageing, reducing their service lifetimes. The uncertainty of the material interaction with the environment compromises their superior strength and stiffness. Validation of new composite materials and structures often involves lengthy and expensive testing programs. Therefore, modelling is an affordable alternative that can partly replace extensive testing and thus reduce validation costs. Durability prediction models are often subject to conflicting requirements of versatility and minimum experimental efforts required for their validation. Based on physical observations of composite macroproperties, engineering and phenomenological models provide manageable representations of complex mechanistic models. This review offers a systematised overview of the state-of-the-art models and accelerated testing methodologies for predicting the long-term mechanical performance of polymers and polymer composites. Accelerated testing methods for predicting static, creep, and fatig ue lifetime of various polymers and polymer composites under environmental factors’ single or coupled influence are overviewed. Service lifetimes are predicted by means of degradation rate models, superposition principles, and parametrisation techniques. This review is a continuation of the authors’ work on modelling environmental ageing of polymer composites: the first part of the review covered multiscale and modular modelling methods of environmental degradation. The present work is focused on modelling engineering mechanical properties.

show abstract

Section: Discussionmentioning

confidence: 99%

Modelling of Environmental Ageing of Polymers and Polymer Composites—Durability Prediction Methods

et al. 2022

View full text Add to dashboard Cite

show abstract

“…(c) Unsupervised discovery of interpretable and parsimonious constitutive laws using only displacement data and physical knowledge [8].…”

Section: Dic Displacement Fieldsmentioning

confidence: 99%

“…The strategy is to use experience in physical modeling to reduce the dependence on data and improve generalizability. The most popular approach has relied on sparse regression to discover a mathematically and physically interpretable form of the underlying constitutive equation [3,8] (in contrast to, e.g., black-box neural networks). The method involves creating an exhaustive library of mathematical functions and then sparsely selecting the combination of those which best explains the data.…”

Section: Data-driven Constitutive Models: Beyond Simulation-based Tra...mentioning

confidence: 99%

“…One of the biggest challenges in realizing data-driven constitutive models for practical application lies in material characterization; e.g., using digital image correlation (DIC) full-field strain maps become available but stress data is hard to obtain. To this end, recent work [8] (Figure 1c) discovered interpretable constitutive models in an unsupervised manner -i.e., without stress data -which presents a promising departure from curve-fitting and supervised learning based on stress data.…”

Section: Data-driven Constitutive Models: Beyond Simulation-based Tra...mentioning

confidence: 99%

See 1 more Smart Citation

What machine learning can do for computational solid mechanics

Kumar,

Kochmann

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Data-driven computational mechanics can be also adopted in the context of inverse problems such as in [13,14] where distance minimization is used to identify material parameters from experimental tests. Also, in [7,3] a machine learning technique is devised able to identify numerical constitutive manifolds from experimental data, while in [15] interpretable hyperelastic constitutive models are discovered from synthetic data sets.…”

Section: Introductionmentioning

confidence: 99%

Data-driven fracture mechanics

Carrara

Lorenzis

Stainier

et al. 2020

Computer Methods in Applied Mechanics and Engineering

104

View full text Add to dashboard Cite

We extend the model-free data-driven paradigm for rateindependent fracture mechanics proposed in Carrara et al. ( 2020), Data-driven Fracture Mechanics, Comp. Meth. App. Mech. Eng., 372 to rate-dependent fracture and sub-critical fatigue. The problem is formulated by combining the balance governing equations stemming from variational principles with a set of data points that encodes the fracture constitutive behavior of the material. The solution is found as the data point that best satisfies the meta-stability condition as given by the variational procedure and following a distance minimization approach based on closest-point-projection. The approach is tested on different setups adopting different types of rate-dependent fracture and fatigue models affected or not by white noise.

show abstract

Unsupervised discovery of interpretable hyperelastic constitutive laws

Cited by 138 publications

References 44 publications

Modelling of Environmental Ageing of Polymers and Polymer Composites—Durability Prediction Methods

Modelling of Environmental Ageing of Polymers and Polymer Composites—Durability Prediction Methods

What machine learning can do for computational solid mechanics

Data-driven fracture mechanics

Contact Info

Product

Resources

About