Finite electro-elasticity with physics-augmented neural networks
نویسندگان
چکیده
In the present work, a machine learning based constitutive model for electro-mechanically coupled material behavior at finite deformations is proposed. Using different sets of invariants as inputs, an internal energy density formulated convex neural network. this way, fulfills polyconvexity condition which ensures stability, well thermodynamic consistency, objectivity, symmetry, and growth conditions. Depending on considered invariants, physics-augmented can either be applied compressible or nearly incompressible behavior, arbitrary symmetry classes. The applicability versatility approach demonstrated by calibrating it transversely isotropic data generated with analytical potential, effective modeling analytically homogenized, rank-one laminate composite numerically homogenized cubic metamaterial. These examinations show excellent generalization properties that networks offer also multi-physical such nonlinear electro-elasticity.
منابع مشابه
Meta-Learning with Memory-Augmented Neural Networks
Despite recent breakthroughs in the applications of deep neural networks, one setting that presents a persistent challenge is that of “one-shot learning.” Traditional gradient-based networks require a lot of data to learn, often through extensive iterative training. When new data is encountered, the models must inefficiently relearn their parameters to adequately incorporate the new information...
متن کاملAPAC: Augmented PAttern Classification with Neural Networks
Deep neural networks have been exhibiting splendid accuracies in many of visual pattern classification problems. Many of the state-of-the-art methods employ a technique known as data augmentation at the training stage. This paper addresses an issue of decision rule for classifiers trained with augmented data. Our method is named as APAC: the Augmented PAttern Classification, which is a way of c...
متن کاملMemory Augmented Neural Networks with Wormhole Connections
Recent empirical results on long-term dependency tasks have shown that neural networks augmented with an external memory can learn the long-term dependency tasks more easily and achieve better generalization than vanilla recurrent neural networks (RNN). We suggest that memory augmented neural networks can reduce the effects of vanishing gradients by creating shortcut (or wormhole) connections. ...
متن کاملPhysics of Neural Networks
The material basis of our thinking, intelligence and creativity are 1013-1014 nerve cells (neurons) which in our brain are densely packed into a grey sub stance weighing about 1.5 kg. Each neuron has — like the root of a tree — highly branched dendrites that collect information from about 10000 other nerve cells. This huge network, which on a microscopic scale looks rather homogeneous and diso...
متن کاملComputational physics: Neural networks
2 Networks of binary neurons 5 2.1 Neural information processing is noisy . . . . . . . . . . . . . 5 2.2 Stochastic binary neurons and networks . . . . . . . . . . . . . 11 2.2.1 Parallel dynamics: Little model . . . . . . . . . . . . . 13 2.2.2 Sequential dynamics . . . . . . . . . . . . . . . . . . . 13 2.3 Some properties of Markov processes . . . . . . . . . . . . . . 14 2.3.1 Eigenvalue s...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Computer Methods in Applied Mechanics and Engineering
سال: 2022
ISSN: ['0045-7825', '1879-2138']
DOI: https://doi.org/10.1016/j.cma.2022.115501