arXiv subjects

Nicolò Grilli

Publications and source records attributed to Nicolò Grilli.

3 recordsLinked to original sources

arXiv2024

SEA: State-Exchange Attention for High-Fidelity Physics Based Transformers

Current approaches using sequential networks have shown promise in estimating field variables for dynamical systems, but they are often limited by high rollout errors. The unresolved issue of rollout error accumulation results in unreliable estimations as the network predicts further into the future, with each step's error compounding and leading to an increase in inaccuracy. Here, we introduce the State-Exchange Attention (SEA) module, a novel transformer-based module enabling information exchange between encoded fields through multi-head cross-attention. The cross-field multidirectional information exchange design enables all state variables in the system to exchange information with one another, capturing physical relationships and symmetries between fields. Additionally, we introduce an efficient ViT-like mesh autoencoder to generate spatially coherent mesh embeddings for a large number of meshing cells. The SEA integrated transformer demonstrates the state-of-the-art rollout error compared to other competitive baselines. Specifically, we outperform PbGMR-GMUS Transformer-RealNVP and GMR-GMUS Transformer, with a reduction in error of 88% and 91%, respectively. Furthermore, we demonstrate that the SEA module alone can reduce errors by 97% for state variables that are highly dependent on other states of the system. The repository for this work is available at: https://github.com/ParsaEsmati/SEA

cs.LG↗

arXiv2021

Crystal plasticity model of residual stress in additive manufacturing

Selective laser melting is receiving increasing interest as an additive manufacturing technique. Residual stresses induced by the large temperature gradients and inhomogeneous cooling process can favour the generation of cracks. In this work, a crystal plasticity finite element model is developed to simulate the formation of residual stresses and to understand the correlation between plastic deformation, grain orientation and residual stresses in the additive manufacturing process. The temperature profile and grain structure from thermal-fluid flow and grain growth simulations are implemented into the crystal plasticity model. An element elimination and reactivation method is proposed to model the melting and solidification and to reinitialise state variables, such as the plastic deformation, in the reactivated elements. The accuracy of this method is judged against previous method based on the stiffness degradation of liquid regions by comparing the plastic deformation as a function of time induced by thermal stresses. The method is used to investigate residual stresses parallel and perpendicular to the laser scan direction, and the correlation with the maximum Schmid factor of the grains along those directions. The magnitude of the residual stress can be predicted as a function of the depth, grain orientation and position with respect to the molten pool.

cond-mat.mtrl-sci↗

arXiv2020

Characterisation of slip and twin activity using digital image correlation and crystal plasticity finite element simulation: Application to orthorhombic $α$-uranium

Calibrating and verifying crystal plasticity material models is a significant challenge, particularly for materials with a number of potential slip and twin systems. Here we use digital image correlation on coarse-grained $α$-uranium during tensile testing in conjunction with crystal plasticity finite element simulations. This approach allows us to determine the critical resolved shear stress, and hardening rate of the different slip and twin systems. The constitutive model is based on dislocation densities as state variables and the simulated geometry is constructed from electron backscatter diffraction images that provide shape, size and orientation of the grains, allowing a direct comparison between virtual and real experiments. An optimisation algorithm is used to find the model parameters that reproduce the evolution of the average strain in each grain as the load is increased. A tensile bar, containing four grains aligned with the load direction, is used to calibrate the model with eight unknown parameters. The approach is then independently validated by simulating the strain distribution in a second tensile bar. Different mechanisms for the hardening of the twin systems are evaluated. The latent hardening of the most active twin system turns out to be determined by coplanar twins and slip. The hardening rate of the most active slip system is lower than in fine-grained $α$-uranium. The method developed in the present research can be applied to identify the critical resolved shear stress and hardening parameters of other coarse-grained materials.

cond-mat.mtrl-sci↗