Periodic Reporting for period 1 - CATALOG (Computational catalog of multiscale materials: a plugin library for industrial finiteelement codes)
Reporting period: 2018-01-01 to 2019-06-30
The CATALOG project is devoted to transfer cutting edge multiscale simulation technology to industrial Finite Element software through a non-invasive and easy-to-use plugin.
True multiscale material modeling technology based on standard techniques via simulations of representative volumes of the material underlying scale is the most rigorous and general strategy to account for the multiscale analysis of non-linear complex materials. However, its inherent multiplicative cost across the scales renders this approach unbearable from a computational standpoint.
Accelerating such multiscale technology through model order hyper-reduction (HPROM) techniques, and assuming a small, engineering affordable, approximation error, has proven to remarkably speed up the computations and break the barrier that prevented the uptake of multiscale technology within FE commercial software. Speedups of up to 4-5 orders of magnitude have been observed in the project for multiscale analysis concerning microstructural RVEs with several millions of integration points, i.e. the reduced order multiscale modelling turns out to be 4-5 orders of magnitude faster than the multiscale modeling of the full representation of the RVE (high fidelity multiscale analysis). This outperforming capability allows to conduct multiscale analysis of average macroscopic discretizations (thousands of FE) considering remarkably fine RVE discretizations at the lower scale (hundreds of thousands of FE) within hours of computation whereas a multiscale high-fidelity analysis would take several years or even decades.
To this end, it has been developed the technology necessary to interact with established industrial FE software such that multiscale simulations can be performed by exploiting existing user material modelling capabilities. The idea is to identify a material with interest in the multiscale simulation field and develop a computational engine which can be plugged in, through user defined routines, at the integration point level of commercial FE software making them feasible for multiscale material modeling.
Such technology is expected to represent a true revolution within multiscale simulation software where scientists and engineers will be able to accurately study the impact of the microscale material description into the structural scale and allow for virtual design of modern materials without the need of expensive and time-consuming trial-and-error experimentation.
True multiscale material modeling technology based on standard techniques via simulations of representative volumes of the material underlying scale is the most rigorous and general strategy to account for the multiscale analysis of non-linear complex materials. However, its inherent multiplicative cost across the scales renders this approach unbearable from a computational standpoint.
Accelerating such multiscale technology through model order hyper-reduction (HPROM) techniques, and assuming a small, engineering affordable, approximation error, has proven to remarkably speed up the computations and break the barrier that prevented the uptake of multiscale technology within FE commercial software. Speedups of up to 4-5 orders of magnitude have been observed in the project for multiscale analysis concerning microstructural RVEs with several millions of integration points, i.e. the reduced order multiscale modelling turns out to be 4-5 orders of magnitude faster than the multiscale modeling of the full representation of the RVE (high fidelity multiscale analysis). This outperforming capability allows to conduct multiscale analysis of average macroscopic discretizations (thousands of FE) considering remarkably fine RVE discretizations at the lower scale (hundreds of thousands of FE) within hours of computation whereas a multiscale high-fidelity analysis would take several years or even decades.
To this end, it has been developed the technology necessary to interact with established industrial FE software such that multiscale simulations can be performed by exploiting existing user material modelling capabilities. The idea is to identify a material with interest in the multiscale simulation field and develop a computational engine which can be plugged in, through user defined routines, at the integration point level of commercial FE software making them feasible for multiscale material modeling.
Such technology is expected to represent a true revolution within multiscale simulation software where scientists and engineers will be able to accurately study the impact of the microscale material description into the structural scale and allow for virtual design of modern materials without the need of expensive and time-consuming trial-and-error experimentation.