An FE-DMN method for the multiscale analysis of thermomechanical composites

Gajek, Sebastian; Schneider, Matti; Böhlke, Thomas

doi:10.1007/s00466-021-02131-0

Cited by 27 publications

(16 citation statements)

References 76 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Even more striking, the size of the considered unit cells could be chosen to be extremely small, and still be representative [26][27][28]. Such a property is extremely desirable, in particular when it comes to nonlinear material properties of composites, and turned out to be critical when undertaking a number of scientific studies, including fatigue [29,30], fracture [31,32] and thermomechanically coupled problems [33,34].…”

Section: State Of the Artmentioning

confidence: 99%

A sequential addition and migration method for generating microstructures of short fibers with prescribed length distribution

Mehta

Schneider

2022

Comput Mech

Self Cite

View full text Add to dashboard Cite

We describe an algorithm for generating fiber-filled volume elements for use in computational homogenization schemes. The algorithm permits to prescribe both a length distribution and a fiber-orientation tensor of second order, and composites with industrial filler fraction can be generated. Typically, for short-fiber composites, data on the fiber-length distribution and on the volume-weighted fiber-orientation tensor of second order is available. We consider a model where the fiber orientation and the fiber length distributions are independent, i.e., uncoupled. We discuss the use of closure approximations for this case and report on identifying the describing parameters of the frequently used Weibull distribution for modeling the fiber-length distribution. We discuss how to integrate these procedures in the Sequential Addition and Migration algorithm, developed for fibers of equal length, and work out algorithmic modifications accounting for possibly rather long fibers. We investigate the capabilities of the introduced methodology for industrial short-fiber composites, demonstrating the rather low dispersion of the effective elastic moduli for the generated unit cells.

show abstract

Section: State Of the Artmentioning

confidence: 99%

A sequential addition and migration method for generating microstructures of short fibers with prescribed length distribution

Mehta

Schneider

2022

Comput Mech

Self Cite

View full text Add to dashboard Cite

show abstract

“…DMNs were shown to work for modeling interface damage [50], strain localization [51], thermomechanically coupled materials [52] and porous materials [49]. Moreover, fully coupled FE-DMN methods were realized [47,48,52,53].…”

Section: State Of the Artmentioning

confidence: 99%

“…This so-called offline training is typically based on linear elastic precomputations obtained from full-field simulations on a high-fidelity representation of the microstructure [44,45], giving rise to the contributions J s . Once the parameters p are identified, the deep material network may be used as a high-fidelity fullfield surrogate model for inverse parameter identification or for concurrent multiscale simulations [48,52,53]. For this purpose, equation (2.3) is solved for nonlinearities that arise from a time discretization of inelastic constitutive laws.…”

Section: Basic Conceptsmentioning

confidence: 99%

See 1 more Smart Citation

Training deep material networks to reproduce creep loading of short fiber-reinforced thermoplastics with an inelastically-informed strategy

et al. 2022

Self Cite

View full text Add to dashboard Cite

Deep material networks (DMNs) are a recent multiscale technology which enable running concurrent multiscale simulations on industrial scale with the help of powerful surrogate models for the micromechanical problem. Classically, the parameters of the DMNs are identified based on linear elastic precomputations. Once the parameters are identified, DMNs may process inelastic material models and were shown to reproduce micromechanical full-field simulations with the original microstructure to high accuracy. The work at hand was motivated by creep loading of thermoplastic components with fiber reinforcement. In this context, multiple scales appear, both in space (due to the reinforcements) and in time (short- and long-term effects). We demonstrate by computational examples that the classical training strategy based on linear elastic precomputations is not guaranteed to produce DMNs whose long-term creep response accurately matches high-fidelity computations. As a remedy, we propose an inelastically informed early stopping strategy for the offline training of the DMNs. Moreover, we introduce a novel strategy based on a surrogate material model, which shares the principal nonlinear effects with the true model but is significantly less expensive to evaluate. For the problem at hand, this strategy enables saving significant time during the parameter identification process. We demonstrate that the novel strategy provides DMNs which reliably generalize to creep loading.

show abstract

“…In decades, chopped carbon fiber (CCF) [ 1–8 ] reinforced epoxy production sound lightweight, excellent mechanical properties which have enjoyed the significant applications of automotive and national defense industry. Generally, the CCF reinforced polymer structures usually suffer a load that far away from failure intensity, and one of fiber's main functions is to transfer the load from one fiber to another.…”

Section: Introductionmentioning

confidence: 99%

A dual‐scale homogenization method to study the properties of chopped carbon fiber reinforced epoxy resin matrix composites

Zheng

Qiu

et al. 2022

Polymer Composites

View full text Add to dashboard Cite

In this work, a dual‐scale homogenization (DSH) procedures including the representative volume element (RVE) and Unit cell (UC) with Halpin–Tsai models in the UC scale and space direction‐sensitive Hotelling expansion (SDSHE) models in RVE scale separately. These two methods are introduced and implemented to predict the response of effective elastic properties in chopped fiber reinforced epoxy matrix structures. A post‐processing scheme of subdomain in UC scale based on Gauss theorem is proposed in SDSHE method, which is used to extract the average strain and stress in UC three‐dimensional structure. In addition, the X‐ray computed tomography was used to determine the fourth order tensor orientation angle and the probability distribution information of chopped fiber length in testing samples. Homogenization properties in RVE scale, and inclusion volume fraction (VF), thickness of interface and orientation tensors are designed in UC scale and compared. The excellent correlations with experiments prove the effectiveness of the proposed models. Furthermore, the algorithm proposed in this paper realizes the homogenization of the two scales in the process of numerical analysis, and it will effectively take into account many problems that cannot be achieved by other algorithms, such as the path strain of the model, the thickness of the interface, and so on.

show abstract

An FE-DMN method for the multiscale analysis of thermomechanical composites

Cited by 27 publications

References 76 publications

A sequential addition and migration method for generating microstructures of short fibers with prescribed length distribution

A sequential addition and migration method for generating microstructures of short fibers with prescribed length distribution

Training deep material networks to reproduce creep loading of short fiber-reinforced thermoplastics with an inelastically-informed strategy

A dual‐scale homogenization method to study the properties of chopped carbon fiber reinforced epoxy resin matrix composites

Contact Info

Product

Resources

About