Abstract
Virtual clustering analysis (VCA) is an effective data-driven numerical homogenization method. In its original version (Tang et al. in Comput Mech 62(6):1443–1460, 2018), a homogeneous reference material is selected, and the governing system of heterogeneous material is reformulated into a Lippmann–Schwinger equation. For effective treatment of representative volume element (RVE) where long fibers penetrate the matrix, we propose to introduce an inhomogeneous reference material in this work. Since analytical expressions are not available in general, the interaction tensor and reference strain field are computed numerically. We validate the consistency with the original version of VCA in the first example, and demonstrate the effectiveness of the proposed strategy by several numerical examples. While better resolving the boundary traction, the proposed virtual clustering analysis for long fiber reinforced composites (VCA-L) algorithm correctly predicts homogenized properties in numerical tests.
Similar content being viewed by others
References
Matouš K, Geers MG, Kouznetsova VG, Gillman A (2017) A review of predictive nonlinear theories for multiscale modeling of heterogeneous materials. J Comput Phys 330:192–220
Feyel F, Chaboche J-L (2000) FE2 multiscale approach for modelling the elastoviscoplastic behaviour of long fibre SIC/TI composite materials. Comput Methods Appl Mech Eng 183(3–4):309–330
Kouznetsova V, Geers MG, Brekelmans WM (2002) Multi-scale constitutive modelling of heterogeneous materials with a gradient-enhanced computational homogenization scheme. Int J Numer Methods Eng 54(8):1235–1260
Feyel F (2003) A multilevel finite element method (FE2) to describe the response of highly non-linear structures using generalized continua. Comput Methods Appl Mech Eng 192(28–30):3233–3244
Wagner GJ, Liu WK (2003) Coupling of atomistic and continuum simulations using a bridging scale decomposition. J Comput Phys 190(1):249–274
Park HS, Karpov EG, Klein PA, Liu WK (2005) Three-dimensional bridging scale analysis of dynamic fracture. J Comput Phys 207(2):588–609
Tang S, Hou TY, Liu WK (2006) A pseudo-spectral multiscale method: interfacial conditions and coarse grid equations. J Comput Phys 213(1):57–85
Tang S (2008) A finite difference approach with velocity interfacial conditions for multiscale computations of crystalline solids. J Comput Phys 227(8):4038–4062
Eshelby JD (1957) The determination of the elastic field of an ellipsoidal inclusion, and related problems. Proc R Soc Lond Ser A Math Phys Sci 241(1226):376–396
Hashin Z, Shtrikman S (1963) A variational approach to the theory of the elastic behaviour of multiphase materials. J Mech Phys Solids 11(2):127–140
Hill R (1965) A self-consistent mechanics of composite materials. J Mech Phys Solids 13(4):213–222
Budiansky B (1965) On the elastic moduli of some heterogeneous materials. J Mech Phys Solids 13(4):223–227
Mori T, Tanaka K (1973) Average stress in matrix and average elastic energy of materials with misfitting inclusions. Acta Metall 21(5):571–574
Liu Z, Moore JA, Aldousari SM, Hedia HS, Asiri SA, Liu WK (2015) A statistical descriptor based volume-integral micromechanics model of heterogeneous material with arbitrary inclusion shape. Comput Mech 55(5):963–981
Li L, Wen P, Aliabadi M (2011) Meshfree modeling and homogenization of 3D orthogonal woven composites. Compos Sci Technol 71(15):1777–1788
Moulinec H, Suquet P (1998) A numerical method for computing the overall response of nonlinear composites with complex microstructure. Comput Methods Appl Mech Eng 157(1–2):69–94
Berkooz G, Holmes P, Lumley JL (1993) The proper orthogonal decomposition in the analysis of turbulent flows. Annu Rev Fluid Mech 25(1):539–575
Michel J-C, Suquet P (2003) Nonuniform transformation field analysis. Int J Solids Struct 40(25):6937–6955
Roussette S, Michel J-C, Suquet P (2009) Nonuniform transformation field analysis of elastic–viscoplastic composites. Compos Sci Technol 69(1):22–27
Liu Z, Bessa M, Liu WK (2016) Self-consistent clustering analysis: an efficient multi-scale scheme for inelastic heterogeneous materials. Comput Methods Appl Mech Eng 306:319–341
Liu Z, Fleming M, Liu WK (2018) Microstructural material database for self-consistent clustering analysis of elastoplastic strain softening materials. Comput Methods Appl Mech Eng 330:547–577
Yu C, Kafka OL, Liu WK (2019) Self-consistent clustering analysis for multiscale modeling at finite strains. Comput Methods Appl Mech Eng 349:339–359
Liu Z, Kafka OL, Yu C, Liu WK (2018) Data-driven self-consistent clustering analysis of heterogeneous materials with crystal plasticity. In: Advances in computational plasticity. Springer, Berlin, pp 221–242
Shakoor M, Kafka OL, Yu C, Liu WK (2019) Data science for finite strain mechanical science of ductile materials. Comput Mech 64(1):33–45
Kafka OL, Yu C, Shakoor M, Liu Z, Wagner GJ, Liu WK (2018) Data-driven mechanistic modeling of influence of microstructure on high-cycle fatigue life of nickel titanium. JOM 70(7):1154–1158
Li H, Kafka OL, Gao J, Yu C, Nie Y, Zhang L, Tajdari M, Tang S, Guo X, Li G et al (2019) Clustering discretization methods for generation of material performance databases in machine learning and design optimization. Comput Mech 64(2):281–305
Tang S, Zhang L, Liu WK (2018) From virtual clustering analysis to self-consistent clustering analysis: a mathematical study. Comput Mech 62(6):1443–1460
Zhang L, Tang S, Yu C, Zhu X, Liu WK (2019) Fast calculation of interaction tensors in clustering-based homogenization. Comput Mech 64(2):351–364
Zhu X, Zhang L, Tang S (2021) Adaptive selection of reference stiffness in virtual clustering analysis. Comput Methods Appl Mech Eng 376:113621
Yang Y, Zhang L, Tang S (2022) A comparative study of cluster-based methods at finite strain. Acta Mech Sin 38(4):1–12
Cheng G, Li X, Nie Y, Li H (2019) FEM-cluster based reduction method for efficient numerical prediction of effective properties of heterogeneous material in nonlinear range. Comput Methods Appl Mech Eng 348:157–184
Nie Y, Cheng G, Li X, Xu L, Li K (2019) Principle of cluster minimum complementary energy of FEM-cluster-based reduced order method: fast updating the interaction matrix and predicting effective nonlinear properties of heterogeneous material. Comput Mech 64(2):323–349
Nie Y, Li Z, Cheng G (2021) Efficient prediction of the effective nonlinear properties of porous material by FEM-cluster based analysis (FCA). Comput Methods Appl Mech Eng 383:113921
Wulfinghoff S, Cavaliere F, Reese S (2018) Model order reduction of nonlinear homogenization problems using a Hashin–Shtrikman type finite element method. Comput Methods Appl Mech Eng 330:149–179
Ri J-H, Hong H-S, Ri S-G (2021) Cluster based nonuniform transformation field analysis: an efficient homogenization for inelastic heterogeneous materials. Int J Numer Methods Eng 122(17):4458–4485
Ferreira BP, Pires FA, Bessa M (2022) Adaptivity for clustering-based reduced-order modeling of localized history-dependent phenomena. Comput Methods Appl Mech Eng 393:114726
Tang H, Chen Z, Xu H, Liu Z, Sun Q, Zhou G, Yan W, Han W, Su X (2020) Computational micromechanics model based failure criteria for chopped carbon fiber sheet molding compound composites. Compos Sci Technol 200:108400
Chen Z, Tang H, Shao Y, Sun Q, Zhou G, Li Y, Xu H, Zeng D, Su X (2019) Failure of chopped carbon fiber sheet molding compound (SMC) composites under uniaxial tensile loading: Computational prediction and experimental analysis. Compos A Appl Sci Manuf 118:117–130
Tang H, Zhou G, Chen Z, Huang L, Avery K, Li Y, Liu H, Guo H, Kang H, Zeng D et al (2019) Fatigue behavior analysis and multi-scale modelling of chopped carbon fiber chip-reinforced composites under tension-tension loading condition. Compos Struct 215:85–97
Chen Z, Huang T, Shao Y, Li Y, Xu H, Avery K, Zeng D, Chen W, Su X (2018) Multiscale finite element modeling of sheet molding compound (SMC) composite structure based on stochastic mesostructure reconstruction. Compos Struct 188:25–38
Sherburn M (2007) Geometric and mechanical modelling of textiles. Ph.D. thesis, University of Nottingham United Kingdom
Aliabadi MF (2015) Woven composites, vol 6. World Scientific, Singapore
Gao J, Mojumder S, Zhang W, Li H, Suarez D, He C, Cao J, Liu WK (2022) Concurrent n-scale modeling for non-orthogonal woven composite. Comput Mech 70(4):853–866
He C, Ge J, Lian Y, Geng L, Chen Y, Fang D (2022) A concurrent three-scale scheme FE-SCA2 for the nonlinear mechanical behavior of braided composites. Comput Methods Appl Mech Eng 393:114827
Han X, Gao J, Fleming M, Xu C, Xie W, Meng S, Liu WK (2020) Efficient multiscale modeling for woven composites based on self-consistent clustering analysis. Comput Methods Appl Mech Eng 364:112929
Acknowledgements
Yang Yang and Shaoqiang Tang were supported partially by NSFC under grant numbers 11832001, 11890681, and 11988102.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendices
SOM algorithm
In SOM, we have a data set \(\varvec{X}\), with each data \(\varvec{x} \in \mathbb {R}^{m}\), and designate k clusters. We prescribe a norm \(\Vert \cdot \Vert \) for the m-dimensional vector space, and a distance \(d_{i,j}\) for the integer indices \(i,j=1,\ldots ,k\). In this paper, we take \(\Vert \cdot \Vert \) as Euclidean norm, and \(d_{i,j}=|i-j|\). An iterative algorithm with N iterations is as follows.
-
1.
Determine a spatial arrangement of the k clusters. In all our tests, we use \(1\times k\) for this arrangement.
-
2.
Initialize the weight vector \(\varvec{w}_{j}\) for each cluster, \(j=1,\ldots ,k\). It is one possible way to pick k data from the input data randomly and use them as the weight vectors of the k clusters.
-
3.
Draw a sample x from the input data with a certain probability, and find the winning cluster i(x) at time-step n ruled by
$$\begin{aligned} i(\varvec{x})=\arg \underset{1\le j \le k}{\min }\Vert \varvec{x}(n) - \varvec{w}_{j}\Vert . \end{aligned}$$ -
4.
Adjust the weight vectors of all affected clusters
$$\begin{aligned} \varvec{w}_{j}(n+1) = \varvec{w}_{j}(n) + \eta (n)h_{j,i(\varvec{x})}(n)\left( \varvec{x}(n)-\varvec{w}_{j}(n)\right) , \end{aligned}$$where \(\eta (n)\) is a learning rate parameter, and \(h_{j,i(\varvec{x})}\) is the neighborhood function centered around the winning cluster i(x). We choose a Gaussian function
$$\begin{aligned} h_{j,i(\varvec{x})}(n) = \exp \left( - \frac{d^{2}_{j,i(\varvec{x})}}{2k^{2}\exp (-6 n(\log {k})/N)} \right) , \end{aligned}$$and a decreasing learning function
$$\begin{aligned} \eta (n) = 0.1 \exp (-\frac{3n}{N}). \end{aligned}$$ -
5.
Go back Step 3 until it reaches N.
Relationship with the finite element method
As described in Sect. 2.6, the RVE is usually discretized by regular pixels/voxels. Given that the finite element method is adopted as the direct numerical simulation method, two procedures are conducted to give a suitable transfer and comparison of data between these two types of discretizations.
The first procedures is to generate a FEM mesh compatible with the regular pixels/voxels. In this work, CPE4 element is adopted in the computation of finite element method. Furthermore, the element is square, which can be converted by the pixel.
The second procedure is the element average technique. Each pixel/voxel only contains one sampling point, corresponding to the Gauss sampling points in the finite element method. To transfer the data from the Gauss sampling points to the single sampling point, element average technique is used to compress the data. That is, for a pixel/voxel, a quantity in the single sampling point corresponds to averaging this quantity in its corresponding element.
Constitutive relation and coordinate transformation for the transversely isotropic fiber material
Through the fiber is transversely isotropic in its three-dimension local coordinate, we only use its material property in two-dimension local coordinate. Based on this requirement, the constitutive relation in its two-dimension local coordinate is given in the matrix form as follows:
where
in which the superscript l means that the relevant quantities are evaluated in the local coordinate system; the subscripts 1, 2 and 3 are corresponding to the longitudinal, in-plane transverse and out-of plane transverse directions; \(***\) denotes an entry determined by symmetry; and
Then these constitutive quantities (especially the local stiffness \(\varvec{C}^{l}\)) need transformation from the local coordinate system to the global coordinate system. The coordinate transformation of the stiffness can be expressed as follows:
Here the superscript g denotes that the corresponding quantities are evaluated in the global coordinate system, and \([\varvec{T}_{c}]\) is the transformation matrix
where \(\theta _{ij}\) (\(i,j=1,2\)) is the angle between the jth axis of the local coordinate system and the ith axis of the global coordinate system. Further, the constitutive relation in the global coordinate system is
where
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Yang, Y., Liu, T., Aliabadi, M.H. et al. Virtual clustering analysis for long fiber reinforced composites. Comput Mech 71, 1139–1159 (2023). https://doi.org/10.1007/s00466-023-02290-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-023-02290-2