Abstract
This paper presents a general Convolution Hierarchical Deep-learning Neural Networks (C-HiDeNN) computational framework for solving partial differential equations. This is the first paper of a series of papers devoted to C-HiDeNN. We focus on the theoretical foundation and formulation of the method. The C-HiDeNN framework provides a flexible way to construct high-order \(C^n\) approximation with arbitrary convergence rates and automatic mesh adaptivity. By constraining the C-HiDeNN to build certain functions, it can be degenerated to a specification, the so-called convolution finite element method (C-FEM). The C-FEM will be presented in detail and used to study the numerical performance of the convolution approximation. The C-FEM combines the standard \(C^0\) FE shape function and the meshfree-type radial basis interpolation. It has been demonstrated that the C-FEM can achieve arbitrary orders of smoothness and convergence rates by adjusting the different controlling parameters, such as the patch function dilation parameter and polynomial order, without increasing the degrees of freedom of the discretized systems, compared to FEM. We will also present the convolution tensor decomposition method under the reduced-order modeling setup. The proposed methods are expected to provide highly efficient solutions for extra-large scale problems while maintaining superior accuracy. The applications to transient heat transfer problems in additive manufacturing, topology optimization, GPU-based parallelization, and convolution isogeometric analysis have been discussed.
Similar content being viewed by others
References
Liu WK, Li S, Park HS (2022) Eighty years of the finite element method: birth, evolution, and future. Arch Comput Methods Eng 29:4431–4453
Liu WK, Jun S, Zhang YF (1995) Reproducing kernel particle methods. Int J Numer Methods Fluids 20(8–9):1081–1106
Liu WK, Jun S, Li S, Adee J, Belytschko T (1995) Reproducing kernel particle methods for structural dynamics. Int J Numer Methods Eng 38(10):1655–1679
Chen JS, Liu WK, Hillman MC, Chi SW, Lian Y, Bessa MA (2017) Reproducing kernel particle method for solving partial differential equations. In: Stein E, de Borst R, Hughes TJR (eds) Encyclopedia of computational mechanics, 2nd edn. Wiley, Hoboken, pp 1–44
Belytschko T, Lu YY, Gu L (1994) Element-free Galerkin methods. Int J Numer Methods Eng 37(2):229–256
Nayroles B, Touzot G, Villon P (1992) Generalizing the finite element method: diffuse approximation and diffuse elements. Comput Mech 10(5):307–318
Liu M, Liu G (2010) Smoothed particle hydrodynamics (sph): an overview and recent developments. Arch Comput Methods Eng 17(1):25–76
Liu W-K, Li S, Belytschko T (1997) Moving least-square reproducing kernel methods (I) methodology and convergence. Comput Methods Appl Mech Eng 143(1–2):113–154
Liu WK, Chen Y, Uras RA, Chang CT (1996) Generalized multiple scale reproducing kernel particle methods. Comput Methods Appl Mech Eng 139(1–4):91–157
Li S, Liu WK (1996) Moving least-square reproducing kernel method part ii: Fourier analysis. Comput Methods Appl Mech Eng 139(1–4):159–193
Hughes TJ, Cottrell JA, Bazilevs Y (2005) Isogeometric analysis: CAD, finite elements, NURBS, exact geometry and mesh refinement. Comput Methods Appl Mech Eng 194(39–41):4135–4195
Bazilevs Y, Calo VM, Cottrell JA, Evans JA, Hughes TJR, Lipton S, Scott MA, Sederberg TW (2010) Isogeometric analysis using t-splines. Comput Methods Appl Mech Eng 199(5–8):229–263
De Lorenzis L, Wriggers P, Hughes TJ (2014) Isogeometric contact: a review. GAMM-Mitteilungen 37(1):85–123
Belytschko T, Organ D, Gerlach C (2000) Element-free Galerkin methods for dynamic fracture in concrete. Comput Methods Appl Mech Eng 187(3–4):385–399
Duarte CA, Oden JT (1996) An hp adaptive method using clouds. Comput Methods Appl Mech Eng 139(1–4):237–262
Duarte CA, Oden JT (1996) H-p clouds-an h-p meshless method. Numer Methods Partial Differ Equ Int J 12(6):673–705
Oden JT, Duarte C, Zienkiewicz OC (1998) A new cloud-based hp finite element method. Comput Methods Appl Mech Eng 153(1–2):117–126
Melenk JM, Babuška I (1996) The partition of unity finite element method: basic theory and applications. Comput Methods Appl Mech Eng 139(1–4):289–314
Griebel M, Schweitzer MA (2000) A particle-partition of unity method for the solution of elliptic, parabolic, and hyperbolic PDEs. SIAM J Sci Comput 22(3):853–890
Chen J-S, Wang H-P (2000) New boundary condition treatments in meshfree computation of contact problems. Comput Methods Appl Mech Eng 187(3–4):441–468
Chen J-S, Han W, You Y, Meng X (2003) A reproducing kernel method with nodal interpolation property. Int J Numer Methods Eng 56(7):935–960
Wagner GJ, Liu WK (2000) Application of essential boundary conditions in mesh-free methods: a corrected collocation method. Int J Numer Methods Eng 47(8):1367–1379
Wagner GJ, Liu WK (2001) Hierarchical enrichment for bridging scales and mesh-free boundary conditions. Int J Numer Methods Eng 50(3):507–524
Han W, Wagner GJ, Liu WK (2002) Convergence analysis of a hierarchical enrichment of Dirichlet boundary conditions in a mesh-free method. Int J Numer Methods Eng 53(6):1323–1336
Huerta A, Fernández-Méndez S (2000) Enrichment and coupling of the finite element and meshless methods. Int J Numer Methods Eng 48(11):1615–1636
Huerta A, Fernández-Méndez S, Liu WK (2004) A comparison of two formulations to blend finite elements and mesh-free methods. Comput Methods Appl Mech Eng 193(12–14):1105–1117
Liu WK, Han W, Lu H, Li S, Cao J (2004) Reproducing kernel element method. Part i: theoretical formulation. Comput Methods Appl Mech Eng 193(12–14):933–951
Li S, Lu H, Han W, Liu WK, Simkins DC (2004) Reproducing kernel element method part ii: globally conforming Im/Cn hierarchies. Comput Methods Appl Mech Eng 193(12–14):953–987
Lu H, Li S, Simkins DC Jr, Liu WK, Cao J (2004) Reproducing kernel element method part iii: generalized enrichment and applications. Comput Methods Appl Mech Eng 193(12–14):989–1011
Hornik K, Stinchcombe M, White H (1989) Multilayer feedforward networks are universal approximators. Neural Netw 2(5):359–366
Cai S, Mao Z, Wang Z, Yin M, Karniadakis GE (2022) Physics-informed neural networks (pinns) for fluid mechanics: a review. Acta Mech Sin 37:1727–1738
Raissi M, Perdikaris P, Karniadakis GE (2021) Physics informed learning machine. Mar. 30. US Patent 10,963,540
Lee K, Trask NA, Patel RG, Gulian MA, Cyr EC (2021) Partition of unity networks: deep hp-approximation. arXiv:2101.11256
Jin P, Zhang Z, Zhu A, Tang Y, Karniadakis GE (2020) Sympnets: intrinsic structure-preserving symplectic networks for identifying Hamiltonian systems. Neural Netw 132:166–179
Hernández Q, Badías A, González D, Chinesta F, Cueto E (2021) Structure-preserving neural networks. J Comput Phys 426:109950
Cheng L, Wagner GJ (2022) A representative volume element network (RVE-net) for accelerating RVE analysis, microscale material identification, and defect characterization. Comput Methods Appl Mech Eng 390:114507
Cuomo S, Di Cola VS, Giampaolo F, Rozza G, Raissi M, Piccialli F (2022) Scientific machine learning through physics-informed neural networks: where we are and what’s next. J Sci Comput 92(3):88
Zhang L, Cheng L, Li H, Gao J, Yu C, Domel R, Yang Y, Tang S, Liu WK (2021) Hierarchical deep-learning neural networks: finite elements and beyond. Comput Mech 67(1):207–230
Zhang L, Lu Y, Tang S, Liu WK (2022) Hidenn-td: reduced-order hierarchical deep learning neural networks. Comput Methods Appl Mech Eng 389:114414
Ammar A, Mokdad B, Chinesta F, Keunings R (2006) A new family of solvers for some classes of multidimensional partial differential equations encountered in kinetic theory modeling of complex fluids. J Nonnewton Fluid Mech 139(3):153–176
Chinesta F, Ladeveze P, Cueto E (2011) A short review on model order reduction based on proper generalized decomposition. Arch Comput Methods Eng 18(4):395–404
Kazemzadeh-Parsi MJ, Ammar A, Duval JL, Chinesta F (2021) Enhanced parametric shape descriptions in PGD-based space separated representations. Adv Model Simul Eng Sci 8(1):1–28
Lu Y, Jones KK, Gan Z, Liu WK (2020) Adaptive hyper reduction for additive manufacturing thermal fluid analysis. Comput Methods Appl Mech Eng 372:113312
Lu Y, Li H, Saha S, Mojumder S, Al Amin A, Suarez D, Liu Y, Qian D, Kam Liu W (2021) Reduced order machine learning finite element methods: concept, implementation, and future applications. Comput Model Eng Sci 129(3):1351–1371. https://doi.org/10.32604/cmes.2021.017719
Park C, Lu Y, Saha S, Xue T, Guo J, Mojumder S, Apley D, Wagner G, Liu W (2023)Convolution hierarchical deep-learning neural network (c-hidenn) with graphics processing unit (GPU) acceleration. Comput Mech
Li H, Knapik S, Li Y, Park C, Guo J, Mojumder S, Lu Y, Chen W, Apley D, Liu W, Convolution Hierarchical Deep-Learning Neural Network Tensor Decomposition (C-HiDeNN-TD) for high-resolution topology optimization. Comput Mech (2023). https://doi.org/10.1007/s00466-023-02333-8
Belytschko T, Liu WK, Moran B, Elkhodary K (2014) Nonlinear finite elements for continua and structures. Wiley, Hoboken
Bessa M, Foster J, Belytschko T, Liu WK (2014) A meshfree unification: reproducing kernel peridynamics. Comput Mech 53(6):1251–1264
Li S, Liu WK (2007) Meshfree particle methods. Springer, Berlin
Hughes TJ (2012) The finite element method: linear static and dynamic finite element analysis. Courier Corporation, North Chelmsford
Wendland H (1999) Meshless Galerkin methods using radial basis functions. Math Comput 68(228):1521–1531
Wang J, Liu G (2002) A point interpolation meshless method based on radial basis functions. Int J Numer Methods Eng 54(11):1623–1648
Tian R (2013) Extra-dof-free and linearly independent enrichments in GFEM. Comput Methods Appl Mech Eng 266:1–22
Petersen P, Raslan M, Voigtlaender F (2021) Topological properties of the set of functions generated by neural networks of fixed size. Found Comput Math 21:375–444
Bartlett PL, Ben-David S (2002) Hardness results for neural network approximation problems. Theor Comput Sci 284(1):53–66
Blum AL, Rivest RL (1992) Training a 3-node neural network is NP-complete. Neural Netw 5(1):117–127. https://doi.org/10.1016/S0893-6080(05)80010-3
Judd JS (1987) Learning in networks is hard. In: Proceedings of 1st international conference on neural networks, San Diego, California, IEEE
Kolda TG, Bader BW (2009) Tensor decompositions and applications. SIAM Rev 51(3):455–500
Sidiropoulos ND, De Lathauwer L, Fu X, Huang K, Papalexakis EE, Faloutsos C (2017) Tensor decomposition for signal processing and machine learning. IEEE Trans Signal Process 65(13):3551–3582
Papalexakis EE, Faloutsos C, Sidiropoulos ND (2012) Parcube: sparse parallelizable tensor decompositions. In: Machine learning and knowledge discovery in databases: European conference, ECML PKDD 2012, Bristol, UK, September 24–28, 2012. Proceedings, Part I 23, Springer, pp 521–536
Song J (2001) Optimal representation of high-dimensional functions and manifolds in low-dimensional visual space (in Chinese). Chin Sci Bull 46(12):977–984
Lu Y, Blal N, Gravouil A (2018) Adaptive sparse grid based HOPGD: toward a nonintrusive strategy for constructing space–time welding computational vademecum. Int J Numer Methods Eng 114(13):1438–1461
Lu Y, Blal N, Gravouil A (2018) Multi-parametric space–time computational vademecum for parametric studies: application to real time welding simulations. Finite Elem Anal Des 139:62–72
Lu Y, Blal N, Gravouil A (2019) Datadriven HOPGD based computational vademecum for welding parameter identification. Comput Mech 64(1):47–62
Badrou A, Bel-Brunon A, Hamila N, Tardif N, Gravouil A (2020) Reduced order modeling of an active multi-curve guidewire for endovascular surgery. Comput Methods Biomech Biomed Eng 23(sup1):S23–S24
Blal N, Gravouil A (2019) Non-intrusive data learning based computational homogenization of materials with uncertainties. Comput Mech 64(3):807–828
Rozza G, Huynh DB, Patera AT (2008) Reduced basis approximation and a posteriori error estimation for affinely parametrized elliptic coercive partial differential equations: application to transport and continuum mechanics. Arch Comput Methods Eng 15(3):229
Rozza G, Veroy K (2007) On the stability of the reduced basis method for stokes equations in parametrized domains. Comput Methods Appl Mech Eng 196(7):1244–1260
Sirovich L (1987) Turbulence and the dynamics of coherent structures. I. Coherent structures. Q Appl Math 45(3):561–571
Amsallem D, Farhat C (2008) Interpolation method for adapting reduced-order models and application to aeroelasticity. AIAA J 46(7):1803–1813
Kerfriden P, Goury O, Rabczuk T, Bordas SP (2013) A partitioned model order reduction approach to rationalise computational expenses in nonlinear fracture mechanics. Comput Methods Appl Mech Eng 256:169–188
Amsallem D, Zahr M, Choi Y, Farhat C (2015) Design optimization using hyper-reduced-order models. Struct Multidiscip Optim 51(4):919–940
Ryckelynck D (2009) Hyper-reduction of mechanical models involving internal variables. Int J Numer Methods Eng 77(1):75–89
Scanff R, Néron D, Ladevèze P, Barabinot P, Cugnon F, Delsemme J-P (2022) Weakly-invasive LATIN-PGD for solving time-dependent non-linear parametrized problems in solid mechanics. Comput Methods Appl Mech Eng 396:114999
Falcó A, Hackbusch W, Nouy A (2019) On the Dirac–Frenkel variational principle on tensor banach spaces. Found Comput Math 19:159–204
Hackbusch W (2012) Tensor spaces and numerical tensor calculus, vol 42. Springer, Berlin
Schaback R, Wendland H (2001) Characterization and construction of radial basis functions. In: Multivariate approximation and applications, pp 1–24
Acknowledgements
S. Tang and L. Zhang would like to thank the support of the National Natural Science Foundation of China (NSFC) under contract Nos. 11832001, 11988102, and 12202451.
Author information
Authors and Affiliations
Corresponding authors
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A Brief review of HiDeNN-FEM
The idea of HiDeNN-FEM is to use Deep-learning Neural Networks (DeNN) (We added the “e” instead of just using DNN because we prefer the acronym HiDeNN) to reconstruct the FE shape function by constraining the weights and biases with mesh coordinates, which reads
where \({\mathcal {F}}_{{I}}\) stands for the fully connected DeNN structures with weights \(\varvec{w}\), biases \(\varvec{b}\), and the activation function \(\varvec{{\mathcal {A}}}\). \({\mathcal {N}}_{{I}}\) denotes the FE shape function for the node at position \(\varvec{x}_I^*\). Assuming a domain \(\varOmega \) is discretized by n points, we can write the HiDeNN-FEM approximation as
where \(u_I\) is the discretized nodal solution of the problem, \(u^h\) is the approximated solution function. Considering the vector notation \(\varvec{{\mathcal {N}}}=[{\mathcal {N}}_1,\dots ,{\mathcal {N}}_{n}]\) and \(\varvec{{u}}=[{u}_1,\dots ,{u}_{n}]^T\), the equation (40) can be simplified as
The detailed construction of such shape functions using DeNN can be found in [38]. Figure 22 illustrates a detailed architecture of the partially connected DeNN, as an example of HiDeNN-FEM. It should be noticed that the FE shape function is only one of the choices, the HiDeNN structure allows one to easily switch from one to another by releasing the constraints on the weights \(\varvec{w}\) and biases \(\varvec{b}\).
Since the HiDeNN shape function serves the same role as FE shape function, the derivatives and integration of the shape function can be implemented in exactly the same way as FEM. Finally, the solution of HiDeNN-FEM can be obtained through an optimization problem in which both the displacement field and the mesh coordinates are simultaneously optimized. The problem reads
where \(\varOmega \) denotes the entire domain, \(\partial \varOmega \) denotes the boundary, \(\varPi \) denotes the potential (when it exists) or the residual of the problem. Hence HiDeNN-FEM provides a new way to adaptively modify the mesh and reduces to FEM when the mesh is fixed. Moreover, it is shown that the HiDeNN-FEM has a function approximation space and gives more accurate results than FEM due to the flexibility of the HiDeNN framework [39].
Below, we summarize the key features of the HiDeNN-FEM method [38]:
-
Partially connected DeNN with constrained weights and biases
-
A flexible framework for function approximation and solving PDEs with automatic r-h-p adaptivity
-
More accurate than traditional FEM
-
Reduces to FEM when freezing the mesh coordinates
The developed novel Convolution HiDeNN method can be seen as an enhancement for highly smooth solutions and improved convergence rates without increasing the degrees of freedom (DoFs).
Appendix B Derivation of radial basis interpolation function
Let us consider the following 1D case for illustration purposes.
Then we can define the following interpolation as part of the approximation centering around the i-th node in the element domain.
where the supporting node set of W is \(A^i_s\) with a given patch size s. Figure 23 illustrates a 1D convolution element with patch size \(s=1\) and the two-node shape functions for \(N_i\). Now the question is how to compute the W based on the given supporting nodes.
Assuming the nodal solution value for the 4 nodes in Fig. 23 is \([u_1,u_2,u_3,u_4]\), we illustrate the radial basis interpolation procedure for the part centering around \(i=2\). In this case, the parametric coordinates for the support nodes are \(\{-3,-1,1\}\). Then we can consider the radial basis interpolation \(u^{i=2}({\xi })\) has the following form
where \(\varvec{\Psi }_a \) is a defined kernel function, which can be the reproducing kernel or cubic spline kernel [3, 4] with the dilation parameter a, \({\textbf{p}}(\xi )\) is the polynomial basis vector of p-th order, \({\textbf{k}}=[k_1, k_2, k_3]^T\) and \({\textbf{l}}=[l_1, l_2, l_3]^T\) are the coefficient vector that helps to enforce the reproducing condition and Kronecker delta property. We give here a specific example for \(\varvec{\Psi }_a \) and \({\textbf{p}}(\xi )\) using a cubic spline kernel and a second-order polynomial.
and
Now we can compute \({\textbf{k}}\) and \({\textbf{l}}\) by enforcing the below conditions.
Solving the above equations gives the solution to \({\textbf{k}}\) and \({\textbf{l}}\), which reads
with
and
Finally, the radial basis interpolation with the computed coefficients reads
where \({W}^{{\xi }_i}_{a,j}\) is obtained by identifying the corresponding coefficient of \(u_j\). By analogy, we can compute the other convolution patch functions W with the support \(A^{i=3}_s\). Detailed mathematical derivation and analysis of the radial basis interpolation can be found in [77].
Appendix C Illustration of a 1D convolution element
For a better understanding of the convolution shape function \(\tilde{{N}}_k\), we illustrate here a 1D convolution approximation with \({N}_{{i}}\) chosen to be linear and patch size \(s=1\), see again Fig. 23 for the supporting nodes of one convolution element.
In this case, the FE shape function nodal support set \(A^e=\{2,3\}\), and then \(A^{i=2}_s=\{1,2,3\}\), \(A^{i=3}_s=\{2,3,4\}\). The equation (53) becomes
where \(A_s=\bigcup _{i\in A^e} A^{i}_s =\{1,2,3,4\}\). Therefore, there are in total 4 convolution shape functions for \(s=1\). If \(s=2\), we can expect 6 shape functions, as shown in Fig. 5.
Appendix D Modification of the natural coordinates for irregular meshes
The modification of the natural (parametric) coordinates of the patch nodes can be done according to the distance ratio in the physical domain. As shown in Fig. 24, if the mesh is irregular in the physical space, the corresponding patch nodes in parametric space are adapted. The neighbouring nodes \(x_1\) and \(x_4\) are clearly far from the element in the physical space. In order to reflect this distancing information in parametric domain. The coordinates \(\xi _1\) and \(\xi _4\) are modified as below
that is
where the right-hand-side of the above equation is the physical distance ratio and left-hand-side is the one in parametric space. By enforcing this condition, we are able to keep the nonuniform distribution of the nodes in parametric space. Here, the original physical element domain is \([x_2, x_3]\) and the corresponding parametric element domain \([\xi _2, \xi _3]\) remains the same as before, i.e., \([-1,1]\).
From our experience this modification is necessary and can help achieve the optimal convergence rates when irregular meshes are used. Similar concept can be used in 2D/3D irregular meshes.
Appendix E Error estimate of C-HiDeNN interpolation
Following the procedure in the previous work on the meshfree reproducing property [8, 27], we can derive the error estimate of C-HiDeNN interpolation, based on the reproducing property.
Theorem 1
Let us assume an exact solution \(u(x)\in C^{p+1}\)\((\bar{\varOmega })\cap H^{p+1} (\varOmega )\), where \(\varOmega \) is a bounded open set in \({\mathbb {R}}^n\). Define an interpolation operator \({\mathcal {I}}\)
where the interpolation operator satisfies the reproducing property
In the refinement, the mesh is given and refined at the same scale. The shape function \(\tilde{N}_I(x)\) can be regarded as a scaled reference shape function \(\tilde{N}^0(\xi )\) in the parent domain with \(\xi =(x-x_I)/h\). Suppose the boundary \(\partial \varOmega \) is smooth enough, then the following interpolation estimate holds
where h is the mesh size and \(C_1\) is a constant independent of h.
Proof
We have
where \(\varLambda (x)=\{ I| x\in \text {supp}\{\tilde{N}_I\}\cap \varOmega \}\) is an index set supporting all the shape functions covering x.
Take Taylor expansion of \(u(x_I)\) at x:
where \(D^m u\) represents the m-th order derivative of u(x), and \(0<\theta <1\) depends on x and \(x_I\). Substituting (61) into (60) yields
According to the reproducing property, we have
particularly, for \(y=x\)
So the deviation between u(x) and its approximation \({\mathcal {I}}u\) is
It deduces the following estimate
Denote the support size of shape function \(\tilde{N}_I(x)\) as rh, the maximum value of \(|\tilde{N}_I(x)|\) is M. Note that the absolute value of C-HiDeNN shape function is generally less than 1 (in the element domain). The following estimate holds
We can conclude that \(\exists \ C_0, \text {with}\ 0< C_0 < \infty \), such that
Next, we consider the derivative of \(u(x)-{\mathcal {I}} u\). Taking the derivative of \({\mathcal {I}} u\) yields
The derivative of \(u(x)-{\mathcal {I}} u\) is
Taking the derivative of (64) with respect to x yields
For \(y=x\), we have
This leads to
The reference shape function \(\tilde{N}^0(\xi )\) is defined in the parent domain with \(\xi =(x-x_I)/h\), independent of the element size h. Thus
The estimate for the derivative of \(u(x)-{\mathcal {I}} u\) becomes
where K is the maximum value of \(|D^1_\xi \tilde{N}^0(\xi )|\). We can conclude that \(\exists \ C_1 \text {with}\ 0< C_1 < \infty \), such that
. \(\square \)
For a linear second-order elliptic boundary value problem, we can use Céa’s inequality [27], therefore
where \(V^h\) represents the C-HiDeNN interpolation space.
Appendix F Illustration of C-PGD/TD solution procedure
The C-PGD solution can be solved through the so-called alternating fixed point algorithm as traditional PGD method [40], or through a minimization problem [44]. Let us consider a 3D Poisson problem as below
where \(\nabla \) denotes the gradient operator, b is the body source term with the assumption that \(b(x,y,z)=b_x(x)b_y(y)b_z(z)\). The domain \(\varOmega \) is a regular domain, \(\varvec{n}\) is the normal vector on the surface. The weak form of the problem can be obtained by multiplying both sides of the above equation by the test function \(\delta u\), which reads
Assuming the solution can be accurately approximated by M modes and \(M-1\) modes are already computed, the following decomposition for u and \(\delta u\) read
and
For notation simplification, we can omit the superscript M in the above equations. Therefore,
and
The convolution approximation is then applied to each of the separated 1D functions \(u_x(x), u_y(y), u_z(z)\), which reads
where \(\tilde{\varvec{N}}\) is the convolution shape function vector formed by the 1D convolution functions in a patch domain. \(\varvec{u}_x, \varvec{u}_y, \varvec{u}_z\) are the associated nodal solution vectors in three directions. Similarly, the same convolution approximation can be applied to \(\delta {u}_x, \delta {u}_y, \delta {u}_z\) and \({u}_x^{(m)}, {u}_y^{(m)}, {u}_z^{(m)}\).
The \(u_x(x), u_y(y), u_z(z)\) can be computed by alternatively fixing two of the functions. For example, we can assume \(u_y, u_z\) are given by assumed values, then \(\delta u_y=0\) and \(\delta u_z=0\), \(\delta u(x,y,z)= \delta u_x u_y u_z \), the weak form (81) for solving \(u_x\) under the decomposition reads
with
and
where \(\tilde{\varvec{B}}_x=\partial \tilde{\varvec{N}}_x/\partial x\), \(\tilde{\varvec{B}}_y=\partial \tilde{\varvec{N}}_y/\partial y\), \(\tilde{\varvec{B}}_z=\partial \tilde{\varvec{N}}_z/\partial z\) Therefore, the Eq. (87) becomes
The final discretized form for solving \(u_x\) is
where \(\tilde{\varvec{K}}_{xx}=\int _{\varOmega _x} \tilde{\varvec{B}}_x^T\tilde{\varvec{B}}_x \ dx\), \(\tilde{\varvec{K}}_{yy}=\int _{\varOmega _y} \tilde{\varvec{B}}_y^T\tilde{\varvec{B}}_y \ dy\), \(\tilde{\varvec{K}}_{zz}=\int _{\varOmega _z} \tilde{\varvec{B}}_z^T\tilde{\varvec{B}}_z \ dz\), \(\tilde{\varvec{M}}_{xx}=\int _{\varOmega _x} \tilde{\varvec{N}}_x^T\tilde{\varvec{N}}_x \ dx\), \(\tilde{\varvec{M}}_{yy}=\int _{\varOmega _y} \tilde{\varvec{N}}_y^T\tilde{\varvec{N}}_y \ dy\), \(\tilde{\varvec{M}}_{zz}=\int _{\varOmega _z} \tilde{\varvec{N}}_z^T\tilde{\varvec{N}}_z \ dz\), with \(\varOmega =\varOmega _x\times \varOmega _y\times \varOmega _z\).
Rearranging the above equation leads to the following linear system of equations
with
By solving the above equation, an estimate of \(\varvec{u}_x\) and therefore \(u_x\) can be obtained. Similarly, we can solve for \(u_y\) and \(u_z\). This alternative fixed point procedure should be repeated until the convergence the product of \(u_x u_y u_z\). This is the solution for one C-PGD mode and can be used for computing incrementally all the modes by varying M from 1 to a given number. The total number of modes can be determined by the convergence criterion: \(\Vert \sum _{m=1}^{M-1} u^{(m)}_x(x) u^{(m)}_y(y) u^{(m)}_z(z) -u^{(M)}_x(x) u^{(M)}_y(y) u^{(M)}_z(z)\Vert \) is small enough.
By adopting the TD definition in [39], the C-TD can use the same solution procedure to give a rough estimate number of modes and then optimize all the current modes together. The advantages in doing so is to reduce the necessary number of modes. In our work, we are interested in the final convergent accuracy of C-PGD/TD with comparison to traditional FEM and PGD/TD. The number of modes is ignored. Hence, we do not distinguish the C-PGD and C-TD in this paper as they both can give a better accuracy than traditional methods with the proposed convolution approximation.
Appendix G Error bound analysis of different methods
Assuming a convex potential energy \(\varPi \) exits for a problem, the numerical solution to the problem can be obtained by solving the following minimization problem
where \(V^h\) denotes a given approximation space. Now, consider two approximation spaces \(V^h_1\), \(V^h_2\) with \(V^h_1 \subset V^h_2\). Their corresponding solutions by solving the above problem are respectively \(u^h_1\) and \(u^h_2\). It is easy to know that the optimized potential energy has following relationship
where \(u^{\text {Ext}}\) is the exact solution. By the convexity of the problem, we have
By analogy, denoting the approximation spaces of TD with infinite modes, C-TD with infinite modes, FEM, C-FEM, C-HiDeNN, DeNN respectively by \(V^{\text {TD}}\), \(V^{\text {C-TD}}\), \(V^{\text {FEM}}\), \(V^{\text {C-FEM}}\), \(V^{\text {C-FEM}}\), \(V^{\text {C-HiDeNN}}\), \(V^{\text {DeNN}}\), with the following relationship
The optimized potential energy follows
Therefore, we have
Here, since we are only interested in the approximation space of TD/PGD, we do not distinguish the two methods in this analysis.
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
Lu, Y., Li, H., Zhang, L. et al. Convolution Hierarchical Deep-learning Neural Networks (C-HiDeNN): finite elements, isogeometric analysis, tensor decomposition, and beyond. Comput Mech 72, 333–362 (2023). https://doi.org/10.1007/s00466-023-02336-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-023-02336-5