Abstract
Phononic metamaterials have the capability to manipulate the propagation of mechanical waves. The traditional finite element (FE) analysis-based methods for predicting phonon dispersion curves are computationally expensive for structure optimization that may require thousands of design evaluations, especially when applied to high-resolution metamaterial models with a large number of elements. To address this issue, this paper presents two physics-embedded deep convolutional neural networks to predict the phonon dispersion curves of 2D metamaterials: (1) a transfer learning-based convolutional neural network (TLCNN) and (2) a physics-guided convolutional neural network (PGCNN). The physics knowledge is embedded into the two proposed models by modifying the loss function of the convolutional neural network (CNN). A comparative study among CNN, TLCNN and PGCNN is conducted to understand the relative merits. The effectiveness of the physics-embedded methods is evaluated by comparing the predicted normalized eigenfrequencies with those obtained by direct numerical simulations (DNS) using FE simulation. Furthermore, a comparison of the computational costs, which include computing time and memory usage, is presented among DNS, CNN, TLCNN and PGCNN. It is demonstrated that the proposed TLCNN and PGCNN have the potential to improve prediction accuracy with a limited amount of input data. However, the computational costs of the “offline” model training are still significant. Among the three methods, PGCNN shows the best prediction accuracies on both the training and test sets.
Similar content being viewed by others
Abbreviations
- CNN:
-
Convolutional neural network
- DL:
-
Deep learning
- DNN:
-
Deep neural network
- FE:
-
Finite element
- ML:
-
Machine learning
- NN:
-
Neural network
- PGCNN:
-
Physics-guided convolutional neural network
- RVE:
-
Representative volume element
- TL:
-
Transfer learning
- TLCNN:
-
Transfer learning-based convolutional neural network
References
Frenzel T, Köpfler J, Jung E, Kadic M, Wegener M (2019) Ultrasound experiments on acoustical activity in chiral mechanical metamaterials. Nat Commun 10(1):1–6
Yu X, Zhou J, Liang H, Jiang Z, Wu L (2018) Mechanical metamaterials associated with stiffness, rigidity and compressibility: a brief review. Prog Mater Sci 94:114–173
Wu L, Wang Y, Zhai Z, Yang Y, Krishnaraju D, Lu J, Wu F, Wang Q, Jiang H (2020) Mechanical metamaterials for full-band mechanical wave shielding. Appl Mater Today 20:100671
Joannopoulos JD, Johnson SG, Winn JN, Meade RD (2011) Photonic crystals. Princeton University Press, Princeton
Gonella S, To AC, Liu WK (2009) Interplay between phononic bandgaps and piezoelectric microstructures for energy harvesting. J Mech Phys Solids 57(3):621–633
Li Y, Baker E, Reissman T, Sun C, Liu WK (2017) Design of mechanical metamaterials for simultaneous vibration isolation and energy harvesting. Appl Phys Lett 111(25):251903
Kushwaha MS, Halevi P, Dobrzynski L, Djafari-Rouhani B (1993) Acoustic band structure of periodic elastic composites. Phys Rev Lett 71(13):2022
Kushwaha MS, Halevi P, Martinez G, Dobrzynski L, Djafari-Rouhani B (1994) Theory of acoustic band structure of periodic elastic composites. Phys Rev B 49(4):2313
Sigalas M, Economou EN (1993) Band structure of elastic waves in two dimensional systems. Solid State Commun 86(3):141–143
Sigalas M, Kushwaha MS, Economou EN, Kafesaki M, Psarobas IE, Steurer W (2005) Classical vibrational modes in phononic lattices: theory and experiment. Zeitschrift für Kristallographie-Crystalline Materials; 220(9–10): 765–809
Pennec Y, Djafari-Rouhani B (2016) Fundamental properties of phononic crystal. Phononic crystals. Springer, Berlin, pp 23–50
Schriemer HP, Cowan ML, Page JH, Sheng P, Liu Z, Weitz DA (1997) Energy velocity of diffusing waves in strongly scattering media. Phys Rev Lett 79(17):3166
Li X, Ning S, Liu Z, Yan Z, Luo C, Zhuang Z (2020) Designing phononic crystal with anticipated band gap through a deep learning based data-driven method. Comput Methods Appl Mech Eng 361:112737
Wang Z, Xian W, Baccouche MR, Lanzerath H, Li Y, Xu H (2022) Design of phononic bandgap metamaterials based on Gaussian mixture beta variational autoencoder and iterative model updating. J Mech Des 144(4):041705
Wang Z, Zhuang R, Xian W, Tian J, Li Y, Chen S, Xu H (2022) Phononic metamaterial design via transfer learning-based topology optimization framework. In: International design engineering technical conferences and computers and information in engineering conference. 2022. American Society of Mechanical Engineers
Hussein MI, Hulbert GM, Scott RA (2006) Dispersive elastodynamics of 1D banded materials and structures: analysis. J Sound Vib 289(4–5):779–806
Sigmund O, Søndergaard Jensen J (2003) Systematic design of phononic band–gap materials and structures by topology optimization. Philos Trans R Soc Lond Ser A Math Phys Eng Sci. 361(1806):1001–1019
Kobayashi F, Biwa S, Ohno N (2004) Wave transmission characteristics in periodic media of finite length: multilayers and fiber arrays. Int J Solids Struct 41(26):7361–7375
Sigalas M, Garcıa N (2000) Theoretical study of three dimensional elastic band gaps with the finite-difference time-domain method. J Appl Phys 87(6):3122–3125
Wang Y, Li F, Wang Y, Kishimoto K, Huang W (2009) Tuning of band gaps for a two-dimensional piezoelectric phononic crystal with a rectangular lattice. Acta Mech Sin 25(1):65–71
Sigalas MM, Economou EN (1992) Elastic and acoustic wave band structure. J Sound Vib 158(2):377–382
Tanaka Y, Tomoyasu Y, Tamura S-I (2000) Band structure of acoustic waves in phononic lattices: two-dimensional composites with large acoustic mismatch. Phys Rev B 62(11):7387
Hamian S, Yamada T, Faghri M, Park K (2015) Finite element analysis of transient ballistic–diffusive phonon heat transport in two-dimensional domains. Int J Heat Mass Transf 80:781–788
Leamy MJ, DiCarlo A (2009) Phonon spectra prediction in carbon nanotubes using a manifold-based continuum finite element approach. Comput Methods Appl Mech Eng 198(17–20):1572–1584
Hussein MI (2009) Reduced Bloch mode expansion for periodic media band structure calculations. Proc R Soc A: Math Phys Eng Sci 465(2109):2825–2848
Krattiger D, Hussein MI (2018) Generalized Bloch mode synthesis for accelerated calculation of elastic band structures. J Comput Phys 357:183–205
Zhao J, Li Y, Liu WK (2015) Predicting band structure of 3D mechanical metamaterials with complex geometry via XFEM. Comput Mech 55(4):659–672
Varnek A, Baskin I (2012) Machine learning methods for property prediction in chemoinformatics: Quo Vadis? J Chem Inf Model 52(6):1413–1437
Wang Z, Xu H,Li Y (2020). Material model calibration by deep learning for additively manufactured alloys. In: International symposium on flexible automation. 2020. American Society of Mechanical Engineers
Seeger M (2004) Gaussian processes for machine learning. Int J Neural Syst 14(02):69–106
Yao X, Wang Y, Zhang X, Zhang R, Liu M, Hu Z, Fan B (2002) Radial basis function neural network-based QSPR for the prediction of critical temperature. Chemom Intell Lab Syst 62(2):217–225
Xu L, Hoffman N, Wang Z, Xu H (2022) Harnessing structural stochasticity in the computational discovery and design of microstructures. Mater Des 223:111223
Bastek J-H, Kumar S, Telgen B, Glaesener RN, Kochmann DM (2022) Inverting the structure–property map of truss metamaterials by deep learning. Proc Natl Acad Sci 119(1):e2111505119
Meyer PP, Bonatti C, Tancogne-Dejean T, Mohr D Graph based metamaterials: deep learning of structure-property relations. Mater Desi, p 111175
Ji Q, Chen X, Liang J, Fang G, Laude V, Arepolage T, Euphrasie S, Martínez JAI, Guenneau S, Kadic M (2022) Deep learning based design of thermal metadevices. Int J Heat Mass Transf 196:123149
Qian X, Yang R (2021) Machine learning for predicting thermal transport properties of solids. Mater Sci Eng R Rep 146:100642
Wang T, Zhang C, Snoussi H, Zhang G (2020) Machine learning approaches for thermoelectric materials research. Adv Func Mater 30(5):1906041
Jin Y, He L, Wen Z, Mortazavi B, Guo H, Torrent D, Djafari-Rouhani B, Rabczuk T, Zhuang X, Li Y (2022) Intelligent on-demand design of phononic metamaterials. Nanophotonics
Liu Z, Jiang M, Luo T (2020) Leverage electron properties to predict phonon properties via transfer learning for semiconductors. Sci Adv 6(45):eabd1356
Sadat SM, Wang RY (2020) A machine learning based approach for phononic crystal property discovery. J Appl Phys 128(2):025106
Miao X-B, Dong H, Wang Y-S (2021) Deep learning of dispersion engineering in two-dimensional phononic crystals. Eng Optim, pp. 1–15
Ouyang Y, Yu C, He J, Jiang P, Ren W, Chen J (2022) Accurate description of high-order phonon anharmonicity and lattice thermal conductivity from molecular dynamics simulations with machine learning potential. Phys Rev B 105(11):115202
Roscher R, Bohn B, Duarte MF, Garcke J (2020) Explainable machine learning for scientific insights and discoveries. IEEE Access 8:42200–42216
Wang J-X, Wu J-L, Xiao H (2017) Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data. Phys Rev Fluids 2(3):034603
Zhao W (2017) Research on the deep learning of the small sample data based on transfer learning. In: AIP conference proceedings. 2017. AIP Publishing LLC.
Xu Y, Weng H, Ju X, Ruan H, Chen J, Nan C, Guo J, Liang L (2021) A method for predicting mechanical properties of composite microstructure with reduced dataset based on transfer learning. Compos Struct, p 275
Yamada H, Liu C, Wu S, Koyama Y, Ju S, Shiomi J, Morikawa J, Yoshida R (2019) Predicting materials properties with little data using shotgun transfer learning. ACS Cent Sci 5(10):1717–1730
Liu Z, Wu CT, Koishi M (2019) Transfer learning of deep material network for seamless structure–property predictions. Comput Mech 64(2):451–465
Jha D, Choudhary K, Tavazza F, Liao WK, Choudhary A, Campbell C, Agrawal A (2019) Enhancing materials property prediction by leveraging computational and experimental data using deep transfer learning. Nat Commun 10(1):5316
Wang D, Lu Z, Xu Y, Wang ZI, Santella A, Bao Z (2019) Cellular structure image classification with small targeted training samples. IEEE Access 7:148967–148974
Li X, Zhang Y, Zhao H, Burkhart C, Brinson LC, Chen W (2018) A transfer learning approach for microstructure reconstruction and structure-property predictions. Sci Rep 8(1):13461
Bostanabad R (2020) Reconstruction of 3D microstructures from 2D images via transfer learning. Comput-Aided Design, 128.
Bostanabad R, Zhang Y, Li X, Kearney T, Brinson LC, Apley DW, Liu WK, Chen W (2018) Computational microstructure characterization and reconstruction: review of the state-of-the-art techniques. Prog Mater Sci 95:1–41
Yang Z, Li X, Catherine Brinson L, Choudhary AN, Chen W, Agrawal A (2018) Microstructural materials design via deep adversarial learning methodology. J Mech Des 140(11):111416
Kim Y, Kim Y, Yang C, Park K, Gu GX, Ryu S (2021) Deep learning framework for material design space exploration using active transfer learning and data augmentation. npj Comput Mater; 7(1):140
Li X, Dan Y, Dong R, Cao Z, Niu C, Song Y, Li S, Hu J (2019) Computational screening of new perovskite materials using transfer learning and deep learning. Appl Sci 9(24):5510
Zhang R, Liu Y, Sun H (2020) Physics-guided convolutional neural network (PhyCNN) for data-driven seismic response modeling. Eng Struct 215:110704
Zhu Y, Zabaras N, Koutsourelakis P-S, Perdikaris P (2019) Physics-constrained deep learning for high-dimensional surrogate modeling and uncertainty quantification without labeled data. J Comput Phys 394:56–81
Sun L, Gao H, Pan S, Wang J-X (2020) Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data. Comput Methods Appl Mech Eng 361:112732
Raissi M, Perdikaris P, Karniadakis GE (2019) Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys 378:686–707
Tao F, Liu X, Du H, Yu W (2020) Physics-informed artificial neural network approach for axial compression buckling analysis of thin-walled cylinder. AIAA J 58(6):2737–2747
Silakorn P, Jantrakulchai N, Wararatkul N, Wanwilairat S, Kangkachit T, Techapiesancharoenkij R, Rakthanmanon T, Hanlumyuang Y (2022) Top-of-line corrosion via physics-guided machine learning: a methodology integrating field data with theoretical models. J Petrol Sci Eng 215:110558
Hong SH, Ou J, Wang, Y (2022) Physics-guided neural network and GPU-accelerated nonlinear model predictive control for quadcopter. Neural Comput Appl, pp 1–21
Biswas R, Sen MK, Das V, Mukerji T (2019) Prestack and poststack inversion using a physics-guided convolutional neural network. Interpretation 7(3):SE161–SE174
Biswas R, Sen MK, Das V, Mukerji T (2019) Pre-stack inversion using a physics-guided convolutional neural network. In: SEG international exposition and annual meeting. 2019. OnePetro
Daw A, Thomas RQ, Carey CC, Read JS, Appling AP, Karpatne A (2022) Physics-guided architecture (PGA) of LSTM models for uncertainty quantification in lake temperature modeling. In: Knowledge-guided machine learning. 2022, Chapman and Hall/CRC, 399–416
Yu Y, Yao H, Liu Y (2020) Structural dynamics simulation using a novel physics-guided machine learning method. Eng Appl Artif Intell 96:103947
Daw A., Thomas RQ, Carey CC, Read JS, Appling AP, Karpatne A (2020) Physics-guided architecture (PGA) of neural networks for quantifying uncertainty in lake temperature modeling. In: Proceedings of the 2020 SIAM international conference on data mining. 2020. SIAM.
Raissi M, Wang Z, Triantafyllou MS, Karniadakis GE (2019) Deep learning of vortex-induced vibrations. J Fluid Mech 861:119–137
Xu K, Darve E (2022) Physics constrained learning for data-driven inverse modeling from sparse observations. J Comput Phys 453:110938
Raissi M, Babaee H, Givi P (2019) Deep learning of turbulent scalar mixing. Physical Review Fluids 4(12):124501
Sun L, Wang J-X (2020) Physics-constrained bayesian neural network for fluid flow reconstruction with sparse and noisy data. Theor Appl Mech Lett 10(3):161–169
Jin H, Mattheakis M, Protopapas P (2022) Physics-informed neural networks for quantum eigenvalue problems. arXiv preprint arXiv:2203.00451
Chehimi M, Saad W (2022) Physics-informed quantum communication networks: a vision towards the quantum internet. arXiv preprint arXiv:2204.09233
Yao H, Gao Y, Liu Y (2020) FEA-Net: A physics-guided data-driven model for efficient mechanical response prediction. Comput Methods Appl Mech Eng 363:112892
Gao Y, Yao H, Wei H, Liu Y (2020) Physics-based deep learning for probabilistic fracture analysis of composite materials. In: AIAA Scitech 2020 Forum
Zobeiry N, Reiner J, Vaziri R (2020) Theory-guided machine learning for damage characterization of composites. Compos Struct 246:112407
Zhang E, Dao M, Karniadakis GE, Suresh S (2022) Analyses of internal structures and defects in materials using physics-informed neural networks. Sci Adv 8(7):eabk0644
Zhou T, Jiang S, Han T, Zhu S-P, Cai Y (2023) A physically consistent framework for fatigue life prediction using probabilistic physics-informed neural network. Int J Fatigue 166:107234
Danoun A, Prulière E, Chemisky Y (2022) Thermodynamically consistent Recurrent Neural Networks to predict non linear behaviors of dissipative materials subjected to non-proportional loading paths. Mech Mater 173:104436
Zhang R, Liu Y, Sun H (2020) Physics-informed multi-LSTM networks for metamodeling of nonlinear structures. Comput Methods Appl Mech Eng 369:113226
Karami M, Lombaert H, Rivest-Hénault D (2023) Real-time simulation of viscoelastic tissue behavior with physics-guided deep learning. Comput Med Imaging Graph 104:102165
Goodfellow I, Bengio Y, Courville A (2016) Deep learning. MIT press, Cambridge
Nair V, Hinton GE (2010) Rectified linear units improve restricted Boltzmann machines. In: Icml.
Gao Y, Mosalam KM (2018) Deep transfer learning for image-based structural damage recognition. Computer-Aided Civil and Infrastructure Engineering 33(9):748–768
El-Sayed MA, Estaitia YA, Khafagy MA 2013 Automated edge detection using convolutional neural network. Int J Adv Comput Sci Appl. 4(10).
Yaoming M, Ruibao T (1991) Elastic constants and phonon dispersion curves of tetragonal La2CuO4 single crystal. Chin Phys Lett 8(4):195
Hou X-H, Xu X-J, Meng J-M, Ma Y-B, Deng Z-C (2019) Elastic constants and phonon dispersion relation analysis of graphene sheet with varied Poisson’s ratio. Compos B Eng 162:411–424
Quiroga J, Mujica L, Villamizar R, Ruiz M, Camacho J (2017) Estimation of dispersion curves by combining effective elastic constants and SAFE method: A case study in a plate under stress. J Phys: Conf Ser. 2017. IOP Publishing
Bertoldi K, Boyce MC (2008) Wave propagation and instabilities in monolithic and periodically structured elastomeric materials undergoing large deformations. Phys Rev B 78(18):184107
Chan Y-C, Ahmed F, Wang L, Chen W (2021) METASET: exploring shape and property spaces for data-driven metamaterials design. J Mech Des 143(3):031707
Vogiatzis P, Chen S, Wang X, Li T, Wang L (2017) Topology optimization of multi-material negative Poisson’s ratio metamaterials using a reconciled level set method. Comput Aided Des 83:15–32
Huntington HB (1958) The elastic constants of crystals. Solid state physics 7:213–351
Wang Z, Xian W, Baccouche MR, Lanzerath H, Li Y, Xu H. A Gaussian mixture variational autoencoder-based approach for designing phononic bandgap metamaterials. In: International design engineering technical conferences and computers and information in engineering conference. 2021. American Society of Mechanical Engineers.
Paszke A, Gross S, Massa F, Lerer A, Bradbury J, Chanan G, Killeen T, Lin Z, Gimelshein N, Antiga L (2019) Pytorch: an imperative style, high-performance deep learning library. Adv Neural Inf Process Syst 32
Mosteller F, Tukey JW (1968) Data analysis, including statistics. Handb Soc Psychol 2:80–203
Acknowledgements
HX gratefully acknowledge the financial support of the start-up funding from the University of Connecticut. YL gratefully acknowledges financial support from the Air Force Office of Scientific Research through the Air Force's Young Investigator Research Program (FA9550-20-1-0183; Program Manager: Dr. Ming-Jen Pan), the National Science Foundation (CMMI-1934829 and CAREER-2046751), and 3M’s Non-Tenured Faculty Award.
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.
Appendix
Appendix
1.1 Hyperparameters related to the structure of three models
See Table 3.
1.2 Training result of the structure-elasticity model (the source model of the TL model)
Figure 7 shows the accuracy of the structure-elasticity model (source model). The predicted elastic stiffness constants against the ground truth values obtained by FE simulation are plotted. The \({R}^{2}\) values are shown in Table 4. The structure-elastic stiffness constants model shows satisfactory prediction accuracies.
1.3 Convergence study for choosing \(\boldsymbol{\alpha }\)
A convergence study is conducted to determine the \(\alpha \) value in the loss function of PGCNN. We test a series of \(\alpha \) values with the same training and test sets. We compare the performances of these models by the metrics discussed in Sect. 4.3. Figure 8 shows the performances of PGCNN with different \(\alpha \) values. We observe the highest accuracy occurring when \(\alpha =10\). Therefore, we choose 10 as the value of \(\alpha \) in Eq. 19.
1.4 Convergence test on the choice of the characteristic length \({\varvec{l}}\)
When using reduced-order metamaterial unit sample, Eq. 18 does not hold and will be replaced by:
where \(\epsilon \) represents the residual error of using reduced-order metamaterial unit sample, \(l\) represents the characteristic length, which is defined as the side length of the metamaterial unit in pixel. When using the 250 \(\times \) 250 pixels metamaterial unit samples for the calculation of \(\overline{{\varvec{K}} }\), \(\overline{{\varvec{M}} }\) and \({\varvec{v}}\), \(\epsilon \) should be equal to 0. A convergence test is conducted for choosing the characteristic length \(l\) which balances the computational cost and the induced residual error. We randomly select 50 metamaterial unit samples from the dataset and calculate their corresponding \(\overline{{\varvec{K}} }\), \(\overline{{\varvec{M}} }\) and \({\varvec{v}}\) matrices with different characteristic length \(l\). The averaged residual error \(\epsilon \) of 50 samples are calculated and the averaged memory usage for the storage of \(\overline{{\varvec{K}} }\), \(\overline{{\varvec{M}} }\) and \({\varvec{v}}\) are recorded as well. As shown in Figure 9, with the characteristic length \(l\) increases, the memory usage for storing the physics-related matrices increases. The curve of residual error converges at \(l=20\). To balance the memory usage and the accuracy, we choose \(l=20\) as the characteristic length of the reduced-order metamaterial unit samples.
1.5 A zoom-in view of the predicted and ground truth phonon dispersion curves
The phonon dispersion curves of one representative structure are separated into three parts by different frequency regions: low frequency region, medium frequency region and high frequency region, as shown in Fig. 10.
1.6 Comparison among multiple purely data-driven CNN models
See Table 5.
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
Wang, Z., Xian, W., Li, Y. et al. Embedding physical knowledge in deep neural networks for predicting the phonon dispersion curves of cellular metamaterials. Comput Mech 72, 221–239 (2023). https://doi.org/10.1007/s00466-023-02328-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-023-02328-5