Abstract
The multiple scale dynamics of a periodic chain composed of nonlinear mass-in-mass cells is studied. Based on a continuous approach of the one-dimensional chain, dispersion equation is obtained which provides the general form of solutions of the linearized system. Taking into account a single harmonic of the chain around a 1:1 resonance with a targeted mode, fast dynamics of the system leads to the detection of the slow invariant manifold and its stability borders. Slow dynamics permits to predict singularities and equilibrium points leading to frequency responses. The developments predict periodic and non-periodic responses and permit to tune parameters of the chain for the aim of localization of vibrating energy and design of periodic or non-periodic responses.
Similar content being viewed by others
References
Veselago V (1968) The electrodynamics of substance with simultaneously negative values of \(\varepsilon\) and \(\nu\). Sov Phys USPEKHI-USSR 10:509, 04
Thompson G (1955) Unusual waveguide characteristics associated with the apparent negative permeability obtainable in ferrites. Nature 175:1135–1136
Bertoldi K, Vitelli V, Christensen J, van Hecke M (2017) Flexible mechanical metamaterials. Nat Rev 2:17066
Chen Y, Hu G, Huang G (2016) An adaptive metamaterial beam with hybrid shunting circuits for extremely broadband control of flexural waves. Smart Mater Struct 25:105036, 10
Brillouin L (1946) Wave propagation in periodic structures. McGraw-Hill Book Company Inc
Wang X (2014) Dynamic behaviour of a metamaterial system with negative mass and modulus. Int J Solids Struct 51:1534–1541, 04
Huang T-Y, Shen C, Jing Y (2016) Membrane- and plate-type acoustic metamaterials. J Acoust Soc Am 139(6):3240–3250
Baguet S, Nguyen V, Grenat C, Lamarque C-H, Dufour R (2019) Nonlinear dynamics of micromechanical resonator arrays for mass sensing. Nonlinear Dyn 95:01
Jezequel L, Lamarque C-H (1991) Analysis of non-linear dynamical systems by the normal form theory. J Sound Vib 149:429–459, 09
Flosi J, Lamarque C-H, Ture Savadkoohi A (2022) Periodic solutions of approximated normal forms and simplification of normal transform by using Gröbner basis. Appl Math Model 109:808–818
Lazarov B, Thomsen J (2009) Using high-frequency vibrations and non-linear inclusions to create metamaterials with adjustable effective properties. Int J Non-linear Mech 44:90–97, 01
Bilal OR, Foehr A, Daraio C (2020) Enhancement of deep-subwavelength band gaps in flat spiral-based phononic metamaterials using the trampoline phenomena. J Appl Mech 87
Peng H, Pai P (2014) Acoustic metamaterial plates for elastic wave absorption and structural vibration suppression. Int J Mech Sci 89:350–361, 12
Mitchell S, Pandolfi A, Ortiz M (2013) Metaconcrete: designed aggregates to enhance dynamic performance. J Mech Phys Solids Submitted: 01
Zhong R, Zong Z, Pai P, Ruan X (2019) Multi-stopband negative stiffness composite column design for vibration absorption. Thin-Walled Struct 144:106330, 11
da Silveira Zanin C, Ture Savadkoohi A, Baguet S, Dufour R, Hurel G (2022) Nonlinear vibratory energy exchanges in a meta-cell. Int J Non-Linear Mech 146:104148
Cveticanin L, Zukovic M (2017) Negative effective mass in acoustic metamaterial with nonlinear mass-in-mass subsystems. Commun Nonlinear Sci Numer Simul 51:89–104, 10
Rodrigues GS, Weber HI (2017) Elastic metamaterials analysis: simple and double resonators
Cveticanin L, Zukovic M, Cveticanin D (2018) Influence of nonlinear subunits on the resonance frequency band gaps of acoustic metamaterial. Nonlinear Dyn 93:08
Romeo F, Rega G (2006) Wave propagation properties in oscillatory chains with cubic nonlinearities via nonlinear map approach. Chaos Solitons Fractals 27:606–617, 02
Manevitch L (2001) The description of localized normal modes in a chain of nonlinear coupled oscillators using complex variables. Nonlinear Dyn 25:95–109, 07
Nayfeh A (1973) Perturbation methods, 01
Ture Savadkoohi A, Lamarque C-H, Weiss M, Vaurigaud B, Charlemagne S (2016) Analysis of the 1:1 resonant energy exchanges between coupled oscillators with rheologies. Nonlinear Dyn 86:2145–2159
Labetoulle A, Ture Savadkoohi A, Gourdon E (2022) Detection of different dynamics of two coupled oscillators including a time-dependent cubic nonlinearity. Acta Mech 233:259–290
Starosvetsky Y, Gendelman O (2008) Strongly modulated response in forced 2dof oscillatory system with essential mass and potential asymmetry. Physica D 237:1719–1733, 08
Lamarque C, Gendelman O, Ture Savadkoohi A, Etcheverria E (2011) Targeted energy transfer in mechanical systems by means of non-smooth nonlinear energy sink. Acta Mech 221:175–200, 09
Charlemagne S, Lamarque C-H, Ture Savadkoohi A (2017) Vibratory control of a linear system by addition of a chain of nonlinear oscillators. Acta Mech 228:09
Gendelman O, Musienko A, Vakakis A, Bergman L (2003) Dynamic interaction of a semi-infinite linear chain of coupled oscillators with a strongly nonlinear end attachment. Physica D 178:1–18, 04
Lazarov B, Jensen J (2007) Low-frequency band gaps in chains with attached non-linear oscillators. Int J Non-linear Mech 42:1186–1193, 12
Vakakis A, King M, Pearlstein A (1994) Forced localization in a periodic chain of non-linear oscillators. Int J Non-linear Mech 29:429–447, 05
Bacigalupo A, Lepidi M, Gnecco G, Vadalà F, Gambarotta L (2019) Optimal design of the band structure for beam lattice metamaterials. Front Mater 6:2, 01
Campana M-A, Ouisse M, Sadoulet-Reboul E, Ruzzene M, Neild S, Scarpa F (2020) Impact of non-linear resonators in periodic structures using a perturbation approach. Mech Syst Signal Process 135:106408, 01
Wu L, Li Y (2021) Harnessing bulging or sloshing modes to design locally resonant liquid-solid metamaterials. J Sound Vib 510:116280, 06
Barnhart M, Xu X, Yangyang C, Zhang S, Song J, Huang G (2018) Experimental demonstration of a dissipative multi-resonator metamaterial for broadband elastic wave attenuation. J Sound Vib 09
Nimmagadda C, Matlack K (2018) Thermally tunable band gaps in architected metamaterial structures. J Sound Vib 439:09
Bitar D, Ture Savadkoohi A, Lamarque C-H, Gourdon E, Collet M (2020) Extended complexification method to study nonlinear passive control. Nonlinear Dyn 99:01
Acknowledgements
The authors would like to thank the following organizations for supporting this research: (i) The “Ministère de la transition écologique” and (ii) LABEX CELYA (ANR-10-LABX-0060) of the “Université de Lyon” within the program “Investissement d’Avenir” (ANR-11-IDEX-0007) operated by the French National Research Agency (ANR).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1
1.1 Continuous approach of the system with modal projection
Let us consider the eigenfunction of the system which is defined in Eq. 23. We normalize this function via
leading to \(\tilde{\rho }_k=\sqrt{2/L}\). We define:
\(\omega _k\) being defined in Eq. 20.
We can project second equation of Eq. 7 on the \(k^{th}\) mode as:
New complex variables of Manevitch [21] are introduced as:
We are looking for the SIM and set \(\nu =\omega _k\).
Taking the first harmonic of Eq. 53 expressed with complex variables provides:
We seek for fixed points (see Eq. 30) so we can write:
We introduce complex variables in polar domain as:
with \(\mathcal {N}_1,\mathcal {N}_2 \in \mathbb {N}^2\) and \(\vartheta _1,\vartheta _2 \in \mathbb {Z}^2\). Finally, this approach leads to the SIM expression:
The SIM does not depend to space variable x because of the modal projection. The expression is compared to numerical results using Runge Kutta algorithm in free and forced cases using the same methodology as in Sect. 4.
In the free case shown in Fig. 24 for the first mode (\(k=1\)) and for the same physical parameters listed in configuration 2 in Table 1 and initial deformation defined as \(u_j(\tau =0)=0.13\cos (\omega _1(j-1))\), we can observe that the numerical results (in red line) are not really following the SIM (in green line) defined in Eq. 58 except in the part close to \(\mathcal {N}_1(\tau )=\mathcal {N}_2(\tau )=0\), when all masses are in small displacements and system behave in a classical manner. We can observe that there are jumps in the numerical curve creating steps like aspect on the curve. In fact, these jumps correspond to the crossing of some cells through one bifurcation point. Indeed, each cell is at a different level of energy characterized by different amplitudes of \(N_1(x,\tau )\) and \(N_2(x,\tau )\), as defined in first part in Eq. 27. For this reason, each cell is susceptible to cross through bifurcation points at different instants than another cells. When projecting on the mode, if some of the cells are in the unstable zone, then when a cell faces a bifurcation point, there will be a jump in the numerical curve obtained after modal projection. This phenomena is the cause of the gap between the numerical integration and the SIM that is appearing “step by step” like on the numerical curve. This is confirmed by superposing time history of numerically obtained \(\sqrt{\frac{2}{L}}\mathcal {N}_2(\tau )\) of projected discrete system with different numerically obtained discrete \(N_2(x=j,\tau )\) time histories of twenty five first cells of the chain in Fig. 25. This figure illustrates well the influence of each cell on the modal projection results.
Same results are observable in the forced case: when forcing is added to the system, the numerical results are also less in agreement with the SIM except in the part close to \(\mathcal {N}_1(\tau )=\mathcal {N}_2(\tau )=0\), when all masses are in small displacements and system behave in classical manner, as shown in Fig. 26 for excitation around the third modal frequency \(\omega _3\). This time, jumps happen around the first bifurcation point \((N_{21},N_{11})\) for some cells creating the visible non-regularities on the numerical curve obtained after modal projection of discrete system in Fig. 26. This is confirmed in Fig. 29 where time history of numerically obtained \(\sqrt{\frac{2}{L}}\mathcal {N}_2(\tau )\) of projected discrete system is compared to numerically obtained discrete \(N_2(x=j,\tau )\) time histories of twenty five first cells of the chain. Moreover, when some cell amplitudes of the chain are in the unstable zone of the SIM while others are behaving in a classical manner, Fig. 29 permits to understand that the modal projection is less effective because it has to traduce two different types of behaviors at the same time. When all cells are behaving in a classical manner close to \(N_1=N_2=0\) or at high amplitudes, spatial projection will provide good results, but does not have advantages in the analytical study compared to study in Sect. 3 while considering only one mode. We can remark for example from Fig. 29 that the numerical curve obtain after modal projection won’t be in agreement with the SIM when the distribution of the number of cells behaving in a classical manner close to \(N_1=N_2=0\) and the number of cell having a modulated response is equitable. The SIM obtained without projection of the continuous system provides better results (see Figs. 27, 28).
For these reasons, the initial study has been conducted without modal projection.
Appendix 2: Discrete consideration of the system
1.1 General analytical determination of discrete system modes
In this part, we are providing the dynamical study in a discrete way. Let us consider the L-cell discrete system described in Eq. 2. We express the system in the matrix form. The study is limited to the order \(\varepsilon ^0\). Equation 2 has the matrix form:
Following the same procedure as described in Eq. 14,15, the linearized system reads:
\({\textbf {Id}}_{L}\) being the LxL identity matrix. In order to use modal projection, we are interested by the eigenvalues and eigenvectors of the matrix \({\textbf {M}}_L\).
Let us evaluate eigenvalues of \({\textbf {M}}_L\). For this purpose, we define the tri-diagonal matrix:
We define \(\alpha =2-\lambda\) and using Gauss pivoting methodology, we decompose:
where:
and:
Finally, by re-injecting in Eq. 62 the previous simplified determinants, the following simple expression is obtained:
Let us now evaluate \(D_L=\det ({\textbf {B}}_L)\). In order to obtain non-trivial solutions of the system, we have to verify \(\det ({\textbf {A}}_L)=0\). Therefore, Eq. 65 leads to a second degree characteristic polynomial and we set the general expression of the solutions of the constant recursive sequence:
where \(X_1\) and \(X_2\) are the solutions of the characteristic polynomial.
We have the firsts relations:
Solving the system up to higher orders shows that \(X_1\) and \(X_2\) can’t have real expressions. Therefore, \(X_1\) and \(X_2\) must be conjugated complexes and we use polar expressions for the two unknowns:
with \(\eta ,\zeta \in \mathbb {R}^2\). Solving Eq. 67 leads to the two expressions:
We can now also remark that:
The solution of the second relation leads to \(\alpha =0\) or \(\alpha =2 \cos (\zeta )\). Since \(\alpha =2-\lambda\), we have the expression of the eigenvalues:
\(D_L\) reads:
We can finally recompose the expression of det(\({\textbf {A}}_L\)):
\(\det ({\textbf {A}}_L)\)=0 if \(\zeta =\frac{2k\pi }{L}\), \(k=0,1,\dots ,L-1\).
According to Eq. 71, we have \(L+1\) equations but we only need L solutions.
The solution \(\lambda =2\) can be eliminated as general solution for every chain because it can’t verify the conditions of periodicity without being linked to the number of cells of the chain (except for the solution \(\lambda =0\)). However, the eigenvalue \(\lambda =2\) can be solution when \(\frac{k}{L}=\frac{1}{4} \implies \zeta =\frac{\pi }{2} \implies \lambda =2\) which in this case verifies periodicity conditions. We can conclude that all solutions can be found solving:
with \(\zeta =\frac{2k\pi }{L}\), \(k=0,1,\dots ,L-1\).
Moreover, we can remark that:
Therefore, we only have \(\left\lfloor \dfrac{L}{2}\right\rfloor +1\) different eigenvalues, \(L\in \mathbb {N}^*\). The first eigenvalue is \(\lambda =0\), corresponding to the rigid body dynamic of the chain.
The treatment of \({\textbf {M}}_L.{\textbf {X}}=\lambda .{\textbf {X}}\) leads to the L eigenvectors. We can identify two cases:
-
L is uneven, implying that the only eigenvalue associated to only one eigenvector is \(\lambda _0\). Every other eigenvalues have two associated eigenvectors.
-
L is even, implying that 2 eigenvalues \(\lambda _0\) and \(\lambda _{\left\lfloor L/2\right\rfloor }\) have both one associated eigenvector. Every other eigenvalues have 2 associated eigenvectors. We note \({\textbf {P}}\) the orthonormal transfer matrix and we adopt the following decomposition:
$$\begin{aligned} \begin{array}{lcc} {\textbf {P}}=(\underbrace{{\textbf {P}}_0}_{\lambda _0}, \underbrace{{\textbf {P}}_1, {\textbf {P}}_2}_{\lambda _1},\ldots ,\underbrace{{\textbf {P}}_{2j-1},{\textbf {P}}_{2j}}_{\lambda _j},\ldots ,\underbrace{{\textbf {P}}_L}_{\lambda _{\left\lfloor L/2\right\rfloor } \hbox { if}\, L\, \hbox {is pair}})\\ \end{array} \end{aligned}$$(76)In all cases, we can use for the first eigenvector:
$$\begin{aligned} \begin{array}{lcc} {{\textbf {P}}}_0 = \sqrt{\frac{1}{L}} \begin{pmatrix} 1&\,&\dots&\,&\,&1 \end{pmatrix} ^T \end{array} \end{aligned}$$(77)
On the contrary, during the last mode of the chain with an even number of cells, all cells present equal amplitudes while each cell is out of phase with its neighbors. We can use:
This modal decomposition is caused by the periodicity conditions of the chain. Indeed, the decomposition of the physical behavior according to each eigenvalue needs two different eigenvectors and this joins the continuous approach: we showed that the general expression of principal displacement was \(U_l(x)=A_k\cos (\omega _k x) + B_k\sin (\omega _k x)\) (see Eq. 21). The determination of \(A_k\) and \(B_k\) can’t be done until an excitation is applied. Therefore, the general form of expression of deformation has to be described using \(\cos (\omega _kx)\) and \(\sin (\omega _kx)\) components. The same process is happening in the discrete study. Indeed, for each double eigenvalue \(\lambda _k=\omega _k^2\), the 2 eigenvectors can be assimilated to \(\sqrt{2/L}\cos (\omega _kx)\) and \(\sqrt{2/L}\sin (\omega _kx)\), as shown on Figs. 30 and 31. On the first figure, \({\textbf {P}}_1\) and \({\textbf {P}}_2\) are represented in red and black ‘+’ symbols respectively while \(\sqrt{2/L}\sin (\omega _1x)\) and \(\sqrt{2/L}\cos (\omega _1x)\) are plotted in red and black respectively. On the second figure, \({\textbf {P}}_5\) and \({\textbf {P}}_6\) are presented in red and black ‘+’ symbols while \(\sqrt{2/L}\sin (\omega _3x)\) and \(\sqrt{2/L}\cos (\omega _3x)\) are plotted in red and black respectively. Eigenvectors are obtained using “eig” function of Matlab\(^\circledR\) and system characteristics are configuration 1 in Table 1. Both discrete eigenvectors and continuous eigenfunctions have been normalized.
Modal frequencies obtain from continuous and discrete approach are compared for different sizes of chain in Table 2. When the number of cells is high (\(L=100\) cells for example), there is a good agreement between \(\omega _k\) defined in the Eq. 20 and the corresponding discrete pulsation \(\sqrt{\lambda _k}\). As expected, when the number of cells decreases, as well as when the considered harmonic increases, results obtained from discrete and continuous approaches depart.
1.2 The SIM of the discrete chain
We can now work from second equation of Eq. 59. \({\textbf {U}}\) and \({\textbf {V}}\) and their derivatives have the modal decomposition:
with \(n\in \mathbb {N}\). In our case, we only consider one \(j^{th}\) eigenvalue:
We project the second equation of Eq. 59 on the 2 eigenvectors associated to the \(j^{th}\) eigenvalue:
We can now introduce new complex variables of Manevitch [21] defined as:
with \(\nu =\sqrt{\lambda _j}\).
After keeping the first harmonic of the system, the complex variables are expressed in polar domain as:
The methodology of the study is the same as in Sect. 3.
1.3 Application to a simple case
We consider a chain with parameters reported in configuration 1 of Table 1. We suppose that initial deformation of the chain is defined as
\({\textbf {U}}(\tau =0)=\mathrm {u}_1(\tau =0){\textbf {P}}_1\) and that there is no other external excitation. Therefore, \({\textbf {U}}\) and \({\textbf {V}}\) are expressed only using \({\textbf {P}}_1\): we are able to set \(\mathrm {u}_2(\tau )=0\) and \(\mathrm {v}_2(\tau )=0\). We use the function “Eig” from Mathematica\(^\circledR\) in order to obtain exact eigenvalues and eigenvectors and we normalize the vectors of the modal base. So, we are here investigating the following case:
We obtain the following relation by injecting Eq. 84 in Eq. 81 and choosing \(j=1\):
Noticing that we’ve set \(L=10\), we can see that we obtain exactly the same equation as Eq. 53 during projection of the continuous system. The rest of the study will lead to the same equation of the SIM. Both continuous and discrete approaches provide similar results.
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
Flosi, J., Lamarque, CH. & Ture Savadkoohi, A. Different dynamics of a periodic mass-in-mass nonlinear chain during a single mode excitation. Meccanica 58, 67–95 (2023). https://doi.org/10.1007/s11012-022-01617-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11012-022-01617-2