Access the full text.
Sign up today, get DeepDyve free for 14 days.
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 2019, VOL. 25, NO. 5, 447–462 https://doi.org/10.1080/13873954.2019.1659374 ARTICLE Port-Hamiltonian modelling of nonlocal longitudinal vibrations in a viscoelastic nanorod a b,c Hanif Heidari and H. Zwart a b School of Mathematics and Computer Sciences, Damghan University, Damghan, Iran; Department of Applied Mathematics, University of Twente, Enschede, The Netherlands; Department of Mechanical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands ABSTRACT ARTICLE HISTORY Received 18 March 2019 Analysis of nonlocal axial vibration in a nanorod is a crucial subject in Accepted 20 August 2019 science and engineering because of its wide applications in nanoe- lectromechanical systems. The aim of this paper is to show how these KEYWORDS vibrations can be modelled within the framework of port- Descriptor port-Hamiltonian; Hamiltonian systems. It turns out that two port-Hamiltonian descrip- viscoelastic; nonlocal tions in physical variables are possible. The ﬁrst one is in descriptor vibration form, whereas the second one has a non-local Hamiltonian density. In addition, it is shown that under appropriate boundary conditions these models possess a unique solution which is non-increasing in the corresponding ‘energy’, i.e., the associated inﬁnitesimal generator generates a contraction semigroup on a Hilbert space, whose norm is directly linked to the Hamiltonian. 1. Introduction The micro and nanoscale physical phenomena have diﬀerent properties from macro- scale [1–3]. Carbon nanotubes (CNTs) are allotropes of carbon. They have diameters as small as 1 nm and lengths up to several centimetres. CNTs have amazing mechanical and electrical properties such as high electrical conductivity, chemical stability, high stiﬀness and axial strength [4]. These excellent properties have led to wide practical application of CNTs in NanoElectroMechanical Systems (NEMS). Due to novel proper- ties and vast applications of CNTs in industry, there is a lot of research on static, buckling and vibration analysis of CNTs using the local and the nonlocal models [5]. For example, Li et al. investigate dynamics and stability of transverse vibrations of nonlocal nanorods [6]. Nonlocal longitudinal vibration of viscoelastic-coupled double- nanorod systems is studied by Karlicic et al. [7]. Heidari investigates controllability and stability analysis of a nanorod [2]. Many electrical, mechanical and electromechanical systems can be suitably modelled in port-Hamiltonian (pH) framework. This modelling exposes fruitful information on physical characteristics of the system such as the relation between the energy storage, dissipation, and interconnection structure [8,9]. CONTACT Hanif Heidari heidari@du.ac.ir School of Mathematics and Computer Sciences, Damghan University, Damghan, Iran © 2019 Informa UK Limited, trading as Taylor & Francis Group 448 H. HEIDARI AND H. ZWART This information is of great interest in analysing and simulating complex network system. Over the last years, many researchers worked on port-Hamiltonian systems, extending the theory and/or solving applied control problems, see, e.g., Jeltsema and Doria-Cerezo [10], Macchelli and Melchiorri [11], and Ramirez et al. [12]. For an overview and more details we refer the reader to [9]. To the best of our knowledge, in spite of a large amount of research on vibration of nanorod and pH systems, there is only little research on pH modelling of vibration of nanorods. In [13] we studied the problem, but there a pH formulation was found using non-physical variables. Therefore, pH modelling of vibration of an elastic nanorod using physical variables is considered in this paper. The rest of paper is organized as follows. In Section 2, a short review on nonlocal theory and governing equations are given. Section 3 presents the ﬁrst port-Hamiltonian formulation. This is in descriptor form, the existence of its solutions is done in Section 4.In Section 5, a second pH formulation is given. The relation between the two formulations is discussed in Section 6. We end the paper with the conclusions and discussion on future works. 2. Model formulation In this section, we recall from [7] the mathematical modelling of vibration in nanorods. We consider a nanorod with length , and cross-sectional area A which is depicted in Figure 1. In our case, the cross-sectional area is constant along the x-coordinate, but in general it could have arbitrary shapes along this x-coordinate. We assume that the material of a nanorod is elastic and homogeneous. Also, we consider the free longitudinal vibration of the nanorod in the x-direction. An inﬁnitesimal element of length dx is taken at a typical coordinate location x. Further, we take that the force N is the resultant of an axial stress σ acting internally on A, where σ is assumed to be uniform over the xx xx cross-section. The stress resultant N may vary along the length, and is also a function of time, i.e., N ¼ Nðx; tÞ. Using our assumptions, we ﬁnd that Nðx; tÞ¼ σ ðx; tÞdA ¼ σ ðx; tÞA: (1) xx xx In addition, an axially distributed force F is assumed, having dimensions of force per unit length, which results from external sources, either internally or externally applied. The equilibrium of forces in the x-direction is Figure 1. Schematic of the present study. MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 449 @N @ w N þ N þ dx Fdx ¼ dm; (2) @x @t where dm ¼ ρAdx is the mass of an inﬁnitesimal element and w is the displacement in the x-direction. Substituting dm ¼ ρAdx and simplifying (2) gives @N @ w ¼ F þ ρA : (3) @x @t Next, we model the stress–strain relation. Based on nonlocal Eringen’s theory, it is assumed that the stress at a point is related to the strain ( ) at all other points in the xx domain. The nonlocal constitutive equation for an elastic medium is as follows @ σ @ xx σ μ ¼ E þ τ ; (4) xx xx d xx @x @t where E is the elastic modulus, μ is the nonlocal parameter (length scales) [7] and τ is the viscous damping coeﬃcient of the nanorod. We remark that we assume all para- meters to be constant. We consider the following standard relation between the strain and w, see [14], @w ¼ : (5) xx @x By substituting Equations (1) and (5) into (4), the stress resultant for the nonlocal theory is obtained as 2 2 @ N @w @ w N μ ¼ EA þ τ ; (6) @x @x @x@t @ w where the last term, , is the strain rate in the nanorod. Finally, we consider an @x@t external force @w 2 2 F ¼ a w þ b (7) @t in which the parameter a is the stiﬀness coeﬃcient of the viscoelastic layer and the last term represents uniform damping, see [7]. In the following sections, we show that the Equations (3), (6), and (7) can be written in a port-Hamiltonian form. In some papers, one can ﬁnd one (scalar) equation describing the motion. To write the Equations (3), (6) into one equation, we have ﬁrst diﬀerentiate Equation (6) with respect to x and next use (3), to get the following equation of motion 2 2 2 2 3 @ w @ @ w @ w @ w ~ ~ F þ ρA μ F þ ρA ¼ EA þ τ ; (8) 2 2 2 2 2 @t @x @t @x @x @t which is mentioned in [7]. 3. Descriptor port-hamiltonian formulation As many physical models, our model can also be written in a port-Hamiltonian form. However, it is not the standard formal as for instance studied in [8], but there will 450 H. HEIDARI AND H. ZWART appear a non-invertible operator in front of the derivative of the state, i.e., it is of descriptor form. Hence, we show that for a suitable state z our model can be written as dz @ E ¼ P ðÞ H z þ PðÞ H z RðÞ H z (9) 1 e 0 e 0 e dt @x 2 n with E, H , and R bounded operators on the Hilbert space L ðð0; ,Þ; R Þ, P e 0 1 a symmetric n n matrix and P an anti-symmetric n n matrix both consisting only of 1, 0, and 1‘s, and E H and R þ R self-adjoint and non-negative. e 0 The state z that we choose is given by 0 1 @w B ρA C @t B C @ w B C z ¼ : (10) μρA B @t@x C @ A @w @x Equation (3) implies @ w @N @N @w 2 2 z_ ¼ ρA ¼ F ¼ a w b (11) @t @x @x @t @ b ¼ z a z z ; (12) 5 1 2 @x ρA where we used (10) and (7). Using Equations (10) and (11), the time derivative of z is written as 3 2 2 @ w @ @ N @w @ w 2 2 z_ ¼ μρA ¼ μ ðz_ Þ¼ μ a μ b μ ; (13) 3 2 2 2 @t @x @x @x @x @t@x where we used the assumption that the parameters are constants. Using Equation (6), this becomes @w @ w 2 2 z ¼ N ðEA þ a μÞ ðEAτ þ b μÞ 3 d @x @t@x EAτ þ b μ ¼ z ðEA þ a μÞz z : (14) 5 4 3 μρA Using (12) and the above equality, we ﬁnd that 0 1 0 00 0 ρA B 2 b @ C a 00 B C ρA @x B C EAτ μb d 2 B C Ez_ ¼ z; (15) 00 EA μa 1 B μρA C B C @00 00 A ρAμ @ 1 1 0 00 @x ρA ρAμ MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 451 0 1 100 00 B C 010 00 B C B C where E ¼ 001 00 . We write the matrix of the right-hand side as B C @ A 000 10 000 00 a product, 0 1 0 1 a 00 0 0 01 0 0 0 B 1 C 2 @ B C 0 00 0 1 b00 B ρA @x B C B C B C 00 EAτ μb 11 : d B00 00C B C μρA B C @ A 00 1 0 0 2 @ A 00 0 EA þ μa 0 0 10 0 00 0 0 1 @x From this we see that our model (15) can be written in the form (9) with 0 1 0 1 a 00 0 0 B 1 C B C 0 00 0 B C ρA B C B C B C P ¼ 00000 ; H ¼ 00 00 ; (16) B C 1 e B C μρA B C @ A 00000 @ A 00 0 EA þ μa 0 00 0 0 1 0 1 0 1 01 0 0 0 00 0 00 B C B C 10 0 0 0 0 b000 B C B C B C B C P ¼ 00 0 11 ; and R ¼ 00 EAτ þ μb 00 : (17) 0 0 d B C B C @ A @ A 00 1 0 0 00 0 00 00 10 0 00 0 00 It is easy to see that these expressions satisfy the conditions stated below (9). If we assume that z is a classical solution of (9), then H ðtÞ :¼ zðtÞ E H zðtÞdx satisﬁes e e the following equality hi 1 1 T T H ðtÞ¼ ðH zÞ P H z ðH zÞ ðR þ R ÞH zdx: (18) e e 1 e e 0 e 2 0 2 For our model this becomes 1 @w @ w @w 2 2 2 2 2 2 H ðtÞ¼ a wðx; tÞ þ ρA ðx; tÞ þ μρA ðx; tÞ þðEA þ μa Þ ðx; tÞ dx; 2 @t @t@x @x (19) 1 @w @w @ w 2 2 2 2 H ðtÞ¼ ðx; tÞNðx; tÞ b ðx; tÞ þ τ EA þ μb ðx; tÞ dx: (20) e d 2 @t @t @t@x We see that the ﬁrst term represents the change of Hamiltonian (H ðtÞ) through the boundary, whereas the integral term represents internal damping. The above power balance is very standard for port-Hamiltonian systems, see [8,9,15]. However, there are a few diﬀerences between the form (9) with E, P , P and R given 1 0 0 452 H. HEIDARI AND H. ZWART in (16)–(17) and the form studied by Jacob and Zwart [8]. The most obvious one is that E is non-invertible. Moreover, our P is not invertible. First results for port- Hamiltonian systems with a non-invertible term in front of the time derivative can be found in [16], but only for ordinary diﬀerential equations. In Villegas [17, Chapter 6] port-Hamiltonian systems with P non-invertible is treated. We will not follows this, but take a more direct approach. So, in the next section, we show that the model (15) possesses a unique solution which is non-increasing in the Hamiltonian H ðtÞ: 4. Existence of solutions We study the existence of solutions under the assumption that the rod is ﬁxed, i.e., @w @w ð0; tÞ¼ ð,; tÞ¼ 0: (21) @t @t As state space, we take all states with ﬁnite energy but satisfying the constraints 0 1 B C 2 4 B C Z :¼ z ¼ 2 L ðð0; ,Þ; R Þjz is absolutely continuous (22) t t 2 @ A dz dz 2 2 2 L ð0; ,Þ; z ð0Þ¼ z ð,Þ¼ 0; and μ ¼ z : 2 2 2 3 dx dx As inner product on Z , we take the inner product associated to the Hamiltonian (19), i.e., hz ; ~z i :¼hz ; H ~z i; (23) t t t e;1 t where the latter inner product is the standard inner product on L ðð0; ,Þ; R Þ and H e;1 is the upper four by four block of H , i.e., E H e e 0 1 a 00 0 B 0 00 C ρA B C H ¼ : (24) e;1 @ A 00 0 μρA 00 0 EA þ μa Lemma 4.1 Z is a closed subspace of L ðð0; ,Þ; R Þ. (25) Proof. Since all physical parameters in (24) are positive, the norm associated to the inner product (23) is equivalent to the standard norm on L ðð0; ,Þ; R Þ. This directly implies that if the sequence fz ; n 2 Ng converges in Z , then the ﬁrst, third, t;n t and fourth component converge in L ðð0; ,Þ; RÞ. Hence it remains to show that the second component converges as well. By (22) and the convergence of z , the 3;n third component of z , we have that t;n MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 453 ð ð x x 1 1 z ðxÞ¼ z ðτÞdτ ! z ðτÞdτ; x 2½0; ,: 2;n 3;n 3 μ μ 0 0 On the other hand, by assumption z ! z , and combining this with (25) gives that 2;n 2 dz z is absolutely continuous with z ð0Þ¼ 0 and μ ¼ z . Using this equation once 2 2 3 dx more together with the fact that z ð,Þ¼ 0 gives that z satisﬁes the condition of Z , 2;n 2 t and thus Z is closed subspace of L ðð0; ,Þ; R Þ. ⁏ From this lemma, we ﬁnd that Z with the inner product (23) is a Hilbert space. Next, we deﬁne the (candidate) inﬁnitesimal generator associated with our p.d.e. We refer to Chapter 5 and 6 of [8] for more detail on semigroup theory. For z 2 DðAÞ we deﬁne 0 1 0 00 0 ρA B C 2 b d B C a 00 ρA dx B C Az ¼ B C t 2 EAτ μb d 2 B C 00 EA μa 1 5 @ A μρA (26) 00 00 ρAμ ¼: A ; where 1 t DðAÞ¼fz 2 Z jthere exists z 2 H ð0; ,Þ s:t: A 2 Z g: (27) t t 5 1 t Since A is deﬁned implicitly, it is important to know that it is well deﬁned, i.e., the outcome Az is uniquely deﬁned. This is shown next. Lemma 4.2 The operator A with domain DðAÞ is well-deﬁned. Proof. So what we have to show is that the z needed to deﬁne A is unique. Let us z z t t assume that there are two, i.e., z ; ~z 2 H ð0; ,Þ are such that A and A 2 5 5 1 1 z ~z 5 5 Z : Since A is linear, we see that this implies that A 2 Z . So if we can show t 1 1 t z ~z 5 5 that for an arbitrary z 2 H ð0; ,Þ the condition A 2 Z implies that z ¼ 0, then 5 1 t 5 we have shown that A is well-deﬁned. Assume that there exists z 2 H ð0; ,Þ is such that A 2 Z . Using (26) and (22) 5 1 t dz dz dz d z 5 5 5 5 this implies that 2 H ð0; ,Þ, ð0Þ¼ ð,Þ¼ 0 and μ ¼ z . Since μ > 0 this 2 5 dx dx dx dx implies that z ¼ 0. ⁏ Theorem 4.3. The operator A deﬁned in (26) and (27) generates a contraction semigroup on Z . t 454 H. HEIDARI AND H. ZWART (28) Proof. Using Lumer-Phillips Theorem, see e.g. [18, Theorem II.3.15] or [8, Theorem 6.1.7], we have to show two properties of A, namely for all z 2 DðAÞ hAz ; z i þhz ; Az i 0; t t t t 1 1 and for all g 2 Z there exists an f 2 DðAÞ such that ðI AÞf ¼ g: We begin by showing (28). Using the deﬁnition of A and the inner product on Z hAz ; z i þhz ; Az i ¼hAz ; H z iþ hH z ; Az i t t t t t e;1 t e;1 t t 1 1 z z t t ¼hA ; H z iþhH z ; A i 1 e;1 t e;1 t 1 z z 5 5 0 1 0 1 *+ z z t t z z A t t A 1 1 @ A @ A ¼ ; H iþ hH ; ; z z e e 5 5 z z 5 5 0 0 5 t 2 t where the last equality is in L ð0; ,Þ; R Þ. Since A 2 Z , we have that 0 1 0 00 0 ρA 0 1 B 2 b d C a 00 B C ρA dx B C A EAτ μb t d 2 C @ A B z 00 EA μa 1 B μρA C B C 0 1 @00 00 A ρAμ d 1 1 0 00 dx ρA ρAμ 0 1 01 0 0 0 2 d B C 1 b00 B dxC B C ¼ 00 EAτ μb 11 H B C @ A 00 1 0 0 0 10 0 dx d z ¼ P þ P R H ; 1 0 0 e dx see (16) and (17). Since P þ P ¼ 0 and R þ R 0, we ﬁnd that 0 0 0 0 hAz ; z i þhz ; Az i t t t t 1 1 d d z z z z t t t t P H ; H iþhH ; P H 1 e e e 1 e z z z z dx 5 5 5 dx 5 ¼ z z ¼ 0; 2 5 ρA 0 MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 455 where we have used the boundary conditions of z . So we see that (28) holds. Next, we 0 1 0 1 show (29). g f 1 1 B C B C g f 2 2 B C B C Let g ¼ 2 Z be given. Then, we have to ﬁnd f ¼ 2 DðAÞ and @ A @ A g f 3 3 g f 4 4 f 2 H ð0; ,Þ such that f f ¼ g ; (30) 1 2 1 ρA df a f þ s f ¼ g ; (31) 1 2 2 2 dx s f þðEA þ μa Þf f ¼ g ; (32) 3 3 4 5 3 f þ f ¼ g ;; (33) 3 4 4 μρA EAτ þμb where s ¼ 1 þ b and s ¼ 1 þ . Furthermore, we have the conditions, see (22) 2 3 μρA df μ ¼ f ; f ð0Þ¼ 0 ¼ f ð,Þ: (34) 3 2 2 dx By considering Equation (32) and Equation (33), we have s EA þ μa 3 f f þ g 3 5 3 ¼ : (35) f g 4 4 μρA EAþμa Since s þ 0, we ﬁnd μρA μρA f ¼ f þ g ðEA þ μa Þg (36) 3 5 3 4 μρAs þ EA þ μa f ¼ ½ f þ g þ μρAs g : (37) 4 5 3 3 4 μρAs þ EA þ μa From (30) and (31) it follows that a df ð þ s Þf ¼a g þ g : 2 2 1 2 ρA dx Combining this with (34) and using (36) we ﬁnd the following diﬀerential equation in f and f 2 5 d f 0 a f ~g 2 12 2 ¼ þ ; (38) f a 0 f ~g dx 5 21 5 where a ; a are positive constants, and ~g ; ~g are a linear combination of g ; ; g . 12 21 1 2 1 5 The solution of (38) is given by 456 H. HEIDARI AND H. ZWART qﬃﬃﬃﬃﬃ 0 1 coshðλxÞ sinhðλxÞ f ðxÞ a 0 @ A qﬃﬃﬃﬃﬃ f ðxÞ 21 f ð0Þ 5 5 sinhðλxÞ coshðλxÞ 0 qﬃﬃﬃﬃﬃ 1 ð a coshðλðx τÞÞ sinhðλðx τÞÞ a ~ 21 g ðτÞ @ A qﬃﬃﬃﬃﬃ þ dτ; (39) ~g ðτÞ 0 2 sinhðλðx τÞÞ coshðλðx τÞÞ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ where λ ¼ a a and we used the ﬁrst boundary condition of (34). To satisfy 12 21 the second boundary condition, we have to solve rﬃﬃﬃﬃﬃﬃ 0 ¼ f ð,Þ¼ sinhðλ,Þf ð0Þþ 2 5 ð rﬃﬃﬃﬃﬃﬃ coshðλð, τÞÞ~g ðτÞþ sinhðλð, τÞÞ~g ðτÞ dτ: 1 2 0 21 Since λ and , are positive constants, this is solvable and so the f and f are given by 2 5 (39) with ð rﬃﬃﬃﬃﬃﬃ 1 a f ð0Þ¼ coshðλð, τÞÞ~g ðτÞþ sinhðλð, τÞÞ~g ðτÞ dτ: 1 2 sinhðλ,Þ a 0 12 Note that these functions lie in H ð0; ,Þ and f satisﬁes the boundary conditions of (34). Given these solutions, the functions f and f follows from (36) and (37), respectively. 3 4 Equation (30) gives f . Summarizing we see that I A is surjective, and so A generates a contraction semigroup on Z . ⁏ 5. Second hamiltonian formulation In this section we show that there is a second port-Hamiltonian formulation for the nanorod. Therefore, we use the boundary conditions already in the formulation. So we assume that w is zero at x ¼ 0 and x ¼ , for all time, see also (21). Using Equation (3) and (7) this implies that @N @N ð0; tÞ¼ 0; ð,; tÞ¼ 0; for all t 0: (40) @x @x We use these boundary conditions to solve (see (6)) @ N N μ ¼ f ; @x for f 2 L ð0; ,Þ.We ﬁnd, see e.g. [19, Section 7.5] NðxÞ¼ gðx; ζÞf ðζÞdζ (41) with (Green’s function) MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 457 coshðγxÞ cosh ðγ ðζ ,ÞÞ x < ζ gðx; ζÞ¼ ; (42) coshðγðx ,ÞÞ cosh ðγζÞ x > ζ sinhðγ,Þ 2 1 where γ ¼ μ . Choose now the state 0 1 @w @ A ρA z ¼ : (43) @t @w @x Using (3), (6) and (7) we ﬁnd (for τ ¼ 0) 0 1 0 1 ρA ρA B C 2 2 b @ @w B C b @N B a z z þ GðEA Þ C 1 2 _ a z z þ z ¼ @ A ¼ ρA @x @x ; 1 2 ρA @x @ A hi @ 1 @ w @x@t @x ρA where G is the mapping deﬁned by (41) and (42), i.e. ðÞ Gðf ÞðxÞ¼ gðx; ζÞf ðζÞdζ: (44) Hence our model can be written as 0 120 1 3 01 0 a 00 2 @ @ A4@ A 5 1 b z_ðtÞ¼ zðtÞ : (45) 0 0 @x ρA 0 0 00 EA G @x This we can write in the port-Hamiltonian format (9) with E the identity, 0 1 0 1 0 1 000 01 0 000 @ A @ A @ A P ¼ 001 ; P ¼ 100 ; R ¼ 0 b 0 ; (46) 1 0 0 010 00 0 000 and 0 1 a 00 @ A Hz ¼ z: (47) 0 0 ρA 00 EA G Note that since gðx; ζÞ¼ gðζ; xÞ > 0 for all x; ζ 2½0; ,, G is a self-adjoint bounded, strictly positive operator. Using this and the fact that the physical parameters are positive, we ﬁnd that H is a coercive operator on L ðð0; ,Þ; R Þ. Asin[8, Chapter 7] we choose as our state space Z ¼ L ðð0; ,Þ; R Þ with inner product hf ; gi ¼ hf ; HðgÞi; (48) where the latter is the standard inner product on L ðð0; ,Þ; R Þ. For z 2 DðAÞ we deﬁne 458 H. HEIDARI AND H. ZWART 0 120 1 3 01 0 a 00 2 @ @ A4@ A 5 1 b Az ¼ 0 0 z (49) @x ρA 0 0 00 EA G @x with domain 1 1 DðAÞ¼fz 2 Zjz 2 H ð0; ,Þ; z ð0Þ¼ z ð,Þ¼ 0; Gðz Þ2 H ð0; ,Þg: (50) 2 2 2 3 From Lemma 7.2.3 and Theorem 7.2.4 of [8] the following theorem follows. Theorem 5.1. For b 2 R, the operator A with domain DðAÞ as deﬁned in (49) and (50) generates a contraction semigroup on the state space Z. If b ¼ 0, then A generates a unitary group on Z. Note that we have written the domain of A in the standard form, verify e.g. [8, 2 2 Equation (7.22)] or [15]. However, since G maps L -functions onto H -functions, the last condition in (50) is always satisﬁed, and thus could be removed. In the formulation (45), and thus Theorem 5.1, we have assumed that τ ¼ 0. Using (6) we see that for τ 0 our model can be written as 0 1 @ τ B C d @ @ zðtÞ¼½P þ P R HðÞ zðtÞ @ AG 0 0ðÞ HzðtÞ 1 0 0 @x @x @x ρA (51) ¼: ½P þ P R HðÞ zðtÞ G SGðÞ HzðtÞ : 1 0 0 R @x Since τ ; ρ; A are positive constants, and G is a positive operator, the operator S is positive. Using Theorem 5.1 and Theorem 2.2 of [20], we see that under the same boundary conditions as formulated in (50) this model generates a contraction semi- group on the state space Z. 6. Relation between the two formulations In Sections 3 and 5 we have shown that the model of the nanorod as presented in Section 2 allows for two port-Hamiltonian formulations. These formulations are both leading to a well-posed diﬀerential equation, and so it is only natural to ask for the relation between these two. Let us begin by stating that it is not exceptional to have more than one Hamiltonian, see, e.g. [21]. In the study of partial diﬀerential equations, the knowledge of conserved quantities, e.g. Hamiltonian is very useful for gaining insight in the system. Thus, knowing more than one Hamiltonian is seen as a positive fact. For the formulation found in Section 3 the Hamiltonian is given by 1 @w @ w @w 2 2 2 2 2 2 H ðz Þ¼ a wðxÞ þ ρA ðxÞ þ μρA ðxÞ þðEA þ μa Þ ðxÞ dx; e t 2 @t @t@x @x whereas the Hamiltonian associated to the formulation in Section 5 equals MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 459 ð ð ð , , , 1 @w 1 @w @w 2 2 H ðzÞ¼ a wðxÞ þ ρA ðxÞ dx þ EA ðxÞgðx; ζÞ ðζÞdζdx: 2 @t 2 @x @x 0 0 0 Although they have the same unit [J] and are equal in the ﬁrst two terms, for μ0 they diﬀer in the last term(s). Best to see this is by noticing that there is an a in last term of H , whereas this missing in H . Since the last term in H comes from (6), in e 2 2 which the a is absent, we conclude that H and H measure diﬀerent quantities e 2 for μ0. When there is no damping, then both Hamiltonians are constant. Thus, along solutions, we will have that ð ð ð , , , @ w @w @w @w 2 2 μρA ðx; tÞ þðEA þ μa Þ ðx; tÞ dx EA ðx; tÞgðx; ζÞ ðζ; tÞdζdx ¼ c; @t@x @x @x @x 0 0 0 where c is a constant, only depending on the initial condition. As said above this relation does not follow from an equality like (6), but is a property of the complete model. In [8] there is no example with two Hamiltonians, and so it surprising that the model of the nanorod has two. Looking at the derivation of the model once more, we notice that the ﬁrst model cannot be derived when the parameters are spatially depending, see the third equality in (13). The second model has a natural extension to spatially depending coeﬃcients by replacing the left-hand side of (6) by @ @N N μ ; @x @x where μ ¼ μðxÞ > 0. Since this is a Sturm-Liouville operator, existence, uniqueness, and other properties of the diﬀerential equation @ @N NðxÞ μðxÞ ðxÞ¼ f ðxÞ @x @x arewellknown,see,e.g.[19, Section 7.5]. For instance, the solution map will again be a strictly positive operator, and so Equation (45) still hold (with another G). So we feel that the bi-Hamiltonian property only holds in the constant parameter case. 7. Conclusions and further work Concluding we see that we have derived two diﬀerent port-Hamiltonian formulations corresponding to the same diﬀerential equation. Since the original model was build under the assumption of constant parameters, we have kept this assumption throughout this paper. For many port-Hamiltonian systems the step from constant to spatial varying (physical) parameters is easy, see the examples in, e.g. [8]. However, for the model of the nanorod this is less obvious. For the model derived in Section 5 this is possible if one replaces the left-hand side of (6) by, see also the discussion in the previous section, 460 H. HEIDARI AND H. ZWART @ @N N μ : @x @x For the model derived Section 3 this is much less clear. However, this should only be done, when the correct nanorod model for spatial dependent coeﬃcients has been derived. If damping is present, i.e., b > 0, then the time-derivative of both Hamiltonians is non-positive. We believe that in this case both semigroups are strongly stable, i.e., the solutions converge to zero as time goes to inﬁnity. A possible proof could be to apply [18, Corollary V.2.22]. To check if the system is exponentially stable, the eigenvalues need to be calculated/estimated. We have only studied the nanorod under one set of boundary conditions. For standard port-Hamiltonian systems, all boundary conditions could somehow be treated in the same theorem. Here the situation is diﬀerent. For instance, when the boundary conditions wð0; tÞ¼ wð,; tÞ¼ 0 are replaced by no force at the boundary, i.e., Nð0; tÞ¼ Nð,; tÞ¼ 0, then for the formulation in Section 4 the boundary conditions must be removed from the state space, see (22), and enter the domain (27) as boundary conditions on z . In the formulation of Section 5, the expression of g changes, since the diﬀerential equation for N has to be solved under the condition Nð0Þ¼ Nð,Þ¼ 0. Another topic which we like to study in the future is the port-Hamiltonian formulation when one of the boundary conditions is non-zero, i.e., for instance when there is a control at the boundary. Since we had to put the boundary conditions into the state space (Section 3)or use it to reformulate our problem (Section 5), thisproblemisnon-trivial.Itisbynomeans clear that it will lead to a boundary control system, like standard port-Hamiltonian p.d.e.’sdo. Nomenclature Symbol Meaning Unit A cross sectional area m 2 2 a stiﬀness coeﬃcient of the light viscoelastic layer gms 2 1 damping coeﬃcient of the light viscoelastic layer b gs 1 2 E elastic modulus gm s strain - xx ~ axially distributed force gs μ nonlocal parameter m N resultant force of axial stress gms ρ mass density gm 1 2 σ axial stress gm s xx τ viscous damping s w displacement of nanorod in x direction m Acknowledgments We want to express our thanks to Serge Nicaise and Marius Tucsnak for their useful comments which really helped our research on this problem further. MATHEMATICAL AND COMPUTER MODELLING OF DYNAMICAL SYSTEMS 461 Disclosure statement No potential conﬂict of interest was reported by the authors. ORCID Hanif Heidari http://orcid.org/0000-0002-6321-3295 References [1] H. Heidari, H. Zwart, and A. Malek, Analysis of the three dimensional heat conduction model in nano- or micro-scale,in Proceedings of the 19th International Symposium on Mathematical Theory of Networks and Systems (MTNS), 2010, pp. 1877–1882. [2] H. Heidari, Dynamical analysis of an axially vibrating nanorod, Int. J. Appl. Math. 29 (2016), pp. 263–270. doi:10.12732/ijam.v29i2.9. [3] A. Malek, H. Heidari, and M. Vali, Artiﬁcial magnetic nano-swimmer in drug delivery,in 22nd Iranian Conference on Biomedical Engineering (ICBME), IEEE 2015, pp. 331–336. [4] V.N. Popov, Carbon nanotubes: Properties and application, Mater. Sci. Eng. 43 (2004), pp. 61–102. doi:10.1016/j.mser.2003.10.001. [5] M. Simsek, Vibration analysis of a single-walled carbon nanotube under action of a moving harmonic load based on nonlocal elasticity theory, Physica E 43 (2010), pp. 182–191. doi:10.1016/j.physe.2010.07.003. [6] C. Li, C.W. Lim, and J.L. Yu, Dynamics and stability of transverse vibrations of nonlocal nanobeams with a variable axial load, Smart Mater. Struct. 20 (2011), pp. 15–23. doi:10.1088/0964-1726/20/1/015023. [7] D. Karlicic, M. Cajic, T. Murmu, and S. Adhikari, Nonlocal longitudinal vibration of viscoelastic coupled double-nanorod systems, Eur. J. Mech. A-Solid 49 (2015), pp. 183–196. doi:10.1016/j.euromechsol.2014.07.005. [8] B. Jacob and H. Zwart, Linear port-Hamiltonian Systems on Inﬁnite-Dimensional Spaces, Vol. 233, Birkhäuser, Basel, 2012. [9] A. van der Schaft and D. Jeltsema, Port-Hamiltonian Systems Theory: An Introductory Overview, Now Publishers Incorporated, Delft, The Netherlands, 2014. [10] D. Jeltsema and A. Doria-Cerezo, Port-Hamiltonian formulation of systems with memory, Proc IEEE 100 (2012), pp. 1928–1937. doi:10.1109/JPROC.2011.2164169. [11] A. Macchelli and C. Melchiorri, Modeling and control of the Timoshenko beam. The distributed port Hamiltonian approach, SIAM J. Control Optim. 43 (2004), pp. 743–767. doi:10.1137/S0363012903429530. [12] H. Ramirez, B. Maschke, and D. Sbarbaro, Irreversible port-Hamiltonian systems: A general formulation of irreversible processes with application to the CSTR, Chem. Eng. Sci. 89 (2013), pp. 223–234. doi:10.1016/j.ces.2012.12.002. [13] H. Heidari and H. Zwart, Port-Hamiltonian Formulation of Nonlocal Longitudinal Vibration in Nanorod, Submitted, 2018. [14] S. Narendar and S. Gopalakrishnan, Axial wave propagation in coupled nanorod system with nonlocal small scale eﬀects, Compos.: Part B 42 (2010), pp. 2013–2023. doi:10.1016/j. compositesb.2011.05.021. [15] Y. Le Gorrec, H. Zwart, and B. Maschke, Dirac structures and boundary control systems associated with skew-symmetric diﬀerential operators, SIAM J. Control Optim. 44 (2005), pp. 1864–1892. doi:10.1137/040611677. [16] C. Beattie, V. Mehrmann, H. Xu, and H. Zwart, Port-Hamiltonian descriptor system, Math. Control Signals Syst. 30 (2018), pp. 17. doi:10.1007/s00498-018-0223-3. [17] J.A. Villegas, A port-Hamiltonian approach to distributed parameter systems, Ph.d. thesis, University of Twente, 2007. 462 H. HEIDARI AND H. ZWART [18] K.J. Engel and R. Nagel, One-parameter semigroups for linear evolution equations,in Graduate Texts in Mathematics Vol. 194, Springer-Verlag, New York, 2000. [19] A.W. Naylor and G.R. Sell, Linear operator theory in engineering and science,in Applied Mathematical Sciences, 2nd, Vol. 40, Springer-Verlag, New York-Berlin, 1982. [20] H. Zwart, Y. Le Gorrec, and B. Maschke, Building systems from simple hyperbolic ones, Syst. Control Lett. 91 (2016), pp. 1–6. doi:10.1016/j.sysconle.2016.02.002. [21] F. Magri, A simple model of the integrable Hamiltonian equation, J. Math. Phys. 19 (1978), pp. 1156–1162. doi:10.1063/1.523777
Mathematical and Computer Modelling of Dynamical Systems – Taylor & Francis
Published: Sep 3, 2019
Keywords: Descriptor port-Hamiltonian; viscoelastic; nonlocal vibration
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.