Composite Structures 252 (2020) 112480

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Laminated piezoelectric beam element for dynamic analysis of

piezolaminated smart beams and GA-based LQR active vibration control
J. Tian, Q. Guo, G. Shi ⇑
Department of Mechanics, Tianjin University, Tianjin 300354, China

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents an accurate two‐noded laminated piezoelectric beam element for the dynamic analysis and
Laminated piezoelectric beam element active vibration control of laminated composite beams with piezoelectric layers. A refined third‐order shear
Electromechanical coupling deformation plate theory is used to model the kinematics, and a piecewise linear interpolation is used to char-
Dynamic analysis acterize the electrical potential. The equations of motion are derived from Hamilton’s principle. The quasi‐
LQR‐based active vibration control
conforming element technique is adopted to evaluate the explicit element stiffness matrix. The resulting
Optimal weighting matrix
Genetic algorithm
two‐noded piezolaminated beam element is free from shear locking and the numerical integration. Linear
Quadratic Regulator (LQR) control scheme is used for the active vibration control and Genetic Algorism
(QA) is employed to determine the optimal weighting matrix used in the control performance index. The lam-
inated piezoelectric beam element is validated by numerical examples. The numerical results show that present
laminated piezoelectric beam element is very accurate in dynamic analysis of laminated piezoelectric beams
and it is an effective strategy to use QA to merely evaluate the optimal weighting matrix for the vibration con-
trol performance by fixing the weighting matrix for the control input cost as an identity matrix in the LQR‐
based active vibration control.

1. Introduction structures. In 1995, Mitchell and Reddy [6] proposed a hybrid plate
theory for the static analysis of laminated composite plates with piezo-
Piezoelectric materials possess the advantages of both positive and electric layers. The hybrid plate theory combines an equivalent single‐
inverse piezoelectric effects, high sensitivity and quick response. layer approximation for the mechanical field and a layerwise‐type
Hence, piezoelectric layers can be used as both sensors and actuators, approximation for the electric field in laminated piezoelectric plates,
and the smart beams composed of piezoelectric sensors and actuators in which the electromechanical coupling of piezoelectric materials is
have a wide range of applications in vibration control, noise attenua- taken into account. The wording of hybrid here means the combina-
tion, damage detection etc. in aerospace, automotive, infrastructures tion of the equivalent single‐layer theory (SLT) with the layerwise the-
and other industries [1,2]. Piezolaminated thin beam is also an impor- ory (LWT). The hybrid plate theory takes the advantages of the
tant type of smart structures in MEMS [3]. The reliable and accurate computational efficiency of SLT and the advantage of the computa-
prediction of the dynamic response of laminated composite beams tional accuracy of LWT. This type of laminated piezoelectric plate the-
with embedded or surface‐bonded piezoelectric layers presents some ories was also named as coupled layerwise plate theory [1]. Many
challenges since it requires the accurate and efficient modelling of studies showed that the hybrid plate theory is an efficient and accurate
electromechanical coupling in laminated piezoelectric beams [1,2]. plate theory for the global response analysis of laminated piezoelectric
The coupled discrete layer theory (DLT) is capable of accurately beams and plates [1,2,6].
modeling the layered behavior of both the electric field and the Various shear deformable plate theories can be used for the kine-
mechanical field in laminated piezoelectric beams and plates [1,4,5]. matic assumptions in the hybrid plate theory [1]. The first‐order shear
However, DLT is computationally expensive as its displacement vari- deformation plate theory (FSDT) of Reissner‐Mindlin plate theory is
ables are proportional to the number of laminae in a laminated piezo- the most common plate theory used for the SLT in the hybrid plate the-
electric structure. As a result, DLT is not a practical choice for the ory. The FSDT is simple and computational efficient in the dynamic
modeling of dynamic and control problems of laminated piezoelectric analysis of laminated composite beams and plates, but it also has a

⇑ Corresponding author.
E-mail address: [email protected] (G. Shi).

https://doi.org/10.1016/j.compstruct.2020.112480
Received 26 December 2019; Revised 18 April 2020; Accepted 14 May 2020
Available online 19 July 2020
0263-8223/© 2020 Elsevier Ltd. All rights reserved.
J. Tian et al. Composite Structures 252 (2020) 112480

drawback of accuracy as it could result in larger errors in the frequency displacement element method is the most common method to evaluate
prediction of higher vibration modes [7]. Therefore, the higher‐order element stiffness matrix. But these conventional finite elements could
shear deformation plate theories (HSDT) are preferred for the dynamic suffer from the accuracy degradation of the resulting strains and stres-
analysis of laminated piezoelectric beams and plates [1]. Several SLTs ses since the differentiation of the displacement field is used to evalu-
accounting for higher‐order shear deformations were proposed for the ate the element strains. The element formulations of the laminated
analysis of laminated composite plates [1,3,8]. By using three‐ piezoelectric beam elements reported recently [22–25] are still based
dimensional (3D) elasticity, Shi [9] proposed a refined third‐order on the assumed displacement method. Shi et al. [26–28] showed that
shear deformation plate theory, in which the averaged rotation of the assumed strain method which is based on the quasi‐conforming
the cross‐section across the plate thickness is used instead of the rota- element technique [29] is a very good approach to formulate accurate
tion defined at the plate middle surface used in other third‐order shear and efficient shear flexible composite beam and plate elements for the
deformation plate theory. The number of the field variables in the analysis of laminated composite beams and plates. Another salient fea-
refined third‐order shear deformation theory proposed by Shi is the ture of the assumed strain beam and plate elements is that they are not
same as that in Reddy plate theory. Nevertheless, it was shown that only free from shear locking but also free from the time consuming
the refined third‐order shear deformation theory proposed by Shi [9] numerical integration. By using the third‐order shear deformation
is able to characterize the boundary layer solutions of shear deform- beam theory proposed Bickford [30] (it can also be reduced from
able beams [10,11] and yields better results in dynamic analysis of Reddy plate theory [8]), Shi et al. [27,31] developed efficient two‐
laminated beams and plates [7]. Challamel [12] also showed that noded composite beam elements for the static and dynamic analysis
Shi plate theory [9] is the simplest gradient plate theory as it involves of laminated composite beams. The composite beam elements devel-
with only three stiffness parameters instead of four stiffness parame- oped by Shi et al. [27,31] yield accurate results in the static and
ters in other higher‐order plate theories. lower‐mode frequency analyses of composite beams. However, Wang
Carrera [13] proposed a Unified Formulation (CUF) for the model- [7] showed that the composite beam element based on Shi plate theory
ing of laminated structures. CUF allows choosing linear to fourth order can yield more accurate prediction of the higher‐mode frequencies
expansions in the thickness coordinate z to model transverse shear than the beam element based on Bickford‐Reddy beam theory.
deformation in terms of power of z or Legendre polynomials. CUF is The Linear Quadratic Regular (LQR) control scheme is a more effec-
capable of taking account for the transverse normal strain effect and tive strategy for the vibration suppression of various structures than
the interlaminar continuity of transverse shear stresses. Both ESL constant gain velocity feedback (CGVF) control and constant ampli-
and LWT approaches can be used in CUF to model the mechanical tude velocity feedback (CAVF) control [32,33]. In the LQR‐based
behavior of laminated structures. CUF can also be used for the hierar- active vibration control, the control gain for a full state feedback is
chical finite element modeling of laminated structures. Many research- determined by minimizing the functional of a control performance
ers employed CUF to implement a large variety of piezoelectric beam index defined by the linear combination of a quadratic control perfor-
and plate theories [14–16]. A number of laminated piezoelectric beam mance measure with a quadratic control input cost measure. The
elements [17,18] and laminated piezoelectric plate/shell elements weighting matrix used for the vibration control performance in the
[19,20] were developed by researchers using CUF. control performance index plays a very important role in the active
The modeling of transverse normal strain effect and the interlami- vibration control, and it is a difficult task to find the optimal weighting
nar continuity of transverse shear stresses results in more computa- matrices. Most of researchers used a trial and error approach to select
tional cost. The dynamic response of laminated piezoelectric beams feasible weighting matrices [22,25,34]. But on the other hand, some
is type of global response of structures. Wang and Shi [21] demon- studies showed that the genetic algorithm (GA) is an efficient algo-
strated that the influence of the transverse normal strain effect and rithm to evaluate the optimal weighting matrices used in the optimal
the interlaminar continuity conditions on the static response of the control performance index [33,35].
sandwich plate with so‐called hard core is not significant where the The objectives of this paper are two‐folded. The first one is to pre-
Young’s modulus of the sandwich core is in the same scale as that of sent an accurate and efficient two‐noded laminated piezoelectric beam
the face panels of the sandwich. It was also shown that the interlami- element for the dynamic analysis of piezolaminated composite beams
nar continuity condition of transverse shear stresses it not important in and the active vibration control of laminated beams with embedded or
the dynamic response analysis of laminated structures [1,2,6,7]. How- surface‐bonded piezoelectric plies. The second one is to study the effi-
ever, the computational efficiency is more meaningful in the active cient use of GA to evaluate the optimal weighting matrices in LQR‐
vibration control of piezolaminated smart structures since the active based active vibration control of piezolaminated smart beams. The
vibration control involves a large number of computations in the time refined third‐order shear deformation plate theory proposed by Shi
domain. Therefore, the hybrid plate theory incorporated with a simpli- [9] is used for the beam kinematic assumption, and a layerwise zigzag
fied third‐order shear deformation theory for kinematics and a linear approximation is adopted for the electrical field in the piezoelectric
zigzag approximation for potential is not only computationally effi- plies. The element equations are derived from the Hamilton’s principle
cient but also accurate enough in the dynamic analysis of laminated with the electric enthalpy which takes account of the electromechani-
piezoelectric structures. As aforementioned, the refined third‐order cal coupling of piezoelectric materials. The quasi‐conforming element
shear deformation theory proposed by Shi [9] is the simplest gradient technique [26,29] is used to explicitly evaluate the element stiffness
plate theory [12] and capable of yielding better dynamic analysis matrix. The resulting two‐noded laminated piezoelectric beam ele-
results than Reddy plate theory [7]. Hence, Shi plate theory can be ment is free from shear locking and the numerical integration. The
used for the efficient and accurate modeling of laminated piezoelectric two‐noded laminated piezoelectric beam element presented in this
beams and plates. paper is a new development from the composite beam element previ-
A quite number of laminated piezoelectric beam elements were ously developed [27,31], since the composite beam element presented
developed for the dynamic analysis and active vibration control of in this work takes account of electromechanical coupling of piezoelec-
laminated composite beams in the past three decades [1,3]. But tric materials and is based on a refined third‐order shear deformation
researchers are still trying to use different laminated beam/plate the- plate theory for beam kinematics. Furthermore, an effective control
ories and different element methods to develop better piezolaminated scheme for GA‐based vibration active control is presented in this
beam elements [22–25]. Besides the laminated plate theories used for new work. The results of numerical examples show that the two‐
the finite element modeling, the element method employed for ele- noded laminated piezoelectric beam element based on the refined
ment formulation is also a crucial factor for the accuracy and efficiency third‐order shear beam theory yields accurate results of the higher‐
of the resulting laminated piezoelectric beam elements. The assumed order mode frequencies. The numerical results also demonstrate that

2
J. Tian et al. Composite Structures 252 (2020) 112480

an effective strategy for LQR‐based optimal vibration control is to in which w0 and u0 are respectively the deflection and the axial dis-
employ GA to evaluate the optimal weighting matrix used for the placement at the beam reference plane with z = 0. Eq. (2) indicates
vibration control performance and at meanwhile fixing the weighting that transverse normal stress σ zz ¼ 0 is assumed in this plate theory.
matrix for the control input cost as an identity matrix. The advantage using the averaged rotation ϕ
x= ϕ
x (x, t) is that it is
capable of specifying the clamped boundary condition more properly
2. Kinematics and electric potential approximation of laminated than the rotation defined at the middle surface used in other higher‐
piezoelectric beams order shear deformation theories [11].
The in‐plane displacement defined in Eq. (3) leads to that the dom-
The hybrid plate theory for laminated piezoelectric plates [6], inant term of the beam bending strain is expressed in terms of the aver-
which combines an equivalent single layer theory for the kinematics aged rotation ϕ
x of the beam cross‐section. It was shown by Shi et al.
assumption with a layerwise‐type approximation for the electric poten- [27,31] that a higher‐order accuracy of the finite element modeling
tial, is adopted in this study. can be achieved if the bending strain is defined in terms of the curva-
ture of the deflection w0 in the finite element analysis of composite
2.1. The refined third-order shear deformation plate theory and the beams and plates based on the higher‐order shear deformation theo-
corresponding strains ries. To this end, by introducing a new variable γ x defined as
γ x ¼ @w
x
þϕ ð4Þ
@x
A typical laminated piezoelectric beam is illustrated in Fig. 1. The
left side of the figure is the schematic of the piezoelectric ply configu- the in‐plane displacement defined in Eq. (3) can be recast as
ration with i = 1, 2 … m, and the right side of the figure is the sche-
uðx; z; tÞ ¼ u0 ðx; tÞ  zð@w
@x
0
 γ x Þ þ ð14z  3h52 z3 Þγ x ð5Þ
matic of the laminated composite beam with j = 1, 2 … n, where m is
the number of piezoelectric plies and n is the total number of laminae Eqs. (2) and (5) yield the axial strain and transverse shear strain of
with m ≤ n. In Fig. 1, u ¼ uðx; z; tÞ is the in‐plane displacement in the x‐ a beam bending in x–z plane taking the following respective form
direction; w ¼ wðx; z; tÞ is the deflection; φ ¼ φðx; z; tÞ signifies the @u
ɛxx ¼ @x ¼ em  zeb þ ðαz  βz3 Þesg ð6Þ
electric potential.
The refined third‐order shear deformation theory proposed by Shi  
5 5 2
[9] is employed for the kinematics of laminated composite beams in γ xz ¼ @u
@z
þ @w
@x
¼  z es ð7Þ

x and averaged deflection w 4 h2
this study. The averaged rotation ϕ
of
the beam cross‐section through the plate thickness are used in Shi where em is the membrane strain; eb is the bending strain; es is the
plate theory. These averaged displacements are defined by the equiv- respective transverse shear strain of the cross‐section defined in Eq.
alence of the strain energies produced respectively by the stresses (4); esg is the gradient of es ; α and β are two constants defined respec-
across the plate cross‐section and the stress resultants at the plate tively as α ¼ 1=4 and β ¼ 5=ð3h2 Þ for the convenience of equation
cross‐section. For beams, the energy equivalence takes the form derivation. These strain components of a beam cross‐section take the
Z h=2 Z h=2 following forms


x;
σ x udz ¼ M x ϕ τxz wdz ¼ Qx w ð1Þ
; eb ¼ @@xw20  @γ@xx ; es ¼ γ x ; esg ¼ @γ@xx
2
h=2 h=2 em ¼ @u
@x
0
ð8Þ
where h is the beam thickness; σ x is the normal stress in the beam axial It can be seen from Eq. (6) that the axial strain in Shi plate theory is
direction; τxz is the transverse shear stress; Mx and Qx are respectively associated not only with membrane strain em and bending strain eb but
the bending moment and the shear force at the cross‐section under con- also with the gradient of the transverse shear strain of the cross‐section

x and w
sideration; ϕ
are respectively the averaged rotation and aver- esg . It can also be found from Eqs. (6) and (8) that a linear bending
aged deflection of the cross‐section through the beam thickness. By strain can be interpolated in the finite element modeling when the
using the averaged displacements defined in Eq. (1), the deflection deflection w0 can be interpolated by a cubic polynomial trial function
w ¼ wðx; z; t Þ and in‐plane displacement u ¼ uðx; z; t Þ of a beam over the beam element, which is the reason that the in‐plane displace-
reduced from Shi plate theory [9] are of the form ment defined in Eq. (5) could yield the higher‐accuracy beam element
 than the one defined in Eq. (3).
wðx; z; tÞ ¼ wðx; tÞ ¼ w0 ðx; tÞ ð2Þ
@w0 

x þ ð1z  52 z3 Þ
uðx; z; tÞ ¼ u0 ðx; tÞ þ zϕ
x
þϕ ð3Þ 2.2. The approximation of electric potential
4 3h @x

The layerwise approximation is used for the distribution of the elec-

tric potential across the beam thickness. When the thickness of the
piezoelectric plies is very thin compared to the laminated composite
beam, a linear zigzag function can be used for the electric potential
interpolation across the thickness of each piezoelectric ply. For a lam-
inated piezoelectric beam with m layers of piezoelectric plies, the elec-
tric potential at a point with a distance z to the reference plane can be
approximated as
 
m ziþ1  z i z  zi iþ1
ϕðx; z; tÞ ¼ ∑ φ ðx; tÞ þ φ ðx; tÞ ð9Þ
i¼1 ziþ1  zi ziþ1  zi

where φi ðx; tÞ and φi þ 1 ðx; tÞ denote the electric potentials at, respec-
tively, the bottom surface and the top surface of the ith piezoelectric
ply.
The electric field in a piezoelectric ply with the x–z plane as its
main plane is given by
Fig. 1. The schematic of a laminated composite beam with m layers of
embedded piezoelectric plies (m ≤ n, n is the total number of the laminae). E ¼ rϕðx; z; tÞ ð10Þ

3
J. Tian et al. Composite Structures 252 (2020) 112480

Thus, when a piezoelectric ply is polarized in its thickness direction These 2D constitutive equations of piezoelectric materials in matrix
only, Eqs. (9) and (10) yield the electric field in the ith piezoelectric form will be used to define the electric enthalpy of piezoelectric plies
ply in the form in the next subsection.
@ϕðx; z; tÞ φiþ1 ðx; tÞ  φi ðx; tÞ
ðx; tÞÞ
φ
Ez ¼  ¼ ¼ i for zi ⩽ z ⩽ ziþ1
@z ziþ1  zi ziþ1  zi 3.2. Electric enthalpy of laminated piezoelectric beam
ð11Þ
For a rectangular laminated piezoelectric beam element with
where φ

i ðx; tÞ is the electric potential difference between the top and
length l, thickness h, width b, number of total laminae n and number
bottom surfaces of the ith piezoelectric ply and it takes the form
of piezoelectric plies m illustrated in Fig. 1, its electric enthalpy takes

i ðx; tÞ ¼ φiþ1 ðx; tÞ  φi ðx; tÞ for ði ¼ 1; 2 . . . mÞ; zi ⩽ z ⩽ ziþ1
φ ð12Þ the form [36]
Z Z zjþ1
The electric potential difference φ

ðx; tÞ defined above will be used
i b n
He ¼ ∑ ðɛTq σq  ɛTφ σφ Þdzdx
in the finite element modeling of the present laminated piezoelectric 2 j¼1 l zj
beams later. Z Z zjþ1
b n
¼ ∑ ðɛTq Qjq ɛq  ɛTq Qjqφ ɛφ  ɛTφ Qjφq ɛq  ɛTφ Qjφ ɛφ Þdzdx
2 j¼1 l zj
3. The governing equations of laminated piezoelectric beams
¼ U qq þ ðU qφ þ U φq Þ þ U φφ ð17Þ

3.1. Constitutive equations of piezoelectric ply The expression above indicates that the electric enthalpy of a lam-
inated piezoelectric beam is composed of three parts: i.e. strain energy
The constitutive equation of a piezoelectric ply in a laminated density U qq , electric field energy U φφ and electromechanical coupling
piezoelectric beam is presented here. The constitutive equation of nor- energy ðU qφ þ U φq Þ.
mal structural lamina just takes the mechanical part of the constitutive By using ɛ 11 ¼ ɛ xx , γ 13 ¼ γ xz , E 3 ¼ E z and substituting Eqs. (6), (7),
equation of a typical piezoelectric ply. (13) and (14) into Eq. (17), one obtains the strain energy of a lami-
In order to follow the customary constitutive equation of piezoelec- nated composite beam U qq taking the following form
tric materials in terms of x1 , x2 and x3 coordinate, let x1 and x3 denote, n R R zjþ1
respectively, the axial and thickness direction of a laminated piezo- U qq ¼ 2b ∑ l zj
ðɛ xx C0j11 ɛ xx þ γ xz C0j55 γ xz Þdzdx
j¼1
electric beam as illustrated in Fig. 1. Then for a piezoelectric ply with
the kinematics defined in Eqs. (2) and (3), the axial strain ɛ11 ¼ ɛxx as
n R
¼ 12 ∑ l
fem Aj11 em þ eb Dj11 eb þ esg ðα2 Dj11  2αβF j11 þ β2 H j11 Þesg
that defined in Eq. (6) and the transverse shear strain γ 13 ¼ γ xz as j¼1

defined in Eq. (7). Under the aforementioned conditions, the nonzero Bj11 ðem eb þ eb em Þ þ ðαBj11  βE j11 Þðem esg þ esg em Þ
stresses of the piezoelectric ply are axial stress σ 11 as well transverse ðαDj11  βF j11 Þðeb esg þ esg eb Þ þ es Sj11 es gdx
shear stress τ13 , and the electric displacements needed to be studied
ð18Þ
in the present 2D problem are D1 and D3 . The 2D electromechanical
coupled constitutive equations of a piezoelectric ply with σ 33 ¼ 0 in which the generalized elastic constants for the jth layer are defined
and polarized in x3‐direction only can be reduced from the 3D consti- as
tutive equations of piezoelectric materials as Z zjþ1
Aj11 ; Bj11 ; Dj11 ; E j11 ; F j11 ; H j11 ¼ b C0j11 ð1; z; z2 ; z3 ; z4 ; z6 Þdz;
    
σ 11 C011 0 ɛ 11 0 e031 0 Z
zj
¼  ð13Þ zjþ1
5 5z2
τ13 0 C055 γ 13 e015 0 E3 Sj11 ¼b C0j55 ð  2 Þdz ð19Þ
zj 4 h
   " #
D1 0 e015 ɛ 11 k011 0 0 For a laminated piezoelectric beam with m plies of piezoelectric
¼ þ ð14Þ
D3 e031 0 γ 13 0 k033 E3 layers, it follows from Eqs. (11), (14) and Eq. (17) that the electric field
energy U φφ is of the form
In the equations above, the superscript 0 represents the correspond- Z Z ziþ1 Z
b m 1 m
ing constants in 2D constitutive equations reduced from the 3D consti- U φφ ¼  ∑ E z k0i33 E z dzdx ¼ ∑ φ
i Ai31 φ

i dx ð20Þ
tutive equations; such as C011 and C055 are the reduced elastic constants, 2 i¼1 l zi 2 i¼1 l
e031 and e015 are the reduced piezoelectric constants; as well k011 and k033 where Eq. (12) is used and the general electrical parameter Ai31 for the
are the reduced dielectric permittivity constants. These reduced con- ith piezoelectric ply is defined as
stants are related to the corresponding constants in the 3D constitutive
Z ziþ1
equations of piezoelectric materials as following 1
Ai31 ¼ b  k0i33 dz ð21Þ
C213 e233 zi ðziþ1  zi Þ2
C011 ¼ C11  ; C055 ¼ C55 ; k011 ¼ k11 ; k033 ¼ k33 þ ; e015 ¼ e15 ;
C33 C33
It follows from Eq. (17) that the electromechanical coupling energy
C13
e031 ¼ e031  e33 U qφ and U φq take the following form respectively
C33
Z Z ziþ1
b m
where C11 and C55 are the elastic constants; k11 and k33 are dielectric U qφ ¼  ∑ ɛxx e0i31 E z dzdx
2 i¼1 l zi
permittivity constants; e31 , e15 and e33 are piezoelectric constants. Z
Let subscript q denote the mechanical quantities, subscript φ for the 1 m
¼ ∑ fem Ai21 φ
i  eb Bi21 φ
i þ esg ðαBi21  βE i21 Þ

φi gdx ð22Þ
electric related quantities and matrix Qi (i = q, φ) represent the corre- 2 i¼1 l
sponding mechanical or piezoelectric constant matrix, then the consti-
Z Z ziþ1
tutive equations in Eqs. (13) and (14) can be written in the following b m
Uq ¼  ∑ Ez e0i31 ɛ xx dzdx
matrix forms. 2 i¼1 l zi
Z
σq ¼ Qq ɛq  Qqφ ɛφ ð15Þ 1 m
¼ ∑ f
φi Ai21 em  φ
i Bi21 eb þ φ

i ðBi21  E i21 Þesg gdx ð23Þ
2 i¼1 l
σφ ¼ Qφq ɛq þ Qφ ɛφ ð16Þ
with the generalized elastic constants

4
J. Tian et al. Composite Structures 252 (2020) 112480

Z ziþ1
1 element quantities. The equations of motion of the system can be
Ai21 ; Bi21 ; Ei21 ¼ b e0i ð1; z; z3 Þdz ð24Þ
zi ziþ1  zi 31 obtained from Hamilton’s principle as
Z t
δ ðT  H þ WÞdt ¼ 0 ð31Þ
3.3. Hamilton’s principle with the electric enthalpy of piezoelectric plies t0

The equations of notion of laminated piezoelectric beams can be

derived from the Hamilton’s principle with the electric enthalpy. 4. Finite element modeling of laminated piezoelectric beams
It follows from Eqs. (3) and (2) that the axial and transverse veloc-
ities take the form respectively For the displacement field defined in Eqs. (2) and (5), the mechan-
  ical nodal displacement vector qq of a two‐noded beam element is of
@u0 @ @w0 @γ @w0
vx ¼   γ x z þ ðαz  βz3 Þ x ; vz ¼ ð25Þ the form
@t @t @x @t @t     
@w @w
T

Accordingly, Shi plate theory leads to that the kinetic energy of a qq ¼ u01 w01 γ x1 u02 w02 γ x2 ð32Þ
@x 1 @x 2
laminated beam element with varying mass density ρ(z) across the
beam thickness is of the form where the subscripts 1 and 2 denotes the node index of the beam ele-
R R h=2 R R h=2 h 2 @u2 i ment as depicted in Fig. 2.
T e ¼ 2b l h=2 ðv2x þ v2z ÞρðzÞdzdx ¼ 2b l h=2 @w @t
þ @t ρðzÞdzdx Eq. (20) shows that the electric field energy in each piezoelectric
R R  2  2 2 2 ply is associated with the electric potential difference between its
¼ 2b l h=2 @w þ @u@t0 þ z2 @@t@x  2z @u@t0 @@t@x
h=2 0 w0 w0
@t
two surfaces as shown in Eq. (12). Therefore, the electric potential dif-
þ2½ð1  αÞz2 þ βz4  @γ@tx þ 2½ðα þ 1Þz  βz3  @u@t0 @γ@tx
@ 2 w0
@t@x ference at the piezoelectric ply cross‐section associated with the two
 2 o nodes of a beam element can be chosen as the nodal electric degrees
þ½ð1 þ 2α þ α2 Þz2  2ðβ þ αβÞz4 þ β2 z6  @γ@tx ρðzÞdzdx
of freedom for each piezoelectric ply in the laminated piezoelectric
ð26Þ beam element. Then the electric field in each piezoelectric ply can
By defining the following parameters related to the integration be approximated using the two nodal electric parameters of the elec-
with the distribution of density ρ(z) across the laminated beam tric potential difference. The nodal electric degrees of freedom for
thickness the ith piezoelectric ply in a two‐noded piezoelectric composite beam
Z h=2 element is defined by the two electric potential differences as
JA; JB ; JD; JE ; JF ; JH ¼ b ð1; z; z2 ; z3 ; z4 ; z6 ÞρðzÞdz ð27Þ
i1 ¼ φ
φ
iþ1 ðx1 ; tÞ  φ

i ðx1 ; tÞ; φ

i2 ¼ φ

iþ1 ðx2 ; tÞ  φ

i ðx2 ; tÞ ð33Þ
h=2
Accordingly, the vector qϕ of the nodal electric degrees of freedom
then the kinetic energy defined in Eq. (26) can be expressed as:
in a two‐noded piezoelectric composite beam element with m piezo-
nR  2  2 2 2
T e ¼ 12 l J A @w 0
þ J A @u@t0 þ J D @@t@x
w0 electric plies takes the form
@t
T
@ 2 w0 @γ x qφ ¼ φ
21  φ

11 φ
i1  φ

m
12 φ
1 φ
22  φ

i2  φ

m ; ði ¼ 1; 2...mÞ ð34Þ
2J B @u@t0 @t@x
þ 2½ðα þ 1ÞJ B  βJ E  @u@t0 @t
2
ð28Þ
Combining the nodal degrees of freedom defined in Eqs. (32) and
þ2½ð1  αÞJ D þ βJ F  @γ@tx @ 2 w0
@t@x (34), a typical two‐noded piezoelectric composite beam element is
 2 o illustrated in Fig. 2. It should be noted that the number of electric
þ½ð1 þ 2α þ α2 ÞJ D  2ðβ þ αβÞJ F þ β2 J H  @γ@tx dx
degrees of freedom at each node of a piezoelectric composite beam ele-
The integration in the equation above involves the interpolations ment, which is based on the hybrid plate theory, depends on the num-
for transverse velocity @w
@t
0
, in‐plane velocity @u@t0 , velocity of deflection ber of piezoelectric plies in the piezoelectric composite beam.
and transverse shear strain velocity @γ@tx . The detailed interpo-
2
slope @@t@x
w0
4.1. Element stiffness matrix
lations for these velocity components will be presented later.
The distributed transverse load acting on the beam top surface is
In the displacement‐based element formulation, the element strains
denoted by f(x). Let qei ¼ f qe ðx; ziþ1 ; tÞ qe ðx; zi ; t gT (zi ⩽ z ⩽ ziþ1 ) be can be expressed in terms of the strain matric B and nodal displace-
the vector of the electric charge density applied, respectively, to the ment vector qq of the element. The strain components defined in Eq.
top and bottom surfaces of the ith piezoelectric ply, then the element
(8) can be expressed as
external work done by the distributed transverse force and the external
electric charges is of the form
Z mq Z
We ¼ w0 ðx; tÞf ðxÞdA þ ∑ qiT ei
φ q dA ð29Þ
Sf i¼1 Sq

where Sf and Sq denotes, respectively, the surface with applied trans-

verse force and applied electric charge density;
φ ¼ f φðx; ziþ1 ; tÞ
qiT φðx; zi ; t g is the vector composed of the electric
potentials on the top and bottom surfaces of the ith piezoelectric ply;
mq is the number of piezoelectric plies with applied electric charges.
When the applied load and electric charges are uniformly distributed
across the beam width, Eq. (29) can be written as
Z mq Z
T
W e ¼ b w0 ðx; tÞf ðxÞdx þ b ∑ qiφ qei dx ð30Þ
l i¼1 l

Having obtained each of the energy components in element level,

the total electric enthalpy H, kinetic energy T and external work W
of a system can be obtained from summation of the corresponding Fig. 2. Schematic of two-noded laminated piezoelectric beam element.

5
J. Tian et al. Composite Structures 252 (2020) 112480

em ¼ Bm qq ; eb ¼ Bb qq ; esg ¼ Bsg qq ; es ¼ Bs qq ð35Þ nodal parameters qiφ in qφ given in Eq. (44). Substituting Eq. (44) into
The element mechanical strain energy defined in Section 3.2 can be Eq. (20), one obtains
Z
expressed by the strain matrices defined in Eq. (35) and the element 1 m 1 m 1
U φφ ¼ ∑ ½qiT ðBiT Ai Bi Þdxqiφ  ¼ ∑ ½qiT Ki qi  ¼ qTφ Kφφ qφ ð46Þ
nodal mechanical displacement vector defined in Eq. (32). By substi- 2 i¼1 φ l φ 31 φ 2 i¼1 φ φφ φ 2
tuting Eq. (35) into Eq. (18), one can write the element strain energy
where Kφφ is the element electric stiffness of a piezoelectric composite
in terms of the strain matrices and the element nodal displacement
vector qq as beam element, and it is obtained from the summation of the element
R electric stiffness Kiφφ of each piezoelectric ply over all the piezoelectric
U qq ¼ 12 qTq l ½BTm A11 Bm þ BTb D11 Bb þ BTs-g ðα2 D11  2αβF 11 þ β2 H 11 ÞBs-g plies across the element thickness.
B11 ðBTm Bb þ BTb Bm Þ þ ðαB11  βE11 ÞðBTm Bs-g þ BTs-g Bm Þ ð36Þ By substituting Eqs. (35) and (45) into Eq. (23), one obtains
ðαD11  βF11 ÞðBTb Bs-g þ BTs-g Bb Þ þ BTs S11 Bs dxqq R m
U φq ¼ 12 l
∑ qiT
φ ½Bφ A21 Bm  Bφ B21 Bb þ Bφ ðαB21  βE 21 ÞBsg qq dx
iT i iT i iT i i
i¼1
The general stiffness parameters in the equation above take the
R m
form. ¼ 1
2 l
∑ qiT
φ ½Kφm þ Kφb þ Kφsg qq ¼ 2 qφ Kφq qq
i i i 1 T

n i¼1
A11 ; B11 ; D11 ; E11 ; F 11 ; H 11 ; S11 ¼ ∑ ðAj11 ; Bj11 ; Dj11 ; E j11 ; F j11 ; H j11 ; Sj11 Þ ð37Þ ð47Þ
j¼1
where the components of the coupling stiffness matrices are of the
where A11 is the tensile stiffness of the piezoelectric composite beam,
form
B11 is tensile‐bending coupling stiffness, D11 is the bending stiffness, Z
E11, F11 and H11 are the stiffness parameters resulting from the high‐ Kiφm ¼ BiT i
ð48-1Þ
φ A21 Bm dx
order shear deformation. Eq. (36) leads to the mechanical stiffness l

matrix of the present piezoelectric composite beam element as Z

Kqq ¼ Km þ Kb þ Ksg þ Kc þ Ks ð39Þ Kiφb ¼ BiT i
φ B21 Bb dx ð48-2Þ
l

In the equation above, Km is the membrane stiffness matrix, Kb is Z

the bending stiffness matrix, Ks-g is the stiffness matrix of shear strain Kiφs-g ¼ φ ðαB21  βE 21 ÞBsg dx
BiT i i
ð48-3Þ
l
gradient, Kc is the stiffness matrix resulting from coupling terms, and
Ks is the transverse shear stiffness matrix. These components of ele- The coupling stiffness matrix corresponding to Eq. (22) Kqφ can be
ment stiffness matrix take the form respectively obtained from the following relationship
Z
m m
Km ¼ BTm A11 Bm dx ð40Þ KTqφ ¼ Kφq ¼ ∑ Kiφq ¼ ∑ ðKiφm  Kiφb þ Kiφsg Þ ð49Þ
l
i¼1 i¼1
Z
Kb ¼ BTb D11 Bb dx ð41Þ The mechanical part of the equivalent nodal load, Fq , can be easily
l evaluated from the first term in the right‐hand side of Eq. (30). The
Z control forces of piezoelectric actuators are induced by the suitable
Ksg ¼ BTsg ðα2 D11  2αβF 11 þ β2 H 11 ÞBsg dx ð42Þ voltages applied to the actuators. When voltages Vðx; ziþ1 ; tÞ and
l
Z Vðx; zi ; tÞ are applied, respectively, to the top surface and the bottom
surface of the ith piezoelectric ply poling in the thickness direction,
Kc ¼ ½B11 ðBTm Bb þ BTb Bm Þ þ ðαB11  βE11 ÞðBTm Bsg þ BTsg Bm Þ
l the induced electric charge density on its top surface and bottom sur-
 ðαD11  βF 11 ÞðBTb Bsg þ BTsg Bb Þdx face takes the form

Z qei ¼ f qe ðx; ziþ1 ; tÞ qe ðx; zi ; tÞ gT

Ks ¼ BTs S11 Bs dx ð43Þ ¼
ki33
f½Vðx; ziþ1 ; tÞ  Vðx; zi ; tÞ ½Vðx; zi ; tÞ  Vðx; ziþ1 ; tÞg
T
ð50Þ
l hi
ki33 T
The electric field in a piezoelectric ply can be interpolated in the ¼ hi ΔV i
ΔV i

same way as the mechanical displacement field in the finite element

where hi , ki33
and ΔV i are, respectively, the thickness, dielectric permit-
modeling. By setting the electric nodal variables of the ith piezoelectric
ply as tivity constant and voltage difference between the top and bottom sur-
faces of the ith piezoelectric ply. Then the function in the second term
T
qiφ ¼
i1
φ
i2
φ ði ¼ 1; 2 . . . mÞ ð44Þ of Eq. (30) can be expressed as
( )
then the electric field in the ith piezoelectric ply in a two‐noded beam k33 ΔV i ki ki
qφ q ¼ φ
iT ei iþ1
φi

ðx;zi ; tÞ 33 ΔV i
¼ ðφiþ1  φi Þ 33 ΔV i ¼ φ
element can be approximated by the first‐order 1D Lagrangian interpo- hi ΔV i hi hi
lation function with its two nodal electric parameters qiφ defined in Eq. ð51Þ
(44). Since the two electric parameters qiφ in Eq. (44) for each piezo-
(for zi ⩽ z ⩽ ziþ1 and no summation on i).
electric ply are merely two parameters in the nodal electric variables
Accordingly, it follows from Eq. (30), Eq. (45) and Eq. (51) that the
of beam element qφ defined in Eq. (34), the electric “strain” matrix
external work produced by applied voltage takes the form
Biφ for the ith piezoelectric ply could be written as in a 1 × 2 m matrix,
and the electric potential difference in the ith piezoelectric ply is inter- Z Z
mq mq
T ki33
polated by W eφ ¼ b ∑ φ q dx ¼ b ∑
qiT ei
ðBiφ qφ Þ ΔV i dx ¼ qTφ Fφ ð52Þ
i¼1 l i¼1 l hi
φ

i ðxÞ ¼ f 0 0  N 1 ðxÞ  0 0  N 2 ðxÞ    gqφ ¼ Biφ qφ ði ¼ 1; 2 . . . mÞ
ð45Þ Consequently, the equivalent nodal electrical forces Fφ induced by
applied voltage is given by
in which N 1 ðxÞ and N 2 ðxÞ are the linear shape functions associated mq Z
with, respectively, node 1 and node 2 given by Lagrangian interpola- ki33
Fφ ¼ b ∑ BiT φ ΔV i dx ð53Þ
tion, and their positions in Biφ are corresponding to the position of their i¼1 l hi

6
J. Tian et al. Composite Structures 252 (2020) 112480

Z
The global stiffness matrix and nodal force vector of a system can Mw ¼ PTw JA Pw dx ð60Þ
be evaluated by the assembling of the corresponding element quanti- l

ties over the whole system. Z

Mu ¼ PTu JA Pu dx ð61Þ
4.2. The mass matrices consistent with the refined third-order shear l

deformation theory Z
Mwx ¼ PTwx JD Pwx dx ð62Þ
l
The element mass matrix can be obtained from the definition of the
kinetic energy given in Eq. (28). It can be seen from Eq. (28) that there Z
are some coupling terms among velocities of @u , @t , @t@x and @γ@tx in the
0 @w0 @ w0
2 Mγ ¼ PTγ ½ð1 þ αÞ2 J D  2βð1 þ αÞJF þ β2 JH Pγ dx ð63Þ
@t l
kinetic energy given by the refined third‐order shear deformation the-
Z
ory proposed by Shi [9]. For instance, Eq. (28) shows that the in‐plane
2
Muwx ¼  ½PTwx JB Pu þ PTu JB Pwx dx ð64Þ
velocity @u@t0 couples with the rotational velocity @@t@x w0
when the density l

ρ(z) is not symmetric about the beam central plane. These coupling Z
terms will result in the coupling terms in the mass matrix, and these Muγ ¼ ½PTu ððα þ 1ÞJB  βJE ÞPγ þ PTγ ððα þ 1ÞJB  βJE ÞPu dx ð65Þ
l
coupling terms are very important for the accuracy of dynamic analysis
of laminated composites. Therefore, the derivation of the mass matri- Z
ces consistent with the refined third‐order shear deformation theory Mwxγ ¼ ½PTwx ððα  1ÞJD þ βJF ÞPγ þ PTγ ððα  1ÞJD þ βJF ÞPwx dx
l
will be briefly presented here.
ð66Þ
The integration in Eq. (28) involves the interpolations of the in‐
plane velocity @u@t0 , transverse velocity @w
@t
0
, velocity of deflection slope The integrations in Eqs. (60)–(66) can be carried out very easily
@ 2 w0 @γ x since they only involves with the integration of polynomials.
@t@x
and transverse shear strain velocity For the conventional @t
.
Eq. (59) shows that the mass matrix in the present piezoelectric
displacement‐based element formulation, the interpolation functions
composite beam element is composed of seven components, where
for velocities are the same as those used for the mechanical displace-
Mw , Mu and Mwx are, respectively, the usual transverse, axial and
ment field. Since the element strains defined in Eqs. (6) and (7) will
rotary inertia matrices; Mγ is the mass matrix resulting from the
be evaluated by an assumed strain method in this work, the interpola-
higher‐order shear deformation; and Muwx , Muγ and Mwxγ are the cou-
tions of velocity components needed for the integration in Eq. (28) will
2 pling terms of different components in the in‐plane displacement. The
be presented here. The velocity of deflection slope @@t@x
w0
can be indepen- variational consistent mass matrix above can account for the contribu-
dently interpolated from the interpolation of transverse velocity @w@t
0
for tion of the higher‐order shear deformation to the mass matrix. The
the element formulation based on assumed strain method. For the numerical results [27] show that the consistent mass matrix can pro-
mechanical nodal displacement vector qq of a two‐noded beam element vide more accurate results than those given by the lump mass matrix.
defined in Eq. (32), the corresponding nodal velocity vector q_ q is of the
form 5. Finite element equations given by Hamilton’s principle
n  @w0  @γx  @u0  @w0  @γx  oT
@u0 2
@ w 2
@ w
q_ q ¼ q_ q ðtÞ ¼ @t 1 @t 1 @t@x @t 1 @t 2 @t 2 @t@x @t 2 If the total number of elements in a system is M, the Hamilton’s
1 2

ð54Þ principle in finite element modeling takes the form

Z t M
A simple linear interpolation can be used for the approximation of δ ðT  H þ WÞdt ¼ δ ∑ ðT e  H e þ W e Þdt ¼ 0 ð67Þ
t0 k¼1
these velocities as follows.
Using the results for a single element given in the previous two sec-
@u0 @u0 ðx; tÞ
¼ ¼ f N 1 ðxÞ 0 0 0 N 2 ðxÞ 0 0 0 gq_ q ðtÞ ¼ Pu q_ q tions, Hamilton’s principle yields the equations of motion of a lami-
@t @t
nated piezoelectric beam as
ð55Þ

( ) "
#

qφ  q
( )

q
Mqq 0 €
q
q Kqq K
q F
@w0 @w0 ðx; tÞ þ ¼ ð68Þ
¼ ¼ f0 N 1 ðxÞ 0 0 0 N 2 ðxÞ 0 0 gq_ q ðtÞ ¼ Pw q_ q 0 0 €

φ
q

Kφq Kφφ

φ
q

Fφ
@t @t
ð56Þ where the bar over each vector or matrix stands for the corresponding
global vector or matrix of the laminated piezoelectric beam obtained
@2w @ 2 wðx; tÞ through assembling the element quantities over all the elements in
¼ ¼ f0 0 N 1 ðxÞ 0 0 0 N 2 ðxÞ 0 gq_ q ðtÞ ¼ Pwx q_ q
@t@x @t@x the beam. For instance, q
q is the global vector of nodal mechanical dis-
ð57Þ placements; q €

q is the global acceleration vector of nodal mechanical dis-
placements q
φ is the global vector of nodal electric degrees of

q ; q
@γ x @γ x ðx; tÞ freedom composes of the element electric variable vector defined in
¼ ¼ f0 0 0 N 1 ðxÞ 0 0 0 N 2 ðxÞ gq_ q ðtÞ ¼ Pγ q_ q
@t @t
qq is the global mechanical stiffness matrix and K
Eq. (34); K
φq is the glo-
ð58Þ


bal piezoelectric stiffness matrix, Kqφ and Kφq are the global electrome-

By substituting the velocity interpolations specified in Eqs. (55) ‐ chanical coupling matrices and Mqq is the global mass matrix.
(58) into Eq. (28) and carrying the integration, one can write the ele- When the external forces and damping effect are omitted, then Eq.
ment kinetic energy in terms the element nodal velocity vector q_ q as (68) reduces to
T e ¼ q_ Tq ðMw þ Mu þ Mwx þ Mγ þ Muwx þ Muγ þ Mwxγ Þq_ q 
( ) "
#

qφ  q 
Mqq 0 €
q
q Kqq K
q 0
þ ¼ ð69Þ
¼ q_ Tq Mqq q_ q ð59Þ 0 0 €
φ
q
φq K
K
φφ
φ
q 0

where Mqq is the element mass matrix in the present laminated piezo- The equation above indicates that the global vector q
φ of nodal
electric beam element, and each component of Mqq take the form, electric degrees of freedom can be solved in terms of the global vector
respectively,
q of nodal mechanical displacements using the static condensation.
q

7
J. Tian et al. Composite Structures 252 (2020) 112480

Consequently, the characteristic equation to evaluate the natural It is worthwhile to point out that the linear bending strain defined
vibration frequency ω of a electromechanical coupling system of piezo- in Eq. (74) is capable of optimally utilizing the given nodal variables
electric composite beams is of the form illustrated in Fig. 2. The linear interpolation for bending strain will
 lead to the resulting beam element with higher accuracy since the
b  ω2 c
ðK MÞqq ¼ 0 ð70Þ bending strain is the dominant term in the beam bending analysis.
b is the equivalent global stiffness matrix and c
qq . The The interpolation for the transverse shear strain es in Eq. (76) is pur-
in which K M¼M
posely assumed as a constant, which is one‐order lower than the inter-
b
equivalent global stiffness matrix K is given by the static condensation polation given by the two nodal shear strain variables γ 1 and γ 2 . As a
of Eq. (69) as matter of fact, the lower‐order interpolation for the transverse shear
b ¼K
K
qq  K
1 K

qφ K
ð71Þ strain in the quasi‐conforming element is equivalent to the reduced
φφ φq
integration for shear strain energy in the displacement‐based C0 beam
element. In the quasi‐conforming element, the transverse shear strain
6. Strain matrices evaluated by quasi-conforming element γ x and its gradient @γ
@x
x
can be approximated independently. Therefore,
technique the shear locking can be avoided easily by the quasi‐conforming ele-
ment technique [26,28,31].
In the conventional displacement‐based element formulation, the In the quasi‐conforming element, the assumed element strain
element strain matrices in Eq. (35) are evaluated from the displace- parameters are determined in terms of element nodal mechanical dis-
ment interpolations in the element. However, these matrices will be placement vector at element level from the weak form of compatibility
evaluated by the quasi‐conforming element technique [20] in this given in Eq. (72). The weak form of the compatibility for each strain
work. For a quasi‐conforming element, the element strain field is inter- components in Eq. (72) can be enforced independently. By setting
polated directly over the element domain rather than from the differ- the weighting function for each strain component same as its interpo-
entiations of the assumed displacement field, and the compatibility in lation function defined in Eqs. (73)–(76), one obtains the strain matri-
an element domain is satisfied in a weak form. Let a prime signify the ces for the four strain components in Eqs. (73)–(76) as
assumed element strain field, then the element strain energy in Eq.
(36) should be modified to account for the weak form of the compat- Bm ¼ Pm A1 1 1
m Cm ; Bb ¼ Pb Ab Cb ; Bsg ¼ Psg Asg Csg ; Bs
ibility as
¼ Ps A1
s Cs ð77Þ
Z ∼
Z ∼
Z ∼
U qq ¼ U qq þ N ðem  þ e0m Þdx
M ðeb  e0b Þdx þ Pðesg where
l l l
Z ∼ Z Z Z
 e0sg Þdx þ Qðes  e0s Þdx ð72Þ Am ¼ PTm Pm dx ¼ l; Ab ¼ PTb Pb dx; As-g ¼ PTs-g Ps-g dx ¼ l;
l
Z l l l

∼ ∼ ∼ ∼ As ¼ PTs Ps dx ¼ l ð78Þ
where N , M , P and Q are the weighting functions corresponding to their l
relevant strain component. Z Z
A suitable interpolation for element strain field depends on the ele- Cm qq ¼ PTm em dx; Cb qq ¼ PTb eb dx;
ment nodal displacement variables. According to the mechanical nodal Zl Z l

displacements illustrated in Fig. 2, a linear trial function over an ele- Csg qq ¼ PTsg esg dx; Cs qq ¼ PTs es dx ð79Þ
ment can be interpolated for in‐plane displacement u0 as well trans- l l

verse shear strain γ x , and a cubic trial function over the element can l in Eq. (78) is the beam element length. In Eq. (79), qq is the element
be interpolated for deflection w0 . Then for the beam strains defined mechanical nodal displacement vector defined in Eq. (32) and Ck
in Eq. (8), a constant strain can be assumed for the membrane strain (k = m, b, s‐g, s) are the coefficient matrices given by the integration
as well as the gradient of the transverse shear strain, and a linear bend- of the right hand side of the corresponding equation in Eq. (79). The
ing strain is assumed for the bending strain. In order to avoid shear integrations in Eq. (79) for a two‐noded beam element can be carried
locking, a constant value is also assumed for the transverse shear strain out very easily, and one obtains the strain matrices in Eqs. (73)–(76)
es . Accordingly, a suitable element strain field for the strains defined in as follows
Eq. (8) can be approximated as
du0 0 Bm ¼ ½ 1=l 0 0 0 1=l 0 0 0 ð80Þ
em ¼ ≈e ¼ αm ¼ I1 ½αm  ¼ Pm αm ð73Þ
dx m
  
  1=l 0 0 0 1 1 0 0 1 1
d2 w0 dγ x 0 αb1 Bb ¼ f 1 xg ð81Þ
eb ¼  ≈e ¼ αb1 þ xαb2 ¼ ½ 1 x ¼ Pb αb ð74Þ 0 12=l 3
0 1 l=2 0 0 1 l=2 0
dx2 dx b αb2
Bsg ¼ ½ 0 0 0 1=l 0 0 0 1=l  ð82Þ
dγ  
esg ¼ x ≈e0sg ¼ αsg ¼ I1 αsg ¼ Psg αsg ð75Þ
dx Bs ¼ ½ 0 0 0 1=2 0 0 0 1=2  ð83Þ

es ¼ γ x ≈e0s ¼ αs ¼ I1 ½αs  ¼ Ps αs ð76Þ

l in the expression above is the beam element length.
where the variable with superscript prime denotes the approximation The element mechanical stiffness matrix can be obtained through
to the corresponding strain; αm , αbi (i = 1, 2), αsg and αs are the substituting the strain matrices given in Eqs. (80)–(83) into Eqs.
assumed element strain parameters; I1 is the identity matrix with rank (40)–(43) and Eq. (39). Since the integrations in Eqs. (40)–(43)
1; Pm, Pb, Ps‐g and Ps are, respectively, the interpolation function for merely involve polynomials, they can be evaluated analytically,
the membrane strain, bending strain, the gradient of the transverse i.e. the element stiffness matrix presented in this study is given
shear strain and transverse shear strain. Since the membrane and explicitly. Therefore, the present laminated piezoelectric beam ele-
transverse shear strains as well gradient of transverse shear are con- ment is very computationally efficient compared with the beam ele-
stant in the present two‐noded beam element, Pm, Ps‐g and Ps are sim- ments using numerical integration to compute their element stiffness
ply equal to I1 . matrix.

8
J. Tian et al. Composite Structures 252 (2020) 112480

7. Optimal vibration control using LQR with the weighting matrix 7.2. The equation of motion based on reduced-order mode superposition
determined by GA
Under the action of practical loading conditions, the vibration
The mode superposition is used to solve the dynamic equations. responses of most structural systems are mainly contributed from the
The Linear Quadratic Regular (LQR) control scheme is an efficient first few mode shapes of vibrations. Therefore, the dynamic responses
feedback control strategy, hence LQR is adopted for the active vibra- of a system can be efficiently analyzed based on the reduced‐order
tion control in this study. modal superposition as well the modal coordinates, and the resulting
reduced‐order dynamic model of the system possesses the uncoupled
7.1. Equations of motion in forms of finite element equations equations of motion in the modal coordinates.
Let Ψi (i = 1, 2 …mr ) be the eigenvectors (mode shapes) of the first
In order to conveniently model the active vibration control using few lower‐modes of vibration of a system, then the dynamic response
the piezoelectric actuators, the vector of nodal electric degrees of free- of the system q
q can be approximated by the eigenvectors Ψi and the
dom q
φ in Eq. (68) can be decomposed into two group, the group asso- modal coordinates η as follows
ciated with the actuators is assigned as q
Aφ and the other group mr 

q ¼ ∑ Ψ i ηi ¼ Ψ η
q ð92Þ
associated with sensors is named as q
Sφ . The stiffness matrices associ- i¼1

ated with q
φ in Eq. (68) can be decomposed accordingly. Then Eq. 

where Ψ is the reduced‐order modal matrix composed of Ψi (i = 1, 2
(68) can be recast as T
…mr ). By multiplying Ψ from left side, Eq. (89) becomes
2
38 :: 9 2 K
qq
A
3

S 8 q 9 8
9
0 0 >
>
>
q K K qφ >
q > > Fq >
>   
Mqq > q> > qφ
> > > >
> >
>
η þ Ψ
ΨT MΨ€
T K Ψη

¼ ΨT F
T KA q

q  Ψ
qφ
φ
A
ð93Þ
6 7>
> >
> 6 7>
< >
7> < >
= >
>
6 7< :: = 66
A AS 7

=
6 0 0 07
Aφ þ 6 K
A
K


Kφφ 7 qφ ¼ Fφ
A A
ð84Þ
6 7> q > 6 φq φφ
7> > > > The orthogonality of the eigenvectors yields the following results
4 5>
> >
> 4 > >
5> > >
> > >
> >
> > : S>
> ; >
> >
>  
0 0 0 :q€Sφ
;
S
K φq

SA
K φφ


K S
φφ


q φ
: F ;

S
φ ΨT M Ψ ¼ I ð94Þ

T K Ψ
Ψ
¼ Ω2 ¼ diagðω2 Þ ði ¼ 1; 2; . . . mr Þ ð95Þ
When the piezoelectric plies used for actuators have no shared elec- i

troded surfaces with other piezoelectric plies, K
AS is a null matrix.

φφ Matrix I in Eq. (94) is an identity matrix. Then Eq. (93) reduces to
Accordingly, Eq. (84) can be decomposed into the following three
TF
TK
€ þ Ω2 η ¼ Ψ
q  Ψ
A q
η qφ
φ ð96Þ
A
equations
The uncoupled equations of motion in the modal coordinates given

qq q
€
qq q

q þ K
S q

q þ K

A
A
qφ
φ ¼ Fq  Kqφ q ð85Þ
S
M φ above can be solved very conveniently.

A q
A
A ¼ F

A
φq
q þ Kφφ q
K ð86Þ 7.3. Modal state space based on the reduced-order modal coordinates
φ φ

S q
S
S ¼ F

S By using the modal coordinates η to define a state vector
K φq
q þ Kφφ q φ φ ð87Þ
XðtÞ ¼ f η η_ g , Eq. (96) then yield the dynamic equations in terms
T

Because the external electric charges are applied only to the sur- of the state space vector as
faces of the piezoelectric plies used for actuators, Eq. (87) leads to _
XðtÞ ¼ AXðtÞ þ Bq uq ðtÞ þ Bφ uφ ðtÞ ð97Þ
the output electric potential corresponding to sensors as
where uq and uφ are, respectively, the mechanical loading vector and

S 1 K

Sφ ¼ K
S q
q φφ φq
q ð88Þ the electrical control input vector; A is the system state matrix; Bq is
the mechanical loading matrix; Bφ is the electrical control input matrix.
Substituting Eq. (88) into Eq. (85), one obtains the equations of
These vectors and matrices take the form
motion for the forced vibration of electromechanical coupled beams " #
   
without damping effect as follows 0 I 0 0


φ ; A ¼
uq ðtÞ ¼ Fq ; uφ ðtÞ ¼ q A
; Bq ¼ ; Bφ ¼
  Ω Λ ψT
A
ψT K
€
q  KA q

q þ K qq ¼ F qφ
qφ
φ ð89Þ
A
Mq
ð98Þ

where M is the global mass matrix, K is the equivalent global stiffness
matrix accounting for the electromechanical coupling; M and K take The output electric potential vector corresponding to sensors
the form, respectively, as should also be expressed in terms of the state space vector X(t). By
n oT
defining YðtÞ ¼ q
Sφ q
_ Sφ , Eq. (88) leads to

qq ; K ¼ K
M¼M
qq  K
S 1 K

S K
S ð90Þ
qφ φφ φq
YðtÞ ¼ CXðtÞ ð99Þ
Eq. (89) is the equations of motion for a system without damping.
 
For a system with Rayleigh damping CR ¼ aMqq þ bKqq where a and b with
" #
are Rayleigh damping parameters, the equations of motion for the K
S ψ

S 1 K 0
φφ φq
forced vibration with electromechanical coupling effect and Rayleigh C¼ ð100Þ
0
S 1 K
K
S ψ
damping can be written as φφ φq

€
qq Þq

q þ ðaM þ bK
_ q þ K q
q  K

q ¼ F
A q
qφ
φ ð91Þ
A
Mq
7.4. The optimal vibration control using LQR optimal control scheme

in which the mass matrix M and the stiffness matrix K are the same as
that in Eq. (90). Linear Quadratic Regulator is an efficient scheme for optimal vibra-
The equations of motion expressed in Eqs. (89) and (91) can be tion control. In the LQR optimal control scheme, the control gain for a
solved by the direct integration method or the mode superposition full state feedback is determined by minimizing a control performance
[37]. index functional of the linear combination of a quadratic control per-

9
J. Tian et al. Composite Structures 252 (2020) 112480

formance measure with a quadratic input cost measure. If the structure

comes to rest at its initial configuration at the terminal time t f and the
vibration response is used as performance measure, the functional of
the control performance index is of the form
Z
1 tf T
J¼ ½X ðtÞ Qx XðtÞ þ uTφ ðtÞ R uφ ðtÞdt ð101Þ
2 0
where Qx is a weighing matrix for state vector X(t) which is positive
and semi‐definite; R is the weighing matrix for the control input uφ ðtÞ
which is real, symmetric and positive definite. It should be noted that
the index functional given in Eq. (101) is subjected to the constraints
defined in Eq. (97).
By using a Lagrangian multiplier matrix P(t), the performance
index functional in Eq. (101) subjected to constraint of Eq. (97) can
be transformed into a unconstrained index functional as

1 T 1
H¼ X ðtÞQx XðtÞ þ uTφ ðtÞR uφ ðtÞ þ XT ðtÞPT ðtÞ½AXðtÞ
2 2
þ Bq uq ðtÞ þ Bφ uφ ðtÞ ð102Þ

Therefore, the optimal control input uφ ðtÞ can be determined

through Euler‐Lagrange equation of the above functional with uφ ðtÞ
as a variable, that is
@H
¼ R uφ ðtÞ þ BTφ PðtÞ XðtÞ ¼ 0 ð103Þ
@uφ

The equation above leads to the optimal control input uφ ðtÞ taking
the following form
Fig. 3. The flowchart of GA-based evaluation of weighting matrices.
uφ ðtÞ ¼ ½R1 BTφ PðtÞXðtÞ ¼ GA XðtÞ ð104Þ

in which GA is the so‐called feedback gain matrix for optimal active

8. Numerical examples
vibration control.
Lagrangian multiplier matrix P(t) can be taken as the steady state
In this section, several numerical examples are solved to show the
solution of the following nonlinear Ricatti equation [38]
accuracy of the laminated piezoelectric beam element in the dynamic
P_ ¼ P A  AT P þ P Bφ R1 BTφ P  Qx with Pðt f Þ ¼ 0 ð105Þ analysis of laminated piezoelectric beams and the feasibility as well as
effectiveness of the GA‐based evaluation of the optimal weighting
The nonlinear Ricatti equation can be solved numerically. Matlab is matrix used in the LQR active vibration control.
used in this study to solve the nonlinear Ricatti equation.
By substituting Eq. (104) into Eq. (97), one finally obtains the space 8.1. Validation of reliability and computational efficiency
state equations with optimal feedback control as
_
XðtÞ ¼ ðA  Bφ GA ÞXðtÞ þ Bq uq ðtÞ ð106Þ The static analyses of two piezoelectric polymeric polyvinylidene
fluoride (PVDF) beams are solved by the present laminated piezoelec-
The dynamic equation above will be used for the optical active tric beam element, in which both mechanical load and electric load are
vibration control using piezoelectric actuators. considered. The resulting numerical results are compared with the
analytical solutions and other numerical solutions to demonstrate that
7.5. LQR optimal vibration control with weighting matrix Q evaluated by the present laminated piezoelectric beam element is free from shear
GA locking and possesses high computational efficiency.

The weighting matrices Qx and R in the functional of the control 8.1.1. The deflections of simply supported three-layer PVDF beam with
performance index defined in Eq. (101) play very important roles in different aspect ratios
the active vibration control. It is not a trivial task to find the optimal The cross‐ply laminated beam illustrated in Fig. 4 is composed of
Qx and R. Most of researchers used a trial and error approach to select three PVDF plies. The lamination scheme is [0/90/0] with equal thick-
feasible Qx and R. GA is adopted to determine the optimized weighting ness for each ply. The beam length is L ¼ 100 mm and its total thick-
matrices Qx in this work. A flowchart of GA‐based evaluation of ness is h. The laminated piezoelectric beam is simply supported and
weighting matrices is depicted in Fig. 3.
A real coded genetic algorithm along with simulated binary cross-
over and parameter based mutation operators is used to evaluate opti-
mal weighting matrices in the LQR optimal control scheme. Since the
objective function in the present active vibration is to minimize the
vibration control performance index defined in Eq. (101), the fitness
in the present GA‐based optimal evaluation of weighting matrices
can be defined by the inverse of the control performance index. The
fitness value has been calculated with respect to each chromosome.
The parameters used in GA‐based determination of optimal weighting
matrices will be given in the related numerical example later, and
more details can be found from Tian’s thesis [39]. Fig. 4. A three-layer [0°/90°/0°] laminated piezoelectric beam.

10
J. Tian et al. Composite Structures 252 (2020) 112480

Table 1
Nondimensional central deflections of a three PVDF ply laminated beam [0/90/0] with different length to thickness ratios under a sinusoidal mechanical load.

Models Number of elements Aspect ratio (L/h)

10 20 100

Present beam element 4 1.9933 22.0087 1158.46

8 2.0354 22.0349 1160.92
16 2.0550 22.0899 1161.41
Analytical solution of deflection at z = 0 [40] 2.0513 22.0874 1161.30

with vanishing electric potential on its two surfaces. The principle 8.1.2. Deflections of a cantilevered PVDF bimorph beam subjected to an
material properties of the PVDF are as follows [40]. applied potential
This example concerns a cantilevered bimorph beam made of two
C11 ¼ 238:24 GPa; C12 ¼ 3:98 GPa; C13 ¼ 2:19 GPa;
PVDF (β‐phase) plies, where two PVDF plies are bonded together
C33 ¼ 10:64 GPa; C55 ¼ 2:15 GPa
and polarized in opposite directions. The dimensions of the can-
tilevered PVDF bimorph beam in the x–z plane are depicted in
e31 ¼ 0:13 C m2 ; e33 ¼ 0:276 C m2 ; Fig. 5, and the beam width is 5 mm. This PVDF bimorph beam is a thin
e15 ¼ 0:135 C m2 ; k33 ¼ 0:1062 nF m1 beam as its aspect ratio of length to thickness is equal to 100. The
material properties of the β‐phase PVDF are [41]:
A mechanical load of sinusoidal loading q ¼ q0 sinðπx=LÞ applied on
the beam top surface is considered in this numerical example. The C11 ¼ 2:621 GPa; C12 ¼ 1:071 GPa; C13 ¼ 1:071 GPa;
deflections of the laminated PDVF beam with three aspect ratios of
C33 ¼ 2:621 GPa; C55 ¼ 0:775 GPa
length to thickness L/h = 10, 20 and 100 are computed using the pre-
sent laminated piezoelectric beam element. The resulting nondimen-
e31 ¼ 0:046 C m2 ; e33 ¼ 0:0 C m2 ; e15 ¼ 0:0 C m2 ;
sional central deflections of the PVDF beam are tabulated in Table 1.
Three mesh densities are used to study the modeling of the equivalent k33 ¼ 0:106 nF m1
nodal load for the nonlinearly distributed sinusoidal load, where the The static deflection of the cantilevered PVDF bimorph beam under
number of elements is for the whole beam. The nondimensional deflec- the action of an applied electric potential is a benchmark problem to
tion is defined by w^ ¼ C55 wðxÞL=ðq0 hÞ. validate the accuracy and efficiency of a finite element model to ana-
Cheng et al. [40] used a 3D asymptotic approach to analytically lyze the electromechanical coupling of laminated piezoelectric beams.
solve the deflections of the three layered PVDF plate of infinite Therefore, the experimental result, analytical solution and numerically
width with simply supported edges, which has the same central results of the deflections of the cantilevered PVDF bimorph beam at
deflection as that of a three layered PVDF beam with the same different positions are available in the literature. A voltage applied
length and simply supported ends. The transverse normal effect across the thickness of the PVDF bimorph beam is taken as the electric
was considered in the work of Cheng et al. The analytical solutions load. The analytical solution for the static deflection of the can-
of the central deflection at the middle surface, i.e. z = 0, given by tilevered PVDF bimorph beam subjected to electric potential given
Cheng et al. [40] are also listed in Table 1 for comparison. One by Tzou et al. [42] takes the form
can see it from the results in the table that the present laminated
piezoelectric beam element yields accurate results for both thick e31 V a x2
wðxÞ ¼ 0:375
beam with L/h = 10 and thin beam with L/h = 100, which Et 2
demonstrate that the present laminated piezoelectric beam element where e31 is a piezoelectric constant of PVDF; Va is the applied voltage
is free from shear locking. across the thickness of the PVDF bimorph beam; and t is the thickness of
a single PVDF ply.
Five beam elements are used to model the PVDF bimorph beam in
the present case study, and the resulting deflections at five positions of
the cantilevered bimorph beam subjected to a unit voltage across the
bimorph thickness are listed in Table 2. The analytical solutions
[42], the experimental result of tip deflection [42] and the numerical
results obtained from three other finite elements [41,43,44] given in
the table for comparison. All the finite elements listed in Table 2 are
based on certain HSDTs and five elements were used to model the
PVDF bimorph beam.
It can be seen from the table that all the numerical results given by
Fig. 5. Schematic of a cantilevered PVDF bimorph beam. four different element models are quite close to the analytical solu-

Table 2
Deflections of cantilevered bimorph beam subjected to unit voltage given by different methods or models (10−4 mm).

Models Position/(mm)
20 40 60 80 100

HSDT plate with 9 DOFs [43] 0.138 0.552 1.240 2.210 3.450
2 node Hermitian beam [41] 0.138 0.552 1.242 2.208 3.450
3 node beam [44] 0.138 0.552 1.242 2.208 3.450
present 2 node beam 0.137 0.551 1.240 2.207 3.449
Analytical solution [42] 0.138 0.552 1.240 2.208 3.451
Experimental result [42] 3.150

11
J. Tian et al. Composite Structures 252 (2020) 112480

Table 3
Comparison of computational efficiency of different element models for PVDF bimorph beam.

Element models Number of elements Number of nodes Mechanical DOFs Electrical DOFs Total DOFs

HSDT plate with 9 DOFs [43] 5 33 270 10 280

3 node beam [44] 5 11 40 36 76
2 node Hermitian beam [41] 5 6 20 18 38
Present 2 node beam 5 6 20 12 32

tions, and all the numerical results of the tip deflection (x = 100 mm) E 1 ¼ 144:8 GPa, E 2 ¼ 9:65 GPa, G12 ¼ G13 ¼ 4:14 GPa,
are larger than the experimental result. Since all the numerical results G23 ¼ 3:45 GPa, ν12 ¼ 0:3, ρ = 1389.23 kg/m3.
given by the four element models listed in Table 2 are quite close to The aspect ratio of the length to thickness of the laminated beam is
the analytical solutions, it is worthwhile to compare their computa- L/h = 15. Four different boundary conditions are considered in this
tional efficiency. The 2 node Hermitian beam element presented in case study. Twenty equal‐sized beam elements are used to model the
the table [41] is based on the simplified HSDT proposed by Reddy laminated composite beam in order to characterize the higher vibra-
and the Hermitian polynomial for the deflection interpolation. This tion modes. The natural frequencies can be obtained from Eq. (70).
beam element has four mechanical degrees of freedom (DOFs) per The first five nondimensional flexural vibration frequencies (desig-
node. The 3 node beam element [44] in the table is based on a HSDT nated by f(1)–f(5)) are tabulated in Table 4. The nondimensional fre-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of refined sinus model. This beam element has three nodes where the
quency is defined by ω
i ¼ ωi L2 J A =ðE 1 bh3 Þ. Some other numerical
end node has five mechanical DOFs and the middle node has three
mechanical DOFs. The HSDT plate element with 9 DOFs [43] is based results and analytical solutions are also listed in Table 4 for compari-
on a HSDT with quadratic deflection interpolation in the thickness son. The numerical solutions of ABAQUS were obtained from the 8‐
direction and it has nine mechanical DOFs per node. The comparison noded solid elements, and two‐layers of solid elements are used for
of computational efficiency of different finite element models for the each lamina of the laminated beam.
PVDF bimorph beam analysis is given in Table 3. It is fair enough that The fundamental frequencies of the axial vibration, denoted by a
the total DOFs of the HSDT plate element with 9 DOFs used for the (1), of the beams with boundaries of SS, CS and CF predicted by the
modeling of the PVDF bimorph beam is calculated by dividing it by present beam element are also listed in the table. The simply supported
a factor of 2.5 when it is compared with the total DOFs used by beam boundary conditions of 1D beam at x = 0 are defined by
elements. Then one can see it from Table 3 that the equivalent total u0 ð0Þ ¼ w0 ð0Þ ¼ 0. The simply supported boundary conditions at
DOFs of the HSDT plate element with 9 DOFs used to model the PVDF x = 0 corresponding to the axial vibration in 3D modeling of beams
bimorph beam is still more than 100. Therefore, the results in Table 3 are uð0; y; zÞ ¼ wð0; y; 0Þ ¼ 0. The 3D modeling for SS boundaries
clearly show that the present two‐noded element uses the least DOFs to was used in ABAQUS for the computation of the fundamental fre-
achieve the same accuracy for the deflection analysis of the PVDF quency of axial vibration a(1). The 3D analysis result of a(1) for SS
bimorph beam. boundaries given by ABAQUA is not listed in Table 4 since the value
of a(1) obtained from ABAQUS for the laminated beam with 3D mod-
8.2. Free vibration of laminated composite beam with different boundary eling of SS boundary conditions is not between its f(3) and f(4).
conditions One can see it from the results in Table 4 that the present element
which is based on the Shi plate theory yields more accurate results,
The vibration analysis of a laminated composite beam composed of especially for the higher‐mode frequencies than the beam element
four equal thickness plies is considered in this example. The laminated based on the third‐order shear deformation beam theory proposed
composite beam is made of the same composite ply with lamination by Bickford [30]. The major difference between the sixth‐order beam
scheme [0/90/90/0]. The principal material properties of the compos- theory deduced from Shi plate theory and Bickford beam theory lies on
ite ply are as follows the different definitions of the rotation of the cross‐section; the former
employs an averaged rotation across the beam cross‐section and the

Table 4
Nondimensional frequencies of symmetric [0/90/90/0] cross-ply beams with aspect ratio L/h = 15.

Models BCs Nondimensional frequency at various vibration modes

f(1) f(2) f(3) a(1) f(4) f(5)

Present SS 2.4952 8 .4551 15.7208 17.2113 23.2552 30.7395

Shi plate theory CC 4.6228 10.4438 17.2895 – 24.4919 31.7930
(20 elements) CS 3.5313 9.4970 16.5073 17.2110 23.8354 31.1695
CF 0.9163 4.9085 11.5193 17.2110 18.8428 26.3646

Shi et al. [27] SS 2.4979 8.4364 15.5932 – 22.8974 30.0061

Bickford-Reddy beam theory CC 4.6194 10.4162 17.1724 – 24.2001 31.2144
(20 elements) CS 3.5264 9.4736 16.4201 – 23.5591 30.6107
CF 0.9199 4.9054 11.4886 – 18.6886 25.9931

Chandrashekhara et al. [45] SS 2.5023 8.4812 15.7558 – 23.3089 30.8386

analytical solutions CC 4.5940 10.2906 16.9559 – 24.1410 31.2874
CS 3.5254 9.4423 16.3839 – 23.6850 31.0569
CF 0.9241 4.8925 11.4400 – 18.6972 26.2118

ABAQUS SS 2.4862 8.4415 15.7185 – 23.3170 30.9390

(80 × 8 mesh) CC 4.5866 10.3185 17.0647 – 24.2398 31.6197
CS 3.5108 9.4336 16.4144 17.1934 23.7902 31.2835
CF 0.9181 4.8749 11.4333 17.1920 18.7360 26.3304

Note: S stands for simply supported end, C for clamped end and F for free end.

12
J. Tian et al. Composite Structures 252 (2020) 112480

later uses the rotation measured at the beam midplane. Reddy plate surface is used for the actuator and the bottom PTZ layer is used for
theory [8] can also reduce to Bickford beam theory for one‐ sensor. An impulse load with magnitude −0.2 N is applied at the beam
dimensional problems. free end for a duration of 1 ms. The laminated beam is modeled by ten
equal‐sized laminated piezoelectric beam elements. The dynamic
8.3. Free vibration of a laminated piezoelectric beam response of the beam is evaluated by the mode superposition with
the first five vibration modes. LQR vibration control scheme is
The laminated piezoelectric beam illustrated in Fig. 6 is considered adopted. In order to study the influence of weighting matrices Qx with
in this numerical example. The host structure of the beam is made of different values on the vibration control effects, the selection of Qx
steel; and a PZT ply is bonded on its top surface and bottom surface based on the trial and error scheme is adopted in this example.
respectively. The beam length is L = 500 mm, the width is From now on, Q is used to replace the notation for weighting
b = 30 mm, the thickness of steel layer is hs = 2 mm, and the thick- matrix Qx for simplicity. For the sake of computational efficiency,
ness of both top and bottom piezoelectric plies is hp = 0.04 mm. The the weighting matrices Q and R in the control performance index
top and the bottom surfaces of the PZT plies are electrode and the defined in Eq. (101) take as diagonal matrices multiplied by some coef-
interfaces of the PZT plies with the steel structure member are ficients [32–35], that is Q = β1 I and R = β2 I, where and β1 and β2 are
grounded. For steel, Young’s modulus E ¼ 210 GPa Poisson ratio real numbers and I is an identity matrix. It can be seen from Eq. (101)
ν ¼ 0:3, and mass density ρ ¼ 7:9  103 kg=m3 . The material prop- that the weighting matrix for the vibration control Q is the dominant
erties of PZT used in this example are given in Ref. [32]. term for the vibration decay speed. Therefore a large number can be
The laminated piezoelectric beam shown in Fig. 6 is modeled using used for the coefficient β1 of the weighting matrix of the control effect
ten laminated piezoelectric beam elements presented in this study. The Q, and an identity matrix I can be used for the weighting matrix of the
contributions of the PZT plies to both the stiffness matrix and the mass control cost R.
matrix are taken into account in the present frequency analysis. The Figs. 8–10 plot the controlled dynamic responses at the free end of
structural damping is not considered in this case study, as the aim is the beam using LQR active control scheme with the weighting matrix,
to access the effectiveness of the active control. Ten beam elements respectively, Q = 108 I, Q = 109 I and Q = 1010, where R = I is used
are used to model the cantilevered beam illustrated in Fig. 6. The first for all the cases. The averaged control voltages applied to the element
five natural frequencies given by the present beam element are listed near the fixed end are also plotted in Figs. 8–10, in which the averaged
in Table 5. The numerical results given in Ref. [32], where Mindlin control voltage is the average value applied to the two nodes of the ele-
plate theory was used, are also listed in the table. The frequencies ment near the fixed end of the cantilevered beam.
given by the present beam element are smaller than those given by The dynamic responses plotted in Figs. 8–10 show that the vibra-
the element based on FSDT since the refined third‐order shear defor- tions decay very fast under the action of the control force induced from
mation plate theory used in the present element can more accurately the actuators. The vibrations come to rest, respectively, at t = 2.5 s,
model the transverse shear deformation than FSDT. t = 1.2 s and t = 0.7 s corresponding to Q = 108 I, Q = 109 I and
Q = 1010. The controlled vibrations with the LQR control using the
8.4. Active vibration control of laminated piezoelectric beam using LQR same weighting matrices presented by Narayanan and Balamurugan
with different weighting matrices Q and R [32] are plotted by the broken line in the figures for comparison. It
can be seen that the present results agree well with the results given
The laminated cantilevered beam with piezoelectric actuator and by Narayanan and Balamurugan. The response curves depicted in
sensor illustrated in Fig. 6 is used for the active vibration control study Figs. 8–10 show that the larger of weighting matrix Q is, the faster
in this numerical example. As shown in Fig. 6, the PTZ layer on the top the dynamic response decays.

Fig. 6. Laminated cantilevered beam with piezoelectric actuator and sensor.

Table 5
The natural frequencies of the cantilever piezoelectric beam.

Element type Natural frequencies (Hz)

Mode 1 Mode 2 Mode 3 Mode 4 Mode 5

FSDT [32] 6.89 43.29 121.23 237.72 393.58

present 6.67 42.54 119.77 234.28 388.91

13
J. Tian et al. Composite Structures 252 (2020) 112480

Fig. 8. Controlled tip deflection of laminated piezoelectric cantilever beam using LQR active control with Q = 108 I, R = I.

Fig. 9. Controlled tip deflection of laminated piezoelectric cantilever beam using LQR active control with Q = 109 I, R = I.

The control time to make the dynamic response vanish and the 8.5. Active vibration control of laminated piezoelectric beam using LQR
peak voltage in the beam element near the fixed end are summarized with different weighting matrices Q and R
in Table 6. It can be seen from the table that the larger of weighting
matrix Q, the shorter the time used to make the vibration come to rest, The influences of weighting matrices Q and R on the control perfor-
but also the larger of the peak voltage used to suppress the vibration. mance will be further studied in this example. The LQR active vibra-
The vibration control performances using CGVF control and CAVF con- tion control of the laminated piezoelectric beam schematically
trol were also studied by the first author of the present paper [39]. And shown in Fig. 11 is considered in this example. The dimensions of
it showed that LQR is really a good active vibration control scheme. the laminated piezoelectric beam in the x–z plane are depicted in

14
J. Tian et al. Composite Structures 252 (2020) 112480

Fig. 10. Controlled tip deflection of laminated piezoelectric cantilever beam using LQR active control with Q = 1010 I, R = I.

Table 6 rial properties of aluminum are E = 70.3 GPa, ν ¼ 0:345 and

The comparison of LQR vibration control using different weighting matrices Q. ρ ¼ 2:7  103 kg=m3 . The material properties of PZT‐5H are as
Weighting matrices Control time (s) Peak voltage (V) follows:
8
LQR (Q = 10 I, R = I) 2.5 1.07 C11 ¼ 126 GPa; C12 ¼ 79:5 GPa; C13 ¼ 84:1 GPa;
LQR (Q = 109 I, R = I) 1.2 2.63 C33 ¼ 117 GPa; C55 ¼ 23 GPa
LQR (Q = 1010 I, R = I) 0.7 5.16

e31 ¼ 6:5 C m2 ; e33 ¼ 23:3 C m2 ; e15 ¼ 17:0 C m2 ;

k33 ¼ 13 nF m1 ; ρ ¼ 7:5  103 kg=m3

Fig. 11, and the width b = 20 mm. Its main structure is made of alu- In order to compare the frequency results, the same element mesh
minum, and two PZT‐5H plies, which are polarized along the ply thick- as in Ref. [22] is adopted here, in which twenty equal‐sized beam ele-
ness, are placed on, respectively, its top and bottom surfaces to form ments are used to model the laminated piezoelectric beam. The mode
truly collocated actuator and sensor layers as illustrated in Fig. 11. superposition with the first five vibration modes is used to evaluate the
The interfaces of the PZT‐5H plies with the aluminum substrate are dynamic response of the beam. Since the focus in this example is the
grounded, and the top and the bottom surfaces of the piezoelectric study of the influence of the weighting matrices Q and R on the active
plies are electroded. vibration control, the damping effect on the vibration is not consid-
Kapuria and Yasin studied this laminated piezoelectric beam [22], ered. The first five undamped frequencies given by present beam ele-
and the material properties here are taken from their paper. The mate- ment and by Kapuria and Yasin are tabulated in Table 7. In the two

Fig. 11. Schematic of laminated piezoelectric beams with truly collocated actuator and sensor layers.

15
J. Tian et al. Composite Structures 252 (2020) 112480

Table 7
The natural frequencies of the laminated piezoelectric beam given by different theories.

Element models Natural frequencies (Hz)

Mode 1 Mode 2 Mode 3 Mode 4 Mode 5

1 segment electrode [22] 86.0370 533.676 1483.84 2885.85 4727.43

5 segment electrodes [22] 86.5922 540.107 1501.22 2911.35 4742.98
Present 85.9746 530.482 1474.39 2869.17 4701.54

Fig. 12. Tip displacement of piezoelectric laminated cantilever beam using LQR active control with various weighting matrix Q.

Fig. 13. Control voltage for the active vibration control of piezoelectric laminated cantilever beam using LQR control scheme with various weighting matrix Q.

16
J. Tian et al. Composite Structures 252 (2020) 112480

node beam element developed by Kapuria and Yasin, both the kine- by the present beam element are little bit smaller than the results
matics and the electric potential are modeled by layerwise theory; obtained from the twenty beam elements with additional one segment
the beam element has two conventional nodes as well one electric electrode.
node which has no physical coordinate; and a quadratic function The cantilevered laminated beams are subjected to an impulse
across the piezoelectric layers was used for the electric potential. excitation of 0.2 N for 1 ms at its free end. The resulting response
Therefore, the laminated piezoelectric beam developed by Kapuria of the tip deflection is plotted by the black dotted curve in Fig. 12,
and Yasin is really a higher‐order piezolaminated beam element. It in which the amplitude of the tip deflection vibration keeps at
can be found from the results in Table 7 that the frequencies given 1:8  104 m.

Fig. 14. Tip displacement of piezolaminated cantilever beam using LQR active control with various weighting matrix R.

Fig. 15. Control voltage for the active vibration control of piezolaminated cantilever beam using LQR control scheme with various weighting matrix R.

17
J. Tian et al. Composite Structures 252 (2020) 112480

Fig. 16. The evolution of fitness during GA process.

Fig. 17. The dynamic response of the tip displacement of a piezolaminated cantilever beam under the LQR active control with optimal weighting matrix
Q = 2.97 × 107 I (R = I).

The active vibration controls using LQR control scheme with the A fixed Q = 107 I is chosen for the study of the influence of various
different weighting matrices Q and R are studied separately. A fixed R on the control performance. Fig. 14 displays the controlled tip
R = I is chosen for the study of the influence of Q with various values. deflections by LQR control with, respectively, R = 0.1 I, R = 0.5 I,
Fig. 12 plots the controlled tip deflections given by, respectively, R = I and R = 2 I. Fig. 15 shows the control voltages for the LQR
Q = 107 I, Q = 108 I, and Q = 109 I. Fig. 13 plots the control voltage active vibration control using different weighting matrices R with,
curves for the active vibration control corresponding to Q = 107 I, respectively, R = 0.1 I, R = 0.5 I, R = I and R = 2 I. It can be seen
Q = 108 I, and Q = 109 I respectively. It can be found from Figs. 12 that the control voltage decreases as the weighting matrix R increases,
and 13 that the larger Q can make the vibration decay fast, but it but the control time becomes longer as the weighting matrix R
results in higher voltage inputs. increase.

18
J. Tian et al. Composite Structures 252 (2020) 112480

Fig. 18. The control voltage curve of a piezolaminated cantilever beam under the LQR active control with optimal weighting matrix Q = 2.97 × 107 I (R = I).

Table 8 dures illustrated in Fig. 3, the rule of survival of fitness leads to the
The comparison of control performance under different weighting matrixes Q. optimal weighting matrix Q. The evolution of fitness as a function of
generation is displayed in Fig. 16, which shows that the evolution of
Weighting matrices Control time (s) Peak voltage (V) Performance Index J
the maximum fitness of individual chromosome is stabilized at the
Q = 107 I > 0.250 14.16 3.24 × 108 averaged fitness after 120 generations of evolution. Q = 2.97 × 107
Q = 108 I 0.163 22.25 5.66 × 108 I is obtained with the fitness 9.13 × 10−9 after 200 generations of
Q = 109 I 0.115 34.58 1.63 × 109
Optimal matrix
evolution.
(Q = 2.97 × 107 I) 0.212 18.75 1.09 × 108 Q = 2.97 × 107 I and R = I are used for the weighting matrices in
the LQR control performance index defined in Eq. (101). Fig. 17 plots
the controlled dynamic response of the tip displacement of a laminated
piezoelectric cantilever beam with optimal weighting matrix Q = 2.
The controlled vibrations depicted in Figs. 12–15 show that the 97 × 107 I and R = I. Fig. 18 displays the control voltages correspond-
weighting matrices Q and R in the control performance index define ing to the optimal weighting matrix Q = 2.97 × 107 I and R = I. It
in Eq. (101) play important roles in the vibration control effect and can be found from the figures that vibration is vanished at
the control cost, in which Q plays a more influential role. t = 0.212 s, and the maximum control voltage is 18.75 V.
Table 8 summarizes the LQR active control performances with the
8.6. Active vibration control of laminated piezoelectric beam using LQR weighting matrices Q given by, respectively, the trial and error
with the weighting matrix Q determined by GA approach used in Subsection 8.5 and the GA‐based Q presented in this
subsection. It can be found from Table 8 that the Q = 2.97 × 107
The study of the influence of weighting matrices Q and R con- determined by GA yields not only the smallest control performance
ducted in the previous example shows that an identity matrix for R index among the various weighting matrices but also a short control
is a rather good choice, but feasible Q can be varying in a large range. time that compromises the control time and the control cost.
It can be seen from the control performance index defined in Eq. (101)
that optimizing Q by keeping R as an identity matrix is a good strategy 9. Conclusions
to evaluate the optimal weighting matrices. This is because if set
R ¼ β0 I; ðβ0 > 0Þ, then a coefficient βQ can be found that the weight- Based on a new refined third‐order shear deformation plate theory
ing matrix Q=βQ would yield the same value of the control perfor- and a layerwise zigzag approximation for the electrical field in piezo-
mance index as weighting matrices Q and R. Therefore, in the electric plies, a two‐noded laminated piezoelectric beam element is
present study R is fixed as an identity matrix, and using GA to deter- presented. The electromechanical coupling in piezoelectric plies is
mine the optimal Q. taken into account by the electric enthalpy of piezoelectric materials.
An optimal weighting matrix Q can be expressed as Q ¼ αQ I. The The element equations are derived from the Hamilton’s principle with
following parameters are used in the process using GA to determine electric enthalpy. The quasi‐conforming element technique is used to
the optimal value αQ . Range of initial values of αQ : 104–1010; Initial compute the element stiffness matrix. The resulting two‐noded lami-
population: 80; Crossing probability: Pc = 0.8; Mutation probability: nated piezoelectric beam element is free from locking shear and free
Pm = 0.005; Maximum generation: Nend = 200. By using the afore- from the time consuming numerical integration as its element stiffness
mentioned parameters and following the flowchart of the GA proce- matrix is evaluated explicitly. LQR optimal control scheme is

19
J. Tian et al. Composite Structures 252 (2020) 112480

employed for the active vibration control, and the optimal weighting [14] Carrera E, Brischetto S, Cinefra M. Variable kinematics and advanced variational
statements for free vibrations analysis of piezoelectric plates and shells. Comput
matrix for the control performance is determined by GA. A number
Model Eng Sci 2010;65(3):259–341.
of numerical examples are solved to demonstrate the accuracy of the [15] Koutsawa Y, Giunta G, Belouettar S. Hierarchical FEM modelling of piezo-electric
present laminated piezoelectric beam element and the effectiveness beam structures. Compos Struct 2013;95:705–18.
using GA to evaluate the optimal weighting matrix Q in LQR‐based [16] Cinefra M, Lamberti A, Zenkour Ashraf M, Carrera E. Axiomatic/asymptotic
technique applied to refined theories for piezoelectric plates. Mech Adv Mater
optimal vibration control. The following conclusions can be draw from Struct 2015;22(1–2):107–24.
the numerical results given by the present two‐noded laminated piezo- [17] Koutsawa Y, Giunta G, Nasser H, Belouettar S. Static analysis of shear actuated
electric beam element. piezo-electric beams via hierarchical FEM theories. Mech Adv Mater Struct
2015;22(1–2):3–18.
[18] Koutsawa Y, Giunta G, Belouettar S. A free vibration analysis of piezo-electric
1) The two‐noded laminated piezoelectric beam element based on beams via hierarchical one-dimensional finite elements. J Intell Mater Syst Struct
the hybrid plate theory and quasi‐conforming element tech- 2014;25(8):1009–23.
[19] Carrera E, Nali P. Mixed piezoelectric plate elements with direct evaluation of
nique is very computational efficient as the element stiffness transverse electric displacement. Int J Numer Methods Eng 2009;80:403–24.
matrix is given explicitly. [20] Kulikov GM, Plotnikova SV, Carrera E. Hybrid-mixed solid-shell element for stress
2) The use of the refined third‐order shear deformation plate the- analysis of laminated piezoelectric shells through higher-order theories. Analysis
and modelling of advanced structures and smart systems, pp. 45–68, Springer,
ory can improve the prediction accuracy of the higher‐mode Singapore, 2018.
frequencies. [21] Wang X, Shi G. A simple and accurate sandwich plate theory accounting for
3) The electric potential difference between the top surface and transverse normal strain and interfacial stress continuity. Compos Struct
2014;107:620–8.
bottom surface of a piezoelectric ply can be efficiently used as
[22] Kapuria S, Yasin MY. Active vibration control of piezoelectric laminated beams
the electric nodal degrees of freedom for each piezoelectric with electroded actuators and sensors using an efficient finite element involving an
ply in the finite element modeling of laminated piezoelectric electric node. Smart Mater Struct 2010;19:15. 045019.
beams. [23] Elshafei MA, Alraiess F. Modeling and analysis of smart piezoelectric beams using
simple higher order shear deformation theory. Smart Mater Struct 2013;22:14.
4) It is an effective strategy for LQR active vibration control to 035006.
employ GA to merely determine the optimal weighting matrix [24] Khandelwal RP, Anupam C. Static and dynamic control of smart composite
used for the vibration control performance by fixing the weight- laminates. AIAA J 2014;9(52):1896–914.
[25] Djojodihardjo H, Jafari M, Wiriadidjaja S, Ahmad KA. Active vibration suppression
ing matrix for the control input cost as an identity matrix. of an elastic piezoelectric sensor and actuator fitted cantilevered beam
configurations as a generic smart composite structure. Compos Struct
2015;132:848–63.
[26] Shi G, Voyiadjis GZ. Efficient and accurate four-node quadrilateral C0 plate
Declaration of Competing Interest bending element based on assumed strain fields. Int J Numer Methods Eng
1991;32:1041–55.
The authors declare that they have no known competing financial [27] Shi G, Lam KY. Finite element vibration analysis of composite beams based on
higher-order beam theory. J Sound Vib 1999;219:707–21.
interests or personal relationships that could have appeared to influ- [28] Shi G, Lam KY, Reddy JN. Assumed strain quadrilateral C0 laminated plate element
ence the work reported in this paper. based on third-order shear deformation theory. Struct Eng Mech 1999;8:623–37.
[29] Tang L, Chen W, Liu Y. Quasi-conforming element for finite element analysis. J
Dalian Inst Tech 1980;19(2):19–35 [in Chinese].
Appendix A. Supplementary data [30] Bickford WB. A consistent higher order beam theory. Dev Theor Appl Mech
1982;11:137–50.
Supplementary data to this article can be found online at https://doi. [31] Shi G, Lam KY, Tay TE. On efficient finite element modelling of composite beams
and plates using higher-order theories and an accurate composite beam element.
org/10.1016/j.compstruct.2020.112480. Compos Struct 1998;41:159–65.
[32] Narayanan S, Balamurugan V. Finite element modelling of piezolaminated smart
References structures for active vibration control with distributed sensors and actuators. J
Sound Vib 2003;262:529–62.
[33] Schulz SL, Gomes HM, Awruch AM. Optimal discrete piezoelectric patch allocation
[1] Saravanos DA, Heyliger PR. Mechanics and computational models for laminated
on composite structures for vibration control based on GA and modal LQR. Comput
piezoelectric beams, plates, and shells. Appl Mech Rev 1999;152(10):305–20.
Struct 2013;128:101–15.
[2] Kapuria S, Kumari P, Nath JK. Efficient modeling of smart piezoelectric composite
[34] Zabihollah A, Sedagahti R, Ganesan R. Active vibration suppression of smart
laminates: a review. Acta Mech 2010;214:31–48.
laminated beams using layerwise theory and an optimal control strategy. Smart
[3] Luschi L, Iannaccone G, Pieri F. A critical review of reduced one-dimensional beam
Mater Struct 2007;16:2190–201.
models of piezoelectric composite beams. J Intell Mater Syst Struct 2019;30
[35] Roy T, Chakraborty D. Optimal vibration control of smart fiber reinforced
(8):1148–62.
composite shell structures using improved genetic algorithm. J Sound Vib
[4] Heyliger P, Pei KC, Saravanos DA. Layerwise mechanics and finite element model
2009;319:15–40.
for laminated piezoelectric shells. AIAA J 1996;34(11):2353–60.
[36] Shi G. Electric enthalpy of piezoelectric materials and coupling analysis of
[5] Kapuria S, Alam N. Efficient layerwise finite element model for dynamic analysis of
mechanical and electric fields. Acta Mater Compos Sin 2003;20(6):115–20 [in
laminated piezoelectric beams. Comput Methods Appl Mech Eng
Chinese].
2006;195:2742–60.
[37] Bathe K-J. Finite element procedures. Prentice-Hall: Englewood Cliffs, NJ, USA;
[6] Mitchell JA, Reddy JN. A refined hybrid plate theory for composite laminates with
1996.
piezoelectric laminae. Int J Solid Struct 1995;32:2345–67.
[38] Shi G, Atluri SN. Active control of nonlinear dynamic response of space-frames
[7] Wang X. Laminated Composite Plate Theories Accounting for Interlaminar
using piezo-electric actuators. Comput Struct 1990;34:549–64.
Continuity and its Accurate, Efficient Plate Elements. Ph.D. thesis, 2015, Tianjin
[39] Tian J. Accurate Beam Element and Active Vibration Control of Piezoelectric
University, China.
Laminated Beams. M.S. thesis, 2019, Tianjin University, China.
[8] Reddy JN. A simple higher-order theory for laminated composite plates. ASME J
[40] Cheng ZQ, Lim CW, Kitipornchai S. Three-dimensional asymptotic approach to
Appl Mech 1984;51:745–52.
inhomogeneous and laminated piezoelectric plates. Int J Solids Struct 2000;37
[9] Shi G. A new simple third-order shear deformation theory of plates. Int J Solids
(23):3153–75.
Struct 2007;44:4399–417.
[41] Chee CY, Tong L, Steven PG. A mixed model for composite beams with
[10] Shi G, Voyiadjis GZ. A sixth-order theory of shear deformable beams with
piezoelectric actuators and sensors. Smart Mater Struct 1999;8(3):417–32.
variational consistent boundary conditions. ASME J Appl Mech 2011;78
[42] Tzou HS, Tseng CI. Distributed vibration control and identification of coupled
(2):856–75.
elastic/piezoelectric systems: finite element formulation and application. Mech
[11] Wang X, Shi G. The boundary layer solutions induced by displacement boundary
Syst Sig Process 1991;5(3):215–31.
conditions of shear deformable beams and accuracy study of higher-order beam
[43] Beheshti SB, Vidal P. A refined sinus finite element model for the analysis of
theories. ASCE J Eng Mech 2012;138:1388–99.
piezoelectric-laminated beams. J Intell Mater Syst Struct 2011;22(3):203–19.
[12] Challamel N. Variational formulation of gradient or/and nonlocal higher-order
[44] Correia VMF, Gomes MAA, Suleman A, et al. Modelling and design of adaptive
shear elasticity beam. Compos Struct 2013;105:351–68.
composite structures. Comput Methods Appl Mech Eng 2000;185:325–46.
[13] Carrera E. Theories and finite elements for multilayered plates and shells: a unified
[45] Chandrashekhara K, Bangera KM. Linear and geometrically nonlinear analysis of
compact formulation with numerical assessment and benchmarking. Arch Comput
composite beams under transverse loading. Compos Sci Tech 1993;47:339–47.
Methods Eng 2003;10(3):215–96.

20