Vibration Analysis of Structures with Rotation and Reflection Symmetry

Shock and Vibration, Jul 2018

The article applies group representation theory to the vibration analysis of structures with Cnv symmetry, and presents a new structural vibration analysis method. The eigenvalue problem of the whole structure is divided into much smaller subproblems by forming the mass and stiffness matrices of one substructure and than modifying them to form mass and stiffness matrices in each irreducible subspace, resulting in the saving of computer time and memory. The modal characteristics of structures with Cnv symmetry are derived from theoretical analysis. Computation and modal testing are used to verify the validity of the theoretical deductions.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

http://downloads.hindawi.com/journals/sv/1996/123520.pdf

Vibration Analysis of Structures with Rotation and Reflection Symmetry

Shock and Vibration, Vol. Vibration Analysis of Structures with Rotation and Reflection Symmetry Jie Zhao 0 1 0 Jinan , Shandong People's Republic of China 1 Department of Material Engineering Shandong University of Technology The article applies group representation theory to the vibration analysis of structures with Cnv symmetry, and presents a new structural vibration analysis method. The eigenvalue problem of the whole structure is divided into much smaller subproblems by forming the mass and stiffness matrices of one substructure and than modifying them to form mass and stiffness matrices in each irreducible subspace, resulting in the saving of computer time and memory. The modal characteristics of structures with Cnv symmetry are derived from theoretical analysis. Computation and modal testing are used to verify the validity of the theoretical deductions. © 1996 John Wiley & Sons, Inc. - INTRODUCTION Group theory is a mathematical tool in the study of symmetry. Symmetry is possessed by many engineering structures. We can apply a symmetry operation to such a structure that interchanges the positions of various points in it but results in the structure looking exactly the same as before the symmetry operation (the structure is in an equivalent position). When the operation is con­ tinuously repeated, the structure continues to be in an equivalent position. There is one type of symmetry operation for which one point in the body remains fixed. These are the point symmetry operations. Two such operations are Cn' which means rotation by 2rr/ n about an axis of the body in the sense of a right-hand screw, and O'v' which means reflection in a plane containing the en axis. Repeating the symmetry operation C n m times, we get an operation C;~, which means rotation by 2mrr/n about the C n axis [m = 0, 1, . . . , (n - 1)]. In particular C~ is the identity operation E under which the structure is not rotated at all. The collection of E, C~, . . . , and C~-l forms a group in the mathematical sense. This group is called the Cn point group. Suppose there are n reflection planes all con­ taining the Cn axis denoted by 0' (0), 0' (I), . • . , O'(n-I), of which O'(i) has an angle rr/n with its neighbor O'(i+1) (i = 0, 1, . . . , n - 1, and define O'(n) = 0'(0). By adding O'v symmetry operations O'~) across the reflection plane 0' (i) (i = 0,1, . . . , n - 1) to the Cn point group, a new point group Cnv can be obtained (Burns, 1977) . For convenience of analysis, the 2n symmetry operations of the Cnv group are called sequentially the first, second, to the 2nth symmetry operation. A Cnv structure can be defined as the one that remains in an equivalent position when acted on by anyone of the 2n symmetry operations of the Cnu group. According to the 2n symmetric operations of the C nv group, a Cnv structure can be divided into 2n identical substructures denoted by S(K) (K = I, 2, . . . , 2n). Choose one of the substructures arbitrarily as the basic substructure and make it correspond to the first symmetry op­ eration E of the Cnu group and define it as 5(1); S(k) can be obtained by applying the kth symmetry operation on the basic substructures 5(1). A C4v structure (the square plate ABCD) is shown in Fig. I, where the z axis coincides with the C4 axis and the reflection plane u(O) contains the x axis, u O) contains AC, etc. As shown in the figure, 5(2) is obtained by applying the rotation operation q (the second element of the C4v group) on SO); 5(5) is obtained by applying the reflection opera­ tion u~O) (the fifth element of C4v group) on 5(1), etc. Thus the plate is divided into eight identical substructures according to the eight symmetry operations of the C4u group. Many engineering structures, such as the top cover of the supporting structure of a hydraulic turbine, are symmetric under group Cnu . Zhong and Qiu (I978a, b) applied group representation theory to static analysis of structures with CIl or Cnu symmetry; but by applying group theory to vibration analysis of these structures, we can di­ vide the natural eigenvalue problems into a set of much smaller subproblems. This leads to a remarkable saving of computer time and memory and hence increases the efficiency of calculation. Also, modal characteristics of natural vibration, derived from theoretical analysis, can be ex­ pected to provide some theoretical basis for modal testing. Only Cnu structures with no center point with n an even integer are studied in this ar­ ticle. COORDINATE SYSTEM FOR APPLYING GROUP THEORY For the application of group theory to the study of Cnu structures, it is extremely inconvenient or even impossible to use the rectangular coordinate system as shown in Fig. I. Thus, a multicoordi­ nate system (MCS) is established according to the classification of nodal points in a Cnv structure. As shown in Fig. 2, an MCS denoted by cd) is established in SO), where the coordinate system C~6 is used to describe the interior points of C(l), C~li is used to describe the first class symmetric points (i.e., points on the interface between 5(1) and s(n+II), and C~i is used to describe the second class symmetric points (i.e., points on the inter­ face between SO) and s(n+2». The points classifiz S(4) A cation and MCS C~) for SCk) can be obtained by applying the kth symmetry operation of the Cnv group on S(l) and C~). The assembly ofC~) (where k = 1, 2, . . . , 2n) form the MCS denoted by Cs that describes the nodal displacement of the whole structure. The advantage of the MCS Cs is that it ensures that every substructure has identical mass and stiffness matrices; however, in such an MCS, ev­ ery symmetric point (including the first and sec­ ond class symmetric points) is described by two coordinate systems, which is inconvenient. To solve this problem, an MCS Cr shown in Fig. 3 is used to describe the symmetric points. Thus, in the MCS Cr , the coordinate systems for symmetric points of substructures SCk) (k = 1, 2, . . . , n) are identical with that of the MCS Cs and the v axis of the coordinate system for symmetric points of substructures SCk) (k = n + 1, n + 2, . . . , 2n) is of opposite direction to that ofthe Cs . Obviously the C~) (k = 1,2, . . . , n) of Cs are right-hand systems while the C~) (k = n + 1, n + 2, . . . , 2n) are left-hand systems, which are due to the reflection operation. Generally, six displacement components (i.e., the displacements u, v, w along axes x, y, z and the angular displacements Ox, Oy, 0z around axes x, y, z, respectively) are needed to describe a nodal displacement of an engineering structure. Then the relationships between Cs and Cr are as follows for the first class symmetric points, {UO), . {vO), . . ,dn), dn+l), . . , v(n), urn +I), . {wCI), . . ,wCn ), w(n+l), . {8~I), . . ,o~n), o~n+l), . {O(yl) , • {8~I), . ., o(yn)' Yo(n+1) , • . ,o~n), o~n+l), . . , d 2n)}Cs = {u(l) ' . . , ur2n)}Cs = {vO)' . . , w(2n)}Cs = {wCI) ' . ., o x(2n)}Cs = {O(xl) , • ., o y(2n)}Cs = {O(yI) , • ., oCz2n)}Cs = {80z) ' . · ,uCn), uO), . ., urn), -vOl,. · ,w(n), wCI), . · ,o~n), -O~I), . ., o(Yn)' 0 (1) Y" • , urn)}C T • , -vCn)}CT • , wen)}C T . , -o( nx)}Cr ., o Y(n)}C T · , o~n), - 0 ~I), . • , -o(nZ )}CT· For the second class symmetric points, {VO), . . . , v(n), vCn+l), . . ,vCn), -v(n), -vOl, . • ., -urn-I)}CT {O YO) , • ., oCyn)' Yo(n+1) , • • , UC2n)}Cs = {u(1) ' . . , V(2n)}Cs = {vOl ' . . , W C2n)}Cs = {w(1) ' . ., Ox(2n)}Cs = {O x(I) , • ., Oy(2n)}Cs = {eeyl) ,. ., O(z2n)}Cs = {O z(I) ' • ., Ox(n)' -o(nx)' -0(1x) " " , -o(xn-I)}Cr ., OY(n) , o(Yn) , 0 Y0)" • ., _o(yn-I)}Cr ., O(zn)' - o(n) - 0 (zI)" " z' , - 0 z(n- I)} (1) (2) where the component uO) of the displacement vector is the u displacement of a nodal point in substruc­ ture SO) and u(i) (i = 2, 3, . . . , 2n) is that of the corresponding nodal point of substructure SU). BASE VECTORS OF EACH CLASS OF POINTS IN MCS Cs Selecting an interior nodal point in S(I) with corre­ sponding nodal points in S(i) (i = 2, 3, . . . , 2n) determined by the symmetry operations of Cnv group, we can get 2n corresponding nodal points. With the same displacement component of every nodal point we can form a 2n dimensional vector {8(l), 8(2), . . . , 8(2n)}. The vector belongs to the Euclidean space R 2n that can be determined as the representative space of the Cnv group for interior points. The representative space can be reduced into four I-dimensional subspaces as VAl V A2 V BI , V B2 and (n - 2) 2-dimensional irred~cibl~ subspaces as V EKI , V EK2(k = 1,2, . . . , (nI2) 1). The base vectors of each irreducible sub­ space are cpAI = {I, 1,. cpA2 = {I, 1, . . , 1; - 1, - 1,. . ., - 1V cpRI = {I, -1,. cpB2 = {I, -1, . ., -1; 1, -1,. . , -1; -1, 1, . ., -IV . ,IV cpfKI = {I, cos kO, . .. , cos ken - 1)0; cos knO, . .. , cos k(2n - I)()V cpfKI = {a, -sin kO, . .. , -sin ken - 1)0; sin knO, . .. , sin k(2n - l)oV cpfK2 = {a, sin kO, . .. , sin ken - 1)0; sin knO,. . ., sin k(2n - 1)()V cpfKl = {I, cos kO, . . , cos ken - 1)0; -cos knO, . . , -cos knO, . -cos k(2n - l)oV, (3) where k = 1,2, . . . ,(nI2) - 1,0 = (27Tln), and in each base vector the two parts before and after the semicolons are all n dimensions. These are called the base vectors for the interior points. Selecting a symmetric point in S(I) and applying the 2n symmetry operations of the Cnv group on it, we can get 2n corresponding points belonging to the 2n substructures. These are, however, physically n different points that should be de­ scribed by the MCS Cr. Thus, the representative spaces of the Cnv group for symmetric points are n-dimensional. They can be reduced into irreduc­ ible subspaces, and the base vectors of these sub­ spaces are n-dimensional under the MCS Cr. These base vectors must be expanded into 2n­ dimensional vectors according to the relation­ ships between Cs and Cr , i.e., Eqs. (1) and (2). The expanded base vectors for symmetric points can be expressed as base vectors for interior points and their combinations. The base vectors for u, W, OV in the first class points are cpAI, cpBI, cpfKI, and cpfK2; the base vec­ tors for v, Ox, Oz in the first class points are cpA2, cpB2, cpfKI, and cpfK2; and the base vectors for u, W, Oy in the second class points are cpAI, cpB2, cos ( k;) cpfKI + sin ( k;) cpfKI and cos (k;) cpfK2 + sin (k;) cpfK2; The base vectors for v, Ox, Oz in the second class points are where k = 1,2, . . . , (nI2) - 1. Cnv GROUP TRANSFORMATION Because we obtained base vectors for every dis­ placement component (i.e., U, v, w, Ox, Oy, and Oz) of every kind of point (i.e., interior points, first class symmetric points, and second class symmetric points) under the MCS Cs, the 2n­ dimensional vectors WI), 0(2), . . . , 0(2n)Y, in which o(i) is one of the six displacement compo­ nents of a nodal point in Sri) (i = 1, 2, . . . , 2n), can be expressed as the linear combination of base vectors in (3). The coefficients of any base vector are called the generalized displacements related to the corresponding irreducible sub­ space. Denote the global physical displacement of S(k) as where L is the total number of degrees offreedom of S(k) when the finite element discretization is made in S(k). Denote 11,12 ,13 ,14 , and Is to be, respectively, the subscript sets corresponding to: the displace­ ments of the interior points; u, W, and Oy of the first class symmetric points; u, W, and Oy of the second class symmetric points; v, O.n and Oz of the first class symmetric points; and v, Ox, and Oz of the second class symmetric points. Let where <I> is the reducing matrix and is formed by arranging all the base vectors for the interior points sequentially. Thus we have the matrix form of the group transformation from physical displacements to generalized displacements. = <I> , (4) I I i2 0 03 qi2 0 qOill l 2il q03 il q~ I Iil ••• ,,,J ... qki2l qlkl2 ••• ,,,I ••• 0 , l 3il ••• ,,,I • " l 3i2 , l 4il ••• ,I ••• 0 : qOl i3 0 0 qr;i cos knrr qlk312 sin 'W i~2 cos 'W q~j4 sin knrr q,k334 1 ,,,,. . . ••• ,,,,I ••• ,,, .. . q0i42 ••• ,I ••• , ,,,, .. . 0 ••• ,,,,I ... ,t ••• qi044 •• , I,'" 0 l 2i4 0 k4 qi4 0 q02 i5 q03 i5 0 -s m. nkrr qik512 cos 'W q~~2 -s m. nkrr qik534 cos 'W q~r where im Elm (m = 1,2, . . . ,5). Denote the transpose of each row vector of the generalized displacement matrix in turn as follows where k = 1, 2, nl2 - 1, and that of the reducing matrix as <l>j (j = 1, 2, . . . , 2n), then Li. Zheng. and Zhao €fJ J = {t/ljOI' t/ljo2' t/ljoJ' t/ljo4' . t/ljkl , t/ljk2, t/ljkJ, t/ljk4'· . .}. Here '.f1'J,O.I is the l·th element of rpAI, and so on. From Eq. (4) we have 8(j)T = €fJJ[qOl, q02, q03, q04, . . qkl, qk2, qkJ, qk4,. . .J T that is, Therefore, the global displacement vector of Sri) is expressed as a linear combination of the generalized displacement vectors (GDV, i.e., qkl, qk2, qkJ, and qk4, k = 0, 1, 2, nl2 - 1) of each irreducible subspace. DECOMPOSITION OF NATURAL EIGENVALUE PROBLEM Structural mass and stiffness matrices are formed first. For convenience, the mass and stiffness ma­ trices of substructure S(k) are formed first under C~~, which describes the interior points in MCS C~), and then they are transformed into mass matrix M and stiffness matrix K, respectively, under MCS C~). Because all the substructures and the MCS they adopt are identical, their mass matrices and stiffness matrices are identical too. Therefore, the potential energy ofthe entire struc­ ture is in which the terms in the summation are energy coupled in proper sequence, so that it is impossi­ ble to decompose the eigenvalue problem. Using Eq. (5) in Eq. (6) and applying group theory, we can prove that 1 (n/2)-1 V = - L 2 k=O X~(qkITKqkJ + qk2TKqk2 + qkJTKqkJ + qk4TKqk4), (5) (6) (7) in which X k2-- { 2n, n, whenk = 0, whenk= 1,2,. ., !2! - I Compared with Eq. (6), the energy coupling problem in Eq. (7) has lessened greatly. But there are still two problems that must be solved: the first is that all the generalized displacement vec­ tors are constrained, that is, they contain zero components that should be deleted; the second is that there is still coupling between generalized displacement vectors qkl and qk2 in subspace V EKI as well as those between generalized dis­ placement vectors qkJ and qk4 in subspace V EK2 (where k = 1, 2, . . . , nl2 - 1), which cannot be further decoupled. To solve these two prob­ lems, special steps are taken as follows. Delete the zero components of GDV qOI, q02, q03 , and q04, which are in I-dimensional subspaces VAl, V A2, VBI, and V B2, and change them into qAI, qA2, qRl, and qR2, respectively. The corre­ sponding stiffness matrices are changed into KA DI, eKlAet'e,KthBe 1,zaenrdo KcBom, 2preosnpeencttsivoeflyG. kDV q I and qk2, which are in the 2-dimensional subspace VEKI, and combine them into one vector denoted by qEKI . The two corresponding identical stiffness matrices are changed into one and denoted by K EKI · . . Similarly, the two GDV qkJ and qk4 10 2-dlmensional subspace V EK2are combined into qEK2, and the stiffness matrices into K EK2 . Notice that qkl and qk2 as well as qkJ and qk4 are identical in form; therefore, KEKJ = K EK2 . And the potential energy can also be written as V = n(qAITKAlqAI + qA2TKA2qA2 + qBITKBJqBI + qB2TKB2qB2) (n/2)-1 + ~ ~ (qEKITKEKlqEKJ + qEK2TKEK2qEK2) . (8) Here any term in summation is completely inde­ pendent of the others, and there is no energy coupling. In a similar way, the kinetic energy expression is T= n((jAITMA/1AI + qA2™A/lZ + qBITMBlqBI + qBITMB2qB2) (nI2)-1 + ~ ~ + qEK2TMEK2qEK2). (qEKITMEKlqEKI Substituting Eqs. (8) and (9) into the Lagrange equation, we get the generalized structural natu­ ral vibration equations completely independently as follows M AI ijAI + K AI qAI = 0 M A2 ijA2 + K A2 qA2 = 0 M BI ijBI + K BI qBI = 0 M B2 ijB2 + K B2qB2 = 0 M EKI ijEKI + K EKI qEKI = 0 MEK2ijEK2 + KEK2qEKz = 0 where k = 1, 2, . . , n!2 - 1. PRELIMINARY WORK ON MODAL CHARACTERISTICS OF Cnv STRUCTURES From Eqs. (10) to (15) we know that because the mass matrices and stiffness matrices in the generalized eigenvalue problems of the four 1dimensional subspaces V AI_ V B2 are different from each other, frequencies solved for them are all single frequencies (one single frequency has only one modal shape), but for the two 2-dimen­ sional subspaces VEKI and V EK2, the mass and stiffness matrices in the generalized eigenvalue equations are identical; therefore, the frequencies solved for them are all duplicated frequencies (a duplicated frequency has two modal shapes). In the subspace V AI, Eq. (10) can be solved to give qA] and hence qOl. Introducing qOl into Eq. (5), we obtain where j = 1, 2,. . . , 2n. Here all substructures vibrate in phase with equal amplitudes and from o(j) = ''l/''j.02 q02, where j = 1, 2, . . . , 2n. From the property of cpAz and q02 it is known that all the corresponding points between the neighboring substructures vi­ brate in antiphase with equal amplitude and u, w, and OJ' are equal to zero at the symmetric point. In the subspace V BI , o(j) = ''l/''j.OJ q03 ' wherej = 1,2, . . . , 2n. Here, all the correspond­ ing points among substructures as S(2k-1) (k = 1, 2, . . . , n) vibrate in phase with equal amplitudes, and the corresponding points among S(2k) (k = 1, 2,. . ., n) also vibrate in phase with equal-ampli­ tude, but the corresponding points between S(2k-l) and S(2k) (k = 1, 2, . . . , n) vibrate in antiphase with equal amplitude. In other words, a substructure must vibrate in phase with one of its two neighboring substructures and in anti­ phase with the other. Furthermore, for the sym­ metric points between the neighboring substruc­ tures that vibrate in phase, their v, Ox, and Oz are all zero; and for the symmetric points between the neighboring substructures that vibrate in anti­ phase, the values of u, w, and Ov are zero. In the subspace V Bz, • (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) wherej = 1,2, . . . , 2n. Here the modal shapes are similar to those in VBI. In the subspace VEKI, qkl and qk2 can be solved from qEKI, hence where j = 1, 2, . . . , 2n. In the subspace V EK1, because qk) = qkl and qk4 = qk2, therefore, where j = 1, 2, . . . , 2n. The regularity of the modal shapes of dupli­ cated frequencies is more complex and hence only its mathematical expressions are given as Eqs. (20) and (21). It is easy to prove that the modal shapes described in (20) and (21) are orthogonal under Cs • COMPUTATION AND EXPERIMENT To verify the validity of the theoretical deductions above, programs were worked out and compari­ son made between the finite element and group theory methods. The example structure is a hex­ agonal plate with the inner boundary fixed and outer boundary free, as shown in Fig. 4. The finite element discretization is made only in substructure S(l), and the others are obtained by applying the corresponding symmetry opera­ tions on them. For the sake of comparative calcu­ lation, special techniques were used for data input in the finite element method, that is, the nodal data input of S(l) is identical with that of the group VAl (SF) COMP:189 EXPE:209 %ERR:-0.09 VA2 (SF) COMP:1514 EXPE:1371 %ERR:0.15 V Bl (SF) COMP:379 EXPE:450 %ERR-0.16 VB2 (SF) COMP:359 EXPE:368 %ERR:-0.02 VEll&. V E12 (DF) COMP:187 EXPE:180 %ERR:0.04 VE2l&. V E22 (DF) COMP:225 EXPE:258 %ERR:-0.13 theory method. For the other substructures, the nodal data can be obtained automatically by that of 5(0 through the program. The calculation showed that the results of these two algorithms are identical but the finite element method used about 10 min while the group theory algorithm used less than 20 s. The computer stor­ age needed by the group theory method is only one-tenth of that of the finite element method. Having C6v symmetry, the hexagonal plate has eight irreducible subspaces, i.e., four I-dimen­ sional subspaces VAl, V B [, V A2, and VB2; and four 2-dimensional subspaces VEIl, V E [2, V E2[, and V E22. From the theoretical studies above, we can predict that vibrations in VAl, V B [, V A2, and VB" must have single frequencies and modal shapes described by Eqs. (16)-(19) and vibrations in V E[ [ and V E[2 as well as V E,[ and V E22 must have duplicated frequencies and modal shapes described by Eqs. (20) and (21). Computational and experimental results of the first-order frequency and modal shape in each subspace are shown in Fig. 5, in which the fre­ quency unit is Hertz. The results offrequency and modal shape test­ ing also proved the validity of this method. The modal testing especially proves the theoretical prediction of modal characteristics quite well. CONCLUSIONS I. The eigenvalue problem of a structure with Cnv symmetry can be divided into much smaller subproblems by using group theory. 2. Calculation time as well as computer mem­ ory is reduced by using the method pre­ sented here. 3. The modal characteristics of Cnv structures derived from theoretical analysis are veri­ fied by modal testing. This work was part of a project supported by the National Fund of Natural Science. Journal of Engineering The Scientiifc Hindawi Publishing Corporation ht p:/ www.hindawi.com Journal of Sensors Machinery Volume 2014 International Journal of Hindawi Publishing Corporation ht p:/ www.hindawi.com and Passive Advances in Civil Engineering Hindawi Publishing Corporation ht p:/ www.hindawi.com Journal of Robotics Hindawi Publishing Corporation ht p:/ www.hindawi.com Advances ctronics Submit your manuscr ipts VLSI Design Hindawi Publishing Corporation ht p:/ www.hindawi.com Hindawi Publishing Corporation Hindawi Publishing Corporation Navigation and Observation Hindawi Publishing Corporation ht p:/ www.hindawi.com Modelling ulation & Engineering International Journal of Distributed Control Science Engineering Electrical and Computer Aerospace Engineering Burns , G. , 1977 , Introduction to Group Theory with Applications , Academic Press, New York. Zhong , W. , and Cheng, G., 1978a , "Application of Group Theory to Structure Analysis," Journal of Dalian Institute of Technology , Vol. 17 , No. 1 , pp. 21 - 40 . Zhong , W. , and Qiu , C. , 1978b , "Structure Analysis of Symmetric Shell Under Group Cnv ," Journal of Dalian Institute of Technology , Vol. 17 , No. 3 , pp. 1 - 22 .


This is a preview of a remote PDF: http://downloads.hindawi.com/journals/sv/1996/123520.pdf

Baojian Li, Xiaozhong Zheng, Jie Zhao. Vibration Analysis of Structures with Rotation and Reflection Symmetry, Shock and Vibration, DOI: 10.3233/SAV-1996-3409