MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 68 II. TWO DEGREE OF FREEDOM SYSTEMS Chapter one developed the fundamentals of dynamics and control of single degree of freedom mechatronic systems. Since we were dealing with single degrees of freedom, the relationships that were developed were temporal in nature. In this chapter, we develop the fundamentals of dynamics and control of two degree of freedom systems which are spatial in nature. A. Modal Analysis and Modal Design It’s convenient to transform the coordinates of a two degree of freedom system into two modal degrees of freedom. Each modal degree of freedom acts like a single degree of freedom system. This allows us to extend the results obtained in the previous chapter to analyze, design, and control two degree of freedom systems. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 69 A.1 Derivation of equations of motion Consider the undamped two degree of freedom system shown. Applying Newton’s second law of motion to each mass, yields the two equations that govern the dynamics of the system (2.1) m1x1 k2 ( x2 x1 ) k1x1 f1 m2 x2 k2 ( x2 x1 ) k3 x2 f 2 Equation (2.1) can be rewritten in the matrix-vector form as (2.2) m1 0 x1 k1 k2 0 m x k 2 2 2 k2 x1 f1 k2 k3 x2 f 2 or more compactly as (2.3) Mx Kx F where x is called the position vector, F is called the force vector, M is called the mass matrix, and K is called the stiffness matrix. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 70 A.2 The eigenvalue problem Consider the freely vibrating two degree of freedom system, that is let F = 0 in Eq. (2.3). Try solutions in the form (2.4) x e st Substituting Eq. (2.4) into Eq. (2.3) and dividing the result by e st , we get (2.5) [s 2M K ]φ 0 Equation (2.5) admits a nontrivial solution only if the matrix above is singular, in which case (2.6) det[s 2 M K] 0 Substituting the mass and stiffness matrices in Eq. (2.2) into Eq. (2.6), yields Eq (2.7) s 2 m1 k1 k 2 k2 0 det 2 k2 s m2 k 2 k 3 ( s 2 m1 k1 k 2 )( s 2 m2 k 2 k 3 ) k 22 m1 m2 s 4 [m1 (k 2 k 3 ) m2 (k1 k 2 )]s 2 (k1 k 2 k 2 k 3 k 3 k1 ) MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 71 Equation (2.7) is a quadratic equation in terms of s2, admitting the two solutions. Each solution is substituted back into Eq. (2.5) in order to obtain the associated vectors φ1 and φ 2 . Equation (2.5) is called the eigenvalue problem associated with the equations of motion, Eq. (2.3). The complex numbers s are called eigenvalues and the associated vectors, φ1 and φ2 , are called eigenvectors. The eigenvectors are also called natural modes of vibration. Since the eigenvalue problem is a homogeneous equation, the eigenvectors are unique up to a multiplicative constant. Therefore, without loss of generality, we shall normalize the eigenvectors by letting the first entry in each eigenvector equal one, that is (2.8) 1 1 φ 1 , φ 2 1 2 The eigenvectors are found by substituting Eq. (2.8) back into Eq. (2.5). MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 72 Properties of eigenvalues and eigenvectors In Eq. (2.2) the mass matrix and the stiffness matrix were both symmetric matrices. In all two degree of freedom systems composed of mass and stiffness, the equations can be derived so that the mass and stiffness matrices are symmetric. Furthermore, the mass and stiffness matrices have another property. The property is analogous to the sign property of numbers. The mass matrix is positive and the stiffness matrix is nonnegative. Of course, M and K are matrices, so we’ll need to define what we precisely mean by positive and nonnegative within the context of matrices. A symmetric matrix A is defined to be positive if the number yT Ay 0 for all nonzero y. Similarly, a symmetric matrix A is defined to be nonnegative if the number yT Ay 0 for all nonzero y. The mass and stiffness matrices satisfy M0 (2.9) M MT K0 K KT MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 73 The eigenvalues and eigenvectors of systems that satisfy Eqs. (2.9) are said to be normal mode systems. The first normal mode property states that the eigenvalues are pure imaginary. Pre-multiplying Eq. (2.5) by φT , yields (2.10) 0 φT [s 2M K ]φ s 2 (φTMφ) (φTKφ) From Eq. (2.9) (2.11) φTMφ s T 0, φ Kφ 2 so s i0 is pure imaginary, in which 0 is called a natural frequency of oscillation. The eigenvalue problem can now be written as 2 (2.12) 0 Mφ Kφ The second normal mode property states that the eigenvectors are real up to a multiplicative constant. To show this, write the eigenvector in terms of its real and imaginary components as (2.13) φ φ R φ Ii Substitute Eq. (2.13) into Eq. (2.12) and separate the real and imaginary parts to get MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg (2.14) - - 74 02 Mφ R Kφ R , 02 Mφ I Kφ I Notice in Eq. (2.14) that the real part of the eigenvector and the imaginary part of the eigenvector satisfy the same eigenvalue problem. Therefore, the eigenvectors must be real – up to a multiplicative constant. The next normal mode property is the orthogonality property. To show this property, pre-multiply the eigenvalue problem, Eq. (2.12), by the transpose of the eigenvectors, as shown below 2 φ T2 : 01 Mφ 1 Kφ 1 (2.15) 2 φ 1T : 02 Mφ 2 Kφ 2 Next, subtract the two equations from each other, recognizing the properties given in Eq. (2.9), to get (2.16) 2 2 ( 01 02 )φ T2 Mφ 1 0 Equation (2.16) reveals that eigenvectors having distinct eigenvalues are orthogonal with respect to the mass matrix. Substituting Eq. (2.15) into Eq. (2.16) further reveals that eigenvectors having distinct eigenvalues are orthogonal with respect to the stiffness matrix. Indeed, (2.17) φ 1T Mφ 2 0, φ 1T Kφ 2 0 MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 75 Next, let’s normalize the eigenvectors, not in the manner given in Eq. (2.8), but rather according to the following (2.18) φ 1T Mφ 1 1, φ T2 Mφ 2 1 Substituting Eq. (2.15) into Eq. (2.18) further reveals that (2.19) 2 2 φ 1T Kφ 1 01 , φ 1T Kφ 2 02 Finally, the orthogonality properties and the normalization given in Eqs. (2.16) through Eq. (2.19) are written compactly as (2.20) φ Tr Mφ s rs , φ Tr Kφ s 02r rs where rs is the Kronecker-delta symbol in which rs = 0 when r and s are different and rs = 1 when r and s are equal. In summary, the eigenvalues and eigenvectors of two degree freedom systems composed of mass and stiffness are normal in the sense that the eigenvalues are pure imaginary, the eigenvectors are real, and they satisfy orthogonality properties with respect to the mass and stiffness matrices. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 76 A.3 Modal equations of motion Since the eigenvectors form a basis set, any vector x can be expressed as a linear combination of the eigenvectors, whether x is associated with the freely vibrating system or not. We begin by representing x as (2.21) x(t ) φ1q1 (t ) φ2q2 (t ) where q1 (t ) and q2 (t ) are called modal displacements. From Eq. (2.21), the r-th modal displacement can be regarded as the level of participation of the r-th mode of vibration in the overall response. Pre-multiplying Eq. (2.21) by 1T M and 2T M , we get (2.22) q1 (t ) φ 1T Mx(t ), q2 (t ) φ T2 Mx(t ) Equations (2.21) and (2.22) are transformations between the physical displacements x and the modal displacements q1 (t ) and q2 (t ) . Let’s now replace the physical coordinates in Eq. (2.4) with modal coordinates. Substituting Eq. (2.21) into Eq. (2.4) yields Eq (2.23): M[φ 1q1 (t ) φ 2 q2 (t )] K[φ 1q1 (t ) φ 2 q2 (t )] F MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 77 Pre-multiplying Eq. (2.23) by 1T and 2T , and invoking the orthogonality properties, Eq. (2.20), we get the equations that govern the modal motions of the two degree of freedom system 2 q1 (t ) 01 q1 (t ) Q1 (t ) (2.23) 2 q2 (t ) 02 q2 (t ) Q2 (t ) where (2.24) Q1 (t ) φ 1T F, Q2 (t ) φ T2 F are called modal forces. What are Q1 (t ) and Q2 (t ) , and how are they related to the physical forces? The physical forces are expressed in terms of the modal forces as follows (2.25) F Mφ 1Q1 (t ) Mφ 2 Q2 (t ) Equation (2.26) can be verified by substituting Eq. (2.25) into Eq. (2.26) to render an identity. Thus, just as the modal displacements qr represents the degree to which a mode of vibration φr participates in the overall response, so too does a modal force Qr represent the degree to which a mode of force Mφ r participates in the overall force. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 78 In summary, the modal transformations of the displacements and of the force are Eq (2.27) x(t ) φ 1q1 (t ) φ 2 q2 (t ), q1 (t ) φ 1T Mx(t ), q2 (t ) φ T2 Mx(t ) F Mφ 1Q1 (t ) Mφ 2 Q2 (t ), Q1 (t ) φ 1T F, Q2 (t ) φ T2 F The modal displacements act as single degree of freedom system coordinates governed by independent modal equations of motion, Eq. (2.24). Modal displacements are excited by modal forces. The spatial relationship between the modal components of force and motion are as follows. x(t ) φ 1 q1 (t ) F Mφ 1Q1 (t ) (2.28) x(t ) φ 2 q 2 (t ) F Mφ 2 Q2 (t ) MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 79 A.4 Dynamics of a double pendulum As an illustration, consider the double pendulum shown. Summing moments about each of the pivot points yields the two equations governing the motion of the double pendulum (2.29) m ga I 11 ka 2 ( 2 1 ) 1 1 M 1 2 m ga I 22 ka 2 ( 2 1 ) 2 2 M 2 2 Equation (2.29) can be written in the compact matrix1 F1 vector form of Eq. (2.4) letting x , F , 2 F2 (2.30) 0 m1 a 2 0 , 2 I 2 0 m2 a 2 m1 ga 2 ka ka 2 K m2 ga 2 2 ka ka 2 I M 1 0 MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 80 Substituting Eq. (2.30) into Eq. (2.6) yields Eq. (2.31) m1 ga 2 2 2 I ka ka 0 1 2 2 0 det[ 0 M K ] det m2 ga 2 2 2 ka 0 I 2 ka 2 2 3k 3g 3g 3k 3k 3g 4 2 3k 0 0 m m a 2 a m m 2 a 2 2 1 1 Equation (2.31) is a quadratic equation in 02 . Solving Eq. (2.31), 1/ 2 2 3k 3k 3 g a 1 3k 3k 3 g m1 m2 2 01,02 2 m1 m2 a 3 g 3k 3k 3g 2 4 2a m m 2a 2 1 3g 1 3k 3k 3 g 3k 3k 2a 2 m1 m2 a m1 m2 3 g 3k 3k 2a m1 m2 MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 81 Assuming eigenvectors in the form of Eq. (2.5) and substituting Eqs. (2.32) back into Eq. (2.5), we obtain the eigenvectors of the double pendulum (2.30) 3k m 1 3k m2 3k m 2 3k m2 3k 1 m1 1 0 φ1 3k 1 0 1 m2 3k m m1 1 0 φ 2 2 3k 2 0 m1 m1 Finally, let’s normalize the modes according to Eq. (2.20). Doing this yields the normalized modes of vibration, Eq. (2.31) 3 φ 1 2 (m1 m2 )a 1/ 2 1 3 ,φ 2 2 1 m1 m2 (m1 m2 )a MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton 1/ 2 m2 m 1 MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 82 The normalized modes of force are then Eq. (2.35) a2 Mφ 1 3(m1 m2 ) 1/ 2 a 2 m1 m2 m1 , M φ 2 m 2 3(m1 m2 ) 1/ 2 1 1 Equation (2.32) provides us with temporal information concerning the double pendulum and Eq. (2.34) and (2.35) provide us with spatial information concerning the double pendulum. We see that the first mode is a “pendulum” mode wherein the system moves as a rigid body pendulum. The second mode is a “pseudoelastic” mode associated with the spring and gravity. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 83 A.5 Actuator dynamics As mentioned earlier, the dynamics of an actuator can interact with a system, producing a coupled system comprised of the mechanical part and the actuator. Of course, such interactions are undesirable in the sense that they complicate the design process – violating the goal to separate the components of a system into independent parts. The method of mounting an actuator to a system effects the coupling of the actuator to the system. Consider the two degree of freedom system given in Eq. (2.2) and Eq. (2.3). Assume that the actuator harmonically excites the mechanical part through the applied force, in which case (2.36) 0 0 X F it eit F0eit , x p 1 eit Xeit X 2 F0e F0 in which denotes the excitation frequency. Substituting Eq. (2.36) into Eq. (2.3) and dividing by e it yields (2.37) [ 2M K ] X F0 MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 84 Let’s now recall the explicit form of the mass and stiffness matrices, (2.38) 0 m k1 k2 M 1 , K k 0 m2 2 k2 k2 and let’s express the physical parameters in these matrices in terms of the quantities m2 k k , 1 1 , 2 2 m1 m1 m2 where r is the mass ratio, 1 is the frequency of the mechanical part, 2 is the frequency of the mount, and is the frequency of the excitation. Substituting Eq. (2.39) into Eq. (2.37), dividing the first equations by m1 , and dividing the second equations by m 2 , yields the mass and stiffness matrices (2.39) r (2.40) 12 r 22 1 0 M , K 2 0 1 r 2 r 22 r 22 and from Eq. (2.38), we get (2.41) 2 12 r 22 22 r 22 1 X F0 2 2 m 2 2 MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 85 Solving Eq. (2.41) yields the response of the two degree of freedom system (2.42) r 22 F0 X , m2 2 12 r 22 4 2 [12 22 r 22 ] 12 22 Also notice that the mount force between the masses is F0 22 (2.43) FM k2 ( X 2 X1 ) ( 2 12 ) The actuator and the system are uncoupled only if the excitation frequency has no affect on X1. From Eq. (2.42), it follows that the actuator and the sensor can be uncoupled only if the excitation frequency is significantly lower than the frequency of the mechanical part; we impose the necessary condition (2.44) 1 The actuator dynamics and the dynamics of the mechanical part are uncoupled in the following two situations. In the first situation, the actuator prescribes a force and in the second situation the actuator prescribes a displacement. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton MAE 524 course notes – Spring 2002, Copyrighted by L.Silverberg - - 86 Prescribed force (hard mount) Let the actuator be significantly less massive than the mechanical part and let it be “hard mounted,” that is let r << 1 and 2 >> From Eq. (2.43) (2.45) FM = F0. The mount force is equal to the excitation force and independent of the excitation frequency. Prescribed displacement (soft mount) Let the actuator be significantly less massive than the mechanical part and Let it be “soft mounted,” that is let r << 1 and 2 << From Eq. (2.42) F0 . k2 The displacement of the actuator is proportional to the excitation force and independent of the excitation frequency. (2.46) X2 In summary, an actuator that acts on a single degree of freedom system, if it produces a force according to Eq. (2.36), can be uncoupled from the system if the excitation frequency is much lower than the frequency of the mechanical part. Under these conditions a hard mounted actuator prescribes a force and a soft mounted actuator prescribes a displacement. These decoupling conditions are used in design to reduce complexity. MODAL ANALYSIS AND MODAL DESIGN Lecture notes prepared by L.Silverberg and J.Morton