ELEMENTS OF VIBRATION ANALYSIS Second Edition Leonard Meirovitch College of Engineering Virgina Polytectnic Institute and State University COTO KUTUPHANES! M.E.T.U. LIBRARY il Boston, Massachusetts Burr Ridge, linois Dubuque, Iowa Madison, Wisconsin New York, New York San Francisco, California St. Louis, MissouriQas35 To My Wife and to the MSs3 Memory of My Parents z (9&4 as “This book was set in Times Roman by Ea Services Lid. ‘The editor was Anne Murphy the cover was designed ty Kar H. Steinbrenner, the production supervisor was Charles Hess ‘rofet supervision was done by Albeet Harrison, Haley Edo Services. McGraw-Hill 328681 A Dino of The McGrane EI Compares ELEMENTS OF VIBRATION ANALYSIS copyright © 1986, 1975 by McCrsw-Hil, Inc. All ights reserved. Typeset in che United SEE ceric Eacept as penited under the Untod Stats Copyright Act of 1976, 20 Set or th publication may be reproduced or dsebuted in any frm or by any means, oc ee, Pda bare or reuoval sem, witout Ge peor writen permission of the ier. joiri2 1914 15 16 17 18 19 BKMBKM 909876545210 ISBN O-07-O41342-8 . Library of Congress Cataloging in Pubication Data Meirovitch, Leonard. ‘lament of vibration analysis Bibliography: Ineldes index Te Vibration 1 Tie. Qko3s.Ms3 1986 531.32 SBN 0-07-081382-8PREFACE Inthe last several decades, impressive progress has been made in vibration analysis, prompted by advances in technology. On the one hand, the requirement for the tnalysis of neresingly complex systems has been instrumental inthe development of powerful computational techniques. On the other hand, the development of fast. 0, b> 0. In this case the roots are either real and negative or complex conjugates with negative real part, so that x(¢) approaches zero as f — co. Hence, y(@) approaches y, so that the equilibrium position is asymptotically stable. 2a=0, b>0. The roots are pure imaginary, so that the solution x(¢) is Oscillatory. Hence, the motion is bounded and the equilibrium position is stable, 3. a<0, or a2 0, b <0. The roots are either complex conjugates with positive real part or they are both real, with one root being positive and the other being negative. In either case the solution is divergent and the equilibrium position is. unstable. ‘The subject of system stability is discussed in greater detail and in a more rigorous ‘manner in Chap. 9. In the special case in which y_ = 0 the equilibrium position is sai This is a case encountered very frequently in practice. to be rivial. Example 14 The system shown in Fig. 1.12a represents a simple pendulum. It consists of a bob of mass m attached to one end of an inextensible string of16 ELEMENTS OF VIBRATION ANALYSIS Cr) Figure 112 length L, where the other end of the string is fixed at point 0. Derive the ‘equation for the angular motion 6() of the pendulum, identify the equilibrium positions and determine the nature of motion in the neighborhood of the ‘equilibrium positions. To derive the equation of motion, we consider the free-body diagram shown in Fig. 1.126. The forces acting on the bob are the string tension T and the gravity force mg. Note that the string is not capable of carrying transverse forces. Using Newton's second law and summing up forces in the transverse direction, we obtain LF = —mg sin 6= ma, = mod oe where a, ~ LUis the acceleration in the transverse direction. Division of Eq (a) through by mL yields © where £0) © Equation (b) has the same form as Eq, (1.16), except that f does not depend on the angular velocity 6. “To identify the equilibrium positions, we use Eq, (I-17) and write @ g S(0.) = —F-sin 6 which has the solutions 0,20, $n, 42m... © [Although Eq, (¢) indicates that mathematically there is an infinite number of, equilibrium positions, physically there are only two positions Ox=0 O2=% n Of course, the first one is recognized as the trivial solution.[REE RESPONSE OF SINOLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 17 ‘Next, let us use the notation x = @ — @, and write the linearized equation of motion + bx=0 @ where, from the first of Eqs. (1.19), wy In the case of the equilibrium point 8, = 0, we obtain a baf>o 0 so that the equilibrium is stable. We will discuss Eq. (g) for this case extensively later in this chapter. In the case of the equilibrium point 8; = x, we have o so thatthe equilibrium is unstable ‘The above results conform to expectation. Any small deviation from the equilibrium position in which the pendulum hangs down results in oscillation about the equilibrium. On the other hand, any small deviation from the ‘upright equilibrium position tends to increase without bounds. The case in which the pendulum oscillates about the equilibrium position 0; = Oi by far the most important one, which explains why the equilibrium postion 8,2 = = is discussed so seldom, 15 FORCE-FREE RESPONSE OF FIRST-ORDER SYSTEMS Let us consider the spring-damper system of Sec. 13 and assume thatthe external excitation is zero. Setting F(?) = 0 in Eq, (1.12), we obtain the homogeneous equation ex() + kx() =0 (125) Using the approach of Sec. 14, we let the solution of Eq. (125) have the exponential form a(t) = Ae (126) Inserting Eq. (1.26) into Eq, (1.25) and dividing through by Ae, we obtain the characteristic equation stk=0 a2n18 ELEMENTS OF VIBRATION ANALYSIS 0 a Figure 113 Which has the single root : (1.28) so that the general solution of Eq, (1.25) is x() = Ae (1.29) where 4 is a constant of integration and (130) is known as the time constant. ‘The constant of integration A depends on the initial excitation. Letting -x(0) = xo be the initial displacement, we can easily verify from Eq. (1.29) that ‘A = xo, 60 that the solution to the force-free problem is fee >. 0 <0 x (ast) Equation (1.31) indicates that the response decays exponentially with ‘The homogeneous solution (1.31) is plotted in Fig. 1.13 as a function of time. We observe that, after being displaced initially by an amount x, the system returns to the zero equilibrium position without any oscillation. The time constant © provides a measure of the speed of return of the system to equilibrium. Indeed, the fate of return is greater for small time constants and vice versa. Hence, for a stiff spring, or for light damping, the rate of return is fast, and vice versa. 16 HARMONIC OSCILLATOR Let us consider now the second-order system described by Eq. (I-14). Before we discuss the general response, we wish to investigate the force-free case, namely, the ‘case in which the force F(e) is identically zero. Moreover, quite often damping is extremely small, $0 that for all practical purposes it can be ignored. Hence, weFREE RESPONSE OF SINGLE-DEGRES-OF-FREEDOM LINEAR SYSTEMS 19 ‘concern ourselves with the undamped case for which ¢ = 0. Upon dividing Eq (1.14) by m, the differential equation of motion reduces to 5) +03) =0 wf == 32) [As shown in Sec. 14, the solution of Eq. (1.32) has the exponential form x) = Ae" (133) Introducing Eq, (1.33) into Eq, (1.32) and dividing through by Ae", we obtain the characteristic equation S+or=0 (134) ‘which has the solutions tio, (133) where 1= /=1. Inserting 5, and s, into Eq, (1.33), the general solution of Ea, (1.32) can be written as Xt) = Aye + Ane (136) where Ay and Az are constants of integration. Their values depend on the initial displacement x(0) and initial velocity (0). Because the roots s, and s2are'pure imaginary, we conclude from Sec. 1.4 that the solution, Eq, (1.36), must represent stable motion. This stable motion consists of pure oscillation and the quantity a, is known as the natural frequency of oscillation of the undamped system, The reason for the term natural frequency is that a force-free undamped second-order system, when set in motion by some initial conditions, wll always oscillate with the same frequency Solution (1.36) isin terms of complex quantities. Yet, on physical grounds, it can be argued thatthe solution must be real. Hence, the interest lies in reducing the solution to real form. To this end, consider the series Jon 1 cage + eat)? + Cent + 2 Gag? + 2 Cage)? emt mL ti 5 (loa)? +p (lat) + aplenty + pleat)? + et Pe = Float? + Float ot ‘fon ~ Flea? + Fou] = cos cnt + isin og (1372) In a similar manner, it is easy to verify that eon = 608 at — isin gt (1376) Inserting Eqs. (1.37) into Eq, (1.36), introducing the notation Apt r= Acs (Ay ~ Az) = Asing (138)120 ELEMENTS OF VIBRATION ANALYSIS and recalling the trigonometric relation cos «cos f + sin asin, solution becomes £08 (a ~ f), the X(0) = A cos (at — 6) 39) where now the constants of integration are A and @. ‘The constants A and ¢ are referred to as the amplitude and phase angle, respectively. Because A and depend on A, and Az, they can also be regarded as constants of integration depending on the initial conditions x(0) and %(0). Equation (1.39) indicates that the system executes simple harmonic oscillation with the natural frequency «,, for which reason the system itself is called a harmonic ‘oscillator. The motion described by Eq, (1.39) is the simplest type of vibration. The harmonic oscillator represents more of a mathematical concept than a physical reality, Nevertheless, the concept is valid for negligible damping, if the interest lies in the response for atime duration too short for extremely light damping to make its effect fet. ‘The discussion of the nature of harmonic oscillation is perhaps enhanced by the vector diagram shown in Fig. 14a. IFA represents a vector of magnitude A and the vector makes an angle «%t — @ with respect to the vertical axis x, then the projection of the veetor A on x represents the solution x(t) = 4 cos (wt — @). The ‘angle et — ¢ increases linearly with time, with the implication that the vector A. rotates counterclockwise with angular velocity a. As the vector’ rotates, the projection varies harmonically, so that the motion repeats itself every time the vector A sweeps a 2x angle. The projection x(t) is plotted in Fig, .14b as a function of time. 7 “The time necessary to complete one cycle of motion defines the period T given by (140) Figure 114FREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 21 where ©, is measured in radians per second (rad/s) if T is measured in seconds. Physically, T represents the time necessary for one complete oscillation to take place; it is equal to the difference between two consecutive times at which the Oscillator reaches the same state, where the state is to be interpreted as consisting of both position and velocity. As an illustration, the period is measured in Fig. 1.14 between two consecutive peaks, It is also customary to measure the natural frequency in cycles per second (cps). In such a case the natural frequency is denoted by fa, and because one cycle is equal to 2x radians we have ial RonF (aly so that the natural frequency f, and the petiod T are the reciprocals of each other. One eycle per second is a unit generally known as one hertz (H2). Finally, it wll prove interesting to evaluate the constants of integration A and 6 in terms of the initial conditions. Introducing the notation x(0) = xo, X(0) = vo, ‘where xo is the initial displacement and vp the initial velocity, and using Eq. (1.39), it is easy to verify that the response of the harmonic oscillator to the initial conditions is te X(0) = X9 008 wt + asin gt (1.42) Moreover, we conclude that the amplitude A and the phase angle $, when expressed in terms of the initial displacement and velocity, have the values an far(2) omar «aay A lege variety of dynamical systems behave lke harmonic oclators, quite often when restricted to small motions. Asan ilusraion, the simple pendulum of Example 14, when restricted (o small angular motions about the val equi rium B = 0, can be desribed bythe difrental equation 20 = aged G4 a20=0 afmt (1.44) Which represents a harmonic oscillator with the natural frequency o, = /@/L Note that Eq (144) is valid a long as sin O =, whichis approximately tre for surprisingly large values of @. For example, 930° = 03236 ed and sin d= Sin 30° 0.000 are close in vale, Infact, there is es than 5 percent eror in ting @ instead of sin @ for 830° Example 1.5 The semicircular thin shell of radius R shown in Fig. 1.15a is. allowed to rock on a rough horizontal surface. Derive the differential equation ‘of motion for the case of no slip, show that for small motions the shell behaves like harmonic oscillator, and calculate the natural frequency of the oscillator.22 ELEMENTS OF VIBRATION ANALYSS Figure 118 This example provides us with the opportunity to derive the differential cquation of motion for a relatively more complicated system than a spring- 20,51 ~+ Oand s; + ~co. In the sequel, we relate the position ofthe roots sy and 52 in the s plane to the system behavior. Inserting the roots given by (1.49) into (47), we can write the general solution x(t) = Aye" + Ane = Avex [(—0+ JF = Doo] + Az exp (6 = P= Doge] =[Avexp (VE =T oy) + Asexp(—YE=T ole (1.50) Solution (1.50) is in a form suitable forthe cases in which £ > 1. For €> 1 the motion is aperiodic and decaying exponentially with time. The exact shape of the curve depends on Ay and 2, which, in turn, can be evaluated in terms ofthe initial displacement xp and initial velocity vg. The case { > | is known as the overdamped case. Typical response curves for ¥(0) = 0 and #(0) =v are given in Fig. 1.17 In the special case in which ¢ = 1, Eq, (148) has a double root, s, In this case the solution can be shown to have the form (see Prob. 1.29) X() = (Ay + tdae“™ ast) ‘whieh again represents an exponentially decaying response. The constants Ay and ‘Az depend on the intial conditions. The case { = 1 is known as critical damping, and response curves for certain intial conditions are show in Fig 1.18. From the expression { = ¢/2my, we see that for { = | the coefficient of viscous damping has the value cy = 2ma, = 2/km. The importance of the concept should not be overstressed, because critical damping merely represents the bordertine between the cases in which £ > Land ¢ < 1. Ttmay be interesting to note, however, that fora given inital excitation a critically damped system tends to approach the ‘equilibrium position the fastest (see Fig. 1.17). ‘When 0-< { < J, solution (1.50) is more conveniently written in the form (0) = [Ay exp T= gt) + Aa exp (1 T= F ont) (Aye + Agee So as)216 ELENENTS OF VIBRATION ANALYSIS sn | Sta ogee yi o 2 nso Pia 1 ote | co | op : ° 5 7 > o xe.ia Pfr eesti 7 Sola 2 Figure 1.17FREE RESPONSE OF SINGLE-DEGRES-OF-FREEDOM LINEAR SYSTEMS 27 sovin wo : =sbie oape a | seston a he a Q 1 2 5 sin : T oa bat i syrinne (iy =3 rads via o 1 7 y » Figure 118 where 04" (1 2)!0 (1.53) is often called the frequency of the damped free vibration. From Eqs. (1.37), we can write e*H = cos ct + isin @,t, Moreover, using the notation of Eqs. (1.38), Eq. (1.52) reduces to x(t) = Ae“ cos (wat — 8) (asa) ‘which can be interpreted as an oscillatory motion with the constant frequency and phase angle ¢ but with the exponentially decaying amplitude Ae", wheresiP.in esFRE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 29 the constants A and depend on the initial conditions, The case 0<{ <1 is roferred (0 as the underdamped case, and typical response curves are shown in Fig 1.19. The curves + Ae” provide the envelope bounding the oscillatory response. Clearly, as ¢+ ©, x(¢) 0, so that the response eventually dies out, which represents the observed behavior of real systems, Example 1.7 Consider the system of Fig, 1.9 and caleulate its response to the initial conditions x(0) = 0, (0) = vo, for {> 1, [= 1, and 0<¢ <1. For { > 1 we make use of the formula (1.50), and write x)= Ay + AZ=0 A= Ay @ so that now the solution has the form x(0) = 2AyeS"' sin SF =T ont o Differentiating Eq. (b) with respect to time, we obtain 20 = 24 (JF =T 0, cosh JF =T at ~ Cog sinh JET enters" © Letting %(0) = v9, Eq. (€) yields 2A, = @ yea Ic follows that for {> 1 the general solution is 24) e-sink JT aot © VF-10, For = 1, itis easy to show from Ea. (1.51) that A, = Oand Az = v9, 50 that the response is X(t) = vote" Ww For 0<{<1, the intial displacement being equal to zero leads to = m/2 in Eq, (1.54). Moreover, the amplitude is related to the initial velocity by A= vp/ay, So that Eq. (1.54) reduces to x)= et sin cat Oe = Or @ Expressions (e), (f), and (g), corresponding to overdamping, critical damping, and underdamping, are plotted in Figs. 1.17, 1.18, and 1.19, respectively for the indicated values ofthe system parameters { and «and the initial velocity vo.30. ELEMENTS OF VIBRATION ANALYSIS 18 LOGARITHMIC DECREMENT in a given system is not known and must be determined experimentally. We are concerned with the case in which damping is viscous and the system underdamped. As shown in Sec. 1.7, viscous damping. ‘causes the vibration to decay exponentially, where the exponent is linear function ‘of the damping factor [. In this section, we wish to explore ways of determining { from the observation of this decay. ‘A convenient measure of the amount of damping in a single-degree-of-freedom system is provided by the extent to which the amplitude has fallen during one complete cycle of vibration. Let us denote by t, and ¢2 the times corresponding to two consecutive displacements x, and x measured one cycle apart (see Fig. 120), so that, using Eq, (1.54), we can form the ratio le" cos (oats — 6) ka 7 Ae 605 (Catz —@) Because t2 = t; + T, where T'= 2n/a, is the period of the damped oscillation, it follows that cos (wats ~ $) = cos [(csty ~ #) + agT] = c0s [(eety ~ 4) + 2x] = 1005 (wat, — $), $0 that Eq. (1.55) reduces to (158) (1.56) In notation of the exponential form of Eq. (1.56), it is customary to introduce the asy ‘where 5 is known as the logarithmic decrement. Hence, to determine the amount of damping in the system, it suffices to measure any two consecutive displacements x igure 1.20FREE RESPONSE OF SINGLE-DPOREE-OF-FREEDOM LINEAR SYSTEMS 31 ‘and xz one cycle apart, calculate the logarithmic decrement 6 by taking the natural logarithm of the ratio x,/x2, and obtain £ from 6 (1.58) Vanes For small damping, 5 is a small quantity, so that Eq, (1.58) can be approximated by a an (ss) ‘The damping factor { can also be determined by measuring two displacements separated by any number of complete cycles. Letting x, and x;,-, be the amplitudes ‘corresponding to the times fy and t;., = ty +)T, where jis an integer, we conclude that M288 (denty wheat (160) Kiet 2¥s Xe Fe because the ratio between any two consecutive displacements one eycle apart is ‘equal to 7, Equation (1.60), in conjunction with Eq, (157), yields beatin Zh (st) Ta which can be introduced into Eq, (1.58), or Eq. (1.59), to obtain the viscous ‘damping factor © Example 18 It was observed that the vibration amplitude of a damped singledegroe-ofreedom system had fallen by 80 percent after five complete cycies, Assume that the system i viscously damped and ealeulate the damping factor ¢. Letting j= 5, Eq. (1.61) yields the logarithmic decrement ioe Si 10Sr a3 a Ite above values inserted into Eq, (1.58), we must conclude that damping is relatively light. Hence, using Eq. (1.59), we obtain 5 _ 011386 yo soon © Leway-one 19 COULOMB DAMPING. DRY FRICTION Coulomb damping arises when bodies slide on dry surfaces. For motion to begin, there must be a force acting upon the body that overcomes the resistance to motion ‘caused by friction. The dry friction force is parallel to the surface and proportional32 ELEMENTS OF VIBRATION ANALYSIS to the force normal to the surface, where the latter is equal to the weight WV in the case of the mass-spring system shown in Fig. 1.21. The constant of proportionality is the static friction coefficient 1,, a number varying between 0 and 1 depending on the surface materials. Once motion is initiated, the force drops to y4I¥, where is the kinetic friction coefficient, whose value is generally smaller than that of. The friction force is opposite in direction to the velocity, and remains constant in ‘magnitude as long as the forces acting on the mass m, namely, the inertia force and the restoring force due to the spring, are sufficient to overcome the dry friction. ‘When these forces become insufficient, the motion simply stops. Denoting by Fy the magnitude of the damping force, where Fy = 14H, the equation of motion can be written in the form mi + Fysgn (8) + koe = 0 (1.62) ‘where the symbol “sgn” denotes sign of and represents a function having the value ‘+1 if its argument ¥ is positive and the value —1 if its argument is negative. Mathematically, the function can be written as sen) == 1.63) sen (2) a (1.63) Equation (1.62) is nonlinear, but it an be separated into two linear equations, one for positive and another one for negative , as follows: metke= Fy foré>0 +7 (164) mitke=Fy foré<0 (1.648) Although Eqs. (1.64) are nonhomogeneous, so that they can be regarded as representing forced vibration, the damping forees are passive in nature, so that discussion of these equations in this chapter is in order. ‘The solution of Eqs. (1.64) can be obtained for one time interval at a time, ‘depending on the sign of 3. Without loss of generality, we assume that the motion starts from rest with the mass m in the displaced position x(0) = x0, where the initial displacement xo is sufficiently large that the restoring force in the spring ‘exceeds the static friction force. Because in the ensuing motion the velocity is to Figure 121REE RESPONSE OF SINOLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 33, negative, we must solve Eq, (1.646) frst, where the equation can be written in the form Etolx=o}f, of== (165) in which f= F,/k represents an equivalent displacement. Equation (1.65) is ‘subject to the initial conditions x(0) = xo, 3(0) = 0, so that its solution is simply 20) = (9 — fi) 008 048 +f (1.66) which represents harmonic oscillation superposed on the average response fi Equation (1,66) is valid for 0 < ¢ < ty, where fis the time at which the velocity reduces to zero and the motion is about to reverse direction from left to right. Differentiating Eq, (1.66) with respect to time, we obtain i XO = ~ 040% —f) sin ot sn 0 that the lowest nontrivial value satisfying the condtion %(%,) = Os t; = my, at which time the displacement is x(¢;) = —(xo ~ 2). If x(¢,) is suficiently large in ‘magnitude to overcome the static friction, then the mass acquires a positive velocity, so that the motion must satisfy the equation d+ o3x = 0} fy (1.68) where x(t) is subject to the initial conditions x(,) = —(o ~ 2), #(t)) = 0. The solution of Eq, (1.68) is (0) = (0 — Ya) €08 ent — fo (1.69) Compared to (1.66), the harmonic component in solution (1.69) has an amplitude smaller by 2f, and a negative constant component, namely ~f,. Solution (1.69) is valid in the time interval ¢, < ¢ < fa, where fi the next value of time at which the velocity reduces to zero, This value i = 2x, at which time the velocity is ready to reverse direction once again, this time from right to le, The displacement at t= tris x42) = x0 ~ Ah ‘The above procedure can be repeated for > ta, every time switching back and forth between Eqs. (1.642) and (1.648). However, a pattern seems to emerge, rendering this task unnecessary. Over each half-cycle the motion consists of @ constant component and a harmonic component with frequency equal to the natural frequency «, ofthe simple mass-spring system, where the duration of every halfeyeleis equal to m/e. The average value ofthe solutions alternates between fa ‘and ~f,, and at the end of each hallcycle the displacement magnitude is reduced by f= 2Fu/k It follows that for Coulomb damping the decay is linear with time, 1s opposed to the exponential decay for viscous damping. The motion stops abruptly when the displacement at the end of a given hal-cycle is not sufficiently large forthe restoring force in the spring to overcome the static friction. This occurs at the end of the hal-cycle for which the amplitude of the harmonic component is34 ELEMENTS OF VIBRATION ANALYSIS Figure 1.22 smaller than 2f,, Letting n be the number of the half-cycle just prior to the cessation of motion, we conclude that m is the smallest integer satisfying the inequality xo — Qn— Dr <%e (1.70) ‘The plot x(@) versus # can be obtained by combining solutions (1.66), (1.69), etc. ‘Such a plot is shown in Fig. 1.22 : Example 1.9 Let the parameters of the system of Fig. 1.21 have the values 1, k-= 800 Ibjin, and jy = 0.1, and calculate the decay per cycle and the number of half-cycles until oscillation stops ifthe initial conditions are x(0) = Xo = 1.2in, X(0) = 0. ‘The decay per cycle is @ Moreover, n must be the smallest integer satisfying the inequality 12 ~Qn—1) x 020966 < 2 x 0.0966 6 from which we conclude that the oscillation stops after the half-cycle m = with m in the position x(t) = xo — 12fg = 1.2 ~ 12 x 0.0966 = 0.0408 in. PROBLEMS 1L1 Consider two dashpots with viscous damping coticens cy and cy and calulte the equivalent, riscous damping coetilent forthe casein which the dashpocs ar arranged in parallel and in sie, respectively[FREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 35 1.2 Consider the system of Fig. 123 and obtain an expesin or the equivalent spring. Then dive the Gierenal equation of motion. Ys Ke by io TAT ron 13 Consider the system of Prob, 12, let ky = y= SODTbjin (8.7563 x 10* Nn, = 150 fin {2526 + 10° Nj, and m = 15 1b (26269 kg) and calclate the system natural equeny. 4d) booy of ifr erossectional area and mast mis depreied a distance x fom the equim ition, a shown in Fig. 1.24 and then released. Derive the diferential equation of motion and obtain the nator frequency of ebilation. The mass density of the liquid in which the Buoy Boas isp igure 1.24 15 The system shown in Fig 1.28 consitng ofan unknown mass manda sping with unknown spring ‘constant f has ben observed to osilate naturally with the fequeney = 100 rad Determine the ‘mass m and spring constant k knowing that when a mass M09 kyle added the modifed naturel frequency i = 90 rad. ek {U6 Derive he dierenta equation of motion forthe system shown in Fig 126 and obtain the period of ‘sation. Denote the mss density of the gud by p and the ttl length of the column of iui by‘36. ELENENTS OF VIBRATION ANALYSIS Figure 126 1.7 The hinges ofthe rectangular door shown n Fig. 27 are mounted on ine making an sogle = with ‘rapes othe vert. Astute thatthe door as uiform mass distribution and determine the natural frequency of oxilation. Figure 1.27 14 To determine the centroid mast moment of nent cof tre mounted on a wheel the sytem i faspended ona kaifeadge, ss shown in Fig. 28, and the natural period ofosciltion Tis measured. ‘Derive a forts for Ii terms of the mass m, the period ofthe system, andthe radius r fom the canter C tothe knife-egeFREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 37 19 A bead of mass is suspended ona mass sting, as shown in Fig. 129. Assume that che string is subjected othe tension 7; and that this tension doesnot change throughout the motion, and deive the Gierenal equation Tor small motions y rom equilibrium, as well as the natural frequency of celation, eee 110 A conneting rod of mas m = 3 10"? kg and cetrodal mats moment of neta Je = O42 x {0°*kg-m? ie suspended on a knifeadge about the uper inner rurfae of the writin bearing, as shown in Fig 130 When disturbed slighty, the rd was observed to oscillate with the natural frequency rad/s Determine the distance h between the support and the center of mast C. 1.11 A mass attached tothe end ofa mast last lade of lngth Ld exural stifles £1 (28 Fig. 131, Devive the equivalent spring constant of the blade ad waite the equation of motion for the ‘wansverse displacement of m.Caleulate the period T. Y el —-——— 8" i 1L12 A disk ofmass moment of inertia is ttached othe end of a massless uniform shaft of length L and torsional rigty GJ Ge Fig. 132). Derive the equation forthe torsional vibration ofthe dik, and ‘obtain the natural frequency of vibration,[38 ELEMENTS OF VIBRATION ANALYSIS ® LN - 1.13 A massless igi bat i hinged at O, a8 shown in Fig. 138. Determine the natural frequency of (ualaton ofthe system for the parameters ky ~ 2500 [bin (43782 x 10° Njm), ke = $00 jin {USTOL 2 10" Nim), m= Tb sin (17513 kph, a ~ 605 (203m), and b = 100in (2.56). ey 2 Figure 1.33 1.14 circular igi isk ofmas polar moment ofinerts [= 08 kg-m*is mounted on a crear shalt sade of two segments of ferent diameters and lengths, as shown in Fg 1.4. The shat xed at both nds Let the shear modulus ofthe shaft material be C= 80 x 10? N/m? and obtain the natural frequency of angular oseiltion ofthe disk | | wo eeaae ee 1.18 The one-story building shown in Fig. 1.35 canbe modeled in the frst approximation asa singl- agree okecdom system by reaoling the columns as masless and the root as rigid. Desve the ‘Sferental equation of motion and determine the aturl equeacy. Assume thatthe mas M can oly teaolate horontlly, 20 that the columas undergo no rotation atthe fp.[FREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 39 a 2 Zs igure 135 116 Two gears and B of mass moments of inertia I, and a espectvely, ae attached to circular safe of gual tifess GU/L (Fig 136). Deve the dierent equation fr the system and determine {he natural frequency ofthe system forthe ease Ry/Ry = n- Hint: Draw one fe-body diagram for each ger recogniing that the reaction forces on the gear tthe point of contac are qualia magaitude and ‘opposite in ection, and thatthe angular mation of gear Bis mimes the angular motion of gear 4 Figure 1:36 (GD A mass min suspended ona mass beam of bending stiles I through spring of ties ke Shown in Fig 137. Derive the diferent equation of motion and determine the natural frequency of ‘xiation, Figure 1.3740 exeMents OF WaRATION ANALYSIS 1.18 The circular shaft shown in Fi, 138 bas the torsional stiffess GJ(2)= GUL — 46/1]. The Shafts ined atx — Oand ha a igi ik of polar mass moment of inertia equal to attached tothe end iP L.Assume thatthe ase of the shal & negligible, derive the dierenil equation of motion and ‘obtain the natural requency of oscil 77 119 A cantilever beam made of wo sitions asa lumped mass at x = L,38shown in ig. 139. Asume that the mass of the Beam can be ignored, derive the diferent equation of motion and obtain the etod of osailation. bt 1.20 A uniform rig bar of mass mi suspended by two inestesible massless tring of length L (Sox Fig 140). Such «system i tferred toa biflar pendulum. Derive the dilerental equation for the ‘slaton 9 about he vertical ais through the bar cete. Note that the mass moment of neta of the bar about its ceteris le = ya a) Li Li co =FREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS 4 1.2 Obtain the natural frequency of he system shown in Fig. 141. The springs liner and the pulley fhas @ mage moment of ineria 1 about the center 0. Let k-= 2500 Ibn (43782 x 10° Nim), 1600 bins! (6179 Nean-),m-=2:51b-sfia A372 kg) and R= 20% (051m) 7, igure At 1122 Auriform disk ordi rolls without slipping inside crcl tack of radi, 2 shown in Fig 142. Detive the equation of motion for arbitealy large angles 0. Then, sow that inthe neighborhood ofthe trivia equriom @ = 0 the system behaves ikea harmonic oscillator, and determine the natural Frequency of exciton Figure 142 1.23 The pendulum shown in Fig. 43s attached to linear sping of sins kDevve the diferent equation of motion of the system, tea linearize the equation and determine the natural frequency of ‘ulation, Figure 14342 ELEMENTS OF VIBRATION ANALYSIS. 1.24 A uniform bar of total mass msn length I otates withthe constant angular velocity @ about 3 ‘ertcal ant s5 shown in Fig. 144 Denote by @ the angle between the vertical axis and the bar, an: (a) Determine the equilibrium postions as expressed by the constant ange By (@) Derive the ferential equation for small motions 8, about 9. (Demis ition reach uso poston i on throu hae ‘motion 8 be harmonic. (4) Calealte the natral requsey of the oscilation 0, for the stable cases {@) Determine the natural frequency fr very lange and draw coocusions. Figure 144 1.25 The inverted pendulum of Fig. L45 i supported by eae spring fines kas shown. Denote by 0 the ange between the pendulom and the vera trough the hinge O and: (@) Determine the equim positions, as expressed by the angle 6 (@) Derive the diferental equation for small angular motions 0 abost (6) Determine stability citerion based on the requirement thatthe motion 8 be harmonic. (4) Calealate the natura raquency of the oscilation 8, 77 Figuce 1AS 126 An Leshapod maces iid members hinged at point O and has amass m atthe tp. The member ‘supported by a spring of tae fa shown in Fig 146 I is eguied to: (a) Determine the equilibrium poston, as expresed by the angle @y about 0. (@) Derive the diferencia equation for small angular motion 6, about.[FREE RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 43 ( Calelate the astra frequency of olin 2 (@) Determine the height H for which the system Becomes unstable 1.27 The system of Prob, 112i immersed in viscous gud to that there ia damping free esising the motion, Calculate the period of the damped osilation, wbere the period refers tothe harmonic {actor inthe response. 1.28 The simple pendlum of Fig .47 is immerse in vioeous liquid otha therefore fl essing ‘the motion, Derive the equation of motion for arbitrary amplitudes 8, then insriz the equation and ‘obtain the frequency of the damped osilation. igure 147 1.29 Show that solution (1.50) ean also be writen inthe form w) = (Ci cosh PHT one + Casion JET ‘Then et + 1, set Cy = Ay and CaF T y= Ay and prove Ba (151.44 ouexrs oF viaraiON ANALYSIS 1.30 Calculate the fequency ofthe damped oscillation of the system shown in F148 for te values = 400 lyin (LO0S1 x 10* Nm), c= 201b-sfia. (380254 N-s/m), m = 1Olbss?in (1781.27 kg), 2 50 in (1.27 m), and L = 100in (254m). Determine the vale of he eel damping 4 Li GY 3 bu TD: 1 ewe v8 . 131 Consider the sytem of Example 1.7, and determin the response 3) to the initial conditions x)= 9, 5(0)=0 for (> Land 1/3 the response has no peaks and for { = 0 there is a discontinuity at w/c, = 1. In the tndamped case, ( ~ 0, the homogeneous diferenial equation reduces to that of harmonic osilator, lading us tothe conclusion that when the driving frequency approaches the natural frequency a, the response of the harmonic osllator tends to neease indefinitely. In such a case the harmonic oscillators said to approach & resonance condition characterized by violent vibration. However, solution (2.47) is no longer valid at resonance; a new solution of Eq. (2.41) corresponding to @ = a, is obtained later in this section ‘We notice that for light damping, such as when { < 0.05, the maximum of |G(i@)| occurs in the immediate neighborhood of «/c, = 1. Introducing the station [lan = Q, we obtain for smal values of ¢ 1 oud 250) and the curves Gio) versus o/a, are nearly symmetric with respec othe vertical through wfoy~ I in that neighborhood. The symbol Q is known asthe quality {actor becuse in many electrical engineering applications such as the toning tout ofa radio, the interest His in an amplitude st resonance that i s lage as,S54. ELEMENTS OF VIBRATION ANALYSIS posible. Th symbols often refered to asthe “Q” ofthe circuit. The points P, and Pa, where the amplitude of [GG falls to O//2 ae called hapower points because the power absorbed by the resistor in an electric cireuit or by the damper in 2 mechanical system responding harmonically a given frequency is proportional to the square of the amplitude (eee See. 210) The increment of frequency tssoctated with the hal-power points P, and Ps referred to asthe bandwidth of the system, For light damping, its not difficult to show that the bandwidth has the value be = 02 ~ 0, = Loy si) Moreover, comparing Eqs. (2.50) and (2.51), we conclude that ie ayaa asm ‘which can be used as a quick way of estimating ‘At this point let us turn our attention to the phase angle and recall that its expression is given by Eq. (2.40). Figure 24 plots $ versus w/a, for selected values (off, We notice that all curves pass through the point ¢ = 1/2, e/a», = 1. Moreover, for w/oy-< 1 the phase angle tends to zero as —+ 0, whereas for'a/o, > 1 it tends to’. For {= 0 the plot ¢ versus o/a, has a discontinuity at o/a,= 1, jumping from @'=0 for w/ay< | to @ = x foran/ao, >1. This can be easily explained by theFORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS 5 fact that for ( = 0 solution (2.47) reduces to 1 7 0 = eae 53) so that the response is in phase with the excitation for e/a, <1 and 180° out of phase for oo, > 1. Indeed, for oj, < 1 the frequency response is postive, s0 that Ihe response is in the same direction as the excitation, and for w/o, > 1 the frequency response becomes negative so that the response is in a direction opposite to that of the excitation. Equation (253) also shows clearly that the response of a hharmonic oscillator increases without bounds as the driving frequency approaches the natural frequency ©. Finally, let us consider the case of the harmonic oscillator at resonance. cause the velocity term is zero, there is no need to use the complex vector form for the excitation and response. Hence, in this case the differential equation of motion, Eq, (2.41), reduces to 20) + x(0) = 024 cos ot (254) Itis not difficult to verify by substitution that the particular solution of Eq, (2.54) is x)= ont sin og (255) which represents oscillatory response with an amplitude increasing linearly with time. This implies that the response undergoes increasingly wild fluctuations as ¢ ‘pecomes large. Physically, however, the response cannot grow indefinitely, as at a certain time the small-motions assumption implicit in linear systems is violated. Because the excitation isa cosine function and the response isa sine function, there is a 90° phase angle between them, as can also be concluded from Fig. 24. The response (0), as given by Eq. (2.55), is plotted in Fig, 2.5 as a function of time. Figure 25‘56 ELEMENTS OF VIBRATION ANALYSIS 24 ROTATING UNBALANCED MASSES ‘Many mechanical systems can be represented by mathematical models of the type shown in Fig, 2.6a. The model consists of a main mass M — m and two eccentric ‘masses m/2 rotating in opposite directions with the constant angular velocity «. To derive the equation of motion of the system, we consider two free-body diagrams, the first for the right eccentric mass, shown in Fig. 2.6b, and the second for the main ‘mass, shown in Fig, 2.6c, Because the effect ofthe two cocentric masses on the main ‘mass can be inferred from Figs. .6b and c, there is no need fora free-body diagram for the left eccentric mass. Indeed, from Figs. 2.6b and c, we conclude that the reciprocating eccentric masses exert on the main mass two vertical forces F, that add up and two horizontal forces F, that cancel each other out. Because the horizontal forces cancel, the main mass undergoes no motion in the horizontal direction, so that it is only necessary to consider the vertical motion x(). AS ‘demonstrated in Sec. 1.3, by measuring the displacement x(t) from the equilibrium position, the effect of the weight of the masses can be ignored in the equation of motion. We note, however, that F, contains a component equal to mg/2 and the foree in any of the two springs contains a component equal to Mg/?, in addition to the values appearing in the equations of motion to be derived shortly. From Fig. 2.6), we observe that the vertical displacement of the eccentric mass is x(t) + [sin wf, so that the equation of motion in the vertical direction is ma m Se bs) + Iino] sju-m [2( ~ lo? sin of] (256) TTT. © 2 cy) o Figure 26Se Mowe, om Fg, 26 he equation a motn fa the min mas -2r. ey —2F 09 = moto em Substituting F, from Eq, (2.56) into Eq. (2.57) and rearranging, we obtain the equation of motion of the system in the form M&(O) + X(0) + kex(0) = mle? sin wt = Im (mlore') (258) ‘where Im denotes the imaginary part of the expression within parentheses. Hence, rotating eccentric masses exert a harmonic excitation on the system. ‘The solution of Eq, (2.58) can be derived directly from the results of Sec. 2.3 Indeed, from Eg. (2.47), we conclude that the response is x()=Im [Cy tanner] ‘ | [GUa)|sin(or— 4) a re0s0 259) slr = loca ——| 51.00 la | lee Figure 27‘58 ELEMENTS OF VIBRATION ANALYSIS in which the phase angle ¢ is given by Eq. (2.40). Writing the response in the form x0) = X sin (ot — 4) (260) wwe conclude that mi(o x= (2) roto eon Hence, in this particular cae the indicated nondimensional rato is MX o\ (2) rt ee) instead of {G(ia) alone, so that Fig, 23 is not applicable. Plots of (a/o,)"IG(ia)| versus oa, with £ asa parameter are shown in Fig. 2.7. On the other hand, the plot versus oo, remains as in Fig. 24. We note that for o-»0, (v/o,)*IG(i@)| +0, whereas for a -+ c0, (w/o,)"G(ia)} > 1. At the same time, from Eq. (2.40), we conclude that as «@ —> 20, x. Since the mass M — m undergoes the displacement Im x, whereas the mass ‘m undergoes the displacement Im (x + le), it follows that for large driving frequencies o the masses M — m and m move in such a way that the mass center of the system tends to remain stationary. This is true regardless of the amount of damping. Note that Im x = —X sin ot for large o. 2.5 WHIRLING OF ROTATING SHAFTS In many mechanical applications one encounters rotating shafts carrying disks. On ‘occasions some of these shafts experience violent vibration. To explain this phenomenon, let us consider a rotating shaft carrying a single disk. Ifthe disk has Some eccentricity, then the rotation produces a centrifugal force causing the shaft to bend. The rotation of the plane containing the bent shaft about the bearings axis is known as whirling. Figure 28a shows a shaft rotating with the constant angular velocity @ relative to the inertial axes x, y. The shaft carries a disk of total mass m at midspan and is supported elastically at both ends. Because the shaft has distributed mass, the system has an infinite number of degrees of freedom. However, ifthe mass of the shalt i small relative to the mass of the disk, then the motion of the system can be described approximately by the displacements x and y of the geometric center S of the disk. Although this implies a two-degree-of-freedom system, the x and y ‘motions are independent, so that the solution can be carried out as for two systems with one degree of freedom each. ‘As a preliminary to the derivation of the equations of motion, we denote the origin of the inertial system x, y by O and the center of mass of the disk by C, where Cis ata distance e from S, as shown in Fig. 2.8b, The equations of motion involve the acceleration ac of the mass center C, To compute ac, we first write the radiusge 28 vector Fc from 0 to C in terms of cartesian components as follows: (+ Coos oni + (y + esin on 263) where i and j are constant unit vectors along axes x and y, respectively. Then, differentiating Eq, (2.63) twice with respect to time, we obtain the acceleration of C in the form Ae = (& — ea? cos wH)h + (J — eu? sin wr)j 264) To derive the equations of motion, we assume that the only forces acting on the disk are restoring forces due to the elastic supports and the elasticity of the shaft and resisting forces due to viscous damping, such as caused by air friction. The clastic effects are combined into equivalent spring constants k, and ky associated with the x and y directions, respectively. Moreover, we assume that the coefficient of viscous damping is the same in both directions and equal to c. The elastically restoring forces and the viscous damping forces are acting at point S. Considering Eq, (2.64), Newton's second law can be written in terms of x and y components as follows: hex — ef = m( = ec cos at) 265) ky — ef = m{— ec? sin ot) ‘which can be rearranged in the form $4 Wout + 08x = en? cos a se 266) G+ 2ylom) + o2,y = ea? sin ot60 ELEMENTS OF VIBRATION ANALYSIS where 267) & Imo, are viscous damping factors and natural frequencies. The steady-state solution of Eqs. (2.66) can be obtained by the pattern established in See. 2.4. Indeed, following that pattern, we can write simply x(0) = X(@) 0s (or 4.) y= Yo) sin(or— 4) (2.68) where the individual amplitudes are X(o)= (2) Gay) Yo) = (2y Go 69) in which 1G.) = T-Ca FEIT «aay (Ol = ola FP + Calin ‘are magnification factors and eco [nn [eG aaa oO. tan” any are phase angles ‘One special case of interest is that in which the stifness is the same in both directions, kz = ky = k. In tis case, the two natural frequencies coincide and so do the viscous damping factors, or 7 c en ae ™ Ima, Moreover, in view of Eqs. (2.72), we conclude from Eqs. (2.70) and (2.71) that the ‘magnification factors on the one hand and the phase angles on the other hand are the same, or 1G.) = 16,(00)| = 16(40)) = Tear (2730) 1 Xan lo? ae n= 6) = 6 = tan It follows immediately that the amplitudes of the motions x and y are equal to oneFORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 61 another, or X(w) = Yo) =e (2) IGtio)| em) But, from Fig, 8b and Eqs. (2.68), we can write tan 0 =2 = tan (or $) 275) from which we conclude that o= 0-9 276) and that b=o em Hence, in this case the shaft whirls with the same angular velocity as the rotation of the disk, so that the shaft and the disk rotate together as a rigid body. This case is known as synchronous whirl. Is easy to verify that in synchronous whirl the radial distance from 0 to S is constant, or an VPFP=e(2) tataieome a9 so that point S describes a circle about point O. To determine the position of C relative to the whirling plane, we consider Eq. (2.76). The relation between the angles 8, ot, and @ is depicted in Fig. 2.9. Indeed, from Fig. 2.9, we can interpret the phase angle @ as the angle between the radius vectors Fos and rsc. Hence, recalling Eqs. (2.73h), we conclude that @ w/2 for @ > ,. The three configurations are shown in Fig. 2.10, AAs a final remark concerning synchronous whirl, we note from Eqs. (2.73) that, ‘the magnification factor and the phase angle have the same expressions as in the Figure 29(62 ELEMENTS OF VIBRATION ANALYSIS case of the rotating unbalanced masses discussed in Sec. 24. This should come as no surprise as the two phenomena are entirely analogous. ‘Next, let us return (0 the case in which the two stiffnesses are different and consider the undamped case, ¢ = 0. In this case, solutions (2.68) reduce to xQ)=X@) coset — y() = Y(o)sin or (2.19) where of) T= (lone Dividing the frst of Eqs. (2.79) by X(w) and the second by Y(«), squaring and adding the results, we obtain x@) (280) yet est x which represents the equation of an ellipse. Hence, as the shaft whirls, point S describes an ellipse with point O as its geometric center. To gain more insight into the motion, let us consider Eqs. (2.79) and write moat Zune ex Ditferentiating both sides of Eq. (2.82) with respect to time and considering Eqs, (2.79), we obtain XY . FF cost or + ¥? sin? ot But, the denominator on the right side of Eq. (283) is always positive, so that the sign of 6 depends on the sign of XY. By convention, the sign of « is assumed as positive, ie, the disk rotates in the counter-clockwise sense. We can distinguish the following cases: 6 283) 1. © 0, 50 that point $ moves on the ellipse in the same sense as the rotation 2. W_p << Oy OF Uy <0) < Oye. In either of these two cases XY <0, s0 that 'S moves in the opposite sense fromFORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS 63 Figere 241 3. © > Omg and @ > oy, In this case XY > 0, so that S$ moves in the same sense ‘The three cases are displayed in Fig. 2.11, ‘Examining solutions (2.79) and (2.80) for the undamped case, we conclude that the possiblity of resonance exists. In fact, there are two frequencies for whic resonance is possible, namely, © = dye and @ = Oy. Clearly, in the case of resonance, solutions (2.79) and (2.80) are no longer valid. It is easy to verify by substitution that the particular solutions in the two cases of resonance are (0) = Jed Sin ot (= es IO) = “Heer t 608 ot ‘The plot x( versus ¢ resembles that of Fig. 25 n fact, iis the same for A = e. The plot y() versus algo resembles that of Fig, 25 except that yy and sin ey must be replaced by Oy and sin (ot ~ m2), respectively. This is easily explained by the fact that sin (@yf ~ #/2) = 008 dg The two frequencies @ = Oy and © = Gy, are called ertial frequencies. 2.6 HARMONIC MOTION OF THE SUPPORT Another illustration of a system subjected to harmonic excitation is that in which the support undergoes harmonic motion. Considering Fig. 2.2, the differential ode eae 2equation of motion can be shown to have the form mit ek =) + Mx=D 8s)” leading to E+ agi + abs = Kons +02y e286) Letting the harmonic displacement of the support be given by (Vt) = Re (de) (287) the response can be written as pfs lee adage oe li jin) + Bhaja, A | oe Following a procedure similar to that used previously, the response can be written in the form x(0) = X 60s (ot — 61) (289) 4 i : re030 | lh reas g oe can x | $050 | s t as ieee <= Q i 7 3 os Figure 213FORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 65 oo oie, Flew 214 where 1+@lojog? 2 ato\r (wear Stem ) ale) bites) ae and 2x(o/eu)* 4 es i= (@jo,F + Tolar Hence, in this case the indicated nondimensional ratio is xe [: + eT" 1G¢01 9) where the ratio X/4 is known as transmissibility. Curves X/A versus c/a, with £ as a parameter are plotted in Fig. 213. Moreover, curves @ versus «a/@, for various values of { are shown in Fig, 2.14, Again, for ¢'= 0 the response is either in phase with the excitation for w/a, <1 or 180° out of phase with the excitation for ofa, > 1.(66, ELEMENTS OF VIBRATION ANALYSIS 2.7 COMPLEX VECTOR REPRESENTATION OF HARMONIC MOTION ; ‘The representation by complex vectors ofthe harmonic excitation and the response of a damped system to that excitation can be given an interesting gcomettic interpretation by means of a diagram in the complex plane. To this end, we ‘consider the second-order system discussed in Sec. 23. Differentiating Eq. (247) With respect (o time, we obtain (0 = fwAIG(ia)le"" = iwox(0) (2.930) (0) = (1a)? AIG(Ga)le""-# = — 0 x(0) (2.938) Because i can be written as f= cos x/2 +i sin n/2 =e, we conclude that the velocity leads the displacement by the phase angle 2/2 and that itis multiplied by the factor @. Moreover, because —1 can be expressed as —1 = cos + isin x =e, it follows that the acceleration leads the displacement by the phase angle x and that itis multiplied by the factor a”. In view of the above, wecan represent Eq, (2.41) in the complex plane shown in Fig. 215. There is no loss of generality in regarding the amplitude A as a real ‘umber, which is the assumption implied in Fig. 2.15. The interpretation of Fig. 2.15 is that the sum of the complex vectors %(0), 2a,(0), and cx(t) balances wlAe™, which is precisely the requirement that Eq. (2.41) be satisfied. Note that the entire diagram rotates in the complex plane with angular velocity a. Itis clear that considering only the real part ofthe response isthe equivalent of projecting the diagram on the real axis. We can just as easily retain the projections on the imaginary axis, or any other axis, without affecting the nature of the response. In view of this, itis also clear that the assumption that A is real is immaterial. Choosing A as a complex quantity, or considering projections on an axis other than the real axis, would merely imply the addition of a phase angle y to all the vectors in Fig. 2.15, without changing their relative positions. This is equivalent to multiplying both sides of Eq. (241) by the constant factor e. ‘The above geometric interpretation extends to first-order systems as well. In fact, to obtain a figure analogous to Fig. 2.15 all that is necessary is to remove the complex vector £(¢) and to adjust the magnitude of the remaining vectors, which 2ta,it fo Figure 215= [FORCED RESPONSE OF SINGLE-DEGREI-OF-FREEDOM LINEAR SYSTEMS 67 can be achieved by letting 2fa=1 and @? =a. In the process, the rotating trapezoid of Fig, 2.15 becomes a rotating triangle. 28 VIBRATION ISOLATION In many systems of the type shown in Fig, 1.10a, we are interested in transmitting. as little vibration as possible to the base. This problem can become critical when the excitation is harmonic, Clearly, the force transmitted to the base is through springs and dampers. From Fig. 2.15, we conclude that the amplitude of that force is Fe = m{Qlont} + (oP? (294) where the amplitude of the velocity is simply ax. Hence, we have coup) am But from Eq, (247), if we recall that the phase angle is of no consequence, we conclude that ae[e+ 22) Pte 296) Because Ak = Fo is the amplitude of the actual excitation force, the nondimen- sional ratio F/Fo is @ measure of the force transmitted to the base. The ratio can be written as & [: + eT \¢a9) sn ands recognized asthe transmissibility given by Ea. (292). Hence the plots F/Fo versus ofan are the same as the plots X/A versus ca, shown in Fig. 2.13 eis not difficult to show that when ofa, = /2 the fll force is transmitted to the base, F,/Fo = 1. For values e/a, > ./2 the force transmitted tends to decrease with increasing driving frequency «, regardless of C. Interestingly, damping does not alleviate the situation and in fact for ofc, > «/2the larger the damping, the larger the transmitted force. Recalling, however, that in increasing the driving frequency ‘we would have o go through a resonance condition for zero damping, we conclude {hata small amount of damping is desirable, Moreover, the case of 20 damping represents only an idealization which does not realy exist, and in practice «small amount of damping is always present 29 VIBRATION MEASURING INSTRUMENTS ‘There are basically three types of vibration measuring instruments, namely, those ‘measuring accelerations, velocities, and displacements. We shall discuss the firstand the third only. Many instruments consist ofa case containing a mass-damper- spring system of the type shown in Fig. 216, and a device measuring thes displacement of the mass relative to the case. The mass is constrained to move along a given axis. The displacement of the mass relative to the case is generally ‘measured electrically. Damping may be provided by a viscous fluid inside the case. ‘The displacement of the case, the displacement of the mass relative to the case, and the absolute displacement of the mass are denoted by y(0), 2(2), and x(0), respectively, so that x(0) = y(t) + 2(0, The relative displacement 2(0) is the one ‘measured, and from it we must infer the motion y() of the case. Although we wish ultimately to determine (0, itis the response 2(0) which is the variable of interest. Using Newton's second law, we can write the equation of motion m3(0) + LX — HO] + Kx — HO] = 0 (98) which, upon elimination of x(), can be rewritten as malt) + c#(0) + ko(®) = — mye) (299) Assuming harmonic excitation, y(0) = Yoo, Eq, (2.99) leads to mi + ci + kz = Yomote™ (2.100) which is similar in structure to Eq, (2.58). By analogy, the response is 20 = ¥(2) [Gaye @.101) whore the phase angle @ is given by Eq. (2.40). Introducing the notation 2(¢) = Zoe", we conclude that the plot Zo/Yo versus @/a, is identical to that given in Fig. 2.7. The plot is shown again in Fig. 2.17 on a scale more suitable for our purposes here. For small values of the ratio aja», the value of the magnification factor |G()| is nearly unity and the amplitude Zq can be approximated by Bm=Y% (ey (2.102) ZZ gue 236FORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 69 Ley gros Figure 217 Because Yow* represents the acceleration of the case, the measurement Zo is proportional to the acceleration of the case, where the proportionality constant is ers. Hence, ifthe frequency « of the harmonic motion of the case is sufficiently low relative to the natural frequency of the system that the amplitude ratio Zo/ Yo can be approximated by the parabola (w/n,)* (see Fig, 2.17), the instrument can be uused as an accelerometer. Because the range of w/a, in which the amplitude ratio can be approximated by («/c,)? is the same as the range in which [G(ia)| is approximately unity, it will prove advantageous to refer to the plot (Gia) versus for instead of the plot Zo/ Yo versus «/ay,. Figure 2.18 shows plots |G(ia) versus ‘v/er, in the range 0 < w/a, < 1, with ¢ acting as a parameter, From Fig. 2.18 we ‘conclude that the range in which |G(ia)| is approximately unity is very small for light damping, which implies that the natural frequency of lightly damped accelerometers must be appreciably larger than the frequency of the harmonic ‘motion to be measured. To increase the range of utility of the instrument, larger ‘damping is necessary. Its clear from that figure that the approximation is valid for 10s on tot |6tie) 1.00 03 036 08% oF uF tr 38 eo, Figure 218170. ELEMENTS OF VIBRATION ANALYSIS a larger range of c/a», if 0.65 < ¢ <0.70, Indeed, for { = 0.7 the accelerometer can be used in the range 0 < c/o < 0.4 with less than 1 percent error, and the range~ can be extended to c/a, <07 if proper corrections, based on the instrument calibration, are made. "The most commonly used accelerometers are the compression-type piczo- clectric accelerometers. They consist of a mass resting on @ piezoclectric ceramic Crystal, such as quartz, barium titanate or lead zirconium titanate, with the crystal acting both as the spring and the sensor. ‘The accelerometers have preload providing a compressive stress exceeding the highest dynamic stress expected. Any eeesleration increases or decreases the compressive stress in the piezoelectric Gloment, thus generating an electric charge appearing at the accelerometer terminals, Piezoelectric accelerometers have negligible damping and they typically havea frequency range from 0 to 5000 Hz (and beyond) and a natural frequency of 30.000 Fiz They tend to be very light, weighing ess than 1 oz and relatively small, ‘measuring less than 1 in in height. ‘Also from Fig. 217, we notice that for very large values of w/t, the ratio Zol¥o = (otog)"1G(ia)| approaches unity, regardless of the amount of damping, Hence, if the object is to measure displacements, then we should make the natural frequency ofthe systems very Iow relative tothe excitation frequency, in which case the instrument is called a seismometer. For a seismometer, which is an instrument tesigned to measure earth displacements such as those caused by earthquakes or underground nuclear explosions, the requirement for a low natural frequency ldictates that the spring be very soft and the mass relatively heavy, so that, in tssence, the mass remains nearly stationary in inertial space while the ese, being Gttached to the ground, moves relative to the mass. Displacement-measuring instruments are generally undamped. They typically have a frequency range from 10 to 500 Hz and a natural frequency between 2 and 5 Hz. ‘Because seismometers require a much larger mass than accelerometers and the relative motion of the mass in a seismometer is nearly equal in magnitude to the fmotion to be measured, seismometers are considerably larger in size than accelerometers. In view of this, ifthe interest lies in displacements, it may prove more desirable to use an accelerometer to measure the acceleration of the case, and then integeate twice with respect to time to obtain the displacement. “The above discussion has focused on the measurement of harmonic motion. In measuring more complicated motions, not only the amplitude but also the phase angle comes into play. As an example, ifthe motion consists of two harmonics, ot YQ) = ¥, 608 wat + Yo 605 oat 2.103) and the accelerometer output is yt) = Y, C08 (wit ~ 43) + Ya 00s (wat ~ 2) (2.104) where $y and $2 are two distinct phase angles, then the accelerometer fails to reproduce the motion y(), because the two harmonic components of the motfon ae shifted relative to one another. There are two cases in which the accelerometer ‘utput is able to reproduce the motion y(¢) without distortion. The first isthe caseFORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS Tl of an undamped accelerometer, £ = 0, in which case the phase angle is zero. The second is the case in which the phase angle is proportional to the frequency, or dsc dren, 2.105) Indeed, introducing Eqs. (2.105) into Eq. (2.104), we obtain yall so that both harmonics are shifted to the right on the time scale by the same time interval, thus retaining the nature of the motion y(®). To explore the possibility of eliminating the phase distortion, let us consider the case of small o/é%, in which case the phase angle 6 is small, as can be concluded from Eq, (2.40). Then, assuming that the phase angle increases linearly with the frequency, we can write £08 w(t ~ 6) + Ys 608 w(t — 0) (2.106) sing b=co cospxt— fmt —Hea* 2107) Inserting Eqs. 2.107) into Eq, (240), we obtain elon 806 Teja, T= aE a which is satisfied provided c= Vo, f= /2/2=0.107 2.109) In general, any arbitrary motion can be regarded as a superposition of harmonic components, Hence, an accelerometer can be used for measuring arbitrary motions if the damping factor ¢ is either equal to zero or equal to 0.707. 2.10 ENERGY DISSIPATION. STRUCTURAL DAMPING In Sec. 2.3 we have shown that the response of a spring-damper-mass system subjected to a harmonic excitation equal to the real part of F(e) = Ake 2.119) is given by the real part of X(0) = AlGlia)leter“# = Xetoro an where X = AIGtia)| 22 ccan be interpreted as the displacement amplitude. Moreover, we have shown in Sec. 27 that there is no loss of generality by regarding A as a real number. Clearly, because of damping, the system is not conservative, and indeed energy is dis- sipated. Since this energy dissipation must be equal to the work done by the external force, we can write the expression for the energy dissipated per eycle of“72 ELEMENTS OF VIBRATION ANALYSIS vibration in the form Abge= [Fae = | Rear 3) where we recall that only the ral parts of F and x must be considered, Inserting Eqs (2934) and (2.110) into (2113), we obtain AB = Harta [eos tsa (at — 4) = moj A?|Glia)|n-sin & nay From Eqs. (240) and (2.48), itis not difficult to show that sin g = 252 ja(ia = <9, 16400) is) where it is recalled that £ = ¢/2mo, Inserting Eqs. 2.112) and (2.115) into (2.114), we obtain the simple expression Ege = cnaX™ < @116) from which it follows that the energy dissipated per cycle is directly proportional to the damping coefficient c, the driving frequency «, and the square of the response amplitude. i ‘Experience shows that energy is dissipated in all real systems, including those systems for which the mathematical model makes no specific provision for ‘damping. For example, energy is dissipated in real springs as a result of internal friction. In contrast to viscous damping, damping due to internal friction does not depend on velocity, Experiments performed on a large variety of materials show that energy loss per eycle due to internal friction is roughly proportional to the square of the displacement amplitude,t Beye = 2X? any) where ais a constant independent ofthe frequency of the harmonic oscillation. This type of damping, called structural damping, is attributed to the hysteresis phenomenon associated with eyclic stress in elastic materials. The energy loss per (yele of stress is equal to the area inside the hysteresis loop shown in Fig. 2.19. Hence, comparing Eqs. (2.116) and (2.117), we conclude that systems possessing structural damping and subjected to harmonic excitation can be treated as if they ‘were subjected to viscous damping with the equivalent coefficient (2.118) 4 See. Meiovteh, Analytical Methods ix Vibration p02, The Macmillan Co New York, 1967FORCED RESPONSE OF SINGLE-DUGRE-OF-FREEDOM LINEAR SYSTEMS. 73 Hystees aay Area een lor perches x) Figure 219 This enables us to write Eq. (1.14) in the form me) + so + kex(t) = Ake’ (2.119) where consideration has been given to Eqs. (2.110) and (2.118). Because ‘we can rewrite Eq. (2.119) in the form mi(e) + KCL + iy)x(t) = Ake (2.120) where 21) is called the structural damping factor. The quantity {I + i) is called complex: stiffness, ot complex damping. ‘The steady-state solution of Eq. (2.120) is the real part of Ae 0" Tela eh (2.122) and, in contrast to viscous damping, for structural damping the maximum amplitude is obtained exactly for @ = a, ‘One word of caution is in order. The analogy between structural and viscous damping is walid only for harmonic excitation, because the response of a system to harmonic excitation ‘with the driving frequency @ is implied in the foregoing development. 2.11 THE SUPERPOSITION PRINCIPLE Let us consider again the second-order linear system depicted in Fig, |.9a, In Sec. 1.3 wwe have shown that the differential equation for the response x(t) of the system to7H ELEMENTS OF VIBRATION ANALYSIS the arbitrary excitation force F(0) can be written in the form dxf), dx(t) ate ade + kx = FO 2a) Where m, c, and k are the system parameters denoting the mass, the coefficient of viscous damping, and the spring constant, respectively. Quite often x(t) and F(t) are called the output and input of the system, respectively. The relation between the response and the excitation, or output and input, can be given an interesting and somewhat useful interpretation by introducing the linear differential operator ad motes 24 Damiztegtk 2124) ‘This enables us to write Eq. (2.123) in the symbolic form DIX] = FO (2.125) where the juxtaposition of x(t) and (0) in Eq. (2.125) implies the operation D on (in such a way as to produce Eq, (2.123), We note that an operator is linear ifthe differential expression D[x(0)] contains the function x(¢) and its time derivatives to the first and zero powers only. Thus, cross products thereof and terms involving fractional powers of x(t) are precluded. ‘The operator D contains all the system characteristics because it involves all the system parameters, namely, m, c, and k, and it specifies the order of the derivatives multiplying each of these parameters as well. In system-analysis language, D represents the “black box” of the second-order system. Relation (2.125) ‘can be illustrated by means of the block diagram shown in Fig. 2.20, which implies that if the input F(@) is fed into the black box represented by D, then the output is. x0. ‘We can use the operator D to define the concept of linearity of a system. To this tend, we consider two excitations F,(2) and F,(¢) and denote the corresponding responses by x(t) and x(t), so that FO= DEA] FO = DEO] 2.126) [Next we consider the excitation F3(0) as a linear combination of F(t) and Fa(®), namely, FAO = cP) + ea, enn where ¢, and cz are known constants, Then, if the response xs(t) to the excitation y(t) satisfies the relation wal) = exxs(®) + ea) (2.128) Excitation sien Response [ i { wns sron Figure 2.20 L[FORCED RESPONSE OF SINGLE-DEGREE-OF-FREEDOM LINEAR SYSTEMS. 75 the system is linear; otherwise itis nonlinear. Using Eqs. (2.126) and (2.127), the above statement can be written in terms of the operator D as follows: DUxs(] = Dleuxi(t) + e2xa()] = exD[x1(0] + eaDLxs(01] esFil) + e2F (0 = Fs (2.129) Equation (2.129) represents the mathematical statement of the so-called principle of superposition, which clearly applies to linear systems alone. In words, the principle implies that for linear systems the responses to @ given number of distinct excitations ean be obtained separately and then combined to obtain the aggregate response. The superposition principle is a very powerful one, and has no counterpart for nonlinear systems. It is because ofthis principle that the theory of linear systems is so well developed compared to that of nonlinear systems. Note that we have already used the principle in Sec. 2.9 to discuss the subject of phase distortion in accelerometers. We shall use the principle again to derive the response of linear systems to periodic and nonperiodic excitation. 2.12 RESPONSE TO PERIODIC EXCITATIO! FOURIER SERIES In Secs. 2.2 and 2.3, we derived the steady-state response of first and second-order systems to harmonic excitation. By virtue of the fact that the systems considered ‘were linear, any transient response’ due to the initial conditions could be obtained separately and then added to the steady-state response by invoking the super- position principle. Harmonic excitation of any arbitrary frequency a is periodic, ic, it repeats itself at equal intervals of time T= 2x/o, where T is the period of excitation. In vibrations we encounter other types of periodic excitations, not necessarily harmonic. As an example, the function illustrated in Fig. 2.21 is periodic but not harmonic. Any periodic function, however, can be represented by a convergent series of harmonic functions whose frequencies are integral multiples of a certain ‘fundainencal frequency a provided that it satisfies certain conditions to be pointed ‘out shortly. The frequencies representing integral multiples of the fundamental frequency are called harmonics, with the fundamental frequency being the first harmonic. Such series of harmonic functions are known as Fourier series, and can Ae fp § fp 1 Figure 221116 ELEMENTS OF VIBRATION ANALYSIS be written in the form (see App. A) @ an FO = 400+ $ (apc08 pont + b,Sin po!) =F 2130) where pare integers, p = 1, 2,3... and Tis the period. The coefiients ay and by are given by the formulas aH2[" rosacea pana. essa nell sosemed pata. aly and they represent a measure of the participation of the harmonic components 0s pwiot and sin pooot, respectively, in the function f(f). Note that ag/2 represents the average value of f(0), Which in the case of Fig, 2.21 is zero. The Fourier series representation is possible provided the integrals defining a, and b, exist. We shall ‘not pursue the subject ofthe integrals’ existence, because for the physical problems ‘we will be concerned with it can be safely assumed that the integrals do exist.} There are certain cases in which the Fourier seties, Eq. (2.130), can be simplified. One such case is when the function f(0) is an odd function of time, which is defined mathematically by f= -f-9 @1) ‘Considering Eq. (2.132), Eqs. (2.131) yield . 12 oH, _foe0s pont de + | f(e)e0s post «| ol se c0s(-paon aco + | * 05 poe | “i[- i Eee Batok P"" popcoe Peoot dt} =0 peo. @133a) Ff, F (0) sin pont dt + "409 sin po | -iL se-sin(-poonat-0-+ [ H00sin pot] 23[f mamas [7 mans] =} 7 40) sin poot dt p=1,2,... (2133) 4 Fora discusion ofthis subject, ee AE. Taylor, Adsanced Calls, p. 714, Ginn and Ca, New Yor, 1955=] FORCED RESPONSE OF SINGLE-DEGRIN-OF-FREEDOM LINEAR SYSTENS. 77 Hence, when (0) is an odd function of , the Fourier series reduces to the sine series 2n T where the coefficients b, (p = 1, 2,...) are given by Eqs. (2.1338). A second case is the one in which f() is an even function of time, defined as {0 =f-9 2135) Using Eg. 2.135), Eas. 2.131) become a= z"., S(O) 008 peogt dt + i F() 00s part «| 40) = ¥ by sin pont 09 2134) =F] fF 9-9 008 roan do) + JP" se cos posta] Hoe pasar |" se pot a] (0) 608 poogt de P=0,1,2,... (21360) = a S050 Pt + . $0) pot] P sensnt-nassa-o-+ [so soresa] Sie Sis sie Sie 3-2" Fein pinsdr+ [psn pos 4 =0 p=12, (2.1365) mnetion oft, the Fourier series simplifies so that, in the case in which f(0 is an ev to the cosine series Hm to+ S eemepoe one ais where the coefficients a, (p =0, 1,2...) are given by Eqs (2.136a). Expansions (2.134) and (2.137) can be easily explained by observing that sin pat (p = 1,2.) are odd functions of time and eos poset (p = 0, 1,2...) are even functions of time. Hence, Ed, (2.134) states that an odd periodic function cannot have even harmonic components and Eq, (2.137) states that an even periodie function cannot have odd harmonic components. Next, we derive the response of linear systems to periodic excitation. To this end, we recognize ftom Secs. 2.2 and 2.3 thatthe response of a linear system to the excitation elt) = a 008 poogt 2.138)‘78. ELEMENTS OF VIBRATION ANALYSIS is simply Xp(0) = ap1G yl 008 (Peat — #5) (2.139) where |G,| is the magnification factor and ¢, is the corresponding phase angle. “Moreover, the response to the excitation Spall) = by sin pent (2.140) Xp(0) = bfGyl sin (Poot ~ 45) a4 If the excitation isin the form ofa periodic function with a Fourier series expansion in the form of Eq, (2.130), by virtue of the superposition principle, the response can be written as a linear combination of responses to the individual harmonic ‘components. Hence, considering Eqs. (2.139) and (2.141) and recognizing that the response to the constant ay/2 is simply do/2, we can write the response of a mass- damper-spring system to the periodic function given by Eq, (2.130) in the form of the series, x(0) = $9 + a [G,lLa, e0s (pet ~ bp) + bp sin (pevot ~ gp)]_ (2.142) where ao/2 can be identified as the average value of the response. For the mass-damper-spring system of Sec. 2.3, the magnification factor is IG,) (2.143) {= Gaofoy P+ Blpao/on) and the phase angle is (2.144) Tis clear from Eqs. (2.142) and (2.143) that if the value of one of the harmonics pao in the excitation is close to the frequency «, of the undamped oscillation, then this particular harmonic will tend to provide a relatively larger contribution to the response, particularly for light damping. The case of zero damping has interesting implications, Specifically, we conclude from Eqs. (2.142) and (2.143) that if po = «, for a certain p, then a resonance condition exists. Hence, resonance can ‘occur in undamped systems when the excitation is merely periodic, and not neces- sarily harmonic, provided the frequency of one of the harmonic components coin- cides with the system natural frequency. Tn deriving the response ofa linear system to harmonic excitation, we found it advantageous earlier in this chapter to represent harmonic functions in terms of ‘complex vectors. Because periodic functions consist of series of harmonic functions, ‘we can expect the same advantage in representing a periodic function in terms of a series of complex vectors. To this end, we recognize that the Fourier series (2.130)7 FORCED RESPONSE OF SINGLE-DPOREE-OF-FREEDOM LINEAR S¥STEMS 79) can also be written in what is generally known as its complex, or exponential form (Gee App-A) SO= & crm 14s) ‘where C, are complex coefcients having the expressions ean eek roma a) As before, the coefficient Cy represents the average value of (0). Instead of working with the negative frequencies implied in Eq, (2.145), it will prove convenient to represent the excitation in the form f= 44a + Re( S ar) aun where dy i areal coeficient and 4, are in general complex coeficens given by A 2[ sore ethan asp The reason for preferring series (2.147) to (2.145) becomes obvious when we observe that every term in series (2147) has the same form as the complex vector described by Eq. (2.15a). By analogy with the real coefficients a, and b,, the complex coefficient , represents the extent to which the harmonic component with frequency pir contributes to f(). Before attempting to obtain the response of the system to the excitation described by series (2.147), itis perhaps desirable to show that expansions (2.145) and (2.147) are indeed equivalent. Expansion (2.145) can be written as FO= Lem Cot Y Cele Yo Cytron = Cot ¥ (Cem + Cpe e149) where Cj = Cis the complex conjugate of C,. Hence, using the relations e+ eo 2.008 pat, PP — et = 2 sin pat, series (2.149) reduces to $= Co+2 ¥ RoC, 08 pot Im Cysin post) 2.150) where Re C, and Im C, denote the real part and the imaginary part of Cp, respectively. On the other hand, series (2.147) can be written in the form Si) = 440+ ¥ (REA, cos post —Im Aysin pat) (2.151) so that, observing from Eqs. (2.146) and (2.148) that A, = 2C, (p= 0,1, 2,.... we ‘conclude that the series (2.145) and (2.147) are indeed equivalent. Note that, as forthe real form of the Fourier series, the constant excitation A9/2 produces a constant response also equal to 0/2. = ‘The response ofa linear system to an excitation given by Re (Ae) was shown, in Sees. 22 and 2.3 to have the form. x(t) = Re [AG((a)e] = ReLA|G(ia)let"P)] (2.152) Where Gi) is the frequency response, Because the system is linear, the response to the excitation f(t) as given by Eq. (2.147) is simply woatieste($, 04cm) ex where, by analogy, G, is the frequency response associated with the frequency poy. Also by analogy with results of Secs. 2.2 and 2.3, we can rewrite the response (2.153) wo-ierne[ Eroiaemnn] es ‘where |G,l is the magnitude of G, and @, is the phase angle associated with the harmoni¢ of frequency par. ‘Note that for a mass-damper-spring system 1 T= (peofen.)® + Bpao/e. Moreover, |Gpl and @, are given by Eqs. (2.143) and (2.144), respectivoly. G 2155) Example 2.1 Consider the excitation /(f) in the form of the periodic square waye shown in Fig, 2.22, and calculate the response of an undamped single- degree-oF freedom system to that excitation. Solve the problem in two ways: first by considering a trigonometric form and then by considering a complex form of the Fourier series. ‘The mathematical description of the excitation over one period is simply _fA for 0 c, where a phase angle of 180° corresponds to a change inthe sign of the response. As shown in Sec. 2.3, the phase angle can be taken into account automatically by replacing the magnification factor |Gy| by the frequency response : oo (paof. o rence, the response is simply “as cpg ae Fp. sn PL (pmo/@)D O° and we note that the same harmonies that participate in f(t) participate also in (0, with the amplitude of the harmonies with frequencies close to w, gaining in magnitude relative to those with frequencies removed from o,. Tt is clear from Eq, (g) that resonance occurs for Poy = Oy ‘To obtain the response by means of the complex Fourier series, we insert Eqs. (a) into Eqs. (2.148) and obtain the coeficients a-2 [some (fears ["emea) (fr emears [roma = M4 ("apt 4iA cos pooT|T? _ AIA (5 peooT _ | T Pi Thee x o 7 w.Considering Eq. (c), we can write prodd . p=even Taserting Ea () into Ea, (2.159), we obtain the response = (ad) 1 wonne 2. (Beare Yee ys oR pk. Omer] OT 2 which is identical to the response given by Eq. (g), obtained by the approach based on the trigonometric form of the Fourier series. ‘Equation (g) can be used to plot x(t) versus f, but this may not be very illuminating. Pethaps a better understanding of the system behavior can be obtained from a plot in the frequency domain instead of a plot in the time domain. Indeed, considerable information concerning the system behavior is, revealed by plots showing the degree of participation of the various harmonics in the excitation f(@) and in the response x(0). These are plots of the amplitude of the harmonic components of the function in question versus the frequency, ‘where such plots are: known as frequency spectra. Figure 2.23a shows the Wey Boy Boy 100FORCED RESPONSE OF SINGLE-DRGREE-OF-FREEDOM LINEAR SYSTEMS. 83 frequency spectrum for the periodic square wave f(t) of Fig. 2.22, obtained by plotting the coefficients of sin paoot in Eq. (6) versus the frequency «, Because the function f(®) is periodic, its frequency spectrum consists of harmonic components with discrete frequencies, namely, «@ = @, 30, 59, .. For this reason this is a discrete frequency spectrum, In a similar manner, Fig. 2.236 represents the frequency spectrum associated with the response x(0), Eq. (g), for the case = 409. Conforming to expectation, the magnitude of the amplitudes of the harmonics 4%, 3%, and Say in x(t) gains relative to that of their counterparts in f(0, whereas all the others are attenuated. Moreover, the amplitudes corresponding to @ > doo are negative, Figure 2.23b also repre- senls a discrete frequency spectrum. 2.13 THE UNIT IMPULSE. IMPULSE RESPONSE In Sec. 2.12, we studied the response ofa system to a periodic excitation of period T: ‘The question remains as to how to obtain the response ofa system to an arbitrary excitation. Clearly, in this case there is no steady-state response and the entice solution must be regarded as transient, although the part due tothe excitation force ‘may persist indefinitely even in the presence of damping provided, of course, that te excitation persists. Before discussing the response to arbitrary excitations, we consider the response to some special types of forcing functions. ‘A very important function in vibrations is the uni impulse, or the Dirac dela {function (Fig. 224), defined mathematically as 0 fortea (2156) ‘Wenote that while the time interval over which the function is different from zero is by definition taken as infinitesimally small, that is, ein Fig. 2.24 approaches zero in the limit and the amplitude of the function in this time interval is undefined, the area under the curve is well defined and equal to unity. We also note that the units of the Dirac delta function are s~!, which should be immediately clear from the fact, that the valu of th integral in (2.156) is nondimensional. The unt impulse applied at r= 0 is denoted by 8(0). ‘The response of a system to a unit impulse applied at ¢ = 0, with the initial conditions equal to zero, is called the impulse response ofthe system and is denoted ee igure 224‘84 ELEMENTS OF VRATION ANALYSIS by 9(0. Clearly, the response to a unit impulse applied at a later time t= a is (t — a} it can be obtained by shifting g(t) to the right along the time scale by the. time interval t = a. Example 22 Calculate the impulse response ofthe damper-spring system of See. 13. Inserting x(0) = g(@) and F() = 6(0) in Eq, (1.12), we obtain ed() + katt) = 30) © by definition. Integrating Eq. (a) with respect to time ‘and taking the limit, we obtain where 9(0) = 40) ‘over the interval At lim f Cel + g(a = tim, fi Sb dt = 1 © But, mf. g(t) de ie g(t) lo dm clato) — gO)] = eg(O+) . © iim i g(t) dt = lim kg(Q)e = 0 “The notation 9(0-+) is to be interpreted as a change in displacement at the end of the time increment Ar = ¢. Note that the result in the second of Eqs. (c) was ‘obtained by invoking the mean-value theorem. Inserting Eqs. (c) into Eq. (b), we conclude that 0+) =2 @ ‘The physical interpretation of Eq, (4) is that the unit impulse produces an instantaneous change in displacement, so that we can regard the effect of the pulse applied at ¢ = 0 as being equivalent to an initial displacement (0) = I/e. In Sec. 1.5, however, we considered the response of a first-order system to an initial displacement. Hence, inserting x(t) = g(#) and x9 = 9(0) = Ic in Eq, (1.31), we obtain the impulse response es «) {= oe © a reo where + = c/k is the time constant. Example 23 Calculate the impulse response of the mass-damper-spring system of Sec. 13. Inserting x(0) = g(t) and (0) = 4(¢) into Eq. (1.14), we obtain malo) + eg) + kal = 540) @ seis nuern ner oo Fang sane prs a Bagh 2 wn wt taflarrariomin[soenr where Jim | mg ae = tim mg] im mde) — G(0)] = mal0+) im J) ed dt = lim oo} = him eLo(e) ~ o(0)] = 0 oO tim [tg de=0 ‘The notation 4(0+) is to be interpreted as a change in velocity at the end of the time increment At = ¢. On the other hand, because the change in velocity is finite and the interval of integration 0 << ¢ is extremely short, there is not sufficient time for displacements to develop, so that g(e) = 0. This fat is due to the presence of the mass m, which was absent in the system of Example 2.2. ‘Combining Eqs. (6) and (c), we conclude that 1 40+) i © The physical interpretation of Eq, (is thatthe unit impulse produces an instantaneous change in the velocity, so that we can regard the eflect ofa unit impulse applied at ¢= 0 as being equivalent to the effect ofan inital velocity 9 = /m. We recall, however, that in Example 1. we calculated the response othe system under consideration to an initial velocity In view of thi, i we introduce n= If into Eq, (g) of Example 17, we can write the impulse response inthe form 1 aL ete oye may oe © ° 10 where «4 = (1 — (2)'0y, 2.14 THE UNIT STEP FUNCTION. STEP RESPONSE ‘Another function of great importance in vibrations is the unit step function. The unit step function is depicted in Fig. 2.25 and is defined mathematically as follows: p fort 0 unaffected. There is a close relationship between the unit step function «(¢ ~ a) and the unit impulse 5(t — a). In particular, the unit step function is the integral ofthe unit impulse, or wero fl_ae-oe a oe ee a0 9) 9 2159) ‘The concept of unit step function enables us to return to some results obtained. previously and express them in a more compact manner, Indeed, referring to Example 22, it is clear that the impulse response of a damper-spring system, Eqs. (©) can be written conveniently in the form a) = Feel) (2.160) Moreover, the impulse response of a mass-damper-spring system, Eqs. (e) of Example 23, can be expressed in the compact form =u sin oa) (2.161) at ‘The response of a system to a unit step function applied at ¢ = 0, with the initial conditions equal to zero, is called the step response of the system in question and is denoted by o(¢). To derive the step response ofa linear system, let us consider Eq, (2.125) and write the relation between the unit impulse 4(¢) and the impulse response g(¢) in the symbolic form Diao] = 500) (2.162) where D is a differential operator. Integrating Eq, (2.162) with respect to time and assuming that the differentiation and integration processes are interchangeable weFORCED RESPONSE OF SINGLE-DEGREE-OFFREEDOM LINEAR SYSTEMS 87 obtain fi ptaende=of [oe a]=f aoe ary But, according to Eq, (2158), the right side of Eq. (2.163) is the unit step function applied at ¢ = a = 0. Moreover, from Eq. (2.125), the relation between the unit step function «(t) and the step response o(?) has the symbolic form DLA] = a) (2.164) Hence, comparing Eqs. 2.163) and (2.164), we conclude that «= i ate) ag 2165) of, the step response isthe integral of the impulse response ‘The step response can be used at times to facilitate the response to relatively involved excitations. Indeed, when the excitation consists of a linear combination of step functions, the response can be expressed asa similar linear combination of step responses (see Example 25). Example 2.4 Calculate the step response of a mass-damper-spring system by integrating the impulse response, according to Eq, (2.165). Plot o(t) versus t ‘The impulse response of a mass-damper-spring system is given by Eq. (2.161), so that using Eq. (2.165) the step response is ao- 2 eetsinogae dem tl emtsinoré dé (0 From Eqs. (1.37), however, itis not difficult to show that joet — gle sin cod 5 © 0 that Eq. (@) can be integrated as follows: doggie [lemon emit 1" e-tn-tn gor tioa = Sima; J, H— eKanviouty ag LP erhon tention 9 -= — | © imax |— Co, tas) Go, + ia) Jy Aer some ulgebraic operations, Eq. (¢) yields the step response a= Hiem (oman Sinon)]ad where the unit step function «(¢) accounts automatically for the fact that o(t) = 0 for ¢ <0. The plot ¢) versus ¢ is shown in Fig. 2.26.{88 ELOAENTS OF VIBRATION ANALYSIS jar Figure 226 > Example 2.5 Use the concept of unit step function and calculate the response “x(2) of an undamped single-degree-of-freedom system to the rectangular pulse shown in Fig. 227. Plot x(¢) versus t It is easy to verify that the function F(¢) depicted in Fig. 227 can be expressed conveniently in terms of unit step functions in the form FQ) = Fola(t + 1) ~ lt — 7] @ But the response of an undamped single-degree-ol-freedom system to a unit step function applied at ¢ = 0 can be obtained from Eq, (2) of Example 2.4 by letting {= 0 and a, = a4. The result is At) = 7 (1 C08 2,8) o ‘Moreover, the response to w(t + T) is (t+ 7), obtained from Eq, (#) of the present example by simply replacing ¢ by ¢+T: Similarly, the response to {dt —T) is At —T), Hence, the response to F(), as given by Eq. (a), is ‘simply x(t) = Folalt + T) — ot — 7) = 2 (C1 ~ cost + Mult + 1) [1 = cos out ~ Matt} © “The plot x) versus ¢is shown in Fig. 228 ineFORCED RESPONSE OF SINGLE-DPGREE-OF-FREEDOM LINEAR SYSTEMS $9 Figure 2.28 2.15 RESPONSE TO ARBITRARY EXCITATION. THE CONVOLUTION INTEGRAL Earlier in this chapter, we studied the response of linear systems to harmonic and periodic excitations. Then, in Secs. 2.13 and 2.14, we discussed the response to a unit impulse and to a unit step function. The question remains as to how to obtain the response to arbitrary excitation. ‘There are various ways of deriving the response to arbitrary excitation, ‘depending on the manner in which the excitation function is described. One way is to represent the excitation by a Fourier integral, obtained from a Fourier series through a limiting process consisting of letting the period T approach infinity, so that in essence the excitation ceases to be periodic. We shall discuss this approach in Chap. 11. Another way is to regard the excitation as a superposition of impulses of varying amplitude and time of application. Similarly, the excitation can be represented by a superposition of step functions. We choose to represent the ‘excitation as a series of impulses. Let us consider an arbitrary excitation F(t), such as that depicted in Fig. 229. During the small time increment Ar beginning ata given time t = x, we can regard the function F(0) as consisting of an impulse of magnitude F(c) Ar, as shown by the shaded area in Fig. 2.29. The impulse can be expressed mathematically as AF(1) F(a) Av (2) 2.166) It follows that the function F() can be approximated by a superposition of such impulses as follows: FO =D FO) dcdt—9 (2167)(90 ELEMENTS OF VIBRATION ANALYSIS. Figure 229 where the representation becomes exact as x ~+ 0. But, the response to the impulse described by Eq. (2.166) is Ant) = FOO) Ar oe 9) (2168) so thatthe response to the excitation FC) is x0 =¥ Flatt = 9) Ar 2169) Leting Ar 0, and replacing the summation by integration, we obtain x= fi Flog de *@.70) Which is known as the convolution integral, and expresses the response as @ superposition of impulse responses. . “The impulse response in the integrand in (2.170) is delayed, or shifted, by the time = t. A similar expression can be derived, however, in which the excitation function F(t) is shifted instead of the impulse response. To show this, we let t= 4, —dr = di, Then, considering the limits of integration in (2.170), we ob- serve that when ¢ = 0, = f, whereas when t = ¢, 2 = 0. Inserting these values into (2.170), it is not difficult to show that xo [i re—aoaar am) hich is also referred to as a convolution integral. Recognizing that «in (2.170) and Zin (2.171) are dummy variables of integration, we conclude that the convolution integrals are symmetric in the excitation F(®) and impulse response g(, so that we can write x= f Fedele — 1) dt = fi Fe—natoae ean ‘The question remains as to which function to shift, the excitation F(®) or the impulse response g(t). Logic dictates that the simpler of the two functions be shifted ‘When F(t) is defined for ¢ < 0 the lower limit in the convolution integrals must bbe changed. This case is discussed in Chap. 11 iterFORCED RESPONSE OF SINGLE-DEGRE-OF-FREEDOM LINEAR SYSTEMS 91 ‘The convolution integrals are not always easy to evaluate, and in many cases they must be evaluated numerically ‘The convolution integrals (2.172) can be shown to represent a special case of broader theorem involving two arbitrary functions f(t) and f(t), not necessarily the excitation F(0) and the impulse response g(0). This can be demonstrated very conveniently by means of the Laplace transformation (see App. B, Sec. B.). Example 26 Derive an expression for the response of a mass-damper-spring system to an arbitrary excitation F(t) in terms of the convolution integral ‘Then, consider the undamped case and calculate the response to the one-sided harmonic excitation FQ) = Fosin or a(0) @ by means of the convolution integral, where «(tis the unit step function, ‘The general response ofa linear system can be written in one of the two forms of the convolution integral given by Eq, (2.172), in which F() is any arbitrary excitation and g(t) is the impulse response. In the case of a mass- O and the second i defined for all times. This explains the ference in the two responses. esa os) ) © 2.16 SHOCK SPECTRUM. ‘Many structures are subjected on occasions to relatively large forces applied suddenly and over periods of time that are short relative to the natural period of the structure. Such forces can produce local damage, or they can excite undesirable vibration ofthe structure, Indeed, at times the vibration results in large eyclicstress ‘damaging the structure or impairing its performance. A force of this type has come to be known as a shock. The response of structures to shock is of vital importance in design. The severity of the shock is customarily measured in terms of the maximum. ‘value of the response. For comparison purposes, the response considered is that of ‘an undamped single-degree-of freedom system. The plot of the peak response of a ‘mass-spring system to a given shock as a function of the natural frequency of the system is known as shock spectrum, or response spectrum. 'A shock F() is generally characterized by its maximum value Fo, its duration T, and its shape, of alternatively the impulse Jf F(®) dt. These characteristics depend on the force-producing mechanism and on the properties of the interface material. A reasonable approximation for the force is the half-sine pulse shown in Fig. 230, we propose to derive the associated shock spectrum. ‘The mathematical definition of the half-sine pulse is ro -{Pee O2/0 ‘We must distinguish between the response during the pulse, 0 < ¢ < n/a, and the response subsequent to the termination of the pulse, ¢> n/a. We observe that during the pulse, the halfsine pulse has precisely the same form as the one-sided harmonic excitation of Example 2.6. Moreover, the system considered in that example is the same mast-spring system under consideration here. Hence, theSTA = FORCED RESPONSE OF SINGLE-DBORES-OF-FREEDOM LINFA® SYSTEMS 93, Figure 230 response during the pulse is as given by Eq, (e) of Example 2.6, or fe i sin or sine — 2) de Fo Gay (inet Zeno.) o r/o in Delo ke eae = alle ggg e184 Mi- @ioF 2 @ toe “The response spectrum is simply the plot nu vers [i which both Eqs (2.179) and (2184) rust be considered. Of course, ony the larger ofthe two values Gist be used, We note that for oy =o solution (2179) not valid but foro, > @FORCED RESPONSE OF SINGLE-DEOREE-OF-FREEDOM LINEAR SYSTEMS. 95 Figwe 231 both solutions are valid. It turns out that the maximum response is given by Eq. (2.184) for @, < o and by Eq. (2.179) for «», > co. The response spectrum is shown in Fig. 231 in the form of the nondimensional plot Xu k/Fo Versus c/o. For different pulse shapes, different shock spectra can be anticipated. For a rectangular pulse, or a triangular pulse, the ratio ,/a has no meaning, because there is no a in the definition of these pulses. However, these pulses are defined in terms of their duration 7; so that in these cases the shock spectrum is given by Yas R/Fo Versus T/Tyy OF Xmas k/Fo versus 27/T,, where T, = 2ra, is the natural Period of the mass-spring system, 2.17 SYSTEM RESPONSE BY THE LAPLACE TRANSFORMATION METHOD. TRANSFER FUNCTION ‘The Laplace transformation method has gained wide acceptance as a tool of analysis in the study of linear systems. In addition to providing an efficient method for solving linear differential equations with constant coefficients, the Laplace transformation permits the writing of a simple algebraic expression relating the excitation and the response of systems. In this regard, we are reminded of the ‘operator D used in Sec. 2.11 to demonstrate the principle of superposition for linear systems. However, the operator D was a differential operator that merely permitted writing the relation between the excitation F(@) and the response x(¢) in a compact, form, as well as bringing the inherent characteristics of the system into sharp focus,196. ELEMENTS OF VIBRATION ANALYSIS. but it did not help in any way toward the solution of the response problem. By ‘contrast, the Laplace transformation does provide a method of solution. Signifi- ‘cant advantages of the method are that it can treat discontinuous functions without particular difficulty and that it takes into account initial conditions automatically Sufficient elements of the Laplace transformation to provide us with a working knowledge of the method are presented in App. B. Here we concentrate mainly on using the method to study response problems. "The (one-sided) Laplace transformation of x(®), written symbolically as (6) = x(t) ia defined by the definite integral so)= est) = [7 oso 185) where sis in general a complex quantity referred to asa subsidiary variable, The function e~* is known as the kernel ofthe transformation. Because this i a definite {ntegral, with asthe variable of integration, the transformation yields a function of 5 To solve Eq, (2.123) by the Laplace transformation method itis necessary to evaluate the transforms ofthe derivatives dx/dt and d#x/at. A simple integration by parts leads to ax)” BO gp oayin|” 45 (em 280. [Per B@anenol ss fPerod = sx(9) — x) 2186) ‘where 2(0) is the value of the function x(0) at ¢ = 0. Physically, it represents the initial displacement of the mass m. Similarly, itis not diffeult to show that xl) [Pn HCO) oa ae 3) — sx(0) 80) (2187) were (0) the initial velocity of m. The Laplace transformation of the excitation function is simply FG) = 2FQ)= fg oF dt (2.188) ‘Transforming both sides of Bq, (2.123), and rearranging, we obtain (ms? + 5 + K)X(8) = F(s) + mx(0) + (ms + )x(0) (2.189) In the following discussion, we shall concentrate on the effect of the forcing function, although we could have just as easily regarded the right side of (2.189) as 1a generalized transformed excitation. Hence, ignoring the homogeneous solution, which is equivalent to letting x(0) = (0) =0, we can write the ratio of the transformed excitation to the transformed response in the form Fo) _ ms? +s +k (2.190) 29)FORCED RESPONSE OF SINGLE-DEGREE-OF-FRIEDOM LINEAR SYSTEMS 97 where the function Z(3)is known as the generalized impedance ofthe system. We notice that Zs) contains all the information concerning the system characteristics, in much the same way as D does. By contrast, however, Z(s) is an algebraic expression inthe s domain, namely, a complex plane sometimes referred 10 a8 the Laplace plane, whereas D is a differential operator in the time domain, The reciprocal of Z(s), denoted by 1 Pe) = 705 a9) is called the admittance of the system. The concepts of impedance and admittance were used first in connection with the steady-state response of systems. In the study of systems, we encounter a more general concept relating the transformed response tothe transformed excitation. This general concept is known asthe system function, or transfer function. Fs the special case ofthe second-order system described by Eq. (2123), the transfer function has the form ot 1 Fis) ms? + es +k mis? + 2,5 + a) where and ay are the vsous damping ator and undamped natural equeney of the system. Note that by letting s = iw in G(s) and multiplying by m, we obtain the frequency response G(ia), Eq. (2.46). Equation (2192) can be Fewsen as 33) = G(s)F(s) (2.193) so thatthe transfer funeton can be regarded as an algebraic operator tha operates tn the transformed excitation to yield the wansformed respons. By analogy with the Block diagram of Fig 220, Ea, (2193) ean betepreentd bythe block dagram of Fig 232: In contast o Fig 220, representing a relaton inthe ine domain in terms ofthe diferent operator D, Fig, 232i the Laplace domala in trms of the algbrac operator Gg). There another advanage ofthis ater approach in that it leads fo the solution of the prablem Indeed, to recover the respons x0) Sr ton of 0, defined symbolically by x) = £7) = GNF) (2.194) ‘The operation #~! involves in general a line integral in the complex domain, For our purposes, however, we need not go so deeply into the theory of the Laplace transformation method. Instead we shall look for ways of decomposing 5s) into a combination of functions whose inverse transformations are known. This is done Gs) (2.192) Tanto [ier Transfonmod Ta Le tor Figure 232by the method of partial fractions, as presented in App. B. Also in App. B we discuss ‘a theorem for the inversion of a function 2(5) having the form of a product of two functions of, such as Eq, (2.194). This is Borel’s theorem, which is applicable to the product of any two functions of s, not necessarily G(s) and F(s). ‘Equation (2.194) can be used to derive the response of any linear system with constant coefficients subjected to arbitrary excitation. There are two excitations of particular interest in vibrations, namely, the unit impulse and the unit step function. These functions and the response to these functions were already where the 2 x 1 matrices {x(0)} and (F (0) are the two-dimensional displacement vector and force vector, respectively. Note that rectangular matrices (of which the square matrices are a special case) are denoted by brackets and coluran matrices by braces. In view of Eqs. (3.3) and (3.4), Eqs. (3.2) can be written in the compact 6210. ELEMENTS OF VIBRATION ANALYSIS matrix form [m}{ 200} + Cex} + LEX} = FO} Gs) It is easy to see from Eqs. (3.3) that the off-diagonal elements of the matrices [mr], [eh and [satisfy ma=my=0 caeen=—e kak —k GO with the implication that the matrices are symmetric, as expressed by Gr] =[Emy Cel= Ce" I= 0K" en where the superscript T' designates the transpose of the matrix. Moreover, [m] is diagonal, Equation (3.5) represents a set of independe't equations only when alt three matrices (mi, [¢], and [F] are diagonal. “The solution of Eq. (3.5) for any arbitrary foree vector is difficult to obtain. The difficulty can be attributed broadly to the fact that the two equations represented by (35) are not independent. In the remaining part of this chapter, we discuss the free-vibration case, obtained when the forces F(t) (i = 1, 2) are zero, and the case in which the forces (0) are harmonic. The case in which the forces F(t) are afbitrary will be discussed in Chap. 4, when more adequate tools for treating such problems are introduced. 33 FREE VIBRATION OF UNDAMPED SYSTEMS. NATURAL MODES ‘a In the absence of damping and external forces, the general system of Fig. 3.1a ‘reduces to the special case shown in Fig. 3.2, where the latter is recognized as a ‘conservative system because there is no mechanism for dissipating or adding energy. ‘The differential equations of motion for the system of Fig. 3.2 can be obtained directly from Eqs. (3.2) by letting cy = c2=¢3 = 0 and P,(t) = F,(0) =0. The resulting equations are simply malt) + (ka + Kadea(0) = axa (0) + (ks + bade — kaxa(0) = 0 ae makalt) — kaxa(t) + (ka + kaa) = 0 (0 x00 ee SSC : WOE Witter tia‘TWO-DEGREE-OF FREEDOM SYSTEMS 111 which represent two simultaneous homogencous differential equations of second order, Recalling the third of Eqs. (3.3), we recognize that the coefficients kithek. bt+h=ke —ke=ka=ky es) in Eqs. (38) represent the clements of the stiffness matrix [K], so that Eqs. (38) can bbe writen in the form ma(0) + kav) + kamal) maa(0) + kyaxi(®) + Karel) = 0 G.10) Because Eqs. (3.10) are homogeneous, if x(t) and x,(#) represent a solution, then .2x,(0) and ax,(0) also represent a solution where a is an arbitary constant, Hence, the solution of Eqs. (3.10) can only be obtained within a constant scalar multiplier ‘The interest les in exploring the existence of a special type of solution of Eqs. (G.10), namely, one in which the coordinates x,(t) and x) inerease and decrease in the same proportion as time unfolds. We refer to such motion as synchronous. Because in this case the time dependence of x,(0) and x(t) is the same, if synchronous motion is possible, then the ratio x,(0)/x,(t) must be independent of. time, Hence, the type of motion we are seeking is one in which the ratio between the ‘0 displacements remains constant throughout the motion. Referring to the displace- ment pattern as the system configuration, the implication of the preceding nt is that the shape of the system configuration does not change during , but the amplitude of the displacement pattern does. Denoting the time dependence of x(t) and x(t) by (0, the solution sought can be written in the form MO=4SO oO =f Gan where u, and us play the role of constant amplitudes. Introducing Eqs. (3.11) into Eqs. (3.10), we obtain mite, FQ) + (Kast + Raat) f= 0 3.12) matsF(O + (hiats + kaaa)f(O = 0 — For Eqs. (3.12) to possess a solution, we must have FO) haus + rata _ haste + haat co “Fo mt, alts where 2s a real constant because mi, mz, ks» kya, i, and us are all real constants. Hence, synchronous motion is possible, provided the equations FO+4O=0 G14) and (has = Am)uy + yatta yaty + (kaa — Amaieg = G15) possess solutions,112 ELEMENTS OF VIBRATION ANALYSIS It is not difficult to show that 4 must be not only real but also positive. Indeed, letting the solution of Eq. (3.14) have the exponential form S() = Ae" G16) it follows that s must satisfy the equation e420 ein which has two roots, namely, Sola i Gus) so that solution 3.16) becomes fl) = Ase + dae! = A exp [it + Agexp Jit G19) But if is a negative number, the exponents /—it and —a/—it are real ‘quantities, equal in value but opposite in sign It Follows that, as ¥ <0, the first teem of f() tends to infinity and the second tends to 2or0 exponentially. This, howevesis inconsistent with the concept of an oscillatory system, for which the motion can neither reduce to zero nor inceate without bounds. Hence, the powsblity that 2 is negative must be disearded, and the one that 2is positive must ‘be adopted instead. Letting 2 = w, where w is real, Eq. (3.18) yields 20) so that solution (3.19) becomes J) = Ayelst + Ase @21) where Ay and A; are generally complex numbers constant in value. Recognizing that e and e~!" represent complex vectors of unit magnitude and recalling that they are related to the trigonometric functions cos a and sin wt by et = 05 ot + isin ot 22) we conclude that F(0) = (A + Ap) C08 ot + H(Ay — Aa) sin ot 623) so that solution (3.23) is harmonic with the frequency « and represents the only acceptable solution of Eq. (3.14). This implies that if synchronous motion is possible, then the time dependence is harmonic. But f(t) is known to be a real function, so that, introducing the notation Ay+Az= Coos — (Ar ~ As) =Csing G24) solution (3.23) becomes f(0 = Cos (ot ~ 4) 25)‘Wwo-peoner-oF-rnsepou sysreM 113 where C is an arbitrary constant, is the frequency of the narmonic motion, and 6 is its phase angle, all three quantities being the same for both coordinates, x,(¢) and x00. ‘Next we must verify whether 1 =o? is arbitrary, or can take only certain values. The answer to this question lies in Eqs. (3.15). Inserting 2 = @ in Eqs. G.15), we obtain (kas = muy + kata = 0 yaty + (kaa — o%m,)ug = 0 which represent two simultaneous homogeneous algebraic equations in the unknowns 1; and us, with @? playing the role of a parameter. The problem of dotermining the values of the parameter « for which Eqs. 3.26) admit nontrivial solutions is known as the characteristic-value problem, or the eigenoalue problem. From linear algebra, Eqs. (3.26) possess a solution only ifthe determinant of the coefficients of uy and u, is zero, or aerma [em Ma]ee mn where A(o*), known as the chakacteristic determinant, is a polynomial of second degree in «*, Indeed, expanding Eq. (3.27), we can write Ao) = mymzo0* — (mykaa + maky,)0* + kyskea — Kz =0 (3.28) which represents a quadratic equation in «* called the characteristic equation, ot frequency equation. The equation has the roots 826) of Ama + mak 1 [maka taka? hak ~ Aa Cena a) ee so that there are only two modes in which synchronous motion is possible, one characterized by the frequency «, and the other by the frequency @, where a, and ‘@, are known as the natural frequencies of the system. It remains to determine the values of the constants u, and u;. These values depend on the natural frequencies c», and o>. We denote the values corresponding, to. by uy, and ws, and those corresponding to 2 by uy and ws2. Hence, the first ‘subscript identities the position of the masses and the second subscript indicates Whether the synchronous motion has the frequency «, oF 2. As pointed out carlier, because the problem is homogeneous, only the ratios usy/ix and tsa/ts2 ‘can be determined uniquely. Indeed, inserting a} and o into Eqs. (3.26), we can write simply 8.300) 8308)1114 ELENENTS OF VIBRATION ANALYSIS ‘The implication is that the two expressions for the ratio u,/u given in Eq. (3300) are equal, and a similar statement can be made concerning usz/u,2 in Eq. (3.306). = ‘The ratios usy/t; and uy2/t,2 determine the shape assumed by the system during synchronous motion with frequencies a, and «2, respectively. If one element in each ratio is assigned a certain arbitrary value, then the value of the other element follows automatically. The resulting pairs of numbers, u;; and uz, on the one hand fand uy, and s,; on the other hand, are known as the natural modes of vibration of the system, The modes can be represented by vectors and exhibited in the form of tou ES} ea fl} _ where {u}, and {u}2 are referred to as modal vectors. The natural frequency «and the modal vector {u}, constitute what is known in a broad sense as the first mode of vibration, and «; and {u}, constitute the second mode of vibration. We note that for 1a two-degree-oF freedom system there are two modes of vibration. We shall see in Chap. 4 that the number of modes coincides with the number of degrees of freedom. ‘The natural modes of vibration, ie, the natural frequencies and the modal vectors, represent a property of the system, and they are unique for a given system except for the magnitude of the modal vectors, implying that the mode shape is unique, Dut the amplitude is not. Indeed, because the problem is homogeneous, a modal vector multiplied by a constant scalar represents the same modal vector. Iti often convenient to render a modal vector unique by assigning a given value either to one of the components of the modal vector or to the magnitude of the modal vector. ‘This process is known as normalization and the resulting vector is said to represent normal mode. Clearly, normalization is arbitrary and it does not affect the mode shape, as all the components of the normalized vector are changed in the same proportion. ‘The motion in time is obtained by recalling Eqs. (3.11) and (3.25). Hence, the two possible synchronous motions can be written in the simple vector form {x0} = (us fll) = Cx{uhs 608 (xt ~ $1) {x}z = (W]e = Cafu} 008 (wat — $2) ‘where we note that /(0) and f®) represent the solution (3.25) corresponding to the first and second mode, respectively. We show in Sec. 35 that the motion of the system at any time can be obtained as a superposition of the two natural modes, namely, 32) {x0} =O} + HO} = Cr{u}s 008 (ost ~ ds) + Ca{u}s c0s (wt — 2) (3.33) ‘The amplitudes C, and C3 and the phase angles @, and ¢ are determined by the initial displacements and initial velocities of the masses m, and ma, Ttis convenient to arrange the modal vectors {u}, and {u}, ina square matrix‘WO-DEGREE-OF FREEDOM S¥STINGS 11S ofthe form (= Clu}. wal = ie e| G34) where [1] is known as the modal matrix. Moreover, introducing the vector _ [AO _ [C1 608 (ot — 4) 01 ffoh {eestor an} 2 Eq, (3.33) can be written in the compact matrix form {2} = BLO} 636) Example 3.1 Consider the system of Fig. 3.2, let my =m, mz = 2m, ky = ky =k, ks = 2k and obtain the natural modes of vibration. ‘Using Eqs. (3.9), we obtain the elements of the stifiness matrix [] in the form kiehithe% kpekth=3k k= so that, using Eq. (3.28), we obtain the frequency equation A(o*) = 2mto* — Tk? + 5k? = 0 o which has the roots Iy= —k (a) of _f7, ff _s]k el Oey . so thatthe natural frequencies are : Introducing ar} and a into Eqs. (3.30), we obtain the ratios 2k = (Kfm)m wh oim 1 © 2k — (Sk/2miym oe 05 so that the natural modes are wi={} m={_sh = where the constants uj, and uz were taken as unity arbitrarily. This clearly116 sevens OF vinRATION ANALYSIS Figure 33, does not affect the mode shapes which are plotted in Fig. 33. We note that the second mode possesses a point of zero displacement. Such a point is called a node. 34 COORDINATE TRANSFORMATIONS. COUPLING ‘The system of Fig. 34a can be regarded as an idealized mathematical niodel of an automobile. For simplicity, the body is represented by a rigid slab of total mass m with its mass center C at distances a and b from the springs k, and ka, respectively, ‘where the springs simulate the suspension, The body has a mass moment of inertia "eaten pestion o Figure 34‘Wo-DeGREE-OF-FREEDOM SYSTENS 117 {Ic about the center C. Note that the mass of the ties is assumed to be negligible and ther stipes included inthe stfiness ofthe suspension. Figure 346 shows @ free-body diagram corresponding to the body in dsplaced position, where the displacements consist of the vertical translation x(@) of the center C and the rotation 6(0) about C. The translation x(?) is measured from the equilibrium Position, so that the weight W = mg ofthe automobiles balanced by correspond- ing initial compressive forces in the springs (se Sec. 1.3). The angular displacement 4% is assumed to be smal. ‘There are two equations of motion, a force equation for the translation in the vertical direction and a moment equation forthe rotation about the mass center: ‘To derive the equations of motion, we refer to the fee-body diagram of Fig, 3.46 and considera diferential clement of mass div at a distance & from the mass center C. Then, observing that for small anges 0 the acceleration of the mass element is in the vertical direction and equal to # + 2, the force equation becomes —he— a — G6 = f (+ 0) dm f an sof Edm= me 37a) J Se where m= fray dm is the total mass of the body. Moreover, we note that the simplification onthe right side of Ea. (3.374) was possible becaise Ju & dm = O by the definition of the mass center. Similarly, the moment equation about C reduces to aC — aB)a— ka + BO) = {., SE + 8) dm nsf cane 0 Bama it am where Ic = froay £7 dim is the mass moment of inertia of the body about the mass center, Equations (237) can be rearanged as me (bs + ka — (hua — ho = 0 1b = (esa hab + (yo? + 690 = 0 leading to the matrix form [old +Lcin eee Hef} om Nex conser pint O sch hat en eral fe appa 0 the ap untae ann ca ee Pe 250 te Rae Sisto ant Bow espn bs nese ey, Dg te ei enon of un Done ve ancl ieee ian) and) who) Soc ats outer et 838)118 ELEMENTS OF viBRATION ANALYSIS fates +40) Bulb pain Figure 35 body diagram is now as shown in Fig. 3.5b, where m is the distance from 0 to dm. Using the same approach as before, the equations of motion are simply kyr = 238) — kato + B16) = ji + nb) dm tear =x [am +f dm =mes +e ow kyr = @y8)ay — kao, + bi 8)b1 = ie (1 + 06) dm off nama Fd = mess + Tol where ¢ = (1/m) funy dm isthe distance from O to C and Io = fouy 1? dm is the ‘mass moment of inertia of the body about point 0. Recalling that kya; = kabs, Eas. (3.40) reduce to i, + mel + (ky + ads = 0 at mes + Lol + (kya + kxbD0 = 0 io which can be written in the matrix form (mH SY natalfe- foTWO-DEGREE-OF-FREEDOM SYSTEMS. 119 Examining Eq. 3.39), we conclude that by using as coordinates the translation x() ofthe mass center C and the rotation 6() about C, the equations of motion are coupled through the stiffness terms. Such coupling is referred to as elastic coupling. On the other hand, the equations of motion described by (3.42) are coupled through the mass terms. Such coupling is called inertial coupling. The preceding statements demonstrate thatthe nature of coupling depends only on the choice of coordinates, ic, on how we describe the system mathematically, rather than on the system itself Indeed, we have the freedom to deseribe the motion of the system in terms of any pair of independent coordinates, and the interest les in a pair of coordinates offering the greatest simplification. In particular, the most desirable system of coordinates is that for which the equations of motion are uncoupled both elastically and inertially, ic, for which the mass and stifiness matries are both diagonal. We show in Sec. 35 that such coordinates do indeed exist, and they are kknovin as natural coordinates, or principal coordinates. The general motion of an tundamped linear system can be represented asa linear combination of the natural ‘modes multiplied by these natural coordinates, 35 ORTHOGONALITY OF MODES. NATURAL COORDINATES ‘The modal vectors {u}, and {u}z possess a very yseful property known as orthogonality. We propose to demonstrate first the orthogonality property and then to show how it can be used to uncouple the equations of motion, thus enabling us to solve the equations with the same ease as solving the equations of single- degree-of-freedom systems. Considering Eqs. (3.30), we can write the modal vectors as follows: {wha =a {wha = mie where o} and «are given by Eg. (3.29). Next, form the matrix product 1 7 1 (aFtato) wt |b =m [o ]) =e = sata, BE +B ke = fm kay — oma] G44) Which isa scalar. Inserting «} and 0 from Ea. (3.29) into Eq. (3.44), we conclude that {u}ELm]{up. = 0 (345)120, ELEMENTS OF VIBRATION ANALYSIS so that the modal vectors {u}, and {u}, are orthogonal. Because the matrix product in Eq. G45) contains the mass matrix [ml] as a weighting matrix, this is not an = ordinary orthogonality but orthogonality with respect to [m]. (Next, rewrite Eqs. (3.26) in the matrix form Cau) = o°Dmi a} G49), But, both modes w}, {u}. and a, {u} must satisfy Eq, (3.46), so that we can write Cus = offs (3.47a) Tee} = Dm} Garb) “Multiplying Eq, (3474) on the left by {u}} and considering Eq, (3.45), we obtain (HEC) = 0 G48) so that the modal vectors {ui}, and {us are orthogonal with respect to the stiffness matrix (K] as well. Of course, because the matrices [m] and [k] are symmetric, Eqs. (345) and (348) are valid also when the positions of {u}, and {u}2 are Jnterchanged. This statement can be verified by taking the transpose of Eqs. (3.45) and (3.48), and recalling thatthe transpose of a product of matrices is equal to the product of the transposed matrices in reversed order. It is worth noting that, ‘multiplying Eq. (3.47a) by {u)f and Eq. (2.476) by {u}f, we ean write {u}FUR{ue = oP{uhFCml{upi i= 1,2 G49) where a, (i = 1,2) are the natural frequencies of the system. Ae ‘The orthogonality property can be used to uncouple the equations of motion both elastically and inertally. To justify this statement, we first rewsite the equations of motion, Eqs. (310), in the matrix form [me] (2(} + De{x(0)} = {0} 3.50) ‘Then, we seek a solution of Eq, (3.50) as the linear combination x0} = (aha. + (2420 G51) where qu(t) and q(t) are two functions of time that remain to be determined. Introducing Eg. (3.51) into Eq, (3.50), we obtain Crk thd + {uada(O) + E(u a0 + (whoa) = {0} G52) “Multiplying Eq, (3.52) on the left by {u}7 and considering tne orthogonality of {4} and {u}z with respect to [7] and [k], as well as Eqs. (3.49), we conclude that 4y(0) must satisfy the equation HO + Og, =0 8.534) Similarly, multiplying Eq. (3.52) on the left by {u} and considering Eqs. (3.45), (848), and (3.49), we conclude that q(t) satisfies the equation HO + O40 =0 (3.538)"TWO-DEGREE-OF-FREEDOM SYSTEMS 121 We observe that Eqs. (3.53) represent two independent equations, one for gi) and one for qa(t). This is in contrast with Eqs. (3.10) for x,(#) and xa(t), which are simultaneous equations. Hence, the linear transformation (3.51), when used in conjunction with the orthogonality of the modal vectors, does indeed permit simultaneous elastic and inertial uncoupling of the equations of motion. Coordi- nates such as 4,(t) and g3(0), for which the system of equations of motion are independent, are called natural coordinates, or principal coordinates. Equations (3.53) describe two independent harmonic oscillators (Sec. 1.6, ‘Their solutions are simply 4x0 = Cx £08 (ort — $1) (3.54a) (0) = Cz 008 (wat ~ $2) (3.548) where Cand ;(/= 1,2) are amplitudes and phase angles, respectively, so that Eq. G51) becomes, (210) = Cr{u}s 608 (ove ~ 41) + Cafu} 005 (03842) 55) ‘The amplitudes C, and C, and the phase angles, and ¢ depend on the initial displacements and velocities of m, and ms. Equation (3.55) identical to Ea. (3.33), thus justifying the statement made in Sec. 3.3 that the motion ofthe system at any time can be expressed as a superposition ofthe natural modes of vibration multiplied by the natural coordinates, ‘The orthogonality of the modal vectors and the process of uncoupling the equations of motion will be preseated in a more formal manner in Chap. 4 Example 32 Consider the automobile of Sec. 34, let the system parameters have the values m= 100 Ib-s'/f, Le = 16001b-s?-ft, ky = 2400 Ibjft, ka = 2700 Ibft, a= 4.40 f, b = 5.60, calculate the natural modes of the system and write an expression for the response. To determine the natural coordinates, itis necessary to find the natural ‘modes first. Inserting the values ofthe parameters given into the equations of, motion, Eq. (3.39), we obtain 100 0 Vf}, F100 4560-1) _ fo 0 1600 }af * [4560 131,136} Lef ~ to o leading to the eigenvalue problem 100 0 T{x) F100 4560 T4x} _ f0 -o[D sealfet [ee stelle tt © where X and © are the amplitudes of x(¢) and 0(), respectively. Hence, the system characteristic equation is a $100 — 100c* 4560 4560 131,136 — 160007, = 160,000(w* — 132.9607 + 4050.00) =0 — (c)122 FLEWENTS OF VIBRATION ANALYSIS having the solutions Oh 65.48 = /C6ARF = MORO = 6648 $1922 _ $47.26 (rad/s) = {ecm so that the natural frequencies are «= 6,88 rads and «; = 9.26 rad/s. Inserting «? from Eqs. (d) into the first row of Eq. (b), we obtain =47.26 x 100X, + 5100X, + 4560, = 0 yielding e374 X, 4560 Moreover, introducing 0? from Eqs (d) into the fist row of Eq, (8), we have 85.10 x 100x2 + 5100X; + 45600, = 0 0.0820 rad © from which we obtain ©: _ 3470 Six Bagg 7 07610 dM wo Hence, letting arbitrarily X, =I and X= 1, the natural modes become tf} -{ ota} LE -fomeh [Note that the same results would have been obtained had we used the second, row of Eq. (6) instead of the first. The modes are plotted in Fig. 3.6. yh alt 2926 els Figure 36‘TWO DEGREE-OF-FREEDOM SYSTEMS 123 ‘The natural modes can be verified to be orthogonal with respect to both the mass matrix and the stiffaess matrix. The orthogonality property can be used to uncouple the equations of motion and to determine the natural coordinates, as outlined in Eas. (3.50)-(3.54). It is not really necessary to carry ‘out these steps-Instead, we can insert the computed modal vectors, Eqs. (9), directly into Eq. (3.55) and obtain the system response Gi) ng mea vo ‘The cocflicients C, and Cz and the phase angles ¢ and $2 depend on the initial conditions x(0), (0), (0), and 6(0). Their calculation is shown in Sec. 36. 1 ero 025 — 63) o 3.6 RESPONSE OF A TWO-DEGREE-OF-FREEDOM SYSTEM TO INITIAL EXCITATION It was indicated in Sec. 3.5 that the motion of a system at any time can be regarded ‘as the superposition of the natural modes multiplied by the natural coordinates. More specifically, the motion of a two-degree-of-freedom system can be written in the vector form (3.55). The natural frequencies «; and @; are unique for a given system and the modal vectors {u}, and {u} can be rendered unique through normalization. On the other hand, the amplitudes C, and C; and the phase angles 1 and 4: play the role of constants of integration and their values depend on the initial conditions. Letting the initial conditions have the values x,(0) = X10» X2(0) = X20, 21(0) = B10, ¥2(0) = v2o, and inserting these values into Eq. G55) with r= 0, we obtain X19 = Citas 008 by + Cats cos bz sap = City 608 + Cota 608 bs _ 4 656) Yo = Crews, Sin; + Cxwauy2 sin by t20 = Crores sin bs + Cxoatza sin ds which can be regarded as two pairs of algebraic equations, the first pair in the unknowns C, c0s $4, C2 60s 2 and the second pair in the unknowns C, sin 1, Cz sin $2. Equations (3.56) have the solution Cy008 4; (usaxi0 — tiaXa0) 1 eta 1 lini — taxied det fu] 57)124 seus OF vineATION ANALYSIS 1 @ det fa) 1 Casin ba = aarp ssta0 — Mantio) ising, (iaabvo — 2020) . From Eqs, (3.57), we obtain 5 — Con ggg | teat naeent (Cite = Ha uaao— maya)? + He Ha) ta) ca 58) aa CTE) $2 ta sluunta0 ati) Equations (3.55) and (3.58) define the response of a two-degree-of-freedom system to initial excitation completely. We shall see in Chap. 4 that the response can be ‘obtained in matrix form in a more systematic way. Example 33 Consider the system of Example 3.1 and obtain the response to the initial excitation x,(0) 12, x2(0) = x20 = 0, 40) = 049 = 0, 5,10) = v0 = 0. From_Eqs. (@) of Example 31, we have a= Jim, a= 1.5811 /kjm. Moreover, choosing arbitrarily uy, = 1,us2 = 1 in Example 3.1, we obtained the modal matrix oa-(% S)-[t os] ) which has the determinant 4 det =| og © the initial conditions listed above and the values given by Eqs. (a) 10 Eqs. (3.58), we obtain Insert © Get fu] det [a] o o oe“TWO DEREE-OF-FREEDOM SYSTENS 125 Hence, introducing Eqs. (c) into Eq, (2.55), we obtain the response {thf} svn hr It must be pointed out that the arbitrary choice u11 = 1, u12 = 1 did not affect the final outcome. Indeed, any other choice would have resulted in such values for C and C; as to keep Eq. (d) unchanged. @ 3.7 BEAT PHENOMENON A very interesting phenomenon is encountered when the natural frequencies of a two-degree-oFfreedom system are very close in value. To illustrate the phenom- ‘enon, let us consider two identical pendulums connected by a spring, as shown in Fig. 37a. The corresponding free-body diagrams are shown in Fig. 3.7b, in which the assumption of small angles 8, and 02 is implied. The moment equal the points O and 0’, respectively, yield the differential equations of motion 136), + mgL8, + ka*(@, ~ 83) = 0 mL38, + mgLO, ~ ka(0, ~ 83) = 0 which can be arranged in the matrix form mL? 0 {0 mg + ka* ka? Oy (0) [Pe aueltet [Mt were tf 9 Indicating thatthe system is coupled clastcally. As expected, when the spring stfes Eredces o ero the coupling disappears andthe two pendulums reduce to independent simple pendulums with identical natural frequencies equal to Vail. For k #0, Eq, (3.60) yields the eigenvalue problem fmt? 0 10x) fmol + kat kat 1f0,) _ fo oo mit}le: ka? mgl + ka? {LO3f ~ Lol 3.61)126 enews OF VIBRATION ANALYSIS leading to the characteristic equation ‘mgL + ka? — mL? ka? ee at tot = (gL + ka? — mL")? — (ka)? =0 (3.62) which is equivalent to gL + ka® — @?mL? = +ka? (3.63) Fence, the two natural frequencies are A fone a o & oe [Poe (3.64) ‘The natural modes are gbtained from the equations afml? 0 fOr) , fmol + kat —ka? Vf, —*L ome? } 10.5: —ka mg + ka?) 10,5, (6s) Inserting w} = g/L and a} = g/L. + 2(k/m)(a*/L?) into Eqs. (3.65), and solving for the ratios @,,/O,; and @:2/@,2, we obtain en, On Si ec 366) so that in the first natural mode the two pendulums move like a single pendulum with the spring k unstretched, which cam also be concluded from the fact that the first natural frequency of the system is that of the simple pendulum, « = /a/L. (On the other hand, in the second natural mode the two pendulums are 180° out of phase, The two modes are shown in Fig. 3.8. ‘As was pointed out in Sec. 3.5, the general motion of the system can be expressed as a superposition of the two natural modes multiplied by the associated“TWO-DEOREE-OF-FREEDOM SYSTEMS. 127 natural coordinates, or 2.(0| es Choosing ©, scalar form {20} = 6,{8}} costo 6) + efi} eovwour—o) 069 12 = 1 and using Eqs. (3.66), Eqs. (3.67) can be rewritten in the O,(8) = Cy £08 (wst ~ 1) + Cz 008 (34 ~ $3) ca 0x(2) = Cx 008 (wt ~ $1) ~ Cz 608 (wat ~ #2) Letting the initial conditions be 8,(0) = become Jo, 03(0) = 04(0) = 03(0) = 0, Eqs. (3.68) 0,(0) = Ho 008 wt + 4 008 «at o oto: = tg cos 2S 1 ‘ G6) 83(2) = 3 605 at — 40 cos ast gin 2 OH = Oy sin 2% sin PEO ‘Note that, in deriving Eqs. (3.69), we used the trigonometric relations ¢os (« + 8) = 008 « 00s fF sin asin B, in which a = (0, ~ @;)1/2, B= (2 + @)4/2 ‘Next let us consider the case in which ka? is very small in value compared with ‘mgL. Examining Eq. (3.60), we conclude that this statement is equivalent to saying that the coupling provided by the spring kis very weak. In this case, Eqs, (3.69) can bbe writen in the form O,(0) & Og 008 H9t C08 dy 040) & 09 sin opt sin Oyet Where 0/2 and Oy ae approximated by 3.70) Oy _or- or 1k a’ 2 m Jal m Jal @n Hence, (2) and 63(0) can be regarded as being harmonic functions with frequency re and with amplitudes varying slowly according to 8 cos fogt and 4 sin fog, respectively. The plots ,(0) versus ¢ and 04(0) versus fare shown in Fig. 39, with the slowly varying amplitudes indicated. by the dashed-line envelopes. Geo- metrically, Fig. 39a (or Fig. 396) implies that if two harmonic functions possessing equal amplitudes and nearly equal frequencies are added, then the resulting function isan amplitude-modulated harmonic function with a frequency equal to the average frequency. At fist, when the two harmonic waves reinforce each other, the amplitude is doubled, and later, asthe two waves cancel each other,128 ELEMENTS OF VIBRATION ANALYSSS Figwe 39 the amplitude reduces to zero, The phenomenon is known as the beat phenomenon, land the frequency of modulation «,, which in this particular case is equal to -ka?/m./gL°, is called the beat frequency. From Fig. 39a, we conclude that the time between two maxima is 7/2= 2s/oo,, whereas the period of the amplitude- modulated envelope is T = 4x/op, ‘Although in our particular case the beat phenomenon resulted from the weak ‘coupling of two pendulums, the phenomenon is not exclusively associated with two-degree-olsfreedom systems. Indeed, the beat phenomenon is purely the result of adding two harmonic functions of equal amplitudes and nearly equal frequencies. For example, the phenomenon occurs in twin-engine propeller airplanes, in which the propeller noise grows and diminishes in intensity as the sound waves generated by the two propellers reinforce and cancel each other in turn, ‘We observe from Fig. 39 that there is a 90° phase angle between 0,0) and 84(0). ‘At t=O the first pendulum (right pendulum in Fig. 3.7a) begins to swing with the amplitude @> while the second pendulum is at rest, Soon thereafter the second pendulum is entrained, gaining amplitude while the amplitude of the frst deereases. ‘At t; = n/p the amplitude of the first pendulum becomes zero, whereas the ‘amplitude of the second pendulum reaches 09. At t2 = 2/a¥p the amplitude of the“TWo-DEGREE-OF-FREEDOM SYSTEMS 129 (2) Pare talon (0) Poe tion Figue 410 first pendulum reaches @ once again and that of the second pendulum reduces to zzr0. The motion Keeps repeating itself s0 that every interval of time 7/4 = moon there isa complete transfer of eneray from one pendulum to the other ‘Another example of a system exhibiting the beat phenomenon is the “Wilberforce spring’, consisting of a mass of finite dimensions suspended by a helical spring such thatthe frequency of vertical translation and the frequency of torsional motion are very close in value. In this ease, the kinetic energy’ changes from pure translational in the vertical direction to pure rotational about the vertical axis, as showin in Fig, 3.10. 38 RESPONSE OF A TWO-DEGREE-OF-FREEDOM SYSTEM TO HARMONIC EXCITATION Let us return to the damped system of Sec. 3.2 and write Eq. (3.5) in the expanded form mary + mas + ork + crake + huts + hive = Fi) 72) Imyas + mana + eink + caake + Ria + havea = Fal) ‘where the diagonal mass matrix has been replaced by a more general nondiagonal but symmetric matrix. Next, let us consider the following harmonic excitation: FiQ= Fie Fy(0) = Fre an) and write the steady-state response as xa = Xe" x(t) = Xzc" G74) where X, and X2 are in general complex quantities depending on the driving frequency «and the system parameters. Inserting Eqs. (3.73) and (3.74) into (3.72), we obtain the two algebraic equations (omy, + fees + kay)X1 + (om, + wera + kiadX: 15) (07m: + loers + kiz)X1 + (ema: + locas + kX: Introducing the notation Zylo) = —oFmy + lacy thy j= 1,2 676)130 ELEMENTS OF VIBRATION ANALYSIS where the functions Zw) are known as impedances, Eqs. (3:75) can be written in the compact matrix form [z@KX} =1F} em where [Z(a)] is called the impedance matrix, (X) isthe column matrix of the Uisplacement amplitudes, and {F} is the column matrix of the excitation amplitudes. “The solution of Fa. (3.77 can be obtained by premltiplying both sides ofthe equation by the invers [Z(o)]~* ofthe impedance matrix [Z(«)], withthe result {X} = [Z@T MF} (3.78) where the inverse [Z(c)]~* can be shown to have the form (see App. C) ae) eo) ZO = scrzion [se Z3(@) di Zia) —Z,(0) 79) role in| &? Introducing Bg, (3.79) into (78), and performing the multiplication, we can write Zao)F 1 ~ Zil0Fs =Z(0)Fs + Zula)Fs Zl@Zaso)~ 70) Zs(oVFaal@) ~ ZO) (80) ‘and we note that the functions X,(o) and X,(w) are analogous to the frequency response of Sec. 23. ‘Next let us confine ourselves to the undamped system of Fig. 32. Moreover, let Fy = 0, so that Eqs. (3.76) veld Zulo)=ky-o'm — Zx(0) kaa —0%m— Zxl0)= ks O80 Introducing Eqs. (381) into (380), we obtain (haz = @?m)F 00) om Maa = 0m) = He Xo) = Xx0) oe Gs) %10) = amas — Fm) — A Fora given set of system parameters, Eqs. (3.82) can be used to plot X,(o) versus @ and X,(«) versus @, thus obtaining the magnitude of the response for any excita- tion frequency wo Example 34 Let us consider the system of Example 3. and plot the frequency-response curves. ‘Using the parameter values of Example 3.1, Eqs. (3.82) become (Bk = 2ma?)Fy key X10) = S35 tea se | XO) = Sa — Tmko® + SEel “TER atk aA | : anal t 1 sue) I 1 fF feed see But the denominator of X, and X2 is recognized as the characteristic determinant, which can be written as (0?) = Dnt — Tmkar? +5 = Imo? — eto?) () where A ope 5k © the form @ ‘The frequency response curves X,() versus oon, and Xq(a) versus fen, are plotted in Fig. 3.11. 39 UNDAMPED VIBRATION ABSORBERS When rotating machinery operates at a constant frequency close to resonance, violent vibrations are induced. Assuming that the system can be represented by @ single-degree-ol-freedom system subjected (0 harmonic excitation, the situation ‘may be alleviated by changing either the mass or the spring. At times, however, this132. ELEMENTS OF VIBRATION ANALYSIS is not possible. In such a cas, «second mass and spring can be added tothe system, where the added mass and spring are so designed as to produce a two-degree-of- Fteedom system whose frequency response is zero atthe excitation frequency. We note from Fig. 31a that a point for which the frequency response is zero does exist. The new two-degree-ofrcedom system has two resonant frequencies, but these frequencies generally present no problem because they differ from the operating frequency. Tet us consider the system of Fig. 3.12, where the original single-degree-of- freedom system, referred to as the main system, consists ofthe mass my and the spring k,and the added sytem, referred to asthe absorber, consists ofthe mass ms and the spring ky. The equations of motion ofthe combined system can be shown tobe amass + (hs had — has = Fe sin 8 3 (8 maz — Kam + bax; Letting the solution of Eqs. (3.83) be xQ=Xisinot x0) = Xzsin or 84) wwe obtain two algebraic equations for X, and Xz having the matrix form Pe i Following the pattern of Sec. 38, the solution of Eq. (3.85) can be shown to. be een X= Gre ome — 07m) — 8 a G86) % Fi i ¥ ka om lla = OF ma) — TTT Za Figure 312“TWO-DPOREEOF-AREEDOS SYSTEMS. 133 a @, = x/ky/m; = the natural frequency of the main system alone 0, = [ig] = the natural frequency of the absorber alone sa Fy the sate defection of th mai system w 12/m, = the ratio of the absorber mass to the main mass With this notation, Eqs. (3.86) can be rewritten as : U-(@/o,*, [1 + mleog/0rg)? — (cofea,)? TEL — (coforg)*] — plcrg/,)? From Eq, (3874), we conclude that for @, =i, the amplitude X, of the main ‘mass reduces to zero, Hence, the absorber can indeed perform the task for which it is designed, namely, to eliminate the vibration of the main mass, provided the natural frequency of the absorber is the same as the operating frequency of the machinery. Moreover, for « = a, Eq. (3.87b) reduces to x= (2) 8-8 3.88) 40 that, inserting Eq, (3.88) into the second of Eqs. (3.84), we obtain x (387) x (3878) n= —FEsinot G89) from which we conclude that the force in the absorber spring at any time is kyl) = —F isin ot 890) Hence, the absorber exerts on the main mass a force —F; sin wt which balances ‘exactly. the applied force F, sin a. Because the same effect is obtained by any absorber provided its natural frequency is equal to the operating frequency, there is ‘a wide choice of absorber parameters. The actual choice is generally dictated by limitations placed on the amplitude X; of the absorber motion. Although a vibration absorber is designed for a given operating frequency ©», the absorber can perform satisfactorily for operating frequencies that vary slightly from @. In this case, the motion of m; is not zero, but its amplitude is very small. This statement can be verified by using Eq. (3874) and plotting Xy(o)/xa versus ‘fay, Figure 3.13 shows such a plot for = 0.2 and @, = @,, The shaded area indicates the domain in which the performance of the absorber can be regarded as satisfactory. As pointed out earlier, one disadvantage of the vibration absorber is that two new resonant frequencies are created, as can be seen from Fig. 3.13. To reduce the amplitude at the resonant frequencies, damping can be added, but this results in an inerease in amplitude in the neighborhood of the operating frequency134 sLEMeTs OF VIBRATION ANALYSIS TAM t { t | t { + i | I tT | i i \ | jj | | ft) Soe Figure 313 = @. It should be recalled that any rotating machinery builds up its frequency from rest, o that the system is likely to go through the first resonant frequency. As 4a matter of interest, the plot X/x Versus «ay, corresponding to the main system alone is also shown in Fig. 3.13 PROBLEMS [31 The system of Fig, 3.14 consists of two point masses my and my caried by a weights string “bjs tothe constant tention T, Assume small transverse diplacments y(t) and y() and decive the diferente eqntions of motion. yA yao i qo Rt ure“TWo-DEGREE-OF-FREEDOM SYSTEMS 135 32 Two diss of mass polar moments of inertia, and Ts are mounted on 2 circular massless shaft ‘consting of two segments of torsional stifness GJ, and Gls, respectively (ee Fig 313). Derive the Aierential equations for the rotational motion of the disk. Figure 315 133 Arig bar of mass pr unitength m caries x pont mass M a its right end. The bare supported by ‘wo springs at shown io Fig. 3.16: Derive the deren equation forthe transaton of pot and rotation about A. Assume small motions WD TE he nt gue 816 44 Derive the ferential equation of motion forthe double pendulum shown in Fig. 3.17. The anh 1 and dean be arbitcarily lage AS Deve the ferential equations of motion forthe system shown in Fig. 3.8, Lat the angle 0 be small.136. ELEMENTS OF VIBRATION ANALYSIS hy id by “NE 219, e Figure 3.18 16 Constr the system of Fig. 319 and show that it can be reduced to a single-deges-otfreedom System of equivalent mass may mma/(y + ms) Note that an unrestrained sytem such as this is Known asa semidefite system (te Sex 410) Figere 319 147 The system of Fig 3.20 represents an sso section being tested in a wind tunnel Lethe ii have total mass m and mass moment of inertia le about the mass enter C, and derive the diferental uations of motion. Lillies eis i Few 3 '38 A uniform tia sd is suspended by a string, a shown in Fig. 221. Derive the diferetia equations ‘toto of the system for aesteariy large angles. Dz Figure 321"TWO-DEGREE-OF- FREEDOM SYSTEMS. 137 39 A igdbarof mass per uni length en) = gy( + n/L) supported by two springs, as shown in Fig 4422, sume small motions and deive the diferent equations of motion. rind meal) cal Figure 322 {410 Figure 223 depicts « tworstory building, Assume the horizontal members to be rigid and the ‘lum tobe males and derive the diferent equations for the horizootal translation of the masses. t 7D. 7 gure 323 ‘311 A sig uniform bare supported by two translational springs and one torsion Derive the dittrential equations of motion, ering (Fig. 328) : Figure 3.24