17

Applications of Structural and Dynamic Principles

17.1 17.2 17.3 17.4 Base Conguration Loaded Applications Structural Conguration Loaded Applications Additional Information Base Motion Excitation for Multiple D.O.F. Problems
Direct Solution Modal Solution Integration of Uncoupled Equations

Anthony J. Kalinowski
Naval Undersea Warfare Center

In this chapter, we consider some practical applications of dynamics and vibrations, with specic emphasis on shock isolation and vibration isolation. In both of these isolation situations, we are concerned with the transmission of interaction forces, u(t), existing between two congurations [which are referred to here as the base conguration (BC) and the structural conguration (SC)], as illustrated in Figure 17.1. This example system is a rst-order representation of an idealized physical system that is general enough to represent both shock and vibration design situations. Most of the underlying physical principles impacting the design of either kind of isolation can be explained and illustrated with this simple one-degree-of-freedom model. The governing equations of motion can be generated from the three-part sequence shown in Figure 17.1. In Figure 17.1(a), the BC and SC are in the unloaded condition, and the two congurations are initially separated by a length L. This corresponds to the state where the model is lying horizontal relative to the vertical direction of the gravity eld, or where the model is in the vertical position before the gravity eld (acceleration of gravity g = 386 in./sec2) is taken into account. For vertically oriented models, it is convenient to write the equation of motion relative to an initially gravity-loaded model, as in Figure 17.1(b), where the SC is shown at rest in a compressed state (the linear isolator has a preloaded compressive force of uT = -Kds, where ds = Mg/K is dened positively as the static deection). Next, Figure 17.1(c) corresponds to a body in motion, and the governing equations of motion are obtained by constructing a free-body diagram of the SC and equating the sum of all vertical forces acting on the body to its mass, M, times its acceleration, d 2 z / dt 2 z (where dot notation is used to refer to time differentiation from here on), resulting in the following dynamic equilibrium equation: + u T (x, x ) + Mg - F(t ) = 0 Mz The relation between the total z displacement and the isolator stretch variable x is given by (17.1)

2005 by CRC Press LLC

body in motion structural configuration uT Mg uT isolator L base configuration uT L s +z

F(t) Mz L + (x s)

(a) isolator alone

y(0)

(b) mass attached

y(0)

(c) free body

y(t)

FIGURE 17.1 Multipurpose single-degree-of-freedom system.

z = y + x + (L - d s ) and it follows that = x + y z

(17.2)

(17.3)

) , could be a nonlinear function of the relative displacement x In general, the isolator force, u T (x, x ; however, for the purposes of this introductory development, only linear isolators and relative velocity x will be considered. Thus, the total isolator force acting on the mass M in Figure 17.1(c) is given by ) = Kx + Cx - Kd s u(x, x ) - Kd s u T (x, x (17.4)

) is the dynamic portion of the isolator force not including the static deection force. where u(x, x Substituting Equation (17.2) through Equation (17.4) into Equation (17.1) gives + Cx + Kx = F(t ) - My (t ) Mx subject to initial conditions x(t = 0) = x 0 , (t = 0) = x 0 x (17.6) (17.5)

It is noted that the Mg term and -Kds cancel in the formation of Equation (17.5). The shock and vibration will take place about the static equilibrium position shown in Figure 17.1(b); = 0), the isolator is still compressed an that is, when the vibrating body comes to rest (i.e., x = 0, x amount equal to the static deection ds. The general solution to Equation (17.5) subject to initial conditions [Equation (17.6)] can be obtained by several different methods; however, the method of Laplace transforms is used here because it readily applies to situations where: 1. The boundary conditions are of the initial-value type. 2. The right-hand side is an arbitrary function of time. 3. An equivalence between an impulse-loaded right-hand side (rapidly applied loading) and a suddenly applied initial-velocity problem with no right-hand side can be easily illustrated. Thus, taking the Laplace transform of Equation (17.5) with respect to the Laplace transform variable s results in

2005 by CRC Press LLC

x(s) =

0 (M ) x 0 (sM + C ) x (s) F(s) - My + + (s 2 M + sC + K ) (s 2 M + sC + K ) (s 2 M + sC + K )

(17.7)

Upon taking the inverse transform, this leads to the general solution for displacement and velocity: + hx 0 x x(t ) = e - ht x 0 cos(w dt ) + 0 sin(w dt ) wd (l)]dl e - h(t -l ) sin[w d (t - l)][F(l) - My + Mw d l=0

(17.8)

l =t

(t ) = - hx(t ) + e - ht [- x 0 w d sin(w dt ) + (x 0 + hx 0 )cos(w dt )] x +

(l)]dl e - h(t -l ) cos[w d (t - l)][F(l) - My M l=0

l =t

(17.9)

In the applications to follow, it will be more convenient to work with the variables w n K / M = 2pf n , corresponding to the isolator natural frequency, and z C/(4pMfn), corresponding to the critical damping ratio. The variables wd and h appearing in Equation (17.8) and Equation (17.9) are called the damped natural frequency and decay constant, respectively, and can be expressed in terms of variables wn and z using w d = w n 1 - z 2 and h = wnz. These new variables have the following physical meanings: wn corresponds to the free harmonic vibration (no driver present) of the isolator in the absence of damping; wd corresponds to the damped free harmonic vibration; z determines whether the system is underdamped (z < 1) or overdamped (z > 1), where in the former case the free motion oscillates harmonically with damped natural frequency wd and in the latter case the free motion does not vibrate harmonically; and nally, h corresponds to the rate at which the underdamped system exponentially decays in time. The solutions represented by Equation (17.8) and Equation (17.9) will be used to evaluate the dynamic responses in all the example problems to follow. The forms of the solutions are general and apply to (t )] are given as an analytical expression or when they are either the situation where the drivers [F(t ), y given as a digital representation (e.g., earthquake responses). In the case of digital driver representations, the integrations in Equation (17.8) and Equation (17.9) can easily be performed by numerical integration (e.g., Simpsons rule), and in the case of analytical driver representations, closed-form integrals can be obtained with the aid of integral tables or with the aid of symbolic evaluation packages such as Maple [Redfern, 1994], MATLABs version of Maple [Sigmon, 1994], and Mathematica [Wolfram, 1991]. A computer program using MATLAB script language [Math Works, 1992] was used to generate the results presented here. Given the displacement and velocity versus time from Equation (17.8) and Equation (17.9), back(t ) into Equation (17.4) gives the dynamic portion of the isolator force u(x, x ), substituting x(t) and x which serves as the major ingredient for isolating the structural conguration from the base conguration. of the SC can be obtained by substitution of u T (x, x ) into Equation (17.1). We The total acceleration z will refer to the single-degree-of-freedom model in Figure 17.1(c) for the example problems considered here, and the physical signicance of the ingredients of the model will be different for each usage. We will consider two types of applications: Base conguration loaded: Here the base conguration has a prescribed displacement motion time ) is the basic input load, and the history, y(t) (and therefore the base acceleration d 2 y / dt 2 y structural conguration has a zero-value external forcing function [F(t) = 0]. Some examples of this type of problem are:

2005 by CRC Press LLC

A. Earthquake-resistant structures, where the ground (base conguration) movement from fault slip motion excites buildings or bridges (structural conguration) B. Vehicle suspension, where the ground displacement road prole [y(x) vs. distance, x, in the forward direction of the vehicle traveling at constant velocity V as it moves parallel to the ground] results in the ground acting as the equivalent base conguration, with prescribed displacement road prole shape y(x) = y(Vt) and corresponding apparent base acceleration d2y/ dt2 = V2d2y/dx2 measured perpendicular to the ground, and the vehicle car body (structural conguration) responds to these irregularities in the road C. Electronic component isolation, where the ship superstructure (base conguration) cabinet houses electronic components mounted in cabinets (structural conguration) that are subject (based on preto waterborne shock waves that impart known base motion accelerations, y viously measured experimental data) Structural conguration loaded: Here the structural conguration has a directly applied force time history, F(t), and the base conguration has no motion (i.e., y(t) = 0.0). Some examples of this type of problem are: A. Unbalanced rotating mass, where an eccentric mass me with offset re is rotating at constant angular speed w (simple motor model with an unbalanced offset mass me) and is mounted inside the motor housing (structural conguration) and the base conguration is taken as xed (y(t) = 0), where the explicit harmonic forcing function is F(t) = merew2 cos(wt); and B. Free-falling mass, where a free-falling mass (structural conguration) is prevented from impacting a rigid surface (base conguration) by having an intermediate shock isolator break the fall of the object dropped from height H (e.g., a stunt motorcycle jumps off a ramp or a package containing fragile equipment drops), where the force on the structural conguration is F = -mg for t 0 and F = 0 and t < 0, with initial condition dx / dt = - 2gH at t = 0 and x = 0 at t = 0.

17.1 Base Conguration Loaded Applications

Problem 1: Vehicle Suspension In this example, a vehicle moving with horizontal constant velocity V passes over a roadway that has a ground prole of y(x) = Y0 sin(kx), 0 x NcL, k = 2p/L, where L is the period of the ground swell. Since the relation between horizontal distance traveled and time is given by x = Vt, the vertical BC base motion can be rewritten as y(t ) = Y 0 sin(wt ), (t ) = -w 2Y 0 sin(wt ), y w = 2pV / L (17.10)

(t ) = 0 over the rest of the time duration. If the vehicle applied over the range 0 t NcL/V and y is four-wheeled, then a higher-degree-of-freedom model is needed to represent the response due to the rotational degrees of freedom; therefore, it is assumed that the vehicle is two-wheeled and is being towed while the load is balanced over the axle. Assume that all other forces acting are negligible; therefore, F(t) = 0.0. Consider a vehicle weighing W = 2000 lb and traveling at a speed of V = 60 mph that encounters a two-cycle (Nc = 2) road swell of amplitude Y0 = -1.01 in. and period L = 20 max = A0 = 2 ft. Substituting the data into Equation (17.10) results in a peak base acceleration of y gs and a drive frequency of f = w/2p = 4.4 Hz. Design a spring-damper isolator (i.e., nd K and C) that limits the vehicle (structural conguration) steady state vibration acceleration transmission / y max to a value of 0.4 and limits of maximum relative displacement of the ratio, TA = peak z isolator x max = 3 4 ds . Before solving for K and C, it must be noted that isolator manufacturers often use a terminology other than explicitly stating these constants. Typically, in place of K and C, a natural frequency (fn) versus load (Mg) curve is supplied and the acceleration transmission ratio TA is given at resonance,
2005 by CRC Press LLC

2.5 Transmission Ratio; Peak U/Fmax -or-Zdd/Yddmax

<--f/fn = 1.41 defines isolation zone

2 = 0.00 = 0.30 = 0.50 = 1.00 1.5 Isolation Region -->

<-- Amplification Region 0.5

0.5

1.5 2 2.5 f/fn (drive frequency)/(natural frequency)

3.5

FIGURE 17.2 Transmission for harmonic structure force or base acceleration.

as in Barry, 1993. Later, it will be shown how K and C can be back-calculated from fn and TA. For the present, however, attention is focused on nding the desired fn as the rst step. The key ingredient in this approach is to use the transmission ratio versus drive frequency (b f/ fn) curve, as shown in Figure 17.2. This cur ve is obtained by substituting a solution of the form x(t ) = A sin(wt ) + B cos(wt ) into Equation (17.5), and solving for the A, B constants, resulting in the following acceleration transmission ratio: 1 + (2zb)2 , (1 - b 2 )2 + (2zb)2

TA =

b w /w n = f / f n

(17.11)

It is of interest to note that all the curves in Figure 17.2 pass through the same frequency ratio, b = 2 , and this special value forms the dividing line between isolation and amplication. Therefore, as a design strategy, to get isolation of the base conguration from the structural conguration, the isolator K is selected such that its natural frequency fn results in b values > 2 . Another response ratio that applies when the relative displacements are of particular interest is the displacement magnication ratio, denoted as TD = peakx/ ymax, which can be derived in a manner similar to the above relation, resulting in the expression TD = b2 (1 - b 2 )2 + (2zb)2 (17.12)

2005 by CRC Press LLC

3 = 0.00 = 0.30 = 0.50 = 1.00

2.5

XXX

1.5

0.5

0.5

1.5 2 2.5 3 3.5 f/fn (drive frequency)/(natural frequency)

4.5

FIGURE 17.3 Magnication for harmonic base acceleration.

This is shown in Figure 17.3 plotted against the drive frequency parameter b. At this point, the designer must decide whether to solve for the natural frequency fn that gives the desired transmission ratio TA, and live with the resulting displacement, or solve for the natural frequency fn that gives the displacement magnication ratio TD, and live with the resulting maximum . It is noted that solving for b is equivalent to solving for fn, since the drive frequency, acceleration z f, is known. In this sample problem, it is decided that reducing peak acceleration has priority; therefore, with the aid of Equation (17.11), we solve for the b that gives the desired TA. Thus,
2 2 2 2 2 2 2 (a + 2TA - aTA ) + (a + 2TA - aTA ) - 4TA (TA - 1) 2 2TA

b=

a = (2z)2

(17.13)

where for light damping (z < 0.05), b can be approximated with b 1 + 1 / TA (17.14)

Before solving for b using Equation (17.13) [or Equation (17.14)], you must select a critical damping ratio z. At this point, a z value can simply be selected according to whether light damping or heavy damping is desired, and a moderately heavy value of z = 0.5 is chosen in this example. Therefore, for the
2005 by CRC Press LLC

problem at hand, substituting z = 0.50 and TA = 0.40 into Equation (17.13) gives b = 2.8131. Upon substituting this value of b along with the drive frequency, f = 4.4 Hz, into the b denition [i.e., the second equation in Equation (17.11)], we can solve for the natural frequency, fn = 1.564 Hz, required to limit the acceleration transmission to TA = 0.40. As the nal step, the actual spring constant K and damping constant C can easily be back-calculated from the fn and z values in the following manner. In the sample problem, the 2000 lb load is divided equally over two isolators, so each one must carry a mass of M = 2000 / (2 386) = 2.5906 lb-in./sec2. Therefore, K = (2pfn)2M = 250.2 lb/in., and C = 4pMzfn = 25.46 lb-sec/in. for each of the two isolators. The dynamic response for (t ), y (t ), and z (t ) / y max (i.e., the SC transient acceleration transimportant variables such as x(t), x mission ratio) is computed by evaluating Equation (17.8), Equation (17.9), and Equation (17.1) over the two cycles of input and for an equally long coastdown time duration after the road prole has become at again. Upon observing the Figure 17.4 solution, it is observed that the deection stays within the maximum allowable constraint space of x con = 0.75 d s = 3.00 in.; however, the (t ) / y max = 0.62 overshoots the target steady state transient portion of the peak acceleration ratio z solution. TA = 0.40. The isolator design employed the steady state solution; therefore, it is not unusual that the transient could exceed the steady state limit. It is further noted that by the end of the second response cycle, the acceleration ratio is already approaching the 0.40 target. When the road prole turns at, the acceleration ratio rapidly tails off toward zero, due to the large damping value. To compensate for the overshoot in transient acceleration, a larger b (and thus a smaller fn) can be used on a second pass through the design process [e.g., enter TA = 0.40 (0.40 / 0.62) into Equation (t ) / y max is lowered from 0.62 to 0.45]. (17.13) and repeat the design process, where the peak z
1.5 Xmax = 2.69 in. Vmax = 38.87 in/sec Yddmax = 2.00 gs Peak Zdd/Ydd = 0.62 1 upper X constraint

Dynamic Response Variable

0.5

X(t)/Xmax V(t)/Vmax Zdd(t)/Yddmax Ydd(t)/Yddmax Xcon/Xmax

M 0.5 K y(x) 1 lower X constraint 1.5

F(t) = 0 C V

0.1

0.2

0.3 0.4 0.5 0.6 0.7 time (sec); (fn = 1.56 Hz, Xstatic = 3.997 in., = 0.50)

0.8

0.9

FIGURE 17.4 Roadway base acceleration transmission to structure.

2005 by CRC Press LLC

In building an isolator from scratch, one can simply design it to have the physical properties K and C directly. However, if isolators are to be selected from off-the-shelf stock, a few comments are in order regarding the selection process. Manufacturers often supply a set of load (Mg) versus fn curves for each isolator in a class of isolators. Upon entering such a set with a 1000 lb load, the isolator having a natural frequency nearest the target value, say fn = 1.564 Hz, is selected. Typically there will not be an isolator corresponding to the exact fn desired at the operating load. Thus, picking an isolator with fn on the low side will make the steady state transmission on the low side. In some cases, the manufacturer must be contacted directly to get damping data, and in other cases damping information is given indirectly via a stated transmissibility, TA, at resonance. The b location of the peak resonance TA value in Figure 17.2 in terms of the damping ratio z is given by b2 = 1 (-1 + 1 + 8z 2 ) 4z 2

Substituting this expression into Equation (17.11) results in a relationship between TA and z as plotted in Figure 17.5. For design purposes, the exact curve in inverse form can be approximated by the simpler expression
2 z res 0.5 1/(TA - 1)

(17.15)

102

<-- approaches infinity as -->0.0 Peak Transmission (Ta or Tu) at Damped Natural Frequency (f = fd)

101 XXX

Exact Sol. Approx Sol.

100 102

101 = (Critical Damping Ratio, C/4Mfn)

100

FIGURE 17.5 Transmission at resonance (harmonic force or base acceleration).

2005 by CRC Press LLC

In Figure 17.5, both the exact curve and the Equation (17.15) approximation are plotted side by side. For example, if the manufacturer states that an isolator has a transmission ratio TA = 1.5 at resonance, then substituting this value into Equation (17.15) gives a corresponding critical damping ratio of z = 0.45. Finally, substituting z = 0.45 into C = 4pMzfn provides the numerical value for the isolator damping constant in question. Problem 2: Shock Isolation of Fragile Equipment (t = 0) = 0], that is Consider a 50 lb structural conguration, initially at rest [i.e., x(t = 0) = 0, x subject to a half-sine pulse-type BC acceleration input of the form = A0 sin(wt ), y = 0.0, y 0 t TP ; TP = p /w

(17.16)

TP < t ;

and specically the peak value of the base acceleration input is A0 = 16 gs and the pulse duration is TP = 11 msec. In this problem, the direct force F(t) is taken as zero. The design problem is to synthesize a shock isolator that will limit the acceleration transmitted to the SC to 4 gs; therefore, a target transmission ratio of TA = 4/16 = 0.25 is sought. It is also desired that the maximum displacement be limited to x = 1.0 in. At this point it is noted upon comparing the base motion of the previous application [Equation (17.10)] to the base motion of the current problem [Equation (17.16)] that there is a similarity between the rst and second problems, except for the fact that the current shock problem has a short pulse length where only one-half of a sine wave (Nc = 0.5) is applied. The principle governing shock isolation is different from the corresponding vibration isolation of the previous problem in that the shock isolation process is characterized as a storage device for a sharply increasing acceleration waveform, and the design concept is to attempt to instantaneously absorb the energy and then release it at the natural frequency of the device but at a lower-level mass deceleration. One approach is, for guessed K and C values, to simply substitute these values and the input waveform of Equation (17.16) into Equation (17.7), Equation (17.8), and Equation (17.1) to get the response motion, observe the response, and reiterate the process with new K and C values until a desired response is obtained. We cannot use the steady state method of the previous problem because the steady state assumption of the input waveform would not be valid. An alternative design process is to convert the base input motion into a nearly (t = 0) = V0 , with no explicit driver on the equivalent initial, suddenly applied velocity problem, x right-hand side of Equation (17.5). The extreme case of a rapidly applied loading is a special mathematical function called a delta function, d(t), whose value is at t = 0 and zero for t > 0. The right-hand-side loading can then be represented as d(t ) (t ) - MA - My where = A (17.17)

l = TP

(l)dl y

(17.18)

l=0

Upon substituting Equation (17.17) into Laplace-transformed Equation (17.7), with initial displacement x0 = 0 and F(s) = 0, and noting that the Laplace transform of a delta function is 1.0, it can be have exactly the same form (same denominator); therefore, 0 M term and - MA seen that the x by interchanging the roles of the driver and initial condition, we can let a problem with a zero initial

2005 by CRC Press LLC

 [by Equation (l7.18)] be replaced with an equivalent driver having area A velocity and pulse-type y . The V quantity can driver but an initial velocity of x 0 V0 = - A problem having a zero y 0 therefore be interpreted as a suddenly applied velocity change. The advantage of this approach is that estimates of the maximum response can easily be made and used to back-calculate the isolator properties needed to achieve the desired isolator performance. Thus, for the equivalent initial velocity representation, we immediately get the solution as a special case of Equation (17.8) that simply reduces to - ht sin(w t )/w x(t ) = - Ae d d (17.19)

Upon differentiating Equation (17.19) and solving for the maximum displacement xmax and maxi max , the following result is obtained: mum mass acceleration z - a /w , x max = - Ae n where a = (z / 1 - z 2 )sin -1( 1 - z 2 ) w D max = A z n i (1 - 2z 2 ) with Di = exp(-z t / 1 - z 2 ) sin(t) + 2z cos(t) 1- z2 Di = 2z, 0.5 < z 1.0 (17.20) (17.21)

0 z 0.5

where

(1 - 4z 2 ) 1 - z 2 t = tan -1 z(3 - 4z 2 )

/ y max . As in the previous example, the acceleration transmission ratio is dened as TA = peak z Substituting Equation (17.21) into this TA expression and solving for the natural frequency gives max TA y 2p AD
i

fn =

(general shock)

(17.22)

and, for very light damping (say, z < 0.1), the approximation Di 1.0 can be used. The isolator selection follows the same concept as in the previous problem where the natural frequency needed to limit the transmissibility is determined. For the problem at hand, substituting the given half-sine = 2A /w = 2A T /p, where into Equation (17.18) corresponds to a velocity change of A input y 0 0 P max = A0 . Upon using these data, Equation (17.21) reduces to it is also noted that y fn = TA (for half-sine pulse shock) 4TPDi (17.23)

From this point forward, the selection procedure for obtaining the isolator parameters is conceptually the same, except that the computation of the desired natural frequency is different and is governed by Equation (17.22) or Equation (17.23) as appropriate. For the problem at hand, substituting the target transmissibility TA = 0.25, pulse length TP = 0.011 sec, and designer-selected damping ratio z = 0.2 (using z = 0.2 in the Di expression of Equation (17.21) gives Di = 0.8209) results in a desired natural frequency of fn = 6.92 Hz. For the 50 lb SC (M = 50/386), the fn = 6.92
2005 by CRC Press LLC

and z = 0.2 data translate into spring and damping constants of K = (2pfn)2M = 244.97 lb/in. and C = 4pMzfn = 2.253 lb-sec/in. Also, using Equation (17.20), the estimated maximum deection is xmax = 0.7520 in.; therefore, this isolator design should meet the space constraint imposed on the problem. This xmax value should also be checked against the isolator manufacturers maximum allowable spring deection (sometimes called sway space), which is usually given in the selection catalog. Upon substituting the above K, M, and C design parameters and actual half-sine base (t ) into Equation (17.8), Equation (17.9), and Equation (17.1), the dynamic response for motion y (t ), y (t ) , and z (t ) / y max (i.e., the SC transient acceleration important variables such as x(t), x transmission ratio) is computed. The results are shown in Figure 17.6, where it is noted that the desired 0.25 transmissibility is achieved and the displacement constraints are not exceeded. It should be noted that the curves labeled in the legend correspond to using the actual half-sine input base motion. These results will not be exactly the same as the equivalent suddenly applied velocity (V0) solution that was used to size the isolators, because the half-sine pulse is not exactly the idealized delta function. For illustrative purposes, the equivalent suddenly applied velocity solution is superimposed on the same plot and denoted with the unconnected symbolic markers * for displacement, o for velocity, and + for acceleration ratio. As expected, the velocity comparison will be (t ) input the initial condition is x 0 = 0 , whereas different in the early time [e.g., with the actual y in the equivalent problem the initial velocity is not zero and represents the entire input to the problem]. As a nal comment, the selection of the isolators from the manufacturers catalog follows along the same lines here and therefore will not be repeated; also check that TP << 1/fn when using Equation (17.22) or Equation (17.23).
1 upper X constraint Xmax = 0.75 in. Vmax = 38.25 in/sec Yddmax = 16.00 gs Peak Zdd/Ydd = 0.25 + + + + + + + + + + + + 0.5 + + + + + + + + + + + + +

0.5 + + Dynamic Response Variable + 0 + + + + + + + +

+ +

+ +

+ + X(t)/Xmax V(t)/Vmax Zdd(t)/Yddmax Ydd(t)/Yddmax Xcon/Xmax M K F(t) = 0 C V y(t)

+ 1 +

lower X constraint

1.5

0.05

0.1

0.15

0.2

0.25

time (sec); (fn = 6.92 Hz, Xstatic = 0.204 in., = 0.20)

FIGURE 17.6 Base conguration half-sine acceleration loading.

2005 by CRC Press LLC

17.2 Structural Conguration Loaded Applications

Problem 3: Rotating Machinery Force Transmission (t ) = y (t ) = 0 , and In this class of problems, the base conguration is considered xed; thus y(t ) = y therefore the only system loading comes through the structural conguration loading F(t). Perhaps one of the most common such loading problems is the situation where a piece of equipment with some sort of rotating member is spinning in a steady state mode at drive frequency w and is resting on the structural conguration, where the total mass M of the equipment to be supported by the isolators is M = Ms + me, where Ms is the mass of the structural conguration including the rotating machinery (except for the offset mass me), and me represents the off-center eccentric mass at radius re. The me term is analogous to wet clothes clinging to the spinning drum during the spin-dry cycle of a common washing machine, where excessive vibrations are set up when the clothes are not uniformly distributed around the drum. The radial acceleration, Ar = rew2, results in a reciprocating force that is represented by F(t ) = m ere w 2 cos(wt ) (17.24)

Therefore, by comparing Equation (17.24) to the rst steady state loading example in Equation (17.10), it is seen that the Mw2Y0 sin(wt) driver in the differential equation is just like the current merew2 cos(wt) driver except for a cosine in place of a sine driver function. In fact, the steady state solutions for the transmissibility of an isolator force transmission ratio of TU = peaku(t)/Fmax has the exact same form as Equation (17.11); therefore, TU = TA for this class of problem, where the driver force amplitude varies as w2. It is cautioned that for other harmonically varying forces, ones that have, say, a frequency-independent amplitude, the form of the TU would not be the same as Equation (17.11). It is also noted that although the actual value of the amplitude of F(t) depends on the size of mere, this value cancels out in forming the TU force transmission ratio. As a specic example, consider a structural conguration, whose total weight (including me) is 800 lb, and the equipment is rotating at 540 rpm (i.e., w0 = 2p9 rad/sec), which due to an offset (re = 10.0 in.), an eccentric rotating mass (me = 0.25 lb-sec2/in.), transmits unwanted vibrations to the oor. It is required that the force transmission ratio be no more than TU = 0.2 (sometimes equivalently referred to as 80% isolation). It is also required that the maximum displacement x(t) not exceed 0.5 in. It is further assumed that the SC is to be supported by four isolators and that they are centered so that the dead weight is distributed equally among them. If this centering assumption is not met, rocking modes will be present and more degrees of freedom will be needed to model the system. The motor cannot go from w = 0 up to the operating frequency instantly; therefore, it is assumed that w has a simple linear time ramp: w = w 0t /t c , 0 t tc; w = w 0, tc t (17.25)

where for the problem at hand tc = 10 (2p/w0), which corresponds to 10 periods of the steady state frequency. The rst step in designing the isolator is to select a critical damping ratio z. Because the motor has a variable speed and the rotational frequency varies with time during the startup according to Equation (17.25), the isolator is expected to have a designer-selected amplication of no more than TU = 10 at resonance, should the motor ever be run at or near the resonant frequency. Thus, substituting TU = TA = 10 into Equation (17.15) results in a damping ratio of z = 0.0502. The next step is to solve for the natural frequency that will limit the steady state vibration force transmission ratio (operating at speed f = w0/2p = 9 Hz) to a value of TU = 0.2. This is accomplished by substituting z = 0.0502 and TU = TA = 0.2 into Equation (17.13), which results in b = 2.4792.

2005 by CRC Press LLC

1.5 Xmax = 0.42 in. Vmax = 20.08 in/sec Fmax = 414.2 lbs Peak U/F = 0.28 1 X(t)/Xmax U(t)/Fmax F(t)/Fmax Tu Limit

Dynamic Response Variable

0.5 upper Tu transmission limit

0.5

lower Tu transmission limit F(t) = FoCos(t) me MS

1 K C y(t) = 0

1.5

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

time (sec); (fn = 3.63 Hz, Xstatic = 0.742 in., = 0.05)

FIGURE 17.7 Transient-into-steady-state force transmission applied to foundation.

There are four isolators; therefore, the mass assigned to each isolator is determined from M = (800 / 386)/4 = 0.5181 lb-sec2/in. The natural frequency is computed from fn = f/ b = 9/2.4792 = 3.63 Hz. Finally, the spring and damping constants are computed from K = (2pfn)2M = 269.5 lb/in. and C = 4pMzfn = 1.186 lb-sec/in. (Refer to the discussion on selecting isolators from manufacturers catalogs in the rst suspension isolator design example.) Upon substituting the above K, M, C design parameters and driver F(t) dened by Equation (17.24) and Equation (17.25) into Equation (17.8), Equation (t ) , F(t), (17.9), and Equation (17.4), the dynamic response for important variables such as x(t), x and u(t)/Fmax (i.e., the SC force transmission ratio) is computed. It is noted that during the startup phase, the variable angular velocity w(t) results in an additional tangential acceleration component that is neglected in the analysis. The transient portion of the results is shown in Figure 17.7, where it is noted that the desired 0.20 transmissibility is exceeded (TA = 0.28) during the transient (and also that the displacement constraints are not exceeded). The damping is light in this example; therefore, it takes many more cycles to reach the steady state. The horizontal dotted lines in Figure 17.7 align with the steady state limit reached for u(t)/ Fmax after running the solution out to t = 4.0. This demonstration example also illustrates the need to balance rotating machinery. Instead of having to live with the vibrations, it would be better to kill the source of the vibration and rebalance the equipment (e.g., make re smaller in this example), so that even if the vibration is still present, its magnitude will be small enough that it will have the same effect as if isolators were used to reduce the transmissibility.

2005 by CRC Press LLC

Problem 4: Free-Fall Shock This class of problem is similar to the baseexcited impulse problem, where a falling object (structural conguration) strikes the ground (base conguration) and, due to the impact, a suddenly applied velocity is imparted to the falling object. The ground is idealized as rigid, and it is assumed that the object falls from a height H. The velocity of the object just prior to hitting the ground is V = 2gH . The problem is to design an isolator that limits the maximum force transmissibility TU to a prescribed amount (often referred to as the fragility factor, nondimensional gs). The solution of this problem follows along the same lines as the initial impulse problem described earlier in this section; therefore, only a rough outline will be presented. The governing differential equations have to be modied slightly to account for the fact that the free-falling bodyisolator conguration has no prestretch (static deection ds), and therefore the static deection force term Kds does not cancel with the dead weight term in the derivation of Equation (17.5). Therefore, in solving this problem, x(t) in Equation (17.5) should be viewed as the deection measured from the unstretched equilibrium position. Once the base of the isolator has just touched the ground at impact, the isolator starts to compress and the x(t) solution describes the ensuing motion of the SC. The loading in Equation (t ) = 0 ) and the initial (17.5) becomes simply the dead weight (i.e., F(t) = -Mg for t 0, with y 0 = -V = - 2gH [the sign convention is that the conditions in Equation (17.6) are x0 = 0, x extension (+x) of the isolator is positive]. Thus, with these conditions substituted into Equation (17.8), Equation (17.9), and Equation (17.4), the dynamic response for important variables such (t ) , F(t), and TU = u(t) / Fmax (i.e., the SC force transmission ratio) can be evaluated, as x(t), x where Fmax in this application is simply the dead weight, -Mg. Solving for the exact solution for u(t) as described above, and nding its maximum value in time, one obtains (-W(1 - 2z 2 ) - z) TU = -1 + exp(-z t / 1 - z 2 ) sin(t) + (1 - 2Wz)cos(t) 1- z2 or TU = -2zW with 1 + 0.5 (1 + 1/(2W)2 , 4W
-1

0 z zb

(17.26a)

zb z

(17.26b)

zb =

W=

Vw n , g (17.26c)

(W(1 - 4z 2 ) + 2z) 1 - z 2 t = tan 2 3 (-1 + 3Wz + 2z - 4Wz )

For intermediate damping, Equation (17.26a) corresponds to the nondimensional time value, t = w dt max , where the slope du/ dt = 0.0, and Equation (17.26b) corresponds to the large damping case where the maximum force occurs at the beginning of impact, t = 0.0. When the damping z is zero (or very small), then using Equation (17.26a) with t tan -1(W /(-1)) and solving for W in terms of TU, we can solve for the natural frequency that limits the maximum transmitted force to the desired TU value, where fn
2005 by CRC Press LLC

g T (T - 2) 2pV U U

for light damping, z 0.0, and TU > 2.0

(17.27)

The TU > 2.0 limitation comes from the fact that with zero or low damping, fragility factors lower than 2.0 are not reachable without substantial damping. In the case of heavy damping, the maximum force occurs at the beginning, t = 0.0, and using Equation (17.26b), one obtains: fn gTU 4pVz for heavy damping, z b z (17.28)

where the breakpoint zb is determined by setting the numerator of the arctan expression for t equal to zero. Finally the determination of the fn value in terms of TU for the intermediate damping, 0 z zb, is the most difcult case because the solution involves obtaining roots to a transcendental equation. It is noted that for large values W 10.0, the breakpoint value is zb 0.5. Use the smallest positive root for tan-1 in Equation (17.26c) (e.g., atan2 in MATLAB). The procedure is numerical and can easily be accomplished by substituting the desired damping ratio z and fragility factor TU into Equation (17.26a) and iterating W until the left side equals the right side (this root, W = Wrt, can be found with a simple numerical root nder routine such as the fzero routine in MATLAB). Next, simply convert Wrt to fn using the second of Equations (17.26c), to get f n = gW rt /(2pV ) for intermedia te damping, 0 z z b (17.29)

To illustrate the general drop application for intermediate damping, consider a 50 lb weight (M = 50/g) that is dropped a height H = 6.00 in. (i.e., V = 68.06 in./sec), where it is required that TU be no bigger than 8.0, with a prescribed intermediate damping ratio z = 0.2. Next insert these data into Equation (17.26a), solve for the root Wrt = 8.870, and nally compute the isolator natural frequency fn = 8.0066 Hz with Equation (17.29). The breakpoint value zb must be checked to ensure that the inequality bounds of Equation (17.29) are not violated; thus, substituting W = Wrt = 8.870 into the rst of Equations (17.26c) gives zb = 0.5290, which is above the z = 0.2 value required by the constraint bounds for intermediate damping. These fn and z data translate into K = (2pfn)2M = 327.82 lb/in. and C = 4pMzfn = 2.607 lb-sec/in. Evaluating the dynamic response as described above with Equation (17.8), Equation (17.9), and Equation (17.4) results in the response illustrated in Figure 17.8 which, as indicated, has peaku(t)/ Fmax = 8.0. Note that the isolator force turns into tension after 0.062 sec, which implies that the structural conguration will jump up off the oor (i.e., pogo stick effect) at a later time after the energy is absorbed. It is also noted that the x(t) solution does not settle down to zero, but rather to the static deection value; this is because Figure 17.1(a) rather than Figure 17.1(b) was the coordinate reference conguration for this freefall problem. For comparison purposes, the solution generated with Equation (17.8), Equation (17.9), and Equation (17.4) is compared with a solution to the same problem in Church, 1963 and is indicated by the unconnected symbolic markers * for displacement, o for velocity, and + for the u(t) / Fmax ratio (after adjusting for a different sign convention for positive x). As can be seen, the agreement between the two solutions is perfect.

17.3 Additional Information

The sample problems considered here represent simple single-degree-of-freedom (SDF) models of what is often a more complicated multiple-degree-of-freedom (MDF) system. These simple example problems illustrate the main concepts involved in shock and vibration isolation; however, great care must be taken not to apply these idealized SDF-type models in situations where a MDF model is needed to represent the full picture (e.g., the four-wheel vehicle over a roadway needs a MDF model that allows the front and rear wheels to experience different parts of the roadway prole to allow for any rocking modes that may be present). Perhaps one of the most comprehensive references for shock and vibration isolation
2005 by CRC Press LLC

10 upper Tu transmission limit 8 Xmax = 1.14 in. Vmax = 68.06 in/sec 6 Fmax = 50.0 lbs Peak U/F = 8.00 X(t)/Xmax V(t)/Vmax U(t)/Fmax F(t)/Fmax Tu Limit

4 Dynamic Response Variable + 2 + + +

0+ + 2 + + 6 8 10 +

+ +

+ +

+ +

+ +

+ +

+ 4 M V K

F(t) = Mg incipient impact C y(t) = 0

lower Tu transmission limit +

0.05

0.1 0.15 0.2 time (sec); (fn = 8.01 Hz, Xstatic = 0.153 in., = 0.20)

0.25

FIGURE 17.8 Free-falling structural conguration shock.

issues is Harris, 1988, which not only covers SDF models in detail for all kinds of inputs, but covers MDF models as well. Good design tips are often found directly in the manufacturers design guides, for example, Aeroex, 1994; Barry, 1993; Lord, 1994; and Firestone, 1994. For some more advanced ideas on optimum shock and vibration isolation concepts, see Sevin and Pilkey, 1971. The effects of nonlinearity should also be considered when necessary; for example, the simple Voigt model linear spring-damper isolators considered here eventually turn nonlinear when the deections become large. In such cases the actual ) could be obtained from the manufacturer and Equation (17.1) nonlinear isolator force function u(x, x resolved as a nonlinear differential equation (which could also easily be solved with a MATLAB script le similar to the one used to generate the sample solutions). For the analysis of MDF systems consisting of a collection of rigid bodies interconnected by any Rube Goldberg collection of linkages, pivot points, and so forth, a handy computer program called Working Model exists [Knowledge Revolution, 1994], enabling the user to geometrically construct the conguration to be analyzed on the computer screen. It is not unlike the object-drawing programs that come with practically all modern-day word processors. The key difference is that once the model is drawn on the screen, following the assignment of initial conditions, spring constants, damping coefcients, and loading via pulldown menus, the user is one mouse click away from setting the program to work and getting a graphical display of selected pertinent response variables (displacement, velocity, acceleration) as the solution is in progress. Versions exist for a variety of platforms, including Macintoshes, PCs, and workstations. Depending on the users on-screen drawing skills, an entire analysis of a rather complex MDF system can be completed in a very short time (e.g., 5 to 10 min).

2005 by CRC Press LLC

17.4 Base Motion Excitation for Multiple D.O.F. Problems

In problems 1 and 2, the base excitation was a prescribed vertical motion y(t). However, the model for the physical representation was a simple single D.O.F. For more complex structures, many degrees of freedom will of course be acting, and it is then useful to employ nite element methods to solve the problem. For example, in the Figure 17.6 inset sketch, the base could represent a building foundation, and the structural conguration above it could be a nite element representation of the building structure. The loading could still be the base excitation y(t) and would correspond to the vertical base excitation of, say, some reference earthquake data or to some prescribed shock base motions for shock and vibration testing of mounted equipment. The equations of motion look like Equation (17.5), except that the M, K, and C are now nite element method mass, stiffness, and damping matrices, respectively, and the relative displacement x(t) is replaced by a vector of unknown {Z(t)} total displacements and the y(t) prescribed base motion is replaced by a vector of known driver {Y(t)} nodal displacements. Following Przemieniecki [1969], for a treatment of enforced displacement problems, the full displacement vector is partitioned into an unknown and known partitions, [{Z(t)}, {Y(t)}]T . We start the analysis without damping present and will add its effect on the undamped solution later. Thus the equations of motion become M zz M yz M zy Z K zz + M yy K yz Y K zy Z 0 = K yy F Y y

(17.30)

where Fy corresponds to the yet-unknown forces of constraint causing the known displacements {Y(t)} of the base structure (e.g., this motion could be an earthquake history or a base motion from a shaker table or shock machine). Expanding the top partition, we have: } + [K ]{Z } = {F } [M zz ]{Z zz z where } - [K ]{Y } {Fz } = -[M zy ]{Y zy (17.32) (17.31)

Direct Solution
The direct solution subject to the unknown structure nodal time history {Z(t)} would be to integrate this system of coupled ordinary differential equations with a standard second order in time, integration scheme (see, for example, Bathe [1996] and Cook [1988]), subject to, say, zero displacement and zero velocity initial conditions. Once having the total displacement vector {Z(t)}, the strains and stresses are then post computed. Also, the forces of constraint, {Fz } , are post computed with the lower half of Equation (17.30). In order to account for damping, we simply add a damping term on the l.h.s. of } to the r.h.s. of }, and correspondingly add a term - [C ]{Y Equation (17.31) of the form [C zz ]{Z zy Equation (17.32) and integrate the damped system of differential equations in the same manner.

Modal Solution
The goal of the modal solution is to simplify the solution so that it does not require integrating coupled equations of the entire system equations like that of the above direct approach, but rather obtains a set (one for each mode) of uncoupled second-order differential equations in time, of the form of Equation (17.5) (specically, rewrite Equation (17.5)) by dividing it by M to obtain: + 2w nVx + w2 x n x = F (t ) / M - y
2005 by CRC Press LLC

(17.33)

A second reason is a byproduct of the uncoupled approach and corresponds to the case where the user is given a shock spectrum [Cunniff, 1989; Geers, 1998; Remmers, 1983], which is the response to a single degree of freedom, subject to a base motion {Y(t)} (e.g., like Problem 1 and Problem 2 earlier). With limited space, we can only outline the procedure; the reader is referred to Cook [1988], for detailed derivations of what is called the mode displacement method. The rst step is to temporarily leave off the damping in Equation (17.31) and extract the normal undamped modes of vibration of Equation (17.31), with the r.h.s. set = 0, thus producing I = 1,2,Imax normalized Eigenfunctions {Fi}, stored columnwise in a matrix [F] = [{F1} ,{F2}, {Fi}], and Eigenvalues wI associated with each mode, {Fi}, where Imax is the length of the displacement vector {Z}. The modes can be extracted using any standard nite element computer code with a structural dynamics option such as ABAQUS, or NASTRAN, or for smaller problems using MATLAB, in conjunction with m-le scripts code found in [Hwon, 2000]. Since damping is left out, these are not the actual damped modes of the system (which typically are complex), but rather modes that are used as a change in variable to uncouple the system of coupled equations Equation (17.5). We introduce a change in variable {Z} = [F]{c}, where {c} is dened as the new generalized variable unknown. Next substitute this variable change into the damped version of Equation (17.31) and into the damped version of the driver Equation (17.32). The new differential equations will now be with respect to {c} and its derivatives on time, but so far the equations will still be coupled. The usual step for uncoupling the equations is to premultiply the new transformed equation by [F]T and then take advantage of the orthogonality of the Eigenfunctions. When there is no damping matrix [Czz] or if damping is present and can be represented as Rayleigh damping (also known as proportional damping), i.e., + 2w nVx + w2 x n x = F (t ) / M - y (17.34)

(where a and b are user-dened constants), then we can simplify the equations on c, resulting in the set of uncoupled equations for each mode i. i + 2w i V i c i + w i2 ci = fi (t ) c i = 1,2,imax (17.35)

where ci is the ith row variable in the transformation variable {c}, zi = is the ith row constant in the proportional damping diagonal matrix diag[V i ] [F]T[C zz ][F] = [a[F]T[K zz ][F] + b[F]T[M zz ][F]] } is in place in the r.h.s. and the driver is given by fi (t ) = {Fi } T {Fz } , where the damping term [C zy ]{Y of Equation (17.32). The nal step is to integrate these Imax equations, producing each row entry of the, ci , of the {c} generalized column matrix. Once these have been found, the desired actual displacement vector {Z} can be found by back-substituting into the transformation equation to obtain:
i = I cut

{Z(t )} =

c (t ){F }
i i i =1

(17.36)

where the inequality Icut Imax denotes the notion that in many cases the higher modes in frequency are truncated from the nal construction of the solution at a mode cutoff Icut, for modes that are ordered from low to high frequency.

Integration of Uncoupled Equations

There are three options here worth noting, regarding the integration of Equations (17.35)

2005 by CRC Press LLC

1. Simply integrate Equations (17.35) for each i, using a direct explicit or implicit numerical time integration scheme for ordinary differential equations. 2. Note that, remarkably, the l.h.s. of Equation (17.36) looks just like the simple single-degree-offreedom damped harmonic motion oscillator of Equation (17.33). Therefore, by comparing like terms in these two differential equations, we can directly use Equation (17.8) and Equation (17.9) to obtain the solution for each i, where simple integration quadratures (e.g., Simpsons rule) can be used to obtain the solution in place of solving differential equations. Simply make the following correspondence between Equation (17.8) Equation (17.33), where: x(t) ci(t) ; wn wi; z 0 [i.e., set y = 0 in Equation (17.8)]. zi; F(t)/M fi(t); y 3. Analyst is given a shock spectrum, which represents a single D.O.F. response to the given motion input {Y(t)}, in lieu of the input itself as in options (1) and (2). There is not enough space to treat this in any detail, but briey the shock spectrum maxima can then be related to the maximum of each mode contribution maximum [Remmers, 1983]. The total maximum response is typically performed by adding the contributions of each mode in a root mean square sense (i.e., the square root of the sum of the squares of the contributions of each mode for displacements, and stresses). This assumes all peaks occur at the same instant in time and squaring kills any chance of cancellation. Therefore this approach (e.g., [Geers, 1998]), is far too conservative. In contrast, however, options (1) and (2) do take proper phasing into consideration and therefore are more accurate.

Dening Terms
C: Isolator damping constant (lb-sec/in.). K: Isolator spring constant (lb/in.). M: Net structural conguration mass (lb-sec/in.2). g: Acceleration of gravity (386 in./sec2). u: Isolator force not including static deection (lb). b = f/fn: Ratio of drive frequency to natural frequency (unitless). z: Critical damping ratio = C/(4pMfn) (unitless). ds: Static deection = Mg/K, amount spring deects under dead weight of mass M (in.). wn: Natural frequency (rad/sec). w n = w d 1 - z 2 : damped natural frequency (rad/sec). h: Damping decay constants (sec-1). TU: Force transmission ratio = peaku(t)/ Fmax (unitless). / y max (unitless). TA: Acceleration transmission ratio = peak z TD: Displacement magnication ratio = peakx/ ymax (unitless). BC: Base conguration, refers to the lower isolator attachment point (where base input accelerations are applied). y SC: Structural conguration, refers to the net vibrating mass (upper connection to the isolator).

References
Aeroex. 1994. Aeroex Isolators Selection Guide, Aeroex International, Inc., Plainview, NY. Barry. 1993. Barry Controls Bulletin DOEM1, Barry Controls, Brighton, MA. Bathe, K. 1996. Finite Element Procedures, Prentice Hall, Englewood Cliffs, NJ. Cook, R. D., Malkus, D. S., and Plesha, M. E. 1988. Concepts and Applications of Finite Element Analysis, 3rd ed., John Wiley & Sons, New York. Church, A. H. 1963. Mechanical Vibrations, 2nd ed. John Wiley & Sons, Inc., New York. Cunniff, P. and OHara, G. O. 1989. A procedure for generating shock design values, J. Sound Vib., 134, 154165. Firestone. 1994. Engineering Manual and Design Guide, Firestone Industrial Products Co., Noblesville, IN.
2005 by CRC Press LLC

Geers, T. L. 1998. Shock Aanalysis and design, in Handbook of Acoustics, Cricker, M. J., Ed., John Wiley & Sons, New York, chap. 52. Harris, C. M. 1988. Shock and Vibration Handbook, 3rd ed. McGraw-Hill Book Co., New York. Hwon, Y. W. and Bang, H. 2000. The Finite Element Method Using MATLAB, CRC Press, Boca Raton, FL. Knowledge Revolution. 1994. Working Model Demonstration Guide and Tutorial, San Mateo, CA. Lord. 1994. Lord Industrial Products Catalog, PC-2201H. Lord Industrial Products, Erie, PA. Math Works. 1992. The Student Edition of MATLAB. Prentice Hall, Englewood Cliffs, NJ. Przemieniecki, J. S. 1968, Theory of Matrix Structural Analysis, McGraw-Hill, New York. Redfern, D. 1994. The Maple Handbook: Maple V Release 3, Springer-Verlag, New York. Remmers, G. 1983. Maurice Biot 50th Anniversary Lecture: the Evolution of Spectral Techniques in Navy Design, Shock and Vibration Bull., Part 1. Sevin, E. and Pilkey, W. D. 1971. Optimum Shock and Vibration, Monogram SVM-6. Shock and Vibration Information Center, Naval Research Laboratory, Washington, D.C. Sigmon, K. 1994. MATLAB Primer, 4th ed. CRC Press Inc., Boca Raton, FL. Wolfram, S. 1991. Mathematica: A System for Doing Mathematics by Computer, 2nd ed. Addison-Wesley Publishing Co., Redwood City, CA.

2005 by CRC Press LLC