Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 46, 1, pp. 223-234, Warsaw 2008 REGULAR AND CHAOTIC VIBRATIONS OF A PRELOADED PENDULUM SYSTEM Jerzy Warmiński Andrzej Mitura Lublin University of Technology, Department of Applied Mechanics, Lublin, Poland e-mail: j.warminski@pollub.pl; a.mitura@pollub.pl Krzysztof P. Jankowski Magna Closures of America, Novi, MI, USA e-mail: krystof.jankowski@MagnaClosures.com Vibrations of a single preloaded pendulum, with the impact phenome- non with a stop taken into account, are analysed in the paper. Themo- del, under some simplification, represents a simple one-degree-of-freedom element of a vehicle latch system. The response of the system has been tested numerically for a wide range of parameters andwith various exci- tation amplitudes and frequencies. Elastic and rigid impacts have been studied, and results for both impact models and for regular and chaotic motionshavebeen compared.Tofindall possible solutions to the system, basins of attraction have been calculated as well. The final verification of the model response has been presented for the experimental signals recorded during vehicle side impact tests. It has been shown that the response of the latch element under study may lead to door opening, and that is why it has to be replaced with another design. Key words: vibration, chaotic motion, impact, pendulum system 1. Introduction Vehicle door latch systemsbelong to those elements that shouldbe designed in such a way that required safety conditions are satisfied. These systems should not allow the door opening when subjected to high amplitude/high frequency acceleration signals. They should work properly even in the case of vehicle 224 J. Warmiński et al. crash that can lead to accelerations acting over a short time but with very large amplitudes. However, as pointed byThomsen (2002), low amplitude but high frequency excitations may change apparent characteristics of mechanical systems. The excitation may result in loss of contact between latch elements that are otherwise biased by springs, andnext to an impact phenomenon. Sys- temswith discontinuities are very sensitive to external inputs, andunder some conditions tend to respond with irregular motions (Wiercigroch and Kraker, 2000). Moreover, prediction of their response with analytical methods is limi- ted to a specific class of models. By these reasons, analyses of models with impacts are mainly carried out numerically. The purpose of this paper is to study the behaviour of a simple model of a latch element represented by a single physical pendulum pushed against the support with a torque produced by a preloaded spring. The value of the torque should be large enough to keep the door locked, but it should not make it difficult to open the door. Themodel is tested with respect to a range of external excitations consisting of harmonic acceleration signals with large amplitudes and high frequencies, and also accelerations recorded during a real vehicle crash test. 2. System model Let us consider a model of a latch element that is commonly used in vehicle door latch systems. After some simplifications, the latch element (or the door handle) can be represented by a one-degree-of-freedom pendulum-like model represented in Fig.1. Fig. 1. Physical model of the latch element: (a) rigid impact model, (b) soft impact model The latch lever, represented by the physical pendulum, is pushed against the stop by a spring with the stiffness k0 and the preload torque T0. The pendulummass, itsmomentof inertia, the centre ofmass locationanddamping Regular and chaotic vibrations... 225 at the pivot are denoted by m, I0,L, c0, respectively. The element should not rotate aboveapre-specified thresholdeven if its base is subjected toanexternal acceleration. In thepresentedmodel, theacceleration is assumedas aharmonic function of time a = Ag sinωt, where g denotes the acceleration of gravity, A is the amplitude of dimensionless acceleration expressed with respect to g, ω denotes the frequency of excitation and t is time. It is worth to note that the displacement of the base is expressed by y0 =−(Ag/ω 2)sinωt. The ideal behaviour of the latch system would be if the pendulum stays in contact with the support, independently of the excitation level. However, one can expect that under some conditions the loss of contact between the pendulum and the stop will occur. In such a case, the rotation angle of the pendulummay increase over the safety limit. The threshold angle that should not be exceeded is assumed to be φ=0.1rad. The rigid impact model is governed by the classical Newton impact law (Fig.1a). Because, in the real system, the support is not ideally stiff but rather a deformable body, an elastic impact, which seems to be closer to the real system, is also proposed (Fig.1b). Both types of impacts were studied in the literature (Thomsen, 2000; Wiercigroch and Kraker, 2000; Pavlovskaia and Wiercigroch, 2000). A soft impact with a deformable stop is represented by a mass-less, viscous-elastic element with the stiffness ks, damping coefficient cs and location ls from the pivot. The viscous-elastic support, represented by a spring and a damper, allows for modelling the dissipated energy, deformation of the stop, and the time duration of the contact between both impacting bodies. These quantities depend on the force impulse during the impact and they are not constant in time as in the classical Newton approach. Therefore, thismodel is considered as themain contactmodel in the subsequent analysis. Nevertheless, for comparison and validation of certain results, the rigid impact model will also be tested in parallel. Differential equations of motion of the systems represented in Fig.1 take the form: —For the rigid-impact model I0φ̈+ c0φ̇+k0φ=−T0−mAgLsinωtcosφ (2.1) where,with theNewton impact law, if φ=0 then φ+ =φ− and φ̇+ =−Rφ̇−, Rdesignates the coefficient of restitution, and φ − and φ+ are theangles before and after the impact, respectively. —For the elastic-impact model I0φ̈+ cRφ̇+kRφ=−T0−mAgLsinωtcosφ (2.2) 226 J. Warmiński et al. where cR = { c0 for φ­ 0 c0+ csl 2 s for φ< 0 kR = { k0 for φ­ 0 k0+ksl 2 s for φ< 0 and cR and kR represent the damping and stiffness for both cases: outside of the contact area and during the contact. The natural frequency of the system takes two values: for φ­ 0, ω0 = √ k0/I0 (outside of contact), and for φ< 0, ωR = √ kR/I0 (during contact). It is worth tomention that the proposed soft- impactmodel does not take into account the dynamics of the deformable stop. This assumption can be justified by the fact that after the loss of contact the natural frequency of the stop is much higher than that of the pendulum, and due to damping of the stop, the force quickly declines to zero. To be able to compare both impactmodelswe can assume that the change of velocity during the elastic impact should relate to the coefficient of restitu- tion R as follows φ̇(timpact)=−Rφ̇(0) (2.3) what means that over the impact time timpact the velocity is reduced R times. The time duration of the impact equals to timpact = π/ωd, where ωd =ωR √ 1− ξ2 represents the natural frequency of underdampedvibrations, and ξ= cR/(I0ω 2 R ) denotes the dimensionless damping coefficient. Taking into account (2.3) and the solution for free damped vibrations, one can write R=exp ( − ξπ √ 1− ξ2 ) (2.4) Solving (2.4), we can determine the equivalent dimensionless damping coeffi- cient with respect to the given coefficient of restitution R of the rigid impact ξ=− lnR √ π2+(lnR)2 (2.5) Relationship (2.5) is presented in Fig.2, and is used in the paper for the determination of the damping coefficient of the viscous-elastic impact model. Because mathematical models (2.1) and (2.2) include discontinuous func- tions, it is difficult, or even impossible, to solve the problem analytically. The- refore, Eqs. (2.1) and (2.2) are solved numerically using the Runge-Kutta method of the fourth order. Before starting the main analysis, the proper integration step size has to be tested. Because the system under conside- ration is discontinuous in time, this point is crucial for further numerical analysis. Regular and chaotic vibrations... 227 Fig. 2. Equivalent damping versus restitution coefficient Figure 3a presents a phase space diagram received for the soft-impact model defined by Eq. (2.2). This result was produced after several tests with an integration step ∆t= 2π/(Nω), where N is the number of steps per the oscillation period T =2π/ω. The trajectory presented in Fig.3a is calculated for N = 10000 and ω = ω0, which corresponds to ∆t = 3.1416 · 10 −5 s. This result is in full agreement for soft (2.2) and rigid-impact (2.1) model. For comparison, the solution plotted in Fig.3b was received with N = 1000 (∆t = 3.1416 · 10−4) and the same set of parameters. As we can see, this result is far from the correct one presented in Fig.3a. On the basis of several integration step tests carried out for bothmodels, it has been determined that the integration step shouldnot be larger than 1/10000 of the vibration period. Fig. 3. Phase diagrams of free impacts with different integration steps: (a) N =10000, (b) N =1000 First, the free vibrations of both consideredmodels are analysed.Assuming that the torque T0 =0 and the impact duration time is very short, the period of the pendulum response can be approximated as T ≈ π/ω0. It means that without the preload torque (T0 =0), two impacts per period of free vibrations take place. Comparisons of time histories of the systemwith and without the impact are presented in Fig.4a. 228 J. Warmiński et al. Fig. 4. Time histories of free vibrations with and without the impact for the preload torque T0 =0Nm (a) and T0 =0.1Nm (b) Free vibrations change their nature if the preload torque is greater than zero. The influence of the torque T0 on the free vibrations of the system is presented in Fig.4b. If the initial torque T0 = 0 (Fig.4a), the period of vibrations with the impact is close to half of the period of free vibrations of the pendulum and it is constant. For the torque T0 = 0.1Nm (greater than zero) the period of vibrations with the impact is different, andmoreover, it is decreasing with the decreasing amplitude (Fig.4b). Figure 5 presents a spectrogram of free vibrations shown in Fig.4b with the preload torque T0 = 0.1Nm. This graph illustrates an evident increase of the vibration frequency with the stiffening effect in the time domain. The character of this phenomenon is highlighted by the black solid line seen in Fig.5. Fig. 5. Spectrogram of free impact oscillations with the preload torque T0 =0.1Nm Regular and chaotic vibrations... 229 3. Dynamics of the system under harmonic excitation Systems with discontinuities subjected to external excitations usually have high tendency to produce complex and irregular responses. Therefore, their dynamics has been tested for different accelerations of the base, with various amplitudes A and frequencies ω of the excitation.Numerical calculationswere carried out for the following exemplary parameters of the latch element I0 =8.888 ·10 −4 kgm2 c0 =9.3 ·10 −5 Nms rad k0 =0.370 Nm rad T0 =0.1Nm m=0.1475 kg L=0.065m R=0.8 ks =10 7 N m cs =133.59 Ns m ω=0−1000 rad s A=0−100 ls =0.1m According to (3.1), the range of amplitude of acceleration hasbeenassumedup to 100g, and frequency up to about 160Hz. These are typical values; however, in more critical situations, these values may reach 500g and 300Hz. The natural frequency of the single pendulum system without impact with parameters listed above has values of either ω0 ≈ 20.4rad/s or ωR ≈ 10607 rad/s (when the pendulum is in contact with the support for the elastic impact model). Let us assume that the system is excited with the constant frequency ω = 20rad/s, close to the natural frequency ω0. The bifurcation analysis is performed for both models with the amplitude of the excitation A as the bifurcation parameter. Bifurcation diagrams for rigid and elastic impacts are presented in Figs.6a and 6b, respectively. The results obtained for bothmodels exhibit similar phenomena.Dark are- as that correspond to non-periodic motion occur for bothmodels in the same place. However, pure periodic motion is only observed for the rigid impact model (solid line in Fig.6a). For the soft impact model, due to deformation of the support, the corresponding motion is ”almost” periodic and the line is broader. Dark areas in Fig.6a and Fig.6b represent more complex motion: quasi-periodic or chaotic. An exemplary chaotic attractor for A= 40 is pre- sented in Fig.7. Chaoticmotion evidently takes place for both impactmodels, however, the soft impact (Fig.7b) leads to a more complex fractal structure of the chaotic attractor. An increase of the excitation frequency results in growth of chaoticmotion zones, as it can be seen in the bifurcation diagrams plotted for ω=100rad/s 230 J. Warmiński et al. Fig. 6. Bifurcation diagrams versus excitation amplitude A (ω=20rad/s): (a) rigid impact model, (b) elastic impact model Fig. 7. Chaotic attractor for A=40 and ω=20rad/s: (a) rigid impact, (b) elastic impact (Fig.8a). From the practical point of view, the critical situation arises when the pendulum loses the contact with the support. This may lead to a cascade of impacts and high amplitude oscillations. Fig.6 and Fig.8a show that at the beginning of the diagrams (A< 1), the response of the pendulum is close to zero, but after crossing the critical point, the system begins to oscillate. A zoomed portion of the beginning of the bifurcation diagram in Fig.8a is presented in Fig.8b. It can be seen that for A <∼ 1 the system does not oscillate, but after that it jumps to periodic motion, and later on the cascade of period doubling bifurcations goes to chaos. Because the solution presented in Fig.8 is very sensitive to initial conditions, an additional analysis of the basins of attraction has been carried out. Depending on the initial conditions, the solution can tend either to ze- ro (trivial solution No. I, see Fig.9) or to permanent oscillations (non-trivial solutions No. II and No. III). These three different possible solutions are pre- Regular and chaotic vibrations... 231 Fig. 8. Bifurcation diagrams of the soft impact model: (a) angular velocity of the pendulum versus amplitude A (ω=100rad/s), (b) zoom Fig. 9. Basins of attractions for trivial and non-trivial solutions, ω=100rad/s, A=1 sented as time histories in Fig.10. Summarising, the response of the system for certain initial conditions leads to oscillations even for a small amplitude of the excitation. 4. System response under vehicle impact excitation The final test of the response of the latch element has been carried out for an excitation in form of a short time shock. The lateral acceleration signal was recorded during a real vehicle crash test by one of OEM customers of Ma- gna Closures of America. The acceleration of the base of the latch can reach very large values, with amplitudes up to 500g and frequency in the range of 232 J. Warmiński et al. Fig. 10. Time histories of the pendulum angular displacement rotation angle for various initial conditions (ω=100rad/s, A=1): (a) φ=0.8, φ̇=0, (b) φ=0.56, φ̇=0, (c) φ=0.9, φ̇=0 0-3000rad/s. Recalling the results with harmonic excitation, it has been noti- ced that even for a low excitation amplitude, the pendulum swings out of the allowed range (which for the real mechanism corresponds to angles of about 0.1rad/s). Therefore, one can assume that for the model under consideration, du- ring the side impact excitation will be sufficiently produced to lead to do- or opening. In Fig.11a, the time history of the lateral acceleration recor- ded during the vehicle side crash test is presented. During the short time event (35ms duration), the amplitude of the acceleration reached almost A = 1000 of the gravity acceleration g. Figure 11b presents the respon- se of the latch element model to this input. It can be seen that the pen- dulum angle (in radians in Fig.11b) reached high values. This test showed that during the side impact with given severity the latch subsystem consi- sting of the single pendulum would not work properly and the door would open. Regular and chaotic vibrations... 233 Fig. 11. Vibration of the system under shock excitation: (a) latch lateral accelerationmeasured during the crash test (in m/s2), (b) pendulum response (angular displacement in radians) 5. Conclusions The results obtained for a single pendulum model of a vehicle latch element show that the response of the system may lead to the loss of the contact between the pendulum and its support, and to the impact phenomenon. Har- monic excitations transit the system to high-level regular or chaotic motions. The analysis of basins of attraction shows three different solutions for low acceleration amplitudes (A < 1), one trivial and two nontrivial. An increase of the latch base excitation amplitude causes transition to chaos of the pen- dulum after period doubling bifurcation (as shown in Fig.8b). Two different approaches for the impact phenomenon gave equivalent results, however the soft (elastic) impact exhibits attractors with a more fractal nature. Numeri- cal simulations performed for various physical parameters revealed that the amplitude of the system response cannot be reduced. The acceleration signals obtained from the real vehicle side crash test and applied to the model al- so showed that the behaviour of the single pendulum latch element did not satisfy the imposed safety conditions. Therefore, a modification of the latch subsystemwill be developed in forthcoming investigations, based on adding a second pendulum that will play the role of a counterweight. References 1. Lenci S., Demeio L., Petrini M., 2006, Response scenario and nonsmooth features in the nonlinear dynamics of an impacting inverted pendulum, Jour- nal of Computational and Nonlinear Dynamics, Transactions of the ASME, 1, 56-64 234 J. Warmiński et al. 2. PavlovskaiaE.,Wiercigroch M., 2004,Analytical drift reconstruction for visco-elastic impact oscillators operating in periodic and chaotic regimes,Cha- os, Solitons and Fractals, 19, 151-161 3. Thomsen J.J., 2002, Some general effects of strong high-frequency excitation: stiffening, biasing, and smoothening, Journal of Sound and Vibration, 253, 807-831 4. WiercigrochM.,KrakerB., 2000,AppliedNonlinearDynamics andChaos of Mechanical Systems with Discontinuities, World Scientific, Singapore Drgania regularne i chaotyczne napiętego wstępnie układu mechanicznego z wahadłem Streszczenie W niniejszej pracy przedstawiono drgania pojedyńczego wahadła fizycznego na- piętego wstępnie sprężyną, z uwzględnieniem możliwości zjawiska uderzenia w ogra- nicznik ruchu. Rozważany układ reprezentuje prostymodel jednego z elementówme- chanizmu zamykania drzwi pojazdu samochodowego. Zbadano numerycznie odpo- wiedź układu dla sztywnego i lepko-sprężystegomodelu uderzenia w szerokim zakre- sie wymuszeń harmonicznych.Porównanodla obumodeli uderzenia obszarywystępo- wania ruchu regularnego i chaotycznego. Przeprowadzono również analizę obszarów przyciągania rozwiązań, określającwszystkiemożliwe rozwiązania systemu.Końcową weryfikację odpowiedzi układu przeprowadzono dla sygnału eksperymentalnego, za- rejestrowanegopodczas próbybocznego zderzenia pojazdów.Pokazano, ze odpowiedź mechanizmumoże doprowadzić do otwarcia drzwi pojazdu i dlatego inne rozwiązanie techniczne powinno zostać użyte. Manuscript received June 18, 2007; accepted for print October 23, 2007