Nonlinear Dynamics 29: 235250, 2002. 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Describing Function Analysis of Systems with Impacts and Backlash

R. S. BARBOSA and J. A. TENREIRO MACHADO
Department of Electrotechnical Engineering, Institute of Engineering of Porto, Rua Dr. Antnio Bernardino de Almeida, P-4200-072 Porto, Portugal (Received: 28 February 2002; accepted: 12 March 2002) Abstract. This paper analyzes the dynamical properties of systems with backlash and impact phenomena based on the describing function method. It is shown that this type of nonlinearity can be analyzed in the perspective of the fractional calculus theory. The fractional-order dynamics is illustrated using the Nyquist plot and the results are compared with those of standard models. Keywords: Describing function, backlash, fractional-order systems, modeling.

1. Introduction Fractional Calculus (FC) is a branch of mathematics that deals with the generalization of the operation of differentiation and integration to an arbitrary order. The theory of FC is a welladapted tool to the modeling of many physical phenomena, allowing the description to take into account some peculiarities that classical integer-order models simply neglect. For this reason, the rst studies and applications involving FC were developed in the domain of fundamental sciences, namely in physics [5] and chemistry [24]. Besides the intensive research carried out in the area of pure and applied mathematics [17], FC has found applications in elds such as viscoelasticity/damping [813], chaos [14], fractals [1517], biology [18], electronics [19], signal processing [20], diffusion and wave propagation [2123], modeling and identication [25, 26], electromagnetism [27, 28] and automatic control [2933]. Nevertheless, in spite of the work that has been done in the area, many aspects remain to be investigated. The phenomenon of vibration with impacts occurs in many branches of technology, namely in impact machines, vibration dampers and shakers, where it plays a very useful role. On the other hand, its occurrence is often undesirable, because it causes additional dynamic loads, as well as faulty operation of machines and devices. Despite many investigations that have been carried out so far, this phenomenon is not fully understood yet, mainly due to the considerable randomness and diversity of reasons underlying the energy dissipation involving the dynamic effects [3638]. In this paper, we investigate the dynamics of systems that contain backlash and impacts. It is shown that these nonlinear phenomena can exhibit a fractional-order dynamics revealing that the FC is an adequate tool for the analysis of these systems. In this perspective, this paper investigates the dynamics of systems with backlash and impact phenomena through the describing function method.

236 R. S. Barbosa and J. A. Tenreiro Machado

Figure 1. Basic nonlinear feedback system for describing function analysis.

Bearing these ideas in mind, the present paper is organized as follows. Section 2 introduces the fundamental aspects of the theory of the describing function method. Section 3 studies the describing function of systems with and without impacts. The results are compared with classical models of systems with simple geometric backlash. Finally, Section 4 draws the main conclusions and addresses perspectives towards future research. 2. Describing Function Analysis The Describing Function (DF) is one of the possible methods that can be adopted for the analysis of nonlinear systems [35]. The basic idea is to apply a sinusoidal signal to the input of the nonlinear element and to consider only the fundamental component of the signal appearing at the output of the nonlinear system. Then, the ratio of the corresponding phasors (output/input) of the two sinusoidal signals represents the DF of the nonlinear element. The use of this concept allows the adaptation of the Nyquist stability test to a nonlinear system detection of a limit cycle, namely the prediction of its approximate amplitude and frequency. In this line of thought, we consider the control-loop with one nonlinear element depicted in Figure 1. We start by applying a sinusoid to the nonlinearity input: x(t) = X cos(t). (1)

At steady-state the output of the nonlinear characteristic, y(t), is periodic and, in general, it is nonsinusoidal. If we assume that the nonlinearity is symmetric with respect to the variation around zero, the Fourier series becomes

y(t) =
k=1

Yk cos(kt + k ),

(2)

where Yk and k are the amplitude and the phase shift of the kth harmonic component of the output y(t), respectively. In the DF analysis, we assume that only the fundamental harmonic component of y(t), Y1 , is signicant. Such assumption is often valid since the higher-harmonics in y(t), Yk for k = 2, 3, . . ., are usually of smaller amplitude than the amplitude of the fundamental component, Y1 . Moreover, most systems are low-pass lters with the result that the higher-harmonics are further attenuated. Thus the DF of a nonlinear element, N(X, ), is dened as the complex ratio of the fundamental harmonic component of output y(t) with the input x(t): N(X, ) = Y1 j 1 e , X (3)

Describing Function Analysis of Systems with Impacts and Backlash 237

Figure 2. Static backlash nonlinearity: (a) geometric model, (b) input-output characteristic.

where X is the amplitude of the input sinusoid x(t) and Y1 and 1 are the amplitude and the phase shift of the fundamental harmonic component of the output y(t), respectively. In general, N(X, ) is a function of both the amplitude X and the frequency of the input sinusoid. For nonlinear systems that do not involve energy storage, the DF is merely amplitude-dependent, that is N = N(X). If it is not the case, we may have to adopt a numerical approach because, usually, it is impossible to nd a closed-form solution. For the nonlinear control system of Figure 1, we have a limit cycle if the sinusoid at the nonlinearity input regenerates itself in the loop, that is: G(j ) = 1 . N(X, ) (4)

Note that (4) can be viewed as the characteristic equation of the nonlinear feedback system of Figure 1. If (4) can be satised for some value of X and , a limit cycle is predicted for the nonlinear system. Moreover, since (4) applies only if the nonlinear system is in a steady-state limit cycle, the DF analysis predicts only the presence or the absence of a limit cycle and cannot be applied to analysis for other types of time responses. 3. Analysis of Systems with Backlash and Impact Phenomena In this section, we use the DF method to analyze systems with backlash and impact phenomena. We start by considering the standard static model and afterwards we study the case of dynamic backlash (i.e. with the impact phenomena). Finally, we compare the results of the two types of approximations. 3.1. S TATIC BACKLASH Here, we consider the phenomena of clearance without the effect of the impacts, which is usually called static backlash. The model and its input-output characteristic are shown in Figure 2. By applying a sinusoidal signal x(t) = X sin(t) at the input member, the DF of the static backlash is given by the following expression [34]: 0, X h/2, (5) N(X, ) = k 2kh(X h/2) X/ h 1 Ns j , X > h/2, 2 1 X/ h X2

238 R. S. Barbosa and J. A. Tenreiro Machado

Figure 3. Classical backlash model.

Figure 4. Nyquist plot of 1/N(F, ) for the system of Figure 3, F = {10, 20, 30, 40, 50} N, 0 < < L , M1 = M2 = 1 kg and h = 101 m.

Ns (z) =

2 1 1 1 sin1 + cos sin1 z z z

(6)

The classical backlash model corresponds to the DF of a linear system of a single mass M1 + M2 followed by the geometric backlash having as input and as output the position variables x(t) and y(t), respectively, as depicted in Figure 3. For a sinusoidal input force f (t) = F cos(t) the condition X = h/2 leads to the limit frequency L applicable to this system: L = F 2 h (M1 + M2 )
1/2

(7)

Figure 4 shows the Nyquist plot of 1/N(F, ) = 1/[G(j )N(X)] for the system of Figure 3 for several values of the input force F . This approach to the backlash study is based on the adoption of a geometric model that neglects the dynamic phenomena involved during the impact process. Due to this reason often real results differ signicantly from those predicted by that model.

Describing Function Analysis of Systems with Impacts and Backlash 239

Figure 5. Representation of a central impact, with initial velocities x1 and x2 .

Figure 6. System with two masses subjected to dynamic backlash.

3.2. DYNAMIC BACKLASH In this section, we consider the impact phenomenon based on the laws of conservation of momentum and an empirical manifestation of the conservation of energy. This approach provides information on the net change in velocity of each body involved in the impact, the net impulse, and the energy-exchange processes accompanying the impact. We consider the case of two bodies colliding on surfaces which are normal to the common line connecting their centers of mass and that have velocity components only along this common line. With these restrictions no rotational or sliding effects occur. This specic case is called a central impact and is illustrated in the conceptual model of Figure 5. The proposed mechanical model consists on two masses (M1 and M2 ) subjected to backlash and impact phenomenon as shown in Figure 6. It is straightforward to see that a collision between the masses M1 and M2 occurs when x1 = x2 or x2 = h + x1 . In this case, we can compute the velocities of masses M1 and M2 after the impact (x1 and x2 ) by relating them to the previous values (x1 and x2 ) through Newtons law: (x11 x12 ) = (x1 x2 ), 0 1, (8)

where is the coefcient of restitution that represents the dynamic phenomenon occurring in the masses during an impact. In the case of a fully plastic (inelastic) collision = 0, while in the ideal elastic case = 1. The principle of conservation of momentum requires that the momentum, immediately before and immediately after the impact, must be equal: M1 x1 + M2 x2 = M1 x1 + M2 x2 . (9)

240 R. S. Barbosa and J. A. Tenreiro Machado

Figure 7. Contour plot of energy loss WL vs. the exciting frequency and the coefcient of restitution , for an input force F = 50 N, M1 = M2 = 1 kg and h = 101 m.

From (89), we can nd the sought velocities of both masses after an impact: x1 = x2 = x1 (M1 M2 ) + x2 (1 + )M2 , M1 + M2 x1 (1 + )M1 + x2 (M2 M1 ) . M1 + M2 (10)

(11)

The total kinetic energy loss EL at an impact is determined by EL = 1 2 M1 M2 (x1 x2 )2 . 2 M1 + M2 (12)

The validity of the proposed model is restricted to frequencies of the exciting input force f (t) higher than a cut-off frequency C . This frequency was determined numerically by considerations on the energy processes occurring during the impact process arriving to the approximate expression: C F 2 M2 h
2 1/4

(1 )

(13)

Figures 7 and 8 illustrate the energy and power losses (i.e. WL and PL ) vs. the exciting frequency and the coefcient of restitution . As expected, the energy (and the power) loss decreases as the coefcient of restitution increases. Moreover, as 1 it yields C 0, which is in accordance with Equation (13). On the other hand, there is also a limiting frequency L determined by application of New tons law to mass M2 , that is f (t) = M2 x2 (t). Considering an input signal f (t) = F cos(t)

Describing Function Analysis of Systems with Impacts and Backlash 241

Figure 8. Contour plot of power loss PL vs. the exciting frequency and the coefcient of restitution , for an input force F = 50 N, M1 = M2 = 1 kg and h = 101 m.

Figure 9. Nyquist plot of 1/N(F, ) for the dynamic backlash, F = 50 N and = {0.1, . . . , 0.9}.

and solving for x2 (t) we arrive at an expression for L when the amplitude of the displacement is within the clearance h/2, yielding F L = 2 h M2
1/2

(14)

For the system model of Figure 6 we can calculate numerically the Nyquist diagram of 1/N(F, ) for a sinusoidal input force f (t) = F cos(t) applied to mass M2 while considering as output position x1 (t) of mass M1 .

242 R. S. Barbosa and J. A. Tenreiro Machado

Figure 10. Nyquist plot of 1/N(F, ) for a system with dynamic backlash, F = {10, 20, 30, 40, 50} N and = {0.2, 0.5, 0.8}.

The values of the parameters adopted in the subsequent simulations are M1 = M2 = 1 kg and h = 101 m. Figures 9 and 10 show the Nyquist plots for an input force F = 50 N and = {0.1, . . . , 0.9} and for F = {10, 20, 30, 40, 50} N and = {0.2, 0.5, 0.8}, respectively. The Nyquist charts of Figures 9 and 10 reveal some interesting features. The most obvious is the occurrence of a jumping phenomenon, which is a characteristic of nonlinear systems. This phenomenon is more visible around 0.5, while for the limiting cases, that is 0 and 1, the singularity disappears.

Describing Function Analysis of Systems with Impacts and Backlash 243

Figure 11. Fourier transform of the output displacement x1 (t), over 500 cycles, vs. the exciting frequency and the harmonic frequency 0 , for = {0.2, 0.5, 0.8}.

The frequency for which the jumping phenomena occurs (J ) has the relation: J F h M2
1/2

(15)

Moreover, Figure 10 shows also that for a xed value of the charts are proportional to the input amplitude F .

244 R. S. Barbosa and J. A. Tenreiro Machado

Figure 12. Log-log plots of Re{1/N} and Im{1/N} vs. the exciting frequency , for = 0.5 and F = {10, 20, 30, 40, 50} N.

Figure 11 presents the harmonic content of x1 (t) for an input force f (t) = 50 cos(t), C < < L , and = {0.2, 0.5, 0.8}. The charts demonstrate that the fundamental harmonic of the output has a much higher magnitude than the other higher-harmonic components. This fact supports the application of the describing function method in the prediction of limit cycles for this system. Note also that for large values of the coefcient of restitution (e.g. = 0.8), the high-order harmonic content increases and, by consequence, the accuracy of the prediction diminishes.

Describing Function Analysis of Systems with Impacts and Backlash 245

Figure 13. Log-log plots of Re{1/N} and Im{1/N} vs. the exciting frequency , for F = 50 N and = {0.1, 0.3, 0.5, 0.7, 0.9}.

Figures 12 and 13 illustrate the variation of the Nyquist plots of 1/N(F, ) for the cases of the static and dynamic backlash. Figure 12 shows the log-log plots of Re{1/N} and Im{1/N} vs. for a constant coefcient of restitution = 0.5 and F = {10, 20, 30, 40, 50} N, while Figure 13 depicts the log-log plots of Re{1/N} and Im{1/N} vs. for a constant input force F = 50 N and = {0.1, 0.3, 0.5, 0.7, 0.9}. Comparing the results for the static and the dynamic backlash models we conclude that:

246 R. S. Barbosa and J. A. Tenreiro Machado

Figure 14. Number of consecutive collisions on side A (nA ) vs. the exciting frequency and the coefcient of restitution , for an input force f (t) = 50 cos(t). For the side B (nB ) the chart is of the same type.

Figure 15. Ratio of the number of consecutive collisions on side A (nA ) and side B (nB ), nA /nB , vs. the exciting frequency and the coefcient of restitution , for an input force f (t) = 50 cos(t).

The charts of Re{1/N} are similar for low frequencies (where they reveal a slope of +40 dB/dec) but differ signicantly for high frequencies. The charts of Im{1/N} are different in all range of frequencies. Moreover, for low frequencies, the dynamic backlash has a fractional slope inferior to +80 dB/dec of the static model. A careful analysis must be taken because it was not demonstrated that a DF fractional slope would imply a fractional-order model. In fact, in this study we adopt integer-order models for

Describing Function Analysis of Systems with Impacts and Backlash 247

Figure 16. Number of impacts per excitation period T = 2/ considering three values of = {0.2, 0.5, 0.8} and a constant input force F = 50 N.

the system description but the fractional-order slope is due to continuous/discrete dynamic variation that results due to the mass collisions. Consequently, a complementary perspective for the characterization of the system behavior is given through the analysis of the number of impacts that occur. Figure 14 depicts nA (or nB ), the number of consecutive collisions on side A (or B), vs. the exciting frequency and the coefcient of restitution . A plot of the same type is given

248 R. S. Barbosa and J. A. Tenreiro Machado in Figure 15 and relates the ratio of the number of consecutive collisions on sides A and B, nA /nB , vs. (, ) for an input force f (t) = 50 cos(t). The results show that: For C < < J the system is characterized by an irregular number of impacts and a chaotic dynamics. For J < < L the motion is characterized by a regular behaviour corresponding to one alternate collision on each side of M1 . We observe that a large part of the graph is characterized by a relation 1 < nA < 2 (or 2 > nB > 1) and that the jumping phenomenon (that occurs for = J ) corresponds to a change on the relation from nA /nB = 1/2 up to nA /nB = 1. We can also analyze the number of collisions per period T = 2/ of the exciting frequency . Figure 16 depicts the number of impacts for different values of the coefcient of restitution = {0.2, 0.5, 0.8}. Once again, we note that the graphs show two kinds of regions. The rough regions are characterized by an irregular number of impacts that correspond to the chaotic dynamics. The soft regions are characterized by a regular number of impacts and a periodic motion of the system. Note also that the higher the coefcient of restitution (e.g. = 0.8) the more scattered are the chaotic regions. 4. Conclusions This paper addressed several aspects of the phenomena involved in systems with backlash and impacts. The dynamics of a two-mass system was analyzed through the describing function and compared with standard models. The results revealed that these systems might lead to chaos and to fractional-order dynamics. These conclusions encourage further studies of nonlinear systems in the perspective of the fractional calculus since integer-order dynamical models are not capable to take into account many phenomena that occur. References
1. 2. 3. 4. 5. 6. 7. 8. 9. Oldham, K. B. and Spanier, J., The Fractional Calculus: Theory and Application of Differentiation and Integration to Arbitrary Order, Academic Press, New York, 1974. Samko, S. G., Kilbas, A. A., and Marichev, O. I., Fractional Integrals and Derivatives: Theory and Applications, Gordon and Breach, Amsterdam, 1993. Miller, K. S. and Ross, B., An Introduction to the Fractional Calculus and Fractional Differential Equations, Wiley, New York, 1993. Podlubny, I., Fractional Differential Equations, Academic Press, San Diego, CA, 1999. Hilfer, R., Applications of Fractional Calculus in Physics, World Scientic, Singapore, 2000. Ross, B. , Fractional Calculus and Its Applications, Lecture Notes in Mathematics, Vol. 457, SpringerVerlag, New York, 1974. Samko, S. G. and Ross, B., Integration and differentiation to a variable fractional order, Integral Transforms and Special Functions 1(4), 1993, 277300. Gemant, A., A method of analyzing experimental results obtained from elasto-viscous bodies, Physics 7, 1936, 311317. Stiassnie, M., On the application of fractional calculus for the formulation of viscoelastic models, Applied Mathematical Modelling 3, 1979, 300302.

Describing Function Analysis of Systems with Impacts and Backlash 249

10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. Bagley, R. L. and Torvik, P. J., Fractional calculus A different approach to the analysis of viscoelastically damped structures, AIAA Journal 21(5), 1983, 741748. Bagley, R. L. and Torvik, P. J., On the fractional calculus model of viscoelastic behaviour, Journal of Rheology 30(1), 1986, 133155. Bagley, R. L. and Calico, R. A., Fractional order state equations for the control of viscoelastically damped structures, ASME Journal of Guidance 14(2), 1991, 304311. Fenander, ., Modal synthesis when modeling damping by use of fractional derivatives, AIAA Journal 34(5), 1996, 10511058. Hartley, T. T., Lorenzo, C. F., and Qammer, H. K., Chaos in fractional order Chuas system, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications 42(8), 1995, 485490. Liu, S. H., Fractal model for the ac response of a rough interface, Physical Review Letters 55(5), 1985, 529532. Kaplan, T., Gray, L. J., and Liu, S. H., Self-afne fractal model for a metal-electrolyte interface, Physical Review B 35(10), 1987, 53795381. Mhaut, A., Fractal Geometries: Theory and Applications, Penton Press, London, 1991. Anastasio, T. J., The fractional-order dynamics of brainstem vestibulo-oculomotor neurons, Biological Cybernetics 72, 1994, 6979. Oustaloup, A., Fractional order sinusoidal oscillators: Optimization and their use in highly linear FM modulation, IEEE Transactions on Circuits and Systems 28(10), 1981, 10071009. Ozaktas, H. M., Arikan, O., Kutay, M. A., and Bozdagi, G., Digital computation of the fractional Fourier transform, IEEE Transactions on Signal Processing 44(9), 1996, 21412150. Nigmatullin, R. R., The realization of the generalized transfer equation in a medium with fractal geometry, Physics of the State Solid (b) 133, 1986, 425430. Mhaut, A., Hliodore, F., Cottevieille, D., and Latreille, F., Introduction to wave phenomena and uncertainty in a fractal space, Chaos, Solitons & Fractals 3(5), 1993, 389402. Mainardi, F., Fractional relaxation-oscillation and fractional diffusion-wave phenomena, Chaos, Solitons & Fractals 7(9), 1996, 14611477. Mhaut, A., From dissipative and to non-dissipative processes in fractal geometry: The janals, New Journal of Chemistry 14(3), 1990, 207215. Matignon, D. and dAnra-Novel, B., Some results on controllability and observability of nite-dimensional fractional differential systems, in Proceedings IEEE SMC/IMACS Symposium on Control, Optimization and Supervision, Lille, France, P. Borne, M. Staroswiecki, J. F. Cassae, and S. L. Khattiabi (eds.), IEEE, New York, 1996, pp. 952956. Mathieu, B., Le Lay, L., and Oustaloup, A., Identication of non integer order systems in the time domain, in Proceedings IEEE SMC/IMACS Symposium on Control, Optimization and Supervision, Lille, France, P. Borne, M. Staroswiecki, J. F. Cassae, and S. L. Khattiabi (eds.), IEEE, New York, 1996, pp. 843847. Engheta, N., On the role of fractional calculus in electromagnetic theory, IEEE Antennas and Propagation Magazine 39(4), 1997, 3546. Engheta, N., Fractional paradigm in electromagnetic theory, in Frontiers of Electromagnetics, D. H. Werner and R. Mittra (eds.), IEEE, New York, 1999, pp. 523552. Oustaloup, A., La Drivation Non Entire, Hermes, Paris, 1995. Machado, J. A. T., Analysis and design of fractional-order digital control systems, SAMS Journal Systems Analysis, Modelling, Simulation 27, 1997, 107122. Machado, J. A. T., Discrete-time fractional-order controllers, FCAA Fractional Calculus and Applied Analysis 4(1), 2001, 4766. Vinagre, B. M., Podlubny, I., Hernndez, A., and Feliu, V., Some approximations of fractional order operators used in control theory and applications, FCAA Fractional Calculus and Applied Analysis 3(3), 2000, 231248. Podlubny, I., Fractional-order systems and PI D -controllers, IEEE Transactions on Automatic Control 44(1), 1999, 208213. Phillips, C. L. and Harbor, R. D., Feedback Control Systems, Prentice-Hall, Englewood Cliffs, NJ, 1991. Slotine, J. E. and Li, W., Applied Nonlinear Control, Prentice-Hall, Englewood Cliffs, NJ, 1991. Choi, Y. S. and Noah, S. T., Periodic response of a link coupling with clearance, ASME Journal of Dynamic Systems, Measurement and Control 111(2), 1989, 253259.

26.

27. 28. 29. 30. 31. 32.

33. 34. 35. 36.

250 R. S. Barbosa and J. A. Tenreiro Machado

37. Stepanenko, Y. and Sankar, T. S., Vibro-impact analysis of control systems with mechanical clearance and its application to robotic actuators, ASME Journal of Dynamic Systems, Measurement and Control 108(1), 1986, 916. Azenha, A. and Machado, J. A. T., On the describing function method and the prediction of limit cycles in nonlinear dynamical systems, Journal of Systems Analysis, Modelling and Simulation 33, 1998, 307320.

38.