Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 47, 1, pp. 211-227, Warsaw 2009 TORSIONAL VIBRATIONS OF DISCRETE-CONTINUOUS SYSTEMS WITH LOCAL NONLINEARITIES HAVING SOFT TYPE CHARACTERISTICS Amalia Pielorz Monika Skóra Kielce University of Technology, Faculty of Management and Computer Modelling, Kielce, Poland e-mail: apielorz@tu.kielce.pl; mskora@tu.kielce.pl The paper deals with nonlinear vibration problems ofmulti-mass torsio- nally deformed mechanical systems. The systems consist of shafts con- nected by an arbitrary number of rigid bodies. In the systems, a local nonlinearity having a nonlinear characteristic of a soft type is introdu- ced. The local nonlinearity is described by functions including irrational functions together with two other nonlinear functions. In the considera- tions, thewave approach is applied.Numerical analysis is focused on the investigation of the influence of the local nonlinearity on the behaviour of considered systems and on the determination of application ranges of the irrational functions. Exemplary numerical calculations are given for a three-mass system. Key words: dynamics of mechanical systems, waves, nonlinear oscilla- tions, discrete-continuousmodels 1. Introduction The paper is aimed at the investigation of nonlinear torsional vibrations of discrete-continuousmechanical systemswith a local nonlinearity having a cha- racteristic of a soft type.This local nonlinearity is describedby three functions. The first one includes irrational functions. The considered systems consist of shaftswith circular cross-sections connectedbymeansof rigidbodies.Theybe- long to a certain class of discrete-continuous systems, namely to those where motion of elastic elements is described by means of the classical wave equ- ation, and the local nonlinearities in such systems are justified by engineering 212 A. Pielorz, M. Skóra solutions in many machines and mechanisms (cf. Boiler and Seeger, 1987; Thomson, 1981). Nonlinear vibrations of discrete systems are considered with various types of nonlinearities (cf. Hagedorn, 1981; Szemplińska-Stupnicka, 1990). Similarly to nonlinear discrete systems, various nonlinear functions can also be applied for thedescriptionof nonlinearities indiscrete-continuous systems.Wepropose to use functions which contain quadric and cubic nonlinearities. Some effects of local nonlinearities in discrete-continuous torsionally defor- med systems have been already shown, e.g., in Nadolski and Pielorz (1997), Pielorz (1999), Pielorz and Skóra (2006). In Nadolski and Pielorz (1997), a single gear transmission with nonlinear loads on gear teeth is studied, in Pie- lorz and Skóra (2006) torsional multi-mass systems with a local nonlinearity having a hard type characteristic, while in Pielorz (1999) having a soft type characteristic are considered. Considerations in the present paper are similar to those given in Pielorz (1999). The differences are in the assumed functions describing local nonli- nearities. In Pielorz (1999) four functions are proposed for that, while here three functions, however on different assumptions. The inclusion of irrational functions in one of these functions can play a significant role in the description of local nonlinearities with a softening characteristic. They contain the second and third degree polynomials which are the most often applied in nonlinear dynamics of discrete systems. Themethodproposed for the determination of solutions in arbitrary cross- sections of shafts in the system leads to equationswith a retarded argument of the neutral type. The numerical analysis is performed for a three-mass system and is focusedon the effect of local nonlinearity ondisplacements anddynamic moments. 2. Assumptions and governing equations The discrete-continuous model of a multi-mass torsional system under consi- derations is shown in Fig.1. It is assumed that the x-axis is parallel to the main axis of the system, and its origin coincides with the position of the left- hand end of the first shaft in an undisturbed state at the time instant t=0. The i-th shaft, i=1,2, . . . ,N, is characterised by length li, density ρ, shear modulus G and polar moment of inertia I0i. The i-th rigid body connecting corresponding shafts is characterised by themassmoment of inertia Ji. A sin- gle local nonlinearity is located in the cross-section x=0, and it can represent Torsional vibrations of discrete-continuous... 213 mechanical properties of various elements, such as clutches and gears, having nonlinear characteristics of a soft type. This local nonlinearity is represented by means of the nonlinear moment Msp. The rigid body J1 is loaded by the external loading M(t).Moreover, it is assumed that angular displacements θi and angular velocities ∂θi/∂t of the shaft cross-sections of the i-th elastic element are equal to zero at time instant t=0. Fig. 1. Nonlinear discrete-continuousmodel of a torsional system Damping in the system is described by equivalent external and internal damping taken into account in the cross-sections where rigid bodies are loca- ted. They are expressed by the following moments Mdi(t)= diθi,t(x,t) MDi(t)=GI0iDiθi,xt(x,t) (2.1) where di and Di are coefficients of the equivalent external and internal dam- ping, respectively, and comma denotes partial differentiation. On the above assumptions, the determination of angular displacement θi(x,t) of the shaft cross-sections is reduced to solving N classical wave equ- ations θi,tt− c 2θi,xx =0 i=1,2, . . . ,N (2.2) with zero initial conditions θi(x,0)= θi,t(x,0)= 0 i=1,2, . . . ,N (2.3) and with the following nonlinear boundary conditions: — for x=0 M(t)−J1θ1,tt+GI01(D1θ1,xt+θ1,x)−d1θ1,t−Msp(t)= 0 (2.4) — for x= ∑ i k=1 lk, i=1,2, . . . ,N −1 θi(x,t)= θi+1(x,t) (2.5) −Ji+1θi,tt−GI0i(Diθi,xt+θi,x)+GI0,i+1(Di+1θi+1,xt+θi+1,x)−di+1θi,t =0 214 A. Pielorz, M. Skóra — for x= ∑ N k=1 lk −JN+1θN,tt−GI0N(DNθN,xt+θN,x)−dN+1θN,t =0 (2.6) where c2 =G/ρ. Uponthe introductionof the followingdimensionless quantities (cf.Pielorz, 1999; Pielorz and Skóra, 2006) x= x l1 t= ct l1 θi = θi θ0 di = dil1 J1c Di = Dic l1 li = li l1 Bi = I0i I01 Kr = I01ρl1 J1 Ei = J1 Ji M = Ml21 J1c2θ0 Msp = Mspl 2 1 J1c2θ0 (2.7) problem (2.2)-(2.6) takes the form θi,tt−θi,xx =0 θi(x,0)= θi,t(x,0)= 0 i=1,2, . . . ,N (2.8) and — for x=0 M(t)−θ1,tt+Kr(D1θ1,xt+θ1,x)−d1θ1,t−Msp(t)= 0 (2.9) — for x= ∑ i k=1 lk, i=1,2, . . . ,N −1 θi(x,t)= θi+1(x,t) (2.10) −θi,tt−KrBiEi+1(Diθi,xt+θi,x)+KrBi+1Ei+1(Di+1θi+1,xt+θi+1,x)+ −Ei+1di+1θi,t =0 — for x= ∑N k=1 lk −θN,tt−KrBNEN+1(DNθN,xt+θN,x)−EN+1dN+1θN,t =0 (2.11) where θ0 is a fixed angular displacement and the bars denoting dimensionless quantities are omitted for convenience. The solutions to equations (2.8)1 are sought in the form θi(x,t)= fi(t−t0i−x+x0i)+gi(t−t0i+x−x0i) i=1,2, . . . ,N (2.12) where the functions fi and gi represent the waves caused by external loading M(t) propagating in the i-th shaft in the positive and negative senses of the Torsional vibrations of discrete-continuous... 215 x-axis, respectively. The constants x0i and t0i are the cross-section and the time instant at which the first disturbance caused by the external moment M(t) occurs in the i-th shaft. The functions fi and gi are continuous and equal to zero for negative arguments. For the system shown in Fig.1, the constants x0i and t0i are equal to x0i = i−1 ∑ k=1 lk t0i = i−1 ∑ k=1 lk (2.13) Substituting solutions (2.12) with (2.13) into boundary conditions (2.9)-(2.11) and denoting the largest argument in each boundary condition separately by z, the following set of equations for the unknown functions fi and gi is obtained rN+1,1g ′′ N(z)+rN+1,2g ′ N(z)= rN+1,3f ′′ N(z−2lN)+rN+1,4f ′ N(z−2lN) gi(z)= fi+1(z−2li)+gi+1(z−2li)−fi(z−2li) i=1,2, . . . ,N −1 r11f ′′ 1(z)=M(z)+r12g ′′ 1(z)+r13f ′ 1(z)+r14g ′ 1(z)−Msp(z) (2.14) ri1f ′′ i (z)+ri2f ′ i(z)= ri3g ′′ i (z)+ri4g ′ i(z)+ri5f ′′ i−1(z)+ri6f ′ i−1(z) i=2,3, . . . ,N where r11 =KrD1+1 r12 =KrD1−1 r13 =−Kr−d1 r14 =Kr−d1 ri1 =KrEi(BiDi+Bi−1Di−1)+1 ri2 =Ei[Kr(Bi+Bi−1)+di] ri3 =KrEi(BiDi−Bi−1Di−1)−1 ri4 =Ei[Kr(Bi−Bi−1)−di] ri5 =2KrBi−1EiDi−1 ri6 =2KrBi−1Ei i=2,3, . . . ,N rN+1,1 =KrBNEN+1DN +1 rN+1,2 =EN+1(KrBN +dN+1) rN+1,3 =KrBNEN+1DN −1 rN+1,4 =EN+1(KrBN −dN+1) (2.15) Equations (2.14) are nonlinear differential equations with a retarded ar- gument of the neutral type (cf. Hale, 1977; Muszyński and Myszkis, 1984). They are the same as in Pielorz (1999). The differences lie in Msp. Equ- ations (2.14) are solved numerically by means of the Runge-Kutta method in the given sequence. The same approach is also applied in the discussion of discrete-continuous torsionally deformed systems in Szolc (2003) together with appropriate experimental confirmations. 216 A. Pielorz, M. Skóra It should be pointed out that the introduction of the common argument z in (2.14) enables us to reduce the consideredproblem for amulti-mass discrete- continuous torsionally deformed system to solving such a set of equations for the functions fi and gi which allow us to determine simultaneously all these functions from z ­ 0. The way used above is presented in details in Ponomariev et al. (1957) for a fixed rod impacted by a rigid body.Moreover, inPielorz (1988) somecomparisonsarepresented for special cases of the system shown inFig.1. That paper concerns, namely, the comparison of the approach used in the present paper with the method of the rigid finite elements, and the comparison of the results obtained for the equation of motion being a dissipative wave equation with the solution for the classical wave equation taking into account the equivalent damping in boundary conditions. In all considered cases, a good agreement of the compared results was noticed. The function Msp(t) in (2.14) represents themoment foranonlinear spring, Fig.1. It can be described by various functions. Below, the following nondi- mensional functions are proposed for that Msp(t)=    k1θ1(0, t)−kw(−θ1(0, t)) w for θ1 ¬ 0 k1θ1(0, t)+kw(θ1(0, t)) w for θ1 ­ 0 (2.16) where k1 and kw appear at the linear and nonlinear terms, respectively, and the exponent w is assumed to be greater than 1. Nonlinear parts in (2.16) are expressed by irrational functions which give many possibilities for the description of experimental data. We are interested in the local nonlinearity having a soft type characteristic, i.e. in function (2.16) with kw < 0. It should be pointed out that functions (2.16) are used in Pielorz and Skóra (2006) for the description of local nonlinearities having hard type characteristics (kw > 0). Local nonlinearities and functions of type (2.16) are justified by numerous experimental studies (cf. Boiler and Seeger, 1987). The constants k1 and kw areusuallydetermined experimentally. In the case of a softening characteristic, we are interested in that part of the diagram of the function (2.16) where Msp is an increasing function. Functions (2.16) contain the second and third degree polynomials. In tech- nical literature, in the discussion of dynamics of nonlinear discrete systems, the second and third degree polynomials are usually applied (cf. Hagedorn, 1981; Stewart et al., 1995). Here, this assumption is generalised by introdu- cing irrational functions for discrete-continuous systems. Torsional vibrations of discrete-continuous... 217 Apart from functions (2.16) with kw < 0, two following functions are proposed for the description of the nonlinear moment Msp(t)=Atanh(Bθ1) (2.17) or Msp(t)= { A[−1+exp(Bθ1)] for θ1 ¬ 0 A[1− exp(−Bθ1)] for θ1 ­ 0 (2.18) where the constants A and B are assumed to be connected with the con- stants w, k1 and kw by the relations AB= k1 A= k1θm+kw(θm) w θm = ( − k1 wkw ) 1 w−1 (2.19) From relations (2.19), it follows that all three functions (2.16)-(2.18) describe the same linear case. Besides, functions (2.17) and (2.18) have the asymp- totes Msp = ±Mmax where Mmax are maximum values of functions (2.16) determined by w, kw and k1. 3. Numerical results The main aim of numerical calculations is the investigation of the influence of the local nonlinearity with the soft type characteristic described by func- tions (2.16) including irrational functions on the behaviour of the discrete- continuous systems considered. The externalmoment M(t) appearing in equations (2.14) canbe arbitrary. Similarly to the investigations of nonlinear discrete systems, it is assumed in the form M(t)=M0 sin(pt) (3.1) where p and M0 are dimensionless frequencies and amplitudes of the external loading. The discussion is focused on the solutions in steady states. Nonlinear effects in the considered systems are caused directly by the non- linear moment Msp(t) described by functions (2.16). They are also connected with the amplitude M0 of the external moment and with the external and internal damping. In the numerical analysis, the damping coefficients are as- sumed to be constant and equal to di =Di =0.1, k1 is fixed and kw, together with w, can vary. 218 A. Pielorz, M. Skóra Numerical results given below are exemplary. They are performed for the three-mass torsional systemandare focusedon the effect of nonlinearity on the angular displacements and on the nonlinear dynamicmoment Msp. All results concern solutions in the cross-section x = 0 where the nonlinear discrete element is introduced, though equations (2.14) allow us also to determine angular displacements in other cross-sections. The three-mass torsional system is characterised by the following basic parameters (cf. Pielorz, 1999) Kr =0.05 k1 =0.05 N =2 B2 =1 l1 = l2 =1 di =Di =0.1 E2 =E3 =0.8 (3.2) Diagrams of functions (2.16) with kw = −0.005, w = 1.5, 1.9, 2.25, 2.5, 3.0 and with the linear case are presented in Fig.2. They show how the non- linearity becomes stronger with the increase of w. Fig. 2. Diagrams of functions (2.16) The wave approach applied in the paper together with the Runge-Kutta method give the numerical solutions first in the transient and next in the steady states. In Fig.3, a solution for the angular displacement versus time is presented in the cross-section x = 0 for p = 0.2, w = 1.75, M0 = 0.4 and kw =−0.005. From Fig.3 it follows that the steady state is achieved for t> 200 with the displacement amplitude equal to 8.19033. Further numerical calculations concern amplitude-frequency curves for an- gular displacements and for the dynamicmoment. Such curves play an impor- tant role in engineering problems. Torsional vibrations of discrete-continuous... 219 Fig. 3. Angular displacements in x=0with p=0.2,w=1.75, kw =−0.005 and M0 =0.4 Fig. 4. Amplitude-frequency curves for angular displacements in x=0 for kw =−0.005,M0 =0.05, 0.2, 0.4 and with w=1.5, 1.75, 1.9, 2.25, 3.0 In Fig.4, the amplitude-frequency curves for the angular displacement in thecross-section x=0of the three-mass systemareplottedwith k3 =−0.005, M0 = 0.05, 0.2, 0.4 and for p ¬ 1.56, using functions (2.16) for the de- scription of the nonlinear moment Msp with exponents w = 1.5, 1.75, 1.9, 2.25, 3.0. Nonlinear effects were observed only in the first three resonant re- gions (ω1 = 0.089, ω2 = 0.261, ω3 = 0.376). One would expect that the 220 A. Pielorz, M. Skóra maximal displacement amplitudes should increase with the increase of the amplitude M0 of the external moment and with the increase of the expo- nent w. This is true up to the frequency p, when the solution begins to tend to infinity, i.e., when the escape phenomenon, known in nonlinear dynamics of discrete systems, occurs (cf. Stewart et al., 1995). For M0 =0.05, the solutions for all exponents w are practically the same and their diagrams are identical. For higher values of the amplitude M0 of the external moment one can notice that there exist intervals of the frequency p of the external moment where the solutions can diverge to infinity. The escape phenomenon can be observed in the first, in the first two or in the all three resonant regions, depending on w and M0. Namely, when M0 = 0.2, function (2.16) with w = 1.5 and w=1.75 gives always solutions behaving as a sinusoidal functionwith periods equal to the period of the external moment in the whole interval of p conside- red. The displacements are higher for w = 1.75. The solutions with function (2.16) are of a sinusoidal type with w = 1.9 for p > 0.066, with w = 2.25 for 0.095 < p< 0.183 and p > 0.275, and with w = 3.0 for p > 0.437 after the third resonance.When M0 =0.4, the solutions to equations (2.14) do not diverge to infinity in the whole considered regions for functions (2.16) only with w = 1.5, while the escape phenomenon does not occur with w = 1.75 for p > 0.077 after the first resonant region, with w = 1.9 in the first two resonant regions for 0.099 0.277, and with the both re- maining values of the exponent w after the third resonance for p> 0.444 and p> 0.544, respectively for w=2.25 and w=3.0. Dots in the diagrams plot- ted in Fig.4 denote the extreme values of the frequency pwhere the solutions cease to behave harmonically. One can notice that the angular displacements in these extreme frequencies have similar values for a fixed exponent w. For clarity, the amplitude-frequency diagrams for the dynamic moment Msp obtained by solving equations (2.14) with functions (2.16) are presented inFig.5 andFig.6, separately for M0 =0.05, 0.2 and M0 =0.05, 0.4with the same values of exponents equal to w=1.5, 1.75, 1.9, 2.25, 3.0. The diagrams in both figures include three resonant regions. For the amplitude M0 of the external moment equal to 0.05, the linear case is obtained and the solutions for the assumed values of w coincide with each other. When M0 = 0.2 (Fig.5), the nonlinear effects can be noticed, i.e., the escapephenomenon.Theappropriate intervals of pwhere the solutionsdiverge to infinity are the same as those for angular displacements shown in Fig.4. From Fig.5 it follows that the escape phenomenon occurs when the dynamic moment approaches the appropriatemaximum for functions (2.16) determined by the exponent w, the coefficient kw of the nonlinear term in functions (2.16) Torsional vibrations of discrete-continuous... 221 Fig. 5. Amplitudes of the dynamic moment Msp for kw =−0.005,M0 =0.05, 0.2 and with w=1.5, 1.75, 1.9, 2.25, 3.00 and the coefficient k1 of the linear term.For the kw =−0.005, k1 =0.05, these maximumvalues for functions (2.16) are 0.7407, 0.2189, 0.1499, 0.0916, 0.0609 for w=1.5, 1.75, 1.9, 2.25 and3.0, respectively. Thedots in diagrams inFig.5 denote extreme values of the frequency p of the external moment, where the solutions cease or begin to be harmonic vibrations with periods equal to the period of the external moment. In the case of w = 1.5 and w = 1.75, the maximum of function (2.16) is never achieved, so the solutions in the whole considered region of p are harmonically changing functions in time in the steady state, and they are higher for w = 1.5. The appropriate maximum is achieved for all the remaining values of w: in the first resonant region for w = 1.9, in the first two resonant regions for w = 2.25, and in the three resonant regions with w=3.0 for p< 0.436. From the amplitude-frequency diagrams for the dynamic moment Msp shown in Fig.6 it follows that when M0 = 0.4 the escape phenomenon can also be noticed. The appropriate intervals of pwhere the solutions diverge to infinity are the same as those for angular displacements shown in Fig.4 with M0 =0.4. One can see that the escape phenomenon occurs when the dynamic moment approaches the appropriate maximum for functions (2.16), similarly to the results presented in Fig.5 for M0 =0.2. The dots in diagrams in Fig.6 denote again extreme values of the frequency p of the externalmoment, where the solutions cease or begin to be harmonic vibrations with periods equal to the period of the external moment. In the case of w = 1.5, the maximum of function (2.16) is not achieved, so the solution in the whole considered region 222 A. Pielorz, M. Skóra Fig. 6. Amplitudes of the dynamic moment Msp for kw =−0.005,M0 =0.05, 0.4 and with w=1.5, 1.75, 1.9, 2.25, 3.0 of p in the steady state is a harmonic vibration. In the case of M0 =0.4, the maximum is achieved for all the remaining values of w: in the first resonant region for w = 1.75, in the first two resonant regions for w = 1.9, and in the three resonant regions for w = 2.25 and 3.0 in such a manner that the solutions with these exponents begin to behave as a harmonic vibration after the third resonance, namely for p> 0.444 and p> 0.544, respectively. The diagrams in Figs 4, 5 and 6 show the effect of the exponent w on the angular displacements and on the dynamic moment. From these diagrams it also follows that the amplitude M0 has a significant influence on the appro- priate solutions of the considered nonlinear problem. It seems to be important to find amplitudes of the external moment where the solutions are harmonic vibrations, dependingon the frequency p. For this reason, the application ran- ges of functions (2.16) are investigated for kw =−0.005, −0.004, −0.003 and for w = 1.75, 1.9, 2.25. Suitable curves are plotted in Fig.7. The curves for kw =−0.005 are marked by solid lines, for kw =−0.004 by dashed lines and for kw = −0.003 by dash-dash-dot-dot lines. They determine amplitudes of the externalmoment belowwhich the numerical solutions are harmonic vibra- tions with periods equal to the period of the external moment. The smallest values for M0 are acceptable in the neighbourhood of the resonances. One can also see that for a fixed kw the restrictions for the admissible values for M0 become stronger with the increase of the exponent w. Besides, there exists an interval of p where the admissible values of M0 increase in a linear man- Torsional vibrations of discrete-continuous... 223 ner. This interval occurs between the first and second resonances for each w and kw. Fig. 7. Application ranges of functions (2.16) with kw =−0.005 (solid lines), kw =−0.004 (dashed lines), kw =−0.003 (dash-dash-dot-dot lines) for w=1.75, 1.9, 2.25 Fig. 8. Diagrams of function (2.16) (solid line), function (2.17) (−·−·−) and function (2.18) (dashed line) The main aim of the paper is to investigate the possibility of application of the functions containing irrational functions in the description of the local nonlinearity inmulti-mass torsional systems.However, fromFigs 4-7 it follows 224 A. Pielorz, M. Skóra that there exist situations connected with the escape phenomenon where the use of functions (2.16) is not convenient. Then other nonlinear functions ought to be applied in the discussion of nonlinear discrete-continuous systems, inc- luding hyperbolic tangent functions (2.17) and exponential functions (2.18). Functions (2.16)-(2.18) are connected by relations (2.19), which means that they give the same linear case. Besides, functions (2.17) and (2.18) have the asymptotes Msp = ±Mmax where Mmax are maximum values of functions (2.16). This is seen from diagrams of functions (2.16), (2.17) and (2.18) with kw = −0.005, w = 2.5, k1 = 0.05 shown in Fig.8, together with the linear case. Fig. 9. Amplitudes of the dynamic moment Msp for kw =−0.005,M0 =0.4, w=1.75, 1.9, 2.25, 3.0 using functions (2.16) (solid lines), hyperbolic tangent function (2.17) (dashed lines) and exponential function (2.18) (dash-dot lines) Acomparisonof the results obtainedusing functions (2.16) (solid lines), hy- perbolic tangent function (2.17) (dashed lines) and exponential function (2.18) (dash-dot lines) as Msp is presented in Fig.9 with kw = −0.005, M0 = 0.4 and for w = 1.75, 1.9, 2.25, 3.0. This is done for the amplitude-frequency curves of the dynamic moment for p < 0.97. For p > 0.97, the suitable dia- grams coincide for all three functions describing the local nonlinearity having the characteristic of a soft type. From Fig.9 it follows that for a fixed p, the application of hyperbolic tangent function (2.17) gives usually slightly higher values of the solution than exponential function (2.18). The differences be- tween them become smaller for higher values of the exponent w. Moreover, Torsional vibrations of discrete-continuous... 225 the use of function (2.17) leads to results closer to the results corresponding to function (2.16) than to exponential function (2.18). As it follows from the diagrams in Fig.9, nonlinear functions (2.17) and (2.18) have no restrictions for their application in the description of the local nonlinearity. In the reso- nant regions, where the solutions with function (2.17) diverge to infinity, the solutions with functions (2.17) and (2.18) take the form of a plateau. 4. Final remarks Fromthe considerations in thepaper, it follows that nonlinear functions conta- ining irrational functions can be applied for the description of local nonlineari- ties with soft type characteristics in multi-mass discrete-continuous torsional systems using the wave approach. Appropriate studies on the nonlinear dyna- mics of discrete systemsmainly concern quadric or cubic nonlinearities. Here, problems for nonlinear vibrations of discrete-continuous systems are formula- ted, and the use of irrational functions enables one to assumenot only integers as exponents in local nonlinearities. Numerical calculations concern the effect of parameters occurring in non- linear functions containing irrational functions on the behaviour of the three- mass system. This effect is observed in the first three resonant regions. It appears that the irrational functions have some restrictions on their applica- tion which are connected with the occurrence of the escape phenomenon. The application ranges shown in Fig.7 include cases when the escape phenomenon is observed and where other nonlinear functions ought to be used instead of functions expressed by the irrational functions. In the present paper, a hyper- bolic tangent function and an exponential function are proposed for that. Analogous calculations were also performed for nonlinear vibrations of a two-mass torsional system. Nonlinear effects were observed in the first two resonant regions. These results are not given in the present paper, however conclusions from them coincide with those for the three-mass system. References 1. Boiler C., Seeger T., 1987,Materials Data for Cyclic Loading, Parts A-E, Elsevier, NewYork 2. Hagedorn P., 1981,Non-Linear Oscillations, Clarendon Press, Oxford 226 A. Pielorz, M. Skóra 3. Hale J., 1977, Theory of Functional Differential Equations, Springer-Verlag, NewYork 4. Muszyński J., Myszkis A.D., 1984,Ordinary Differential Equations [In Po- lish:Równania różniczkowe zwyczajne], PWN,Warszawa 5. Nadolski W., Pielorz A., 1997, Nonlinear dynamic loads on gear teeth in discrete-continuousmodel of a single gear transmission,Meccanica, 32, 165-168 6. Pielorz A., 1988, Application of wave method in investigation of drive sys- tems, comparisonwith othermethods,Mechanika Teoretyczna i Stosowana,26, 97-112 7. PielorzA., 1999,Non-linear vibrations of a discrete-continuous torsional sys- tem with non-linearities having characteristic of a soft type, Journal of Sound and Vibration, 225, 2, 375-389 8. PielorzA., SkóraM., 2006, Torsional vibrations of discrete-continuous sys- tems with local nonlinearity having hard type characteristics, Journal of The- oretical and Applied Mechanics, 44, 949-962 9. Ponomariev S.D. et al., 1957,ModernMethods for StrengthCalculations in Machine Design [In Polish: Współczesne metody obliczeń wytrzymałościowych w budowie maszyn], T.I, PWN,Warszawa 10. StewartH.B., Thompson J.M.T., UedaY., LansburyA.N., 1995,Opti- mal escape from potential wells – patterns of regular and chaotic bifurcation, Physica D, 85, 259-295 11. Szemplińska-StupnickaW., 1990,TheBehavior ofNonlinearVibrating Sys- tems, Vol. I, II, Kluwer Academic Publishers, Dordrecht 12. SzolcT., 2003,DynamicalAnalysis ofComplexDiscrete-ContinuousMechani- cal Systems [In Polish:Analiza dynamiczna złożonych, dyskretno-ciągłych ukła- dów mechanicznych], Habilitation Thesis, IFTRREPORTS, 2 13. Thomson W.T., 1981, Theory of Vibration with Applications, Englewood Cliffs: Prentice-Hall Drgania skrętne układów dyskretno-ciągłych z lokalną nieliniowością o charakterystyce typu miękkiego Streszczenie Wpracy rozpatrywane są nieliniowe drganiawielomasowychukładówmechanicz- nych odkształcanych skrętnie. Układy te złożone są z wałów połączonych dowolną liczbą brył sztywnych.W układach tych uwzględniono lokalną nieliniowość z charak- Torsional vibrations of discrete-continuous... 227 terystyką typu miękkiego. Lokalna nieliniowość opisana jest za pomocą funkcji za- wierających funkcje niewymierne oraz za pomocą dwóch innych funkcji nieliniowych. W rozważaniachwykorzystano podejście falowe.W analizie numerycznej skoncentro- wano się na zbadaniuwpływu lokalnej nieliniowościna zachowanie się rozpatrywanych układów oraz na wyznaczeniu zakresów zastosowania funkcji niewymiernych. Przy- kładowe obliczenia numeryczne dotyczą układu trzymasowego. Manuscript received February 27, 2008; accepted for print April 23, 2008