Delay differential equations (DDEs) are widely utilized as the mathematical models in engineering fields. In this paper, a method is proposed to analyze the stability characteristics of periodic DDEs with multiple time-periodic delays. Stability charts are produced for two typical examples of time-periodic DDEs about milling chatter, including the variable-spindle speed milling system with one-time-periodic delay and variable pitch cutter milling system with multiple delays. The simulations show that the results gained by the proposed method are in close agreement with those existing in the past literature. This indicates the effectiveness of our method in terms of time-periodic DDEs with multiple time-periodic delays. Moreover, for milling processes, the proposed method further provides a generalized algorithm, which possesses a good capability to predict the stability lobes for milling operations with variable pitch cutter or variable-spindle speed.
National Natural Science Foundation of China514053435150533611702192Tianjin Research Program of Application Foundation and Advanced Technology15JCQNJC0500015JCQNJC05200Innovation Team Training Plan of Tianjin Universities and CollegesTD12-5043Tianjin Science and Technology Planning Project15ZXZNGX002201. Introduction
Time-delay systems widely exist in engineering and science, where the rate of change of state is determined by both present and past state variables, such as machining processes [1, 2], wheel dynamics [3, 4], feedback controller [5, 6], gene expression dynamics [7], and population dynamics [8, 9]. However, for some of above applications, the time delay in the dynamic system may lead to instability, poor performance, or other types of potential damage. Therefore, it is necessary for engineers and scientists to research the dynamics of these systems to reduce or avoid such problems.
Compared to the finite dimensional dynamics for systems without time delay, time-delay systems have infinite-dimensional dynamics and are usually described by delay differential equations (DDEs). Their stability properties can be analyzed through obtaining the stability charts that show the stable and unstable domains. For example, a stable milling process can be realized by choosing the corresponding parameter from a stability lobe diagram (SLD), which is a function of spindle speed and depth of cut parameters. Thus, more and more attention has been paid on this issue and many analytical and numerical methods have been developed to derive the stability conditions for the system parameters.
By using the D-subdivision method, Bhatt and Hsu [10] determined stability criteria for second-order scalar DDEs. Budak and Altıntaş [11, 12] and Merdol and Altintas [13] proposed a method in frequency domain called multifrequency solution. By employing a shifted Chebyshev polynomial approximation, Butcher et al. [14, 15] presented a new technique to study the stability properties of dynamic systems by obtaining an approximate monodromy matrix. Insperger and Stepan [16–18] proposed a known method called semidiscretization method (SDM), which is based on the discretization of the DDEs and approximates their infinite-dimensional phase space by a finite discrete map in time domain. Bayly et al. [19] carried out a temporal finite element analysis for solving the DDEs, which are written in the form of a state space model and discretizing the time interval of interest into a finite number of temporal elements. Based on the direct integration scheme, Ding et al. [20], Liu et al. [21], and Jin et al. [22] used a full-discretization method to gain stability chart efficiently. Recently, Khasawneh and Mann [23] and Lehotzky et al. [24] presented a numerical algorithm called spectral element method. This method has good efficiency because of its highly accurate numerical quadratures for the integral terms.
SDM is a known and widely used method to determine stability charts for general time-periodic DDEs arising in different engineering problems. In this paper, based on SDM, a generalized method for periodic DDEs with multiple time-periodic delays is proposed to obtain the stability chart of DDEs. The structure of the paper is as follows. In Section 2, the mathematical model is introduced. In Section 3, two typical examples are used to verify the effectiveness of the proposed method. In Section 4, conclusions with a brief discussion are presented.
2. Mathematical Model
The general form of linear, time-periodic DDEs with multiple time-periodic delays can be expressed as(1)y˙t=A0yt+Atyt-∑j=1NBjtut-τjtut=Dyt,where y(t)∈Rn is the state, u(t)∈Rm is the input, A0 is a constant matrix, A(t) and Bj(t), respectively, are n×n and n×m periodic coefficient matrices that satisfy A(t)=A(t+T) and Bj(t)=Bj(t+T), j=1,2,…,N, and τj(t)=τj(t+T)>0, D is an m×n constant matrix, T is the time period, and N is the number of time-delays. Note that (1) can also be written in the form (2)y˙t=A0yt+Atyt-∑j=1NCjtyt-τjtwith Cj(t)=Bj(t)D.
Consider that the period T is divided into k number of discrete time intervals, such that each interval length Δt=T/k. Introduce symbol [ti,ti+1] to represent the ith time interval, i∈Z+. Here, ti means the ith time node and is equal to iΔt. Thus, considering the idea in [2], the averaged delay for the discretization interval [ti,ti+1] is defined as follows:(3)τi,j=1Δt∫titi+1τjtdt=τ0,j-τ1,jci,where(4)ci=k2π∫i2π/ki+12π/ksintdt,i=0,1,…,k-1.
The number of intervals mi,j related to the delay item τi,j can be approximately obtained by(5)mi,j=intτi,j+Δt/2Δt,where int(∗) indicates the operation that rounds positive number towards zero.
Substituting (3) into (2) and solving it as an ordinary differential equation over the discretization period [ti,ti+1] with initial condition x(ti)=xi, the following equation is derived:(6)yt=eA0t-tiyi+∫titeA0t-ξAξyξ-∑j=1NCjξyξ-τi,jdξ.
Substituting t=ti+1 into (6), then it can be equivalently expressed as(7)yti+1=eA0Δtyi+∫titi+1eA0ti+1-ξAξyξ-∑j=1NCjξyξ-τi,jdξ.In [ti,ti+1], A(t), Cj(t), y(t), and y(t-τi,j) are defined as follows: (8)At=Ai+Ai+1-AiΔtt-ti,Cjt=Ci,j+Ci+1,j-Ci,jΔtt-ti,yt=yi+yi+1-yiΔtt-ti,yt-τi,j=βi,jyi-mi,j+αi,jyi+1-mi,j,where Ai=A(ti), Ci,j=Cj(ti), yi=y(ti), αi,j=(mi,jΔt+Δt/2-τi,j)/Δt, and βi,j=1-αi,j. Here, it should be noted that the approximation of y(t-τi,j) in (8) is the same as that for the so-called zero-order SDM in [2].
Substituting (8) into (7) leads to(9)yi+1=Φ0+Fiyi+Piyi+1-∑j=1Nβi,jRi,jyi-mi,j+αi,jRi,jyi+1-mi,j,where(10)Fi=Φ1-2ΔtΦ2+1Δt2Φ3Ai+1ΔtΦ2-1Δt2Φ3Ai+1,Pi=1ΔtΦ2-1Δt2Φ3Ai+1Δt2Φ3Ai+1,Ri,j=Φ1-1ΔtΦ2Ci,j+1ΔtΦ2Ci+1,j.Clearly, Φ0, Φ1, Φ2, and Φ3 can be expressed as follows:(11)Φ0=eA0Δt,Φ1=∫0ΔteA0Δt-sds=A0-1Φ0-I,Φ2=∫0ΔtseA0Δt-sds=A0-1Φ1-ΔtI,Φ3=∫0Δts2eA0Δt-sds=A0-12Φ2-Δt2I,where I denotes the identity matrix. Let M=max(mi.j) and(12)Zi=colyi,yi-1⋯yi-M;then combining (9) and (12), one can be recast into a discrete map as(13)Zi+1=DiZi,where each Di matrix is given by(14)Di=Hi+1Φ0+Fi0⋯000I0⋯0000I⋯000⋮⋮⋱⋮⋮⋮00⋯I0000⋯0I0+∑j=1N0⋯-Hi+1αi,jRi,j-Hi+1βi,jRi,j⋯00⋯00⋯00⋯00⋯0⋮⋮⋱⋮⋮⋮0⋯00⋯00⋯00⋯0,where Hi+1=(I-Pi)-1. The horizontal position of the discrete input matrices Hi+1αi,jRi,j and Hi+1βi,jRi,j in (14) depends on the value of mi,j corresponding to τi,j and they, respectively, begin from the column of 2mi,j-1 and 2mi,j+1 for a single DOF system as opposed to the column of 4mi,j-3 and 4mi,j+1 for a two-DOF one.
Based on (13) and (14), the following mathematical expressions can be established by coupling the solutions of the k successive time intervals in period T: (15)Vk=ΦV0=Dk-1Dk-2⋯D1D0V0,where Φ is the Floquet transition matrix that gives the connection between Vk and V0. According to the Floquet theory, the stability of the system is determined using the following criterion. If the moduli of all the eigenvalues of the transition matrix Φ are less than unity, the system is stable. Otherwise, it is unstable.
Here, it should be noted that the matrix Vi can be reduced because only the delayed positions show up in the governing equation of the milling process. Thus, the size of the approximation vector in (14) could be reduced by removing the delayed values of the velocities, such that the size of vector Vi can be decreased to M+2 for a single DOF system and 2M+4 for a two-DOF system. This can give some additional improvement in the computational time for the proposed method.
3. Verification of Method
There are several numerical and semianalytical techniques to determine the stability conditions for periodic DDEs. However, most of them were developed with the aim of constructing stability charts for milling processes, such as the analysis of the milling system with runout [25], with variable pitch/helix cutter [26–30], with variable-spindle speed [31–33], or with serrated cutter [34, 35]. In order to verify the proposed method, two typical milling operations are chosen and considered. The first is the varying spindle speed process, which can be described by a DDE with time-periodic delay in general. The other is the milling process with variable pitch cutters, which is often characterized by a DDE with multiple delays. Both methods are known means to influence and to prevent chatter vibration in milling.
3.1. Milling with Varying Spindle Speed
Generally, the mathematical models for milling processes with spindle speed variation can be written as (16)y˙t=A0yt+Atyt-Btyt-τt;that is, (2) is degenerated into one with one-time-periodic delay. For a single DOF system in [2], the matrices in (16) have the form (17)A0=01-ωn2-2ζωn,At=Bt=00-Gtm0,where m is the mode mass, ωn is the natural frequency, ζ is the damping ratios, and G(t) is the specific directional factor and has the form(18)Gt=ap∑j=1NgjtsinϕjtKtcosϕjt+Krsinϕjt,where ap is the axial depth of cut, N is the number of teeth, Kt and Kr are the linearized cutting coefficients in tangential and radial directions, gj(t) denotes whether the jth tooth is cutting, and the angular position of tooth j is(19)ϕjt=2π60∫0tΩsds+j2πN,where Ω(s) is the spindle speed and is assumed to change in the form of a sinusoidal wave, which is periodic at a time period T=60/Ω0/RVF, with a nominal value, Ω0, and an amplitude Ω1=RVA×Ω0, as shown in Figure 1. For this sinusoidal modulation, the shape function is modeled as(20)Ωt=Ω0+Ω1sin2πTt=Ω01+RVA·sinRVF·2π60Ω0t,where RVA=Ω1/Ω0 is the ratio of the speed variation amplitude to the nominal spindle speed and RVF=60/(Ω0T) is the ratio of the speed variation frequency to the nominal spindle speed.
Schematic drawing of the sinusoidal modulation of the spindle speed.
To illustrate the effectiveness of the approach method for milling with spindle speed, the method and results in [2] are taken into consideration. Here, it should be noted that the delayed term is approximated by a linear function of time and the periodic coefficient is approximated by a piecewise constant function for the method in [2]. However, for the proposed method, the delayed term y(ξ-τi,j), the state term y(ξ), and the periodic terms A(ξ) and Cj(ξ) in (7) are all discretized by linear interpolation (see (8)). Thus, different policies are utilized in the process of equation approximations for two methods. Figure 2 illustrates the stability charts that correspond to the milling processes for RVA = 0.1 and for four different RVF values using the proposed method and the method in [2]. The parameters are as follows. The cutting processes using a 4-flute tool (N=4) with zero helix angles are considered under half-immersion up-milling. The cutting force coefficients are Kt=800×106N/m2 and Kr=300×106N/m2. The mode mass is m=3.1663kg, the natural frequency is ωn = 400 Hz, and damping ratios is ξ=0.02. It can be seen from Figure 2 that the results obtained via the proposed method in this paper are in close agreement with those in [2].
Comparison of stability charts for milling processes with sinusoidal spindle speed modulation with RVA = 0.1 in the high-speed domain: (a) RVF = 0.5; (b) RVF = 0.2; (c) RVF = 0.1; (d) RVF = 0.05.
Meanwhile, the computational times corresponding to every graphic in Figure 2 are also recorded to evaluate the efficiency of proposed method. Here, considering the assumption q1T=q2τ0 (τ0 is the tooth passing period) with q1 and q2 being relatively prime [2] and the equation q1/q2=RVF/N obtained consequently, if the resolution of τ0 is adopted as 40 and q1=1, the resolutions of period T are 320, 800, 1600, and 3200 for RVF = 0.5, 0.2, 0.1, and 0.05, respectively. For a 200×100 grid of the spindle speed and the depth of cut and a personal computer (Intel(R) Core(TM) i5-2300, 2.8 GHz, 3 GB), the computational times are, respectively, 436 s, 1026 s, 2012 s, and 4132 s corresponding to RVF = 0.5, 0.2, 0.1, and 0.05 for the proposed method as opposed to 1953 s, 4757 s, 9185 s, and 18341 s for the method in [2] using our own codes. Time costs reduce nearly by 70% for every case. Obviously, the low computational cost of our method is illustrated. The reason about the cost reduction can be explained as follows. The matrices Φ0, Φ1, Φ2, and Φ3 in (11) are dependent on spindle speed but on depth of cut. Consequently, they are not needed to calculate in the process of sweeping the range of the depth of cut for the proposed method. However, this is also necessary for the method in [2]. Thus, for a parameter plane formed by the spindle speed and the cutting depth and divided into a Ns×Nd size grid, the method in [2] must be calculated Ns×Nd×k times to obtain a stability chart, but only Ns times for the method in this paper.
3.2. Milling with Variable Pitch Cutter
Considering a system for milling process with variable pitch cutter [29, 30] as shown in Figure 3, the mathematical models can be written as(21)y˙t=A0yt+Atyt-∑j=1NBjtyt-τj;that is, the original system of (2) is degenerated into one with multiple delays in this case. τj is the pitch period corresponding to the pitch angle ψ (see the right graphic of Figure 3), and the matrices A0, A(t), and Bj(t) in (21) have the form(22)A0=00100001-ωnx20-2ζxωnx00-ωny20-2ζyωny,At=∑j=1NBjt,Bjt=00000000-apmxhxx,jt-apmyhxy,jt00-apmxhyx,jt-apmyhyy,jt00,where ζx and ζy are the damping ratios, ωx and ωy are the natural frequencies, and mx and my are the modal masses of the cutter. hxx,j(t), hxy,j(t), hyx,jt,, and hyy,j(t) are the cutting force coefficients for the jth tooth defined as(23)hxx,jt=gϕjtKtcosϕjt+Krsinϕjtsinϕjt,hxy,jt=gϕjtKtcosϕjt+Krsinϕjtcosϕjt,hyx,jt=gϕjt-Ktsinϕjt+Krcosϕjtsinϕjt,hyy,jt=gϕjt-Ktsinϕjt+Krcosϕjtcosϕjt.
Schematic mechanical model of a system for milling process with variable pitch cutter.
To illustrate the performance of the proposed approach on uniform and variable pitch milling tools, the frequency-domain method published in [29] is considered. The comparison of the results using the proposed approach and the method [29] is carried out for both uniform and variable pitch cutter milling, as shown in Figure 4. The main system parameters are down-milling, half-immersion, the number of the cutter teeth which is N=4, the natural frequencies which are ωx=563.6 Hz and ωy=516.21 Hz, the damping ratios which are ζx=0.0558 and ζy=0.025, the modal masses which are mx=1.4986 kg and my=1.199 kg, and the cutting force coefficients which are Kt=679×106N/m2and Kt=256×106N/m2. It can be seen from Figure 4(a) that the proposed method agrees closely with the results of analytical method for the uniform pitch cutter milling. For the variable pitch cutter as shown in Figure 4(b), two methods gain consistent predicting results, except for the high-speed domain at approximately 8500 rpm, where a clear deviation occurred. Reference [27] chose one point (ap = 5 mm and Ω = 8500 rpm) in this deviation and showed its stabilization by time-domain simulations. The reason of this phenomenon is as follows. The proposed method is based on the time-periodic cutting force coefficients (see (23)), rather than the simplified time-averaged ones in [29]. Thus, the stability prediction by our method is more reasonable and has better accuracy.
Comparison of the predicted stability lobes: (a) uniform pitch cutter with ψ= [90°, 90°, 90°, 90°] and (b) variable pitch cutter with ψ= [70°, 110°, 70°, 110°].
4. Conclusion
In this work, an improved semidiscretization algorithm is proposed to obtain the stability char for DDEs with multiple time-periodic delays. Two milling examples, variable-spindle speed milling system with one-time-periodic delay and variable pitch cutter milling system with multiple delays, are utilized to demonstrate effectiveness of the proposed algorithm. Through the comparison with prior works, it is found that the results gained by the presented method in this paper are in close agreement with those existing in the past literature. Moreover, the proposed method also has good computational efficiency. Here, it should be noted that if discussing the milling process only, the proposed method is a generalized algorithm, which can consider the milling processes with variable pitch cutter and variable-spindle speed simultaneously.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (51405343, 51505336, and 11702192), Tianjin Research Program of Application Foundation and Advanced Technology (15JCQNJC05000 and 15JCQNJC05200), Innovation Team Training Plan of Tianjin Universities and Colleges (TD12-5043), and Tianjin Science and Technology Planning Project (15ZXZNGX00220).
AltintasY.2012298New York, NY, USACambridge University Press10.1515/hzhz-2014-0136InspergerT.StépánG.2011178New York, NY, USASpringer Science & Business MediaApplied Mathematical Sciences10.1007/978-1-4614-0335-7Zbl1245.93004StépánG.Delay, Nonlinear Oscillations and Shimmying Wheels63Proceedings of the IUTAM Symposium on New Applications of Nonlinear and Chaotic Dynamics in MechanicsNew York, NY, USASpringer37338610.1007/978-94-011-5320-1_38Zbl0947.70528TakácsD.StépánG.HoganS. J.Isolated large amplitude periodic motions of towed rigid wheels2008521-227342-s2.0-3974915463810.1007/s11071-007-9253-ySunJ.Delay-dependent stability criteria for time-delay chaotic systems via time-delay feedback control200421114315010.1016/j.chaos.2003.10.018MR2033691ParkJ. H.KwonO. M.A novel criterion for delayed feedback control of time-delay chaotic systems200523249550110.1016/j.chaos.2004.05.023MR2089574VerdugoA.RandR.Hopf bifurcation in a DDE model of gene expression200813223524210.1016/j.cnsns.2006.05.001MR2360687Ruiz-HerreraA.Chaos in delay differential equations with applications in population dynamics2013334163316442-s2.0-8487212097910.3934/dcds.2013.33.1633Zbl1318.92042YanJ.ZhaoA.PengL.Oscillation of impulsive delay differential equations and applications to population dynamics200546454555410.1017/S1446181100009664MR2137310BhattS. J.HsuC. S.Stability criteria for second-order dynamical systems with time lag196633113118MR0207245BudakE.AltıntaşY.Analytical prediction of chatter stability in milling, Part I: General formulation1998120223010.1115/1.2801317BudakE.AltıntaşY.Analytical prediction of chatter stability in milling, Part II: application of the general formulation to common milling systems19981201313610.1115/1.28013182-s2.0-0003021259MerdolS. D.AltintasY.Multi frequency solution of chatter stability for low immersion milling2004126345946610.1115/1.17651392-s2.0-12344276621ButcherE. A.BobrenkovO. A.On the Chebyshev spectral continuous time approximation for constant and periodic delay differential equations20111631541155410.1016/j.cnsns.2010.05.037MR2736831Zbl1221.651582-s2.0-77957355390ButcherE. A.MaH.BuelerE.AverinaV.SzaboZ.Stability of linear time-periodic delay-differential equations via Chebyshev polynomials200459789592210.1002/nme.894MR20343402-s2.0-1442313921InspergerT.StepanG.Semi-discretization method for delayed systems200255550351810.1002/nme.505MR1925168InspergerT.StepanG.Updated semi-discretization method for periodic delay-differential equations with discrete delay200461111714110.1002/nme.1061MR2078678InspergerT.Full-discretization and semi-discretization for milling stability prediction: some comments201050765866210.1016/j.ijmachtools.2010.03.0102-s2.0-77955227706BaylyP. V.HalleyJ. E.MannB. P.DaviesM. A.Stability of interrupted cutting by temporal finite element analysis2003125222022510.1115/1.15568602-s2.0-0038399830DingY.ZhuL.ZhangX.DingH.A full-discretization method for prediction of milling stability20105055025092-s2.0-7794949187710.1016/j.ijmachtools.2010.01.003LiuY.ZhangD.WuB.An efficient full-discretization method for prediction of milling stability20126344482-s2.0-8486494074010.1016/j.ijmachtools.2012.07.008JinG.QiH.CaiY.ZhangQ.Stability prediction for milling process with multiple delays using an improved semi-discretization method201639494995810.1002/mma.3543MR34588412-s2.0-84959219077KhasawnehF. A.MannB. P.Stability of delay integro-differential equations using a spectral element method2011549-102493250310.1016/j.mcm.2011.06.009MR28346532-s2.0-80051585496LehotzkyD.InspergerT.StepanG.Extension of the spectral element method for stability analysis of time-periodic delay-differential equations with multiple and distributed delays20163517718910.1016/j.cnsns.2015.11.007MR34344722-s2.0-84951860464WanM.ZhangW.-H.DangJ.-W.YangY.A unified stability prediction method for milling process with multiple delays201050129412-s2.0-7044952291710.1016/j.ijmachtools.2009.09.009SimsN. D.MannB.HuyananS.Analytical prediction of chatter stability for variable pitch and variable helix milling tools200831766468610.1016/j.jsv.2008.03.0452-s2.0-48349138966JinG.ZhangQ.HaoS.XieQ.Stability prediction of milling process with variable pitch cutter201320132-s2.0-8487730922910.1155/2013/932013932013JinG.ZhangQ.HaoS.XieQ.Stability prediction of milling process with variable pitch and variable helix cutters201422822812932-s2.0-8489261030810.1177/0954406213486381AltintaşY.EnginS.BudakE.Analytical stability prediction and design of variable pitch cutters1999121217317810.1115/1.28312012-s2.0-0032690932JinG.ZhangQ.QiH.YanB.A frequency-domain solution for efficient stability prediction of variable helix cutters milling201422815270227102-s2.0-8491006519710.1177/0954406214522614InspergerT.StéfánG.Stability analysis of turning with periodic spindle speed modulation via semidiscretization200410121835185510.1177/10775463040448912-s2.0-11244303640ZatarainM.BediagaI.MuñoaJ.LizarraldeR.Stability of milling processes with continuous spindle speed variation: analysis in the frequency and time domains, and experimental correlation200857137938410.1016/j.cirp.2008.03.0672-s2.0-43849105274SeguyS.InspergerT.ArnaudL.DesseinG.PeignéG.On the stability of high-speed milling with spindle speed variation2010489-128838952-s2.0-7795360890610.1007/s00170-009-2336-9DombovariZ.AltintasY.StepanG.The effect of serration on mechanics and stability of milling cutters20105065115202-s2.0-7795094209110.1016/j.ijmachtools.2010.03.006DombovariZ.StepanG.The effect of helix angle variation on milling stability201213452-s2.0-8486701942910.1115/1.4007466051015