You are on page 1of 11
276 DESIGN AND ANALYSIS OF CNC SYSTEMS igure 6.9: A ballscrow drive system used for teashing CNC design, PI Current Loop ) = Kp ia — Kai) + Ki in Kai) (6.56) Le gy, cst) =f kG 6.57) Vi = Kip ia — Kai) + Kien 6.58) PI Velocity Loop in = (8;Ve - 0) (Kin Auxiliary state zy = + (6,V, —a) > =8,v,-0 (6.60) 3 a ig = Kip (SWe—0) + Kies 661) Rewriting the equation for $1 and ¥; by substituting ig 41 Ki -K,(8,V.—0) + Kun — 2) SF = ia ~ Koi = Kip (SWVe ~ 0) + Kut ~ Kyi 6.62) Vi = Kup (ia — Kai) + Kuti = Kop (Kip (SVe — 0) + Kuta ~ Kai) + Kun (6.63 Vi = KipKipS Ve — KepKipto + KopKita ~ KopKei + Kuen 6.64) 6.4 STATE SPACE MODEL OF FEED DRIVE CONTROL SYSTEMS By substituting V; into # di a KupKiSey _ KorKin + Setaey, — Sue, + ‘The states are combined as follows: ty » « 1 % -§ £0 0 Of 0 -3 4 Kiileaks BK x Kate g |] | | Katee g @ Spe eee Se =p ve Bla] -K, -K 0 Ky Of fat +| Kise 0 (rc a - o 0 0 off s of” a) | o 0 0 oly 0 0 (6.68) Because we are interested in physical variables such as velocity (w), current, (, equivalent current command (i,), motor voltage (Vi) © 1 0 0 0 Olfw 0 0 i ° 1 0 0 offi 0 0 Ye in} =| Kip 0 0 Ke Offzrt+] Kips, 0 {rt Vi | | Kirin —Kipk Ku KpKir Of 2x) | KipkipSy 0 |” ” Xe ° 0 0 0 1 hx oo (6.69) or in standard state space notation je} = [A] ee + [6] fe} = fA) [2] (6.70) 6} =[C] i) +[D] tw ‘The state space model of the machine indicated above represents the con- tinuous part of the physical system. If the system needs to be controlled in discrete time domain, its ZOH equivalent must be considered which leads to state space model of the system in z domain, The digital controller then can be combined to achieve a full state space model of the closed-loop system in discrete time domain, 27 278 DESIGN AND ANALYSIS OF CNC SYSTEMS Fein) San Sheps ape ts = Hee Figure 6.10: MATLAB Simolink block diagram of feed drive system, Example 10. A lead lag controller is designed to control a feed drive system whose MATLAB block diagram is given in Figure 6.10, The parameters of the drive are given as follows: J, =7 x 10~! [kg m*], B, = 0,00612 [Nmv/(rad/s)], K, =0.72 (Nm/A], Kz = 0.8871A/V1, K, = 20/(2n), Ky = 1 (Wimm] and sam- pling time of 7’ = 0.0002 (s] used. The solution procedure is given as follows: ‘The open-loop transfer function of the drive is obtained as follows: _ KKK _ Ko Cl) STs B) aust" om where the gain (Ky) and time constant (r)) are Ko = Si = 382.2 tmm/V] t= f= 0.1144 [a]. The ZOH equivalont of the system is obtained to represent the drive dynamics in the diserete time domain. yp Gul) ayy, Kale G2) = 0-229 = a 2 (6.72) 3 Fes Te) By applying partial fraction expansion rules, G,(s)/s can be transformed as follows: ta, a, Gs Sist1/n) s? 8 stilt 1 C.=tin? = 00” S64 T)t) in {42 (Cs) | tn aie x= lim {eras G > Feria a), (4 Bt =" (3 an) (8 644 STATE SPACE MODEL OF FEED DRIVE CONTROL SYSTEMS 279 ‘The ZOH equivalent of G,(s) is as follows: Gy(e) = a= =a- = Gye) = 22 (b:27* + bo) boz + br (6.74) Ag) ~ Fay tetas 1 hae hay" whore by = Ko[m(l -e"5) ~ Per] = 5.8014e — 005 by = Ko [1 ~ 1,(1 - e-)] = 5.80486 - 005 9983 a, =—(1+e"%) = -1.9988 Note that one of the open-loop poles is on the unit circle , while the other is very close to the unit circle, ie, 2? +a;2 + a9 = @— 1) — 0.9983). Assume that the position control loop is closed by a proportional controller with a gain K, [V/mm]. The proportional controller is designed both in Laplace and discrete time domain as follow. Laplace Domain Design ‘The ZOH is ignored and the closed-loop transfer function of the system is derived as follows: K,Go(s) K,Ko Cul) = TE Gay ~ Ta 81 KR ‘The roots of the __characteristie equation (pip = (-1+ yT=4KpKo) /(2r,)), that is, the poles of the system, start from open-loop poles. Ky =0—> pi =0, pp = —I/x, = 1/0.1144 = ~8.7413; become a complex conjugate when 1~4r.K,Ko=0 — K, = 1/(4s,Ko: (4 x 0.1144 x 382.2) = 0.0065783, which gives identical poles at Pi = po = -1/2t.) = -1/(2 x 0.1144) = 4.8706. If the proportional gain (K,) is increased further, the system has complex poles that indicate an underdamped, oscillatory step response. If we wish to have a damping ratio of ¢ = 0.8, the proportional gain is selected by forcing the desired characteristic 280 DESIGN AND ANALYSIS OF CNC SYSTEMS equation on the closed-loop system. P+ Deans bok (6.75) 1 + pag? t pa 8 +2 x 0.8 x os tah 1 ~ Ty0ax Oe 5.4633? x 0.1144 B22 = 5.4633 [rad/s) = 0.0108, Regardless of the magnitude of proportional gain K,, the system remains stable despite the increased oscillatory behavior of second-order continuous systems. Discrete Time Domain Design ‘The closed-loop transfer function of the system in 2 domain is Gate) = KoGol2 K Ko (boz +61) “HOS TAG) 25 (ai + K,Kobo)e + do + KpKabr Kobo) + Gat KyRabo)? — to > KpRoba) Pia= 5 with poles (ps, pe = (—(ax +X Kobo) /(@rTEpRabal= Way 1 KGRGP) /2) If K, =0 the poles start from the open loop poles of the system at py = 1. and pz = 0.9983. If0 < K, < 0.00658, the poles are on the real axis and the system is damped. The poles will meet at p = pz = 0.9991 when K, = 0.00658. IfK, > 30 The poles will eave the unit circle and the system will be unstable in diserete time domain, Load-Lag Compensator Design in Laplace Domain ‘The controller structure is given as C(s) = K(1 +asT)/(1 +s), where gain K and a, 7 are the parameters of the compensator. A compensator will be designed to achieve 60 deg phase margin at crossover frequency of «, = 60 Hz, From the Bode plot of the plant in s domain, at frequency of 60 Hz (377 rad/s), the phase of the plant G,(s) is found to be -179{deg]. Additional phase lead of ¢ = +59deg = 1.03 [rad] must be added by the lead compensator at ‘og = 60 Hz = 377 [rad/s _Ltsing _ 1+ sin 1.03 T-sing, T—sin1.03 1 1 Je 377VT3.015 = 13.015 x 10-4 65 SLIDING MODE CONTROLLER 281 After the parameters are calculated, the gain of the controller needs to readjusted to ensure a unity gain at frequency 60 Hz. The gain of the system is -22.7dB at y 60 Hz; hence, the gain needs to be increased by +2.7 dB. 20logK = +22.7 + K = 107” — 13. 646 Hence, the lead comper ing controller becomes 149.5693 x 10-%s ~ 13.646 (3525 x 10- 6.5 SLIDING MODE CONTROLLER Sliding mode controller is a typical example of robust, nonlinear control system. An application and design of a sliding mode controller to a feed drive mechanism is given here as an example [10]. Open-loop block diagram of a feed drive system is sown in Figure 6.11. ‘The ballscrew drive system application considered here is controlled in a ‘current mode, and its open loop transfer function between the table position (x ‘mm]) and amplifier command generated in the CNC (u[V1) is as follows: #08) = [KaKeuls) Te] Go (6.76) KiKits yg 1 ney sae) ~ RRO K 1 = [u(s) — ——T.(s)], Tee ~ REO where K = K,Kirg/d is the gain, p ~B/4 is the velocity loop pole, and 72(s) is the torque disturbance caused by friction and the cutting process. In the linear motor drives, the inertia (J) is replaced by the table-workpiece mass, and the disturbance torque (7) is replaced by the cutting force and friction in the linear motor-driven systems. The differential equation of the drive system Figure 6.11: Open oop block diagram ofa feed drive ayrtem pomered undercurrent rode ofthe ampifer Disturbance current Toraue amplifier kr. u KL fe Pee 1 © 18 ce rar SL ase} Ls Ls positon Input Motor Inertaand Gain Viscous Damping DESIGN AND ANALYSIS OF CNC SYSTEMS can be reorganized as follows ¥ stx(a) + Bax(s) = wis) -— Tu RPE) + ROO) =U) — FET 6.77) B 1 + EO =U — ETO x % ao or normalizing the disturbance torque with respect to input, Texlt) + B.x(t) = ut) de), (6.78) where the reflected disturbance at the input is d(¢) = 7.(¢)/(K,K,). The accel- eration of the drive can be separated from Eq. (6.78) as follows: 20) = +B) +ue) do) (6.79) ‘The objective of the feed drive control system is that the controller must. be capable of following the commanded trajectory and velocity with minimum error at high speeds, regardless of the slight variations in the inertia (J.), vis~ cous friction (B.), and external disturbances (4) such as Coulomb friction and cutting forces reflected on the drive input. The conventional drive control sys- ‘tems, such as pole placement and feed-forward control techniques work quite well when the system is modeled accurately, and external disturbances are rejected by having a large bandwidth in linear drives. In addition, large trans- mission ratio in the ballscrew drives reduce the reflected torque on the rotating drive motor shaft significantly. However, when the friction is not modeled and compensated accurately, and when the external forces such as in linear drives are not determined before, conventional control techniques do not exhibit high tracking accuracy in high-speed machining, leading to inaccuracies in con- tour machining operations. The sliding mode controller belongs to a class of nonlinear control strategies, which are robust to such uncertainties and time variation in the drive dynamics. ‘There are two fundamental steps in designing a sliding mode controller: the selections of a sliding surface and a Lyapunov function. For accurate tracking of position and velocity, which are the key requirements from high-speed machine tools, the sliding surface (S) is selected as follows: $=), =x) + -2), (6.80) where 2(1/s] is the desired but achievable bandwidth of the drive, x,, are the reference command and actual positions, and x, x are the reference command and actual velocities of the drive, respectively. The control input (u) must be manipulated in such a way that, after a finite time, both the position and veloc- ity of the drive approach reference command values (x —> x), — %,), hence, forcing the value of position and velocity errors to be zero, that is, S — 0. 65 SLIDING MODE CONTROLLER Inertia (J,) and viscous damping (B,) on the machine drive are assumed to change slowly and insignificantly. The variation in the external disturbance caused by the cutting process and friction is considered to be strong, but with upper (d*) and lower (d~) limits measured on the machine. The external dis- ‘turbance can be tracked by the following simple observer: d= pS, dik) =dik-1) + pxS-T, (6.81) where Ts] is the control period, & is the control interval counter in discrete time domain, p is the parameter adaptation gain (p ~ 0.005), and x is used to limit the integral control of the disturbance as follows: 0 it di) 0 is the control gain to be selected. The control law (u) is obtained from Eq. (6.86) as uk) = Je [AGC — 20h) +, (b)] + Betth) + dik) + KS) 687) Sth) = Alx-(k) — x(k)) + [x (k) —x(k)], where (f) is the control interval counter. The reference position, x,(k), velocity x,(h), and acceleration x,(e) are obtained from the command generation algo: rithm running in the CNC system. The actual position x(k) is measured from the encoder, and the actual velocity =(&) is estimated by taking the derivative of the measured position from the linear drive or measuring directly from a tachogenerator. However, evaluation of velocity and acceleration from discrete position commands and encoder readings may be noisy. The following simple low-pass filter can be used to smooth them: xe(h) = a-(k — 1) + bx-(h) — x(k) Se) = adh — 1) + be) -2-(k =) lee th) = anh D4 xm, where ip (it) is the measured velocity from the tachogenerator. The filter gain is typically set to a ~ 0.6. ‘The disturbance compensation through on-line estimation is useful to min- imize the effects of slowly varying cutting forces and friction. However, when the drive changes the direction of the velocity, especially at the corners and quadrants of the circular paths, the friction force reverses its direction and leaves glitches on the surface. If a major Coulomb friction force is known, it may be more advantageous to precompansate it at the feedforward command generation as follows: Uf = TP Kak) > Hk) > 0 uyeth) = 0 > x(k) =0 Wj, = 77 [(KoKi) alk) <0 where T?, are the magnitudes of Coulomb friction measured in positive and negative directions of the motion. Hence, the overall control signal is obtained as follows: U(h) = tame () +H pel) ‘The sliding mode controller can attenuate the influence of external distur- bances, while providing a good tracking performance with high bandwidth. The implementation of sliding mode controller is shown in Figure 6.12 16 ACTIVE DAMPING OF FEED DRIVES. stored > 4] St igure 6.12: Sliding mode controller with feedforward friction compensstion 6.6 ACTIVE DAMPING OF FEED DRIVES ‘The ballscrew drives exhibit torsional flexibility at the motor shaft-serew cou- pling, serew itself, and nut as shown in Figure 6.13. The axial displacement of the screw is coupled with its torsional flexibility, and the serew may expe- rience lateral flexibility which applies tension and compression loads on the table-guideway interface. The structural vibrations caused by the ballscrew assembly occur typically above the bandwidth frequency of the servodrive, i.e. more than 100 Hz. However they affect the surface finish quality and pre sion positioning accuracy during machining; hence, they need to be avoided. ‘The linear cutting force (Fj) and table mass (m,) are transmitted to motor as a reduced torque. The bandwidth and speed are increased by using two parallel ballserew drives in the most recent, high-speed machine tools. ‘The mechanical drive system is represented by its rigid body motion when the structural dynamic flexibility is neglected, as presented in previous sec- tions. However, it may be important to damp the structural dynamics of the machine that are excited by the cutting load and inertial forces during high- speed motions. ‘Transfer funetions determine the relationship between the forces and pc tions of the table and motor because of the flexibility of the mechanical drives, and they replace the rigid body-based transfer function. If the machine is built, the transfer functions G(s) are measured experimentally by applying white noise to the current amplifier to generate random torque (T,,), and the result- ing angular position of the motor shaft (63) and table position (x2) are measured from rotary and linear encoders, respectively. The direct transfer function at the table (Gz) is measured by applying impact load (F}) and measuring the table vibrations (x;) with the use of the accelerometer or displacement sensor. Alternatively, the transfer functions can be predicted from the finite element ‘model of the drive structure by assuming a damping constant. ‘A simple structural dynamic model of the ball-serew drive system can be approximated by the reflected inertias at the motor (Jn) and leadserew (Ji) connected by a torsional spring (f,) and damping (c;) elements as shown in Figure 6.13. 285 Xe, thle coupling screw. mat interface DESIGN AND ANALYSIS OF CNC SYSTEMS ‘The stiffness of the ballscrew guidewoy drive varies as a function of table position as shown in Figure 6.14. The static stiffness is determined primarily by the equivalent axial stiffness of the ball serew-nut con- tact as outlined in [DIN 69051- \ 6]. The ballscrew drive system, ballserew with bearings and the interme- diate transmission or clutch, has NM: a finite stiffhess that assists in determining the static displace- ment of the table under load Batcren! model during the high-speed position- k ing of the table. Tho ballscrow is m, WW) om, ] supported by thrust bearings at ra two ends. Bearings provide radial Linear drive model guidance to the screw and absorbs Figure 6.18: Torsional-axial vibrations of the positioning accuracy atthe table 3) allscrew affect the feed forces in the axial direc- Zaxajbearngs J oil sttiness Kg tp 1 xp bearing o 025 0507s 40 Sle postion xi. ‘thrust bearings (DIN 69051-61 tre 6.14: Axial stiffness of the ballscrew with single- and double-sided

You might also like