Professional Documents
Culture Documents
net/publication/334118949
Modified IMC Technique for Nonlinear Uncertain Milling CNC Machine Tool
System
CITATIONS READS
7 350
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Anil Kumar Yadav on 02 July 2019.
Abstract
The objective of this paper is to model a nonlinear computer numerical control machine tool (CNCMT) system for milling
process. A modified internal model control (M-IMC) technique is proposed for position control of machine table. The study
of milling process is mainly focused on the cutting mechanics and dynamics. The performance of CNCMT system is degraded
due to the presence of disturbance like cutting force and parametric uncertainty. Considering the limitation of conventional
internal model control (C-IMC) technique that is the poor performance under constant disturbance, an M-IMC technique is
designed for the rejection of disturbance and position control of uncertain nonlinear CNCMT system. The performances of
nonlinear uncertain CNCMT system using M-IMC technique is also compared with C-IMC and Ziegler–Nichols-tuned pro-
portional–integral–derivative control techniques. The proposed M-IMC technique is designed to attain the better disturbance
rejection capability, i.e., regulatory control with less control effort for the regulation of desired position of a machine table.
Keywords Milling CNC machine tool system · Disturbance rejection · Modified IMC and nonlinear system
13
Vol.:(0123456789)
Arabian Journal for Science and Engineering
even or uneven. The parametric uncertainty is due to lack of 2 Problem Formulation and Modeling
precise knowledge of the actual CNCMT system parameters, of a Nonlinear CNCMT System
and is called as uncertain CNCMT system. To achieve the for the Milling Process
desired position of machine table against the external distur-
bances and the parametric uncertainty, the M-IMC, i.e., IMC 2.1 System Description and Problem Formulation
with modified filter is proposed. However, more than 90% of
industrial controllers are PID type because it has simple struc- The components of the practical setup of CNCMT system for
ture, excellent performance, easy to design and inexpensive milling process and mechatronic diagram of CNCMT system,
[24–26]. Therefore, the performance of the nonlinear CNCMT i.e., assembly of DC servomotor, gearbox, machine table and
system using M-IMC technique is also compared with the PID lead screw shaft are shown in Fig. 1a, b, respectively. The cut-
controller. ting force at the tip of machine tool is transmitted to the feed
IMC strategy was introduced by Garcia and Morari [27], drive and spindle, i.e., x-axis and z-axis motors, respectively,
and the design concept of IMC is widely recognized that can as shown in Fig. 1a. Therefore, it is necessary to control or
achieve the robust performance for tracking and disturbance to reject the disturbances on motors of both x-axis and z-axis
rejection from the process output [28, 29]. The large and con- simultaneously. The input in the CNC machine tool is electri-
stant steady-state error (Ess) is the limitation of C-IMC tech- cal power to the power amplifier which gives the suitable volt-
nique, i.e., IMC with conventional filter [30], which fails in age applied to the DC servo motor, and the output variables
proper position alignment when machine table is subjected to might be the velocity and position of the machine table. Here,
constant cutting force. In the design of IMC, low-pass filter two feedback loops of position and velocity are used. The
plays a significant role for servo and regulatory control opera- velocity feedback loop provides the additional damping ratio
tions [31]. The applications of IMC for various systems such which in turn improves the OS, and is employed where OS is
as speed control of heavy-duty vehicle [30, 31], load frequency undesirable like in the milling process. The position control-
control in power system [32], speed control permanent magnet ler generates the control signal to the drive circuitry accord-
synchronous motor [33] and DC–DC boost converter [34] are ing to error generated from desired and actual positions of
discussed. the machine table. The desired actual positions are generated
In this research, two types of control problems such as from computer programs and encoder output, respectively, as
servo and regulatory are discussed. In servo problem, the pur- shown in Fig. 1a. For the regulation of z-axis motion, a sensor
pose of the controller is to track the desired position and speed is mounted to the spindle [16] which sends the feedback signal
of machine table and tool, respectively, whereas in regulatory (i.e., force, torque or vibration) to the control unit, i.e., PI-type
problem, the controller acts in such a way that the disturbance angular motion controller as shown in Fig. 1a.
whatsoever is swiftly minimized or is rejected completely.
The idea of this work is taken from the literature [4, 7, 16]. 2.2 Mathematical Modeling and Open‑Loop
The key contributions of the paper are summarized as follows: Behavior
(1) the explicit modeling of the nonlinear CNCMT system
for motion of both x-axis and z-axis, (2) design of modified 2.2.1 Mathematical Modeling for x‑ Axis Linear Motion
IMC for x-axis and ZN-tuned PI for z-axis controls and (3)
a comparative performance analysis among ZN-tuned PID In Fig. 1b, a DC servo motor is employed to rotate the lead
control, C-IMC and M-IMC techniques is presented. For the screw shaft of CNCMT system, which controls the position
comparative analysis, the performance parameters such as of machine table. Applying the Kirchhoff’s voltage law in the
overshoot (OS), rise time (RT), settling time (ST), Ess, inte- electrical loop of the DC servo motor as shown in Fig. 1b, the
gral of the absolute error (IAE) and integral of the square of relation is as follows:
the error (ISE) are studied to obtain the optimum all-round
performance of CNCMT system. The expected advantage of dia
Ea − Eb = Ra ia + La (1)
the M-IMC technique over C-IMC technique is to improve the dt
regulatory performance by reducing the Ess under constant where Ea is the applied voltage, ia is the armature current,
disturbance. It also decreases the relative variations in IAE Ra and La are the armature resistance and the inductance of
and ISE of the nonlinear CNCMT system with and without the motor, respectively. Eb is the(back electromotive force
)
random disturbance and noise as compared to PID control and (EMF) and is defined as Eb = Kb d𝜃m ∕dt , where Kb is the
C-IMC techniques. MATLAB/Simulink is used to simulate back EMF constant. The mechanical dynamics of motor
the results. shaft and lead screw shaft, i.e., output shaft is as follows:
d2 𝜃m d𝜃
Jm + Bm m + TL = Tm (2)
dt2 dt
13
Arabian Journal for Science and Engineering
Fig. 1 a Components of the practical setup of CNCMT system for milling process and b mechatronic diagram of CNCMT systems
13
Arabian Journal for Science and Engineering
linear velocity of machine table on lead screw shaft, i.e., (7), (9) and (10), the mechatronic coupling of DC servomo-
v(t) = ẋ 0 (t) , and sign(v) is the signum function of v and tor, gear backlash, lead screw shaft and machine table of
defined as sign(v) = 1 , if v > 0 , sign(v) = 0, if v = 0 and CNCMT system is obtained, which is represented by (11)
sign(v) = −1 , if v < 0. and (12).
The hard nonlinearity like gear backlash is discussed in
Kb N dx0 (t) di (t)
[8], which is generally expressed as follows: Ea − ⋅ = Ra ia + La a (11)
p dt dt
⎧ N �T − 𝜌�, Ṫ > 0 and T (t) = N �T (t) − 𝜌� ⎫
⎪ � L � L g � L �⎪ ( ) J B
Tg (t) = ⎨ N TL + 𝜌 , Ṫ L < 0 and Tg (t) = N TL (t) + 𝜌 ⎬ N ⋅ Kt ⋅ ia (t) − GD Fc (t) + Ff (t) − ⋅ ẍ 0 (t) − ⋅ ẋ 0 (t) = p ⋅ m ⋅ ẍ 0 (t)
p p
⎪ No change, otherwise ⎪
⎩ ⎭ (12)
(5) 2.2.2 Mathematical Modeling for z‑Axis Angular Motion
where 𝜌 is the backlash distance and N is the gear ratio. It
is noted that in the simulation, original backlash model is The dynamics of servomotor for z-axis angular motion is
used, whereas in controller design, the backlash model (5) described as follows [4, 16]:
is replaced by (6): 𝜏m (t) = kt i = Je 𝜔(t)
̇ + Be 𝜔(t) + 𝜏c (13)
Tg (t) = NTL (t) + d(TL (t)) (6) where Je and Be are the equivalent inertia and equivalent
where d(⋅) is a nonlinear function on TL (t) and TL (t) , and
̇ damping coefficient, respectively,𝜔(t) is the angular speed
bounded such as ||d(TL (t))|| ≤ 𝛼 where 𝛼 is the corresponding of z-axis motor. 𝜏m is the motor torque, kt is motor torque
upper bound [8]. Generally, the term d(TL (t)) is considered constant, i is current of z-axis motor and 𝜏c is the torque due
with disturbance torque with the uncertain bound. Therefore, to cutting force.
Tg (t) = NTL (t) and 𝜃m = N𝜃0 are considered. Put 𝜃m = N𝜃0 The block diagram of a nonlinear CNCMT system for
( ) ( ) control application of linear position along x-axis and angu-
in Eb = Kb d𝜃m ∕dt , and yield Eb = Kb ⋅ N ⋅ d𝜃0 ∕dt . Put
lar speed about z-axis is given by Fig. 2, which is designed
this value of Eb in (1), and (7) is obtained.
using Fig. 1, and (11)–(13). In Fig. 2, H2 is the gain of tacho-
d𝜃0 di generator, Ka is the gain of power amplifier and u is the
Ea − Kb N = Ra ia + La a (7) controlled input to the motor for motion of x-axis, whereas
dt dt
kai is the gain of current amplifier and u′ is the controlled
Put Tg (t) = NTL (t) in (3), after that put the value of TL
input to the motor for motion of the z-axis. In Fig. 2, a delay
from (2) to (3)
term e−Td s is incorporated which signifies the delays among
( ) rotational motions of motor and lead screw shafts, and linear
d2 𝜃0 d𝜃0 d2 𝜃m d𝜃m
J0 2 + B0 + TD = N Tm − Jm 2 − Bm motion of machine table, and it also improves the modeling
dt dt dt dt accuracy toward practical applications [32].
(8)
Put 𝜃̈m = N 𝜃̈0 in (8) and rearranging them
2.2.3 Open‑Loop Behavior of System for x‑Axis and z‑Axis
2 2 Motions
d 𝜃0 d𝜃 d𝜃
NTm − TD = N 2 Jm 2
+ J0 20 + N 2 Bm 0
dt dt dt From Fig. 2, it is clearly seen that the nonlinear CNCMT
d𝜃0 d 2 𝜃0 d𝜃 system has three inputs, i.e., u, u′ and Fc , and two outputs,
+ B0 ⇒ NTm − TD = J 2 + B 0 (9)
dt dt dt i.e., linear position x0 and angular speed 𝜔 ; therefore, it may
be called as multi-input multi-output (MIMO) system. The
where J and B are the equivalent inertia and damping coef- nominal values and parametric uncertainty of CNCMT sys-
ficient, respectively, and described as J = N 2 Jm + J0 and tem parameters are given in Table 1. In this table, H1 is the
B = N 2 Bm + B0. gain of encoder and Td is the delay time. The open-loop step
In order to establish the relation between the rotational response of actual nonlinear and linearized CNCMT systems
motion of lead screw shaft and linear position of machine with nominal values as given in Table 1 under the influence
table assuming that the work-in is equal to work-out that is of constant Fc as disturbance of 500 N is shown in Fig. 3a, b.
represented as follows: To show the effect of disturbance on open-loop per-
NTm (t) − TD (t) − J 𝜃̈0 (t) − B𝜃̇ 0 (t) = p ⋅ m ⋅ ẍ 0 (t) (10) formance of nonlinear and linearized CNCMT systems,
Fc is applied at 20 s. From Fig. 3a, b, the open-loop step
where m is combined mass of machine table and mass of
responses are unstable because it does not give bounded
workpiece, x0 is the linear position of machine table on lead
output for bounded input. In order to achieve the desired
screw shaft, the pitch p is defined as p = x0 (t)∕𝜃0 (t) . From
13
Arabian Journal for Science and Engineering
Fig. 2 Block diagram of a nonlinear CNCMT system for control application of x- and z-axis motion
Table 1 Nominal values and parametric uncertainty of CNCMT sys- system. In this paper, conventional low-pass filter like
tem 1∕(𝜇s + 1)n is replaced by a filter named
( ) as, modified filter
Symbols with nominal values Parametric uncertainty having structure as 𝜗s3 + 𝜆s2 + 𝜑s + 1 ∕(𝜇s + 1)n , where 𝜗 ,
𝜇 , 𝜆 , 𝜑 and n are filter parameters. In conventional filter, ′ 𝜇′
Ra 0.613 Ω ∈ [0.4, 0.8] maintains the position tracking and disturbance rejection of
La 2.158 mH ∈ [1, 3] the system. The ′ 𝜇′ effectively tackle the speed of response
Kb 0.34 Vs/rad ∈ [0.2, 0.5] because it is a time constant, whereas in modified filter the
Kt 1.01 Nm/A ∈ [0.5, 1.5] additional parameters 𝜗 , 𝜆 and 𝜑 are suitably designed to
kt 1.343 Nm/A ∈ [0.7, 1.9] improve the regulatory performance of the system. The
m 105 kg ∈ [75, 140] parameters 𝜗 , 𝜆 and 𝜑 are contributing in the zeros of the
[ ]
Jm 3.7 ∗ 10−3 Kgm2 ∈ 2 ∗ 10−5 , 5 ∗ 10−4 system; therefore, the modified filter incorporated the fea-
J0 0.00154 Kgm2 ∈ [0.0005, 0.0025] tures of adding zeros in the control system like disturbance
Je 0.0316 Kgm2 ∈ [0.02, 0.05] rejection. Therefore, IMC with modified filter is proposed
[ ]
Bm 1.34 ∗ 10−4 Nms/rad ∈ 0.5 ∗ 10−5 , 3 ∗ 10−3 in this paper, which improves the regulatory control, i.e.,
B0 0.002 Nms/rad ∈ [0.001, 0.003] disturbance rejection. The modified filter is considered as
Be 0.009 Nms/rad ∈ [0.008, 0.01] third order because the order of modeled nonlinear CNCMT
Ka 2 ∈ [1, 3] system is three. The structure, i.e., order of modified filter
kai 6.4841 ∈ [5, 8] can be changed according to order of system or internal
Fco 10 N ∈ [5, 15] model (IM) that is derived from original system.
Fs 20 N ∈ [15, 25] For the design of controller, a linearized model of
Fv 15 N ∈ [10, 20] CNCMT system is required. Taking the Laplace transforms
Vs 0.1 m/s ∈ [0.08, 0.12] of (11) and (12), and for simplicity neglecting the small
𝛿 2 ∈ [1.5, 2.5] value of La [8], yields (14a) and (14b):
Td 0.2 s ∈ [0.1, 0.3] ( )/
Kb ⋅N
𝜌 0.75 – Ia (s) = Ea (s) − s ⋅ X 0 (s) (14a)
p
a R
p 5 mm –
N 6:1 (N2 ∕N1 ) –
H1, H2 1, 1.2 – ( ) J
N ⋅ Kt ⋅ Ia (s) − GD Fc (s) + Ff (s) − ⋅ s2 ⋅ X0 (s)
p
B
speed and position of machine tool and machine table − ⋅ s ⋅ X0 (s) = p ⋅ m ⋅ s2 ⋅ X0 (s) (14b)
p
respectively; classical PI and PID controllers, and C-IMC
and MIMC techniques are designed via linearized system Put the value of Ia (s) from (14a) into (14b), yields
and simulated with original nonlinear system. Here, it is ( ) (15) by
assuming the disturbance torque TD = GD Fc + Ff = 0 , i.e.,
noted that the nonlinearities due to frictional model and disturbances due to cutting and frictional forces are zero.
gear backlash are omitted while considering the linearized
13
Arabian Journal for Science and Engineering
∫
= pNKt ∕(p mRa + Ra J)∕ s +
Ea (s) (p2 mRa + Ra J) u(t) = KP e(t) + KI e(t)dt + KD e(t)
̇ (18)
(16)
From Fig. 2 and (16), the open-loop transfer function of [ ]
TI ∫
1
CNCMT system from delayed output X0 (s) to controlled input u(t) = KP e(t) + e(t)dt + TD e(t)
̇ (19)
U(s) is as follows:
where KP , KI and KD are proportional, integral and derivative
Ka ⋅ N ⋅ Kt ⋅ p/ 2 gains, respectively, TI and TD are the integral and derivative
X0 (s) (p mRa + Ra J) −T s
G(s) = = ( ) ⋅e d (17) time constants, respectively, and u(t) and e(t) are output and
U(s) N 2 K K +BR +K NK H
s s+ input to the controller, respectively. The parameters of pro-
t b a a t 2
(p2 mR +R J)
a a
portional (P), proportional–integral (PI) and PID controllers
The linearized CNCMT system (17) is a second-order are computed using ZN tuning method [25] using relations
plus time delay system which has one pole at the origin;
hence, the open-loop CNCMT system is marginally stable.
13
Arabian Journal for Science and Engineering
as follows: for P controller: KP = 0.5Ku , for PI controller: 3.3 Modified IMC (M‑IMC) Technique
KP = 0.4Ku and TI = Tu ∕1.2, and for PID controller:
Since IMC with F(s) has limitation for the rejection of dis-
KP = 0.6Ku , TI = Tu ∕2, TD = Tu ∕8, KI = KP ∕Ti and KD = KP ⋅ TD .
turbance; therefore, F(s) of (22) is replaced by modified filter
(20)
F � (s) which is as follows:
where Ku and Tu are the ultimate gain and period of the sys-
tem, respectively. 𝜗s3 + 𝜆s2 + 𝜑s + 1
F � (s) = (24)
(𝜇s + 1)n
3.2 Conventional IMC (C‑IMC) Technique
where 𝜗 , 𝜆 and 𝜑 should satisfy the condition as given in
The generalized architecture of conventional IMC is presented (25) for poles p1 , p2 and p3 of IM, and n = 5 or 6 , depend-
in Fig. 4. It has IM Gm (s) , controller C(s) , which has inverse ing upon the requirement to make controller proper. The
IM, i.e., G−1 proposed filter can be modified according to the order and
m (s) and low-pass filter F(s) , the uncertain CNCMT
system G(s) and the cutting force as disturbance d , ẽ is the type of the IM.
matching error between CNCMT system and IM, and Xd (s)
Lim (1 − T(s)) = 0, i = 1, 2 and 3. (25)
and X0 (s) are desired and actual positions of the machine table, s→pi
respectively.
The complete rejection of disturbances, perfect position where T(s) = C(s) ⋅ Gm (s) (26)
tracking and the minimizing the effect of noise is achieved if From (22) and (24)
Gm (s) = G(s) , and Gm (s) has breakup as follows:
1 𝜗s3 + 𝜆s2 + 𝜑s + 1
Gm (s) = G−m (s).G+m (s) (21) C(s) = ⋅ (27)
G−m (s) (𝜇s + 1)n
where G−m (s) is minimum phase term and G+m (s) is non-mini-
mum phase term. G+m (s) contains zeros at right-hand part and Put Gm (s) = G−m (s) ⋅ G+m (s) from (21) and C(s) from (27)
the time delay; its steady-state gain must be unity. Thus, the in (26), we get
steady-state gain of Gm (s) will remain in G−m (s) that is used
to design of C-IMC C(s) which is as follows: 𝜗s3 + 𝜆s2 + 𝜑s + 1
T(s) = G+m (s) ⋅ (28)
(𝜇s + 1)n
1
C(s) = − ⋅ F(s)
Gm (s) (22) For the calculation of filter parameters which are
𝜗, 𝜆and 𝜑 , G+m (s) = e−Td ⋅s is considered as shown in (17).
where F(s) is a low-pass filter. The conventional low-pass From (28), T(s) is as follows:
filter is described as:
( )
e−Td ⋅s 𝜗s3 + 𝜆s2 + 𝜑s + 1
(23)
n
F(s) = 1∕(𝜇s + 1) T(s) = (29)
where 𝜇 is the tuning filter parameter, which maintains (𝜇s + 1)n
position tracking, disturbance rejection and robustness of
From (29), put the value of T(s) in (25) and simplifying
the system, and n is an integer and selected on the basis of
it, we get (30):
C(s) when it becomes proper or semi-proper for physical
[ ( )]
realization. (𝜇s + 1)n − e−Td ⋅s 𝜗s3 + 𝜆s2 + 𝜑s + 1
Lim =0
s→pi (𝜇s + 1)n
(𝜇s + 1)n
⇒ 𝜗s3 + 𝜆s2 + 𝜑s + 1 = (30)
e−Td ⋅s
For Lim , putting s = p1 , p2 and p3 in (30), (31)–(33)
s→p1, p2, p3
are obtained as a result which are as follows:
( )n
𝜇p1 + 1
𝜗p31 + 𝜆p21 + 𝜑p1 + 1 = (31)
e−Td ⋅p1
( )n
𝜇p2 + 1
𝜗p32 + 𝜆p22 + 𝜑p2 + 1 = (32)
Fig. 4 Block diagram of the conventional IMC for CNCMT system e−Td ⋅p2
13
Arabian Journal for Science and Engineering
13
Arabian Journal for Science and Engineering
x 10
-3
According to a proper balance in all the performance
6
specifications as given in Table 2, 𝜇 = 0.5 is selected for
5 C-IMC and M-IMC of nonlinear CNCMT system for further
study in this paper.
4
The C-IMC CC - IMC (s) with 𝜇 = 0.5 and n = 2 , and from
Position (m)
Fig. 7 Response of CNCMT
system with 𝜇 = 0.5, 1and1.5 of
C-IMC techniques
13
Arabian Journal for Science and Engineering
4.2 Comparative Tracking Performance Under Table 3 Parameters of ZN-tuned P, PI, PID and manual PID control-
Constant Disturbance lers
x-/z-axis motion Controller KP KI KD
For the design of P, PI and PID controllers, Ku and Tu are
For x-axis ZN-P 1078.75 – –
evaluated using Routh table (see Appendix 1). The param-
ZN-PI 863 1569.10 –
eters of P, PI and PID controllers are calculated using (20)
ZN-PID 1294.20 3957.80 106.80
and given in Table 3. The response of nonlinear CNCMT
Manual PID 800 1700 90
system with ZN-tuned PID controller is compared with ZN-
For z-axis ZN-PI 1.2 210 –
tuned P, PI and manual PID controllers as shown in Fig. 9a.
The ZN-tuned PID controller gives undesirable OS nearly
200%. Therefore, the PID controller’s parameters are manu-
ally adjusted based on the knowledge and called as manual controller and C-IMC technique. The large Ess in position is
PID controller. From Fig. 9a, the manual PID controller undesirable for position control applications because under
reduces the OS from 200 to 50% and gives a better regula- this condition desired position can never be achieved and
tory performance as compared to ZN-tuned P, PI and PID cannot be tolerated for CNC milling operations like cutting,
controllers. Hence, the manual PID controller is selected for boring, drilling and finishing.
further comparative analysis.
The desired position regulation and disturbance rejection 4.3 Robustness Analysis Under Parametric
of system using PID controller, C-IMC and M-IMC tech- Uncertainty
niques for 𝜇 = 0.5 for comparative performance analysis are
presented in Fig. 9b. In same figure, the desired position is To check the suitability of proposed controller for the
assumed to be unity and a constant cutting force of 500 N uncertain nonlinear CNCMT system, the parametric uncer-
for 20–40 s is applied. The speed response about z-axis of tainty as given in Table 1 is taken into consideration, and
CNCMT system using ZN-PI controller is presented in the response with PID control, C-IMC and M-IMC tech-
Fig. 9c. niques is presented in Fig. 10a–c, respectively. The position
From Fig. 9b, the maximum deviation and constant Ess in response in Fig. 10a–c is plotted after taking nominal, lower
the position of machine table under constant disturbance are (L) and upper (U) limit values of CNCMT system as given
nearly 11.4% using C-IMC, whereas the maximum devia- in Table 1. The performance index of the uncertain nonlin-
tion in position about 9.8% and it is regulated within 4.7 s ear CNCMT system with PID control, C-IMC and M-IMC
using M-IMC. Hence, the improvement in constant distur- techniques is given in Table 4.
bance rejection capability using M-IMC is observed as com- From Table 4 and Fig. 10a–c, it is clearly seen that the
pared to C-IMC technique. The nonlinear CNCMT system effect of parametric uncertainty, i.e., L and U limit values
under constant disturbance with C-IMC does not approach of the nonlinear CNCMT system on servo performance in
the desired position of 1 m; it is saturated at position of terms of OS and ST is very less, i.e., narrow bands of OS
0.886 m. The milling CNCMT system with PID controller and ST. The narrower bands of performance specifications
gives higher servo OS and better regulatory, i.e., disturbance give similar performance under parametric variations, i.e.,
rejection performance as compared to C-IMC and M-IMC the robustness of the nonlinear CNCMT system toward
as shown in enlarged view of Fig. 9b, whereas C-IMC gives the parametric uncertainty is achieved with the proposed
better servo (set position tracking without disturbance) controller.
performance as compared to PID and M-IMC techniques. From Fig. 10a, the position response with PID control-
According to proper balance in servo and regulatory control, ler gives high OS, i.e., nearly 151% while considering the
M-IMC is better for nonlinear CNCMT as compared to PID upper limit value and is undesirable. The Ess in the position
Fig. 8 Response of nonlin-
ear CNCMT system with
𝜇 = 0.1, 0.5and1 of M-IMC
13
Arabian Journal for Science and Engineering
of machine table is neither conversing nor constant after 4.4 Comparative Performance Analysis Under
disturbance using C-IMC with the lower limit values as seen Non‑deterministic Disturbance and Noise
in Fig. 10b. From the enlarged view of Fig. 10c, the effect of
regulatory performance with change in parameters to upper Figure 11a, b shows the position response of the nonlinear
and lower limits of CNCMT system on position of machine machine table without and with non-deterministic distur-
table, the maximum variation in position is seen as low bance and sensor noise, respectively. The corresponding
when the upper limit values of parameters are selected; it is controller output u(t) , i.e., control effort for position response
high when the lower limit of parameters is selected, and it is of nonlinear CNCMT system is shown in Fig. 12a, b. The
nominal with nominal values of system parameters. Hence, non-deterministic disturbance due to cutting force is shown
the proposed control technique is well suited for position in Fig. 13 that signifies uneven shape and size of work-
tracking and disturbance rejection of uncertain CNCMT sys- piece. The noise signal that is also known as sensor noise
tem for less low limit and high upper limit values of system is usually generated during measurement of output position
parameters. from encoder as shown in Fig. 14. The performance index
of controlled nonlinear CNCMT system with and without
non-deterministic disturbances and noise in terms of IAE
and ISE is presented in Table 5.
13
Arabian Journal for Science and Engineering
Fig. 10 Response of uncertain
nonlinear CNCMT system with
a PID control b C-IMC and c
M-IMC techniques
Table 4 Performance indexes of controllers for CNCMT with nomi- The values of IAE and ISE are determined using
nal, L and U limit system parameters IAE = ∫0 |e(t)|dt and ISE = ∫0 e2 (t)dt after taking T = 60
T T
13
Arabian Journal for Science and Engineering
Fig. 11 Response of nonlin-
1.5
ear CNCMT system with all
Position (m)
designed controllers a without 1
and b with non-deterministic
disturbance and noise 0.5
desired PID C-IMC M-IMC
0
0 10 20 30 40 50 60
t (s)
(a)
1.5
Position (m) 1
0.5
desired PID C-IMC M-IMC
0
0 10 20 30 40 50 60
t (s)
(b)
Fig. 12 Output of controllers x 10
4
0 1 2 3 4 5 45 46 47 48 49 50
-10
0 10 20 30 40 50 60
t (s)
(a)
(b)
ting force
Disturbance
40
20
0
0 10 20 30 40 50 60
t (s)
13
Arabian Journal for Science and Engineering
Fig. 14 Noise signal
2
Noise
0
-2 2
0
-4 -2
-6 20 20.2 20.4 20.6 20.8 21
0 10 20 30 40 50 60
t (s)
PID 2.001 0.8226 5.016 × 106 48.35 61.57 2.435 × 1011 2316.3 7384.8 4.8535 × 105
C-IMC 3.90 1.692 5.061 × 104 4.82 1.734 5.82 × 104 23.12 2.48 13.89
M-IMC 4.759 1.99 1.562 × 105 5.695 2.008 1.66 × 105 19.67 0.9045 6.27
Fig. 15 Position response of
the nonlinear CNCMT system
with and without backlash using
M-IMC technique
and M-IMC techniques, whereas the M-IMC technique gives higher OS in servo control, and better regulatory
improves the varying disturbances and sensor noise rejec- performance under constant disturbance rejection as com-
tion capability in terms of percentage relative variations in pared to C-IMC and M-IMC techniques, whereas C-IMC
IAE and ISE of the nonlinear CNCMT system, i.e., variation gives better performance under varying disturbance as
between the conditions without and with disturbance and compared to PID control and M-IMC techniques. A com-
noise as compared to PID control and C-IMC techniques. parative performance analysis of PID control, C-IMC and
Hence, it can be concluded that the nonlinear CNCMT sys- M-IMC techniques vividly show that according to proper
tem with M-IMC gives better non-deterministic disturbance balance in uncertainty, servo and regulatory (under both
and noise rejection capability in terms of percentage relative constant and varying disturbances) performances, M-IMC
variations in IAE, ISE and energy content of control effort as is better in all performances except servo OS as compared
compared to PID control and C-IMC techniques. The least to PID controller and C-IMC technique. The M-IMC has
value of energy content of the control effort signifies that the eliminated the disturbance within desired value, which is
minimum physical energy is required to drive the system. highly desirable for position control applications where the
disturbance and noise signal lead to more energy loss and
degrade the position accuracy.
5 Conclusion The expected advantage of the M-IMC over C-IMC
technique is also achieved. The proposed control technique
In this paper, a milling CNCMT system (for both x-axis and may also be applicable for reconfigurable milling CNCMT
z-axis) along with PI and PID controls, C-IMC and M-IMC systems that enhances the manufacturing efficiency and
techniques is designed, and implemented successfully for also for disturbance rejection of other systems. For better
constant and varying disturbance rejection due to cutting performance, PID controller’s parameters can be obtained
force. The nonlinear CNCMT system with PID controller using various optimization techniques such as genetic
13
Arabian Journal for Science and Engineering
algorithm, particle swam, ant colony and cuckoo search Figure 10a: https: //anilei 007-gmail. tinyt a ke.com/sf/
which are suggested as future scopes of this research. MzQ0MzMyMF8xMDMxNDIwNA
Figure 10b: https: //anilei 007-gmail. tinyt a ke.com/sf/
MzQ0MzMyMV8xMDMxNDI3Nw
Appendix 1 Figure 10c: https: //anilei 007-gmail. tinyt a ke.com/sf/
MzQ0MzMyNV8xMDMxNDI4MQ
Considering the linearized system (17) in series with Ku , Figure 11a, b: https://anilei007-gmail.tinytake.com/sf/
apply first-order Pade approximation for delay term and put MzQ0MzMyN18xMDMxNDI5NQ
the values from Table 1. yields (39): Figure 12a, b: https://anilei007-gmail.tinytake.com/sf/
( ) MzQ0MzMzMl8xMDMxNDMyNA
0.0045 1 − 0.1s
G(s) = ⋅ Ku (39) Figures 13 and 14: https://anilei007-gmail.tinytake.com/
s(0.0055s + 1) 1 + 0.1s
sf/MzQ0MzMzNl8xMDMxNDMyOQ
In order to apply Routh criteria, perform 1 + G(s) = 0 and Figure 15: https : //anile i 007-gmail . tinyt a ke.com/sf/
(40) is obtained. MzQ0MzMzN18xMDMxNDMzMA
( )
0.00055s3 + 0.1055s2 + 1 − 0.00045Ku s + 0.0045Ku = 0
(40)
Simplified (40), gives (41)
References
( )
s3 + 191.82s2 + 1818.2 − 0.82Ku s + 8.2Ku = 0 (41) 1. Altintas, Y.; Verl, A.; Brecher, C.; Uriarte, L.; Pritschow, G.:
Machine tool feed drives. CIRP Ann. Manuf. Technol. 60(2),
Formulate the Routh array for (41) 779–796 (2011)
2. Zhang, X.; Zhang, J.; Zhang, W.; Liang, T.; Liu, H.; Zhao, W.:
s3 1 1818.2 − 0.82Ku Integrated modeling and analysis of ball screw feed system and
s2 191.82 8.2Ku milling process with consideration of multi-excitation effect.
191.82(1818.2−0.82Ku )−8.2Ku
Mech. Syst. Signal Process. 98, 484–505 (2018)
s1 0 3. Altintas, Y.; Aslan, D.: Integration of virtual and on-line machin-
191.82
s0 8.2Ku ing process control and monitoring. CIRP Ann. Manuf. Technol.
66(1), 349–352 (2017)
4. Zhang, K.; Yuen, A.; Altintas, Y.: Pre-compensation of contour
191.82(1818.2−0.82Ku )−8.2Ku errors in five-axis CNC machine tools. Int. J. Mach. Tools Manuf
From s1 = 0 ; = 0 ⇒ Ku
191.82 74, 1–11 (2013)
= 2157.50 . For Tu , formulate the auxiliary equation from 5. Koren, Y.: Control of machine tools. J. Manuf. Sci. Eng. 119,
the s2 row: 191.82s2 + 8.2Ku = 0 ⇒ s2 = −92.23 , put 749–755 (1997)
s = jw, w = 2𝜋f andf = 1∕Tu ; then, Tu = 0.66 is obtained. 6. Jeong, Y.H.; Min, B.K.; Cho, D.W.; Lee, S.J.: Motor current pre-
diction of a machine tool feed drive using a component-based
simulation model. Int. J. Precis. Eng. Manuf. 11(4), 597–606
(2010)
7. Choi, B.K.; Choi, C.H.; Lim, H.: Model-based disturbance attenu-
Appendix 2 ation for CNC machining centers in cutting process. IEEE/ASME
Trans. Mechatron. 4(2), 157–168 (1999)
8. Shi, S.; Lin, J.; Wang, X.; Xu, X.: Analysis of the transient back-
This appendix presents the website addresses of run- lash error in CNC machine tools with closed loops. Int. J. Mach.
ning MATLAB simulation as screenshot video where it is Tools Manuf. 93, 49–60 (2015)
uploaded. 9. Huang, W.S.; Liu, C.W.; Hsu, P.L.; Yeh, S.S.: Precision control
Figure 3a, b: https://anilei007-gmail.tinyt ake.com/sf/ and compensation of servomotors and machine tools via the dis-
turbance observer. IEEE Trans. Ind. Electron. 57(1), 420–429
MzQzOTcxOV8xMDI5Nzg4Mw (2010)
Figure 7: https: //anilei 007-gmail. tinyta ke.com/sf/MzQ0M 10. Miao, E.M.; Gong, Y.Y.; Niu, P.C.; Ji, C.Z.; Chen, H.D.: Robust-
zI5OF8xMDMxNDE0NQ ness of thermal error compensation modeling models of CNC
Figure 8: https: //anilei 007-gmail. tinyta ke.com/sf/MzQ0M machine tools. Int. J. Adv. Manuf. Technol. 69(9–12), 2593–2603
(2013)
zMwNF8xMDMxNDE2Mw 11. Yang, M.; Yang, J.; Ding, H.: A two-stage friction model and its
Figure 9a: https : //anile i 007-gmail . tinyt a ke.com/sf/ application in tracking error pre-compensation of CNC machine
MzQ0MzMxN18xMDMxNDIwMA tools. Precis. Eng. 51, 426–436 (2018)
Figure 9b: https : //anile i 007-gmail . tinyt a ke.com/sf/ 12. Yeh, S.S.; Su, H.C.: Development of friction identification meth-
ods for feed drives of CNC machine tools. Int. J. Adv. Manuf.
MzQ0MzMxMV8xMDMxNDE4Mg Technol. 52(1–4), 263–278 (2011)
Figure 9c: https : //anile i 007-gmail . tinyt a ke.com/sf/ 13. Aslan, D.; Altintas, Y.: Prediction of cutting forces in five-axis
MzQ0MzMxOV8xMDMxNDIwMw milling using feed drive current measurements. IEEE/ASME
Trans. Mechatron. 23(2), 833–844 (2018)
13
Arabian Journal for Science and Engineering
14. Jamaludin, Z.; Brussel, H.V.; Swevers, J.: Friction compensation 25. Ang, K.H.; Chong, G.; Li, Y.: PID Control system analysis,
of an XY feed table using friction-model-based feed forward and design, and technology. IEEE Trans. Control Syst. Technol. 13(4),
an inverse-model-based disturbance observer. IEEE Trans. Ind. 559–576 (2005)
Electron. 56(10), 3848–3853 (2009) 26. Yadav, A.K.; Gaur, P.: Improved self-tuning fuzzy proportional-
15. Sparhama, M.; Sarhanb, A.A.D.; Mardia, N.A.; Hamdia, M.; integral derivative versus fuzzy-adaptive proportional–integral–
Daharic, M.: ANFIS modeling to predict the friction forces in derivative for speed control of nonlinear hybrid electric vehicles.
CNC guideways and servomotor currents in the feed drive system J. Comput. Nonlinear Dyn. 11(6), 061013–061017 (2016)
to be employed in lubrication control system. J. Manuf. Process. 27. Garcia, C.E.; Morari, M.: Internal model control-1: a unifying
28(1), 168–185 (2017) review and some new results. Ind. Eng. Chem. Process Des. Dev.
16. Aslan, D.; Altintas, Y.: On-line chatter detection in milling using 21(2), 308–323 (1982)
drive motor current commands extracted from CNC. Int. J. Mach. 28. Liu, T.; Gao, F.: New insight into internal model control filter
Tools Manuf. 132, 64–80 (2018) design for load disturbance rejection. IET Control Theory Appl.
17. Sencer, B.; Dumanli, A.: Optimal control of flexible drives with 4(3), 448–460 (2010)
load side feedback. CIRP Ann. Manuf. Technol. 66(1), 357–360 29. Shamsuzzoha, M.; Lee, M.: IMC-PID controller design for
(2017) improved disturbance rejection of time-delayed processes. Ind.
18. Fan, S.X.; Ryozo, N.; Fan, D.P.: Double-loop robust tracking Eng. Chem. Res. 46(7), 2077–2091 (2007)
control for micro machine tools. Sci. China Technol. Sci. 54(11), 30. Yadav, A.K.; Gaur, P.: Speed control of an uncertain heavy duty
3054–3063 (2011) vehicle using improved IMC technique. Arab. J. Sci. Eng. 42(7),
19. Kim, D.; Jeon, D.: Fuzzy-logic control of cutting forces in CNC 2981–2991 (2017)
milling processes using motor currents as indirect force sensors. 31. Yadav, A.K.; Gaur, P.: Intelligent modified internal model control
Precis. Eng. 35(1), 143–152 (2011) for speed control of nonlinear uncertain heavy duty vehicles. ISA
20. Jee, S.; Koren, Y.: Adaptive fuzzy logic controller for feed drives Trans. 56(3), 288–298 (2015)
of a CNC machine tool. Mechatronics 14(3), 299–326 (2004) 32. Saxena, S.; Hote, Y.V.: Load frequency control in power systems
21. Molina, A.; Ponce, H.; Ponce, P.; Tello, G.; Ramirez, M.: Artificial via internal model control scheme and model-order reduction.
hydrocarbon networks fuzzy inference systems for CNC machines IEEE Trans. Power Syst. 28(3), 2749–2757 (2013)
position controller. Int. J. Adv. Manuf. Technol. 72(9–12), 1465– 33. Li, S.; Gu, H.: Fuzzy adaptive internal model control schemes
1479 (2014) for PMSM speed-regulation system. IEEE Trans. Ind. Inf. 8(4),
22. Stephens, M.A.; Manzie, C.; Good, M.C.: Control-oriented mod- 767–779 (2012)
eling requirements of a direct-drive machine tool axis. J. Dyn. 34. Kobaku, T.; Patwardhan, S.C.; Agarwal, V.: Experimental evalu-
Syst. Meas. Control 134(5), 054503-1-6 (2012) ation of internal model control scheme on a DC–DC boost con-
23. Bort, C.M.G.; Leonesio, M.; Bosetti, P.: A model-based adaptive verter exhibiting nonminimum phase behavior. IEEE Trans. Power
controller for chatter mitigation and productivity enhancement in Electron. 32(11), 8880–8891 (2017)
CNC milling machines. Robot. Comput. Integr. Manuf. 40, 34–43
(2016)
24. Yadav, A.K.; Gaur, P.: An optimized and improved STF-PID speed
control of throttle controlled HEV. Arab. J. Sci. Eng. 41(9), 3749–
3760 (2016)
13