This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 1

A Linear Quadratic Regulator Based Optimal

Direct Thrust Force Control of Linear
Permanent Magnet Synchronous Motor
M. A. M. Cheema, Student Member, IEEE, John Edward Fletcher, Senior Member, IEEE
Dan Xiao and M. F. Rahman, Fellow, IEEE


Abstract—Linear permanent magnet machines are often Since DTFC is conceptually identical to DTC it
characterized by low inductance and short pole pitch which leads inherits the same disadvantage resulting in larger ripple in the
to a small operational range of load angles. The resultant control response of thrust force caused by the use of hysteresis based
performance using conventional direct thrust force control
techniques is poor with high force ripple. This research improves
flux and thrust force controllers and variable switching
this aspect of direct thrust force control. A novel multiple-input- frequency. This variable switching is caused by the use of
multiple-output (MIMO) state space model, independent of the hysteresis based controllers with a switching table to generate
mover’s speed, having stator flux and thrust force as states, is the reference voltage.
formulated for the linear PMSM. An optimal linear state Duty-ratio based DTC [9]-[10] and Model Predictive DTC
feedback control scheme is then designed using the optimal linear (MP-DTC) [11]-[17] schemes are presented for PMSMs that
quadratic regulator technique. Integral action is added to the
designed control scheme by state augmentation to minimize the
effectively reduce the ripple in the stator flux and torque
steady-state error and reduce the force ripple. Experimental simultaneously. However, these control schemes are
results clearly prove that the proposed optimal control scheme parameter dependent and a parameter mismatch can
results in a faster transient response of speed and force with deteriorate the performance of these schemes. Moreover, this
improved steady-state regulation of force and flux when duty-ratio based DTC and MP-DTC control schemes are
compared to the state of the art. computationally exhaustive and generally result in an
increased as well as variable switching frequency.
Index Terms— Direct Thrust Force Control (DTFC), Linear
Quadratic Regulator (LQR) and Linear Permanent Magnet Space vector modulation (SVM) is one of the popular
Motor (Linear PMSM). methods that can be employed in the implementation of DTC
to reduce the ripple in both torque and flux response and to
I. INTRODUCTION achieve a fixed switching frequency. A method for SVM
he potential application areas for linear machines include based DTC with reference flux vector calculator (RFVC-DTC)
T automation of manufacturing processes, transportation, is reported in [18]-[19]. RVFC-DTC has only one PI
controller to control the torque and necessarily requires the
renewable energy devices and pumping applications. This
broadening application scope for linear machines necessitates tuning of gains according to the controlled motor; however the
robust, simple and fast control schemes for linear drive relation between the gains and torque response is still not
systems. Direct torque control (DTC) for rotational permanent clarified in [18]-[19]. Another approach for SVM based DTC
magnet synchronous motors (PMSMs) [1]-[3] and direct thrust utilizing the decoupled control of torque and flux using two PI
force control (DTFC) for linear PMSMs [4]-[8] have become a controllers is detailed in [20]. Simulation of SVM-based
widespread control schemes due to their simple structure and DTFC is given in [21]. In this research the modulation index
fast dynamic response. The key feature of DTC/DTFC lies in for SVM is calculated by deciding the thrust force error using
the simple structure of the control schemes which makes the a static gain, but the tuning of that gain is not reported.
control performance robust to parameter variation. The main Another SVM based DTFC, derived from [20]; utilizing
disadvantage of DTC, due to the nature of its control structure, decoupled control of thrust force and flux is reported in [22].
is the presence of large ripple in the torque and stator flux. The main limitation of this technique is the tuning of gains for
the PI controllers using conventional techniques hence does
not yield in optimal transient response in terms of rise time
Manuscript received August 14, 2015; revised November 16, 2015; and steady-state error.
accepted December 15, 2015. The linear quadratic regulator (LQR) based control provides
Copyright © 2016 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be a systematic way of designing the control law for optimal
obtained from the IEEE by sending a request to pubs-permissions@ieee.org. performance of linear multivariable systems [23]. In [24]-[27]
M. A. M. Cheema, John Edward Fletcher, Dan Xiao and M. F. Rahman are LQR-based optimal vector control of PMSM in dq-axes
currently with the School of Electrical & Telecommunication Engineering,
The University of New South Wales, Sydney, Australia, 2052. reference frame using linear state feedback is reported.
However, the key limitation of the work in [24]-[27] is that the

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 2

state feedback gains are speed dependent and have to be II. DYNAMIC MODEL OF THE LINEAR PMSM
computed separately for every operating point as one set of The mathematical model of a surface-mount linear PMSM
gains cannot guarantee the optimal control performance for can be expressed in mover’s flux vector (dq), stator flux vector
the whole operational range of the speed. Therefore, the LQR- (xy) and stationary (αβ) frames of reference. These frames and
based vector control of [24]-[27] is computationally extensive. their orientations with respect to each other are illustrated in
So far, there is no literature available related to the the Fig. 1.
application of linear quadratic control theory to DTFC of The α-axis is aligned with the machine’s phase a-axis.
linear PMSM. In this research a direct thrust force control Mathematically, the dynamic model of the surface-mount
scheme for the linear PMSM based on optimal linear quadratic linear PMSM in the xy-reference frame can be expressed in
regulator is presented. A novel multiple-input multiple-output space vector notation as:
(MIMO) state space model independent of the mover speed
⃗⃗⃗⃗
and having stator flux and thrust force as states is formulated ⃗⃗⃗ ⃗⃗ ⃗⃗⃗ (1)
for the linear PMSM.
This model allows direct control of the thrust force and flux where, ⃗⃗⃗ [ ] is the stator voltage vector, ⃗⃗⃗
using linear state feedback law comprising of system states, [ ] is stator flux vector, ⃗⃗ [ ] is the stator
i.e. stator flux and thrust force. Subsequently, an optimal current vector and is the angular speed in electrical rad/s of
linear state feedback control law is synthesized using the the stator flux vector with respect to α-axis as shown in Fig. 1.
linear quadratic regulator approach. Integral action is added to The steady-state space vector diagram of surface-mount
the proposed optimal control scheme by state augmentation. linear PMSM according to (1) is shown in Fig. 1. It is
Experimental results demonstrate the improvement in terms of important to note that in case of a small resistive drop the
steady-state and transient force and flux response of the
resistive drop term ⃗⃗ in (1) can be neglected and the stator
proposed approach under various operating conditions when
voltage vector ⃗⃗⃗ co-insides with ⃗⃗⃗ becoming
compared to the prior technique of [22].
The dynamic model of linear PMSM is given in section II. perpendicular to ⃗⃗⃗ . From Fig. 1, it is clear that ⃗⃗⃗ is aligned
The analysis of thrust force control, formulation of the novel with the x-axis, therefore, and so that (1) can
state space model of linear PMSM and its controllability be expressed in the xy-reference frame as:
analysis is given in section III. The existing state of the art
technique [22] is summarized in section IV. Section V (2)
describes the fundamentals of optimal linear quadratic control (3)
with integral action.
Section VI details the linear quadratic based DTFC as well In (2) and (3), is the magnitude of the stator flux vector
as the state augmentation of the novel state space model to ⃗⃗⃗ (Wb). represents the stator resistance in (Ω).
incorporate integral action. Experimental results and
conclusions are provided in sections VII and VIII respectively. A. Estimation of Stator Flux Magnitude and Thrust Force
The flux linkages for the surface-mount linear PMSM in dq-
axes are given as (Fig. 1):
I s Rs q  axis
y  axis (4)
  axis
Vs
(5)
d s
s 
dt where, , , and are dq-axes flux linkages and currents
js s iq Is s x  axis
respectively. is the permanent magnet flux (Wb) and is
q  Ls iq the machine inductance (H).
d r As illustrated in Fig. 1, the dq-axes components of stator
 s r 
dt flux given by (4) and (5) can be converted to αβ-axes flux
id f Ls id d  axis linkages as:
r

d  Ls id   f [ ] [ ][ ] (6)

As shown in Fig. 1, is the angle (in elec. rad.) between

  axis the mover’s flux vector (produced by mover’s permanent
Fig. 1. Steady-state space vector diagram according to (1). Showing the stator magnet) and the α-axis. The angle is linked to mover’s
and mover’s flux linkages with respect to dq, xy and αβ-frames of reference mechanical displacement (in meters) as:
for surface-mount Linear PMSM. Current is arbitrarily shown, under
maximum force per ampere control becomes zero according to (14).
(7)

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 3

where, is the number of pole pairs, is the pole-pitch in A. Selection of Reference Stator Flux Magnitude
meters. The stator flux magnitude and angle (in elec. The stator flux reference is selected according to
rad., as shown in Fig. 1) can be expressed as:
maximum force per ampere (MFPA) by ensuring , and
for a surface-mount linear PMSM can be expressed in terms of
√ (8) thrust force reference as:

( ) (9) √ ( ) (14)
The thrust force for a surface-mount linear PMSM can
be expressed in terms of as [21]: The prototype surface-mount linear PMSM being used in
this research has a peak rated force of 312 N. According to
(10) (14), when varies from 0 to 312 N, the corresponding
variation in is negligible, i.e. from 0.0846 to 0.0847 Wb
where, is a constant which is used to quantify the end
effects in the linear PMSM [5] and [28]. A typical value of because of low values of and for the prototype linear
for linear PMSM is 0.9 [28]. The load angle between stator PMSM as given in Table I. Therefore, is set to 0.0846
and mover’s flux vector is: Wb in the prototype system for ease.

(11) B. Characteristics of the Co-efficient for Surface-Mount

Linear PMSM with Low Stator Inductance and Short Pole-
III. ANALYSIS OF THRUST FORCE CONTROL IN Pitch
SURFACE-MOUNT LINEAR PMSM According to (12) the maximum force occurs when is
In PMSMs, the load angle is closely linked to torque controlled at radians. However a closer inspection of (12)
generation as suggested by several studies [1]-[2] and [29]- suggests that, when is kept constant at (as explained
[32]. These studies are also analogously valid for linear previously), then for a machine with small value of the pole-
PMSMs. The relation between the electromagnetic thrust force pitch and low stator inductance , the rated maximum force
and the load angle for a surface-mount linear PMSM is of the linear PMSM is typically achieved at values of much
given as [13]: smaller than and therefore, has a small operational range of
. The primary motivation for this research is that the low
(12)
inductance and short pole pitch are the key factors that make
The studies in [31] and [32] detail the linearization of the surface-mount linear PMSMs difficult to control under DTFC
nonlinear relation between the load angle and torque for compared to their rotational counterparts. If is the load
interior PMSMs. The linearization methods discussed in [31]- angle corresponding to the rated maximum thrust force ,
[32] can also be applied to surface-mount linear PMSM by then from (11) is:
linearizing (12) using Taylor’s expansion at load angle as:
( ) (15)
( ) (13)
For example, as described previously, the stator flux for the
where, ( )⁄ | , and ( ). In (13), is
prototype system is Wb, and from (15)
called as the linearization co-efficient and for a linear PMSM
is given as:
it is evaluated at leading to .
radians (i.e., degrees)
TABLE I
PARAMETERS OF LINEAR PMSM PROTOTYPE FOR EXPERIMENT
Hence, the operational range of for the prototype linear
PMSM according to (15) is [- ] radians which
Pole Pairs 3 is a small subset of [ ]. This justifies the assumption that
Max. Current (A) 3.27 for the small range of values for .
Therefore, using (12) and (13), the linearization coefficient
PM Flux (mWb) 84.6
for the prototype linear PMSM with stator flux is set to
(mH) 1.95 can be approximated as:
(ohms) 3.01 , ( ) (16)
(m) 0.0256
From the above discussion it is clear that the small
Mover’s Mass operational range of load angle results in an almost constant
1.25
M (kg) linearization coefficient for the linear PMSM when the
Rated Cont. Force (N) 52 reference flux is selected using (14). Therefore the thrust force
Max Peak Force (N) 312 varies linearly with the load angle within its operational
range as demonstrated in Fig. 2.

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 4

( )
300 (21)
200
Force (N)

100 Equation (21) can be expressed as:

0
-100 ( ) (22)
-200 Exact
-300 Linearized where, is the angular speed of mover’s flux vector in elec.
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 rad/s. From (20) and (22), after simple mathematical
 (elec. rad)
manipulation the linear differential equation governing the
Fig. 2. Thrust force vs load angle for surface-mount linear PMSM of Table I, thrust force regulation at can be expressed as:
the exact curve is according to (12) and the linearized curve is according to
(13) with Wb, according to (14) under (MFPA).
(23)
2100
2060
2020 where:
K (N/elec. rad)

1980
1940
1900 The dynamics of stator flux regulation can be obtained from
1860 (2) as:
1820
1780
1740
1700
(24)
0 25 50 75 100 125 150 175 200 225 250 275 300
Force FT (N)
Equations (23) and (24) can be combined in matrix form to
Fig. 3. Characteristics of according to (17) with Wb. formulate a novel state space model that governs the dynamics
of thrust force and stator flux regulation in the linear PMSM:
It is also important to evaluate the effect of thrust force on
the linearization coefficient . In DTFC; the stator flux tracks [ ] [ ][ ] [ ]* +
the reference flux , and the thrust force tracks
reference force . Therefore, from (12) and (13), is
expressed in terms of and as: * +[ ] (25)

Evidently (25) presents a 2nd order state space model

( ) (17) comprising flux and force as states. This state space
√ ⏟ model provides the linear time invariant governing dynamics
of the flux and force regulation in the linear PMSM and
It is important to note that for the prototype linear PMSM provides the basis for formulating a linear state feedback
control law forcing the states to track the desired trajectories.
is vanishingly small because of low inductance and short pole
It is noteworthy that the state space model presents the force
pitch and thus the effect of force on becomes negligible and
regulation dynamics as a type-0 servo system which does not
(17) reduces to (16). According to (17), the variation in
include any integrator. Equation (25) can be expressed as:
magnitude for the prototype linear PMSM as reference
force and hence the actual force varies from 0 to 312 N ( ) ( ) ( ) ( ) (26)
( ) is less than 1.2% and remains almost constant as
shown in Fig. 3. where:
C. Formulation of a Novel State Space Model of the Linear ( ) [ ] ( ) [ ] ,
PMSM for DTFC
The expression for in terms of xy-reference frame =[ ], =[ ], =* +
variables can be given as [6] and [22]:

(18) and ( ) [ ]

Substituting the value of from (18) into (3) yields: The term ( ) in (26) can be regarded as disturbance. The
dynamic system of (26) has and as state variables. These
(19) states can be estimated using (4) to (10), and are also the
natural choices for the outputs of the system to be controlled.
Since, in DTFC the value of is kept constant at ,
Therefore, the output equation of the system can be given as:
therefore, (19) can be written as:
(20) [ ]=* +[ ] (27)
Since from Fig. 1, therefore, from (13):
Equation (27) can be expressed as:

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 5

( ) ( ) (28) λref vx xy v
+- PI Inverter
where: ( ) [ ] and is a 2×2 identity matrix. vref Fref vy v (VSI)+ LPMSM
SV-PWM
+
- PI + PI
- 
D. Controllability of The Novel State Space Model abc

It is of critical importance to establish the controllability of s i i
the state space model of (26) to rationalize its formulation. vm FT λs id
Estimation of  
The controllability matrix for the 2nd order dynamic system of λs ,s and FT
r

(26) can be given as:  r d Eq. (4) to (10) iq dq

P
 dt
[ ] (29)
Fig. 4. State of the art PI controller based Direct Thrust Control of linear
Substitution of values of and from (26) into (29) yields: PMSM [22].

x(t)  s FT 
T T T
[ ] (30) r(t )  ref Fref   u(t )= vx v y 
 
+_  KI +_ B ++  C

In (30), is independent of systems states, i.e. and Integral Action A

and is independent of the mover’s speed as well. The rank of Plant K

is always 2 which is the same as the order of the system. y(t)  s FT 
T

Therefore, the system described by (26) is always controllable

for all values of the inputs and as well as the Fig. 5. Type-0 servo system with full state feedback and integral action.

disturbance ( ) [23].
The integral and proportional components of the PI
controller are in feed forward path resulting in a comparatively
IV. STATE OF THE ART PI BASED DTFC
slow transient response when zero overshoot is desired. This
In order to reduce the ripple in the thrust force response of performance limitation of the state of the art PI based DTFC is
the linear PMSM a space vector modulation based DTFC illustrated experimentally and is used as a benchmark.
scheme has been proposed recently [22]. This DTFC scheme
is based on the concept of decoupled control of the thrust force V. LINEAR QUADRATIC REGULATOR BASED STATE FEEDBACK
and the stator flux. The block diagram of this scheme is shown CONTROL WITH INTEGRAL ACTION
in Fig. 4. The DTFC control architecture proposed in the
The novel state space model of linear PMSM given by (26)
scheme comprises two PI controllers; one for the thrust force
is a type-0 MIMO servo system which does not involve an
control and second for the stator flux control. is generated
integrator action in thrust force dynamics. The integral action
by using another PI controller having speed error as an input. needs to be incorporated in the system to ensure a zero steady-
It can be observed from Fig. 4 that the flux and force errors state error. Special steps are required to be taken when
are the inputs to the PI controllers to generate the reference designing the state feedback law to include the integrator [23].
voltages and , which are then converted to αβ-reference The block diagram of state feedback control of a general
frame voltage components and . According to Fig. 4, type-0 servo system utilizing the integral action is shown in
and are utilized to generate the control signals for the Fig. 5. The state variables and according to the
voltage source inverter using space vector pulse width dynamics of (26) are also illustrated by Fig. (5).
modulation (SV-PWM) scheme [22]. It can be observed from The integral action is achieved by using state augmentation
Fig. 4 that the electrical speed of the mover is achieved by by inserting the integrator in the feed forward path between
differentiating the mover’s electrical position from the the error comparator and the plant which adds new states to
position sensor. The speed is related to the mechanical the system [23]. The mathematical formulation of linear
speed of the mover, as: quadratic regulator (LQR) based state feedback control with
integral action is detailed in the following subsections.
(31)
A. Mathematical Formulation of Error Dynamics
Moreover, in Fig. 4 shows the reference speed. The The state space model of a general type-0 MIMO servo
main limitation of this approach is that the gains of the PI system utilizing integral action, using state augmentation, as
controllers are tuned using conventional techniques hence do shown in Fig. 5, can be expressed mathematically as:
not yield optimal response in terms of rise time and steady-
state error under various conditions. ̇( ) ( ) ( ) (32)
( ) ( ) (33)
̇ ( ) ( ) ( ) ( ) (34)
where:
[ ] is the (n ) state vector

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 6

[ ] is the (m ) output vector ̅ * + , and ̅ * +, and from (35) and (40) it is

( ) [ ] is the (m ) input vector
evident that the control law for error dynamics of (42) can be
( ) [ ] is the (p ) reference-input vector expressed as:
[ ] is the (p ) vector containing integral of
the tracking errors (augmented states) ( ) ( ) ( ) (43)
(n ) Constant matrix (state transition matrix) ( ) ̅ ( ) (44)
(n ) Constant matrix (input co-efficient matrix)
(p ) Constant matrix (output co-efficient matrix) where: ̅ [ ] , now from (42) and (44) the closed
A linear state feedback control law comprising both the loop error dynamics are given as:
nominal and added states can be given as: ̇( ) (̅ ̅ ̅) ( ) (45)
( ) ( ) ( ) (35) It is evident from (45) that the problem of designing the
where, is a (m ) state feedback gain matrix and is a state feedback law for (36) to track the reference values is now
(m ) integral feedback gain matrix. transformed to a regulator design problem. The gain matrix ̅
The state space dynamic model for the system obtained is to be designed such that the error vector ( ) converges to
after the state augmentation can be written in matrix form by zero at steady-state. The closed loop error dynamics of (45)
combining (32) to (34) as: also allow the utilization of the linear quadratic regulator
design approach for computing the gain matrix ̅ .
̇( ) ( )
[ ̇ ] * +[ ] * + ( ) * + ( ) (36) B. Optimal Linear Quadratic Regulator Design
( ) ( )
The linear quadratic regulator approach provides a
where, is a (p ) identity matrix. The objective is to design
systematic way of computing the feedback gain matrix ̅ for
the state feedback law such that the system becomes
making MIMO error dynamics of (42) asymptotically stable.
asymptotically stable so that ( ), ( ), and ( ) approach
The gain matrix ̅ is computed by minimizing a quadratic
constant values. In the steady-state (36) can be expressed as:
performance index given [23]:
̇( ) ( )
[ ̇ ] * +[ ] * + ( ) * + ( ) (37) ∫ ( ( ) ( ) ( ) ( )) (46)
( ) ( )
Considering ( ) as a vector of step inputs, we where: and are both positive-definite matrices.
have ( ) ( ) (constant) for . Now, by Note that the second term on the right-hand side of (46)
subtracting (37) from (36) the error dynamics can be obtained accounts for the expenditure of the energy of the control
as: signals. The matrices Q and R determine the relative
importance of the error and the expenditure of this energy.
̇( ) ̇( ) ( ) ( )
[ ̇ ̇ ( )] * +[ ] The feedback law designed is thus a compromise between
( ) ( ) ( ) the use of the control effort and the response speed, and at the
same time guarantees a stable system. The optimal gain matrix
* +[ ( ) ( )] (38)
that minimizes the performance index of (46) is given as [23]:
Now, define:
̅ [ ] (47)
( ) ( ) ( )
{ ( ) ( ) ( ) (39) Where is a positive-definite matrix and is the solution of
( ) ( ) ( ) the following reduced-matrix Riccati equation:
̅ ̅ ̅ (48)
From (38) and (39), following expression can be obtained:
̇ ( ) ( ) From (46) and (47), the feedback control law for error
[ ̇ ] * +[ ] * + ( ) (40) dynamics can be computed as:
( ) ( )
( ) ̅ ( ) (49)
Now, define an (( ) ) order error vector as:
From (39) and (49), the feedback control law for the dynamic
( )
( ) [ ] (41) system of (36) which drives the error dynamics of (42) to zero
( )
can be given as:
By using (40) and (41) the error dynamics can be expressed in
̅ ( )
compact form as: ( ) [ ] (50)
( )
̇( ) ̅ ( ) ̅ ( ) (42)
Equation (50) gives the state feedback control law with
where: optimal gain matrix such that the error vector ( ) of (42) is
driven to zero and the system states track their respective
reference values.

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 7

Optimal Control Law of Eq. (55)

λref vx v
+
- ki  +
- Inverter
xy
v (VSI)+ LPMSM
vref Fref vy 
SV-PWM
+
- PI +
- kiF  +
- abc
kλ 
kF s i i
id
λs  r
vm FT Estimation of
λs ,s and FT iq
 r d
Eq. (4) to (10) dq
P
 dt

Fig. 6. Proposed Linear Quadratic Regulator based Direct Thrust Force Control of linear PMSM, the integral action is added by state augmentation.

VI. NOVEL LQR BASED DIRECT THRUST FORCE CONTROL OF

LINEAR-PMSM
The objective of the optimal DTFC scheme proposed in this * + [ ][ ] (55)
paper is that the states and of the novel state space
model (26), which are also the outputs of the system, track the
reference values and .According to (26) and Fig. 5 In (55), and are state feedback gains, whereas
and 6, the input reference vector ( ) can be given as: and are integral gains. The implementation of the
control law of (55) for the system under study is illustrated by
( ) [ ] (51) the diagram of Fig. 6. The feedback gain matrix for (55) is:

The novel augmented state space model of linear PMSM, ̅ [ ] [ ]

including integral action, can be formulated from (26), (36)
and (51) after simple mathematical manipulation and is given The tuning of these gains is performed using optimal linear
as: quadratic regulator approach as explained in the previous
section. From (50) and (55) the state feedback law of (55) can
̅( ) ̅ ̅( ) ̅ ( ) ̅ ( ) * + ( ) (52) be expressed as:

where: ( ) [ ] , ̅( ) [ ] ,
* + ⏟ ̅ [ ] (56)
̅ ( )
̅ = ,̅ , ̅ [ ],
where is a positive-definite matrix and is the solution of the
[ ] [ ] (48) when solved for the system of (52). The block diagram of
∫( ) , and ∫( ) the proposed linear quadratic regulator based DTFC is given
in Fig. 6 and illustrates the structure of the optimal controller
A. Controllability Analysis of the Error Dynamics of (55). In addition, Fig. 5 shows the structure of the proposed
The error dynamic for the system given by (52) can be controller of (55) in state space form. A comparison of Fig. 4
formulated using (42) and (52) in compact form as: and Fig. 6 shows that under the proposed optimal controller,
̅ ̅ voltages and are achieved using the optimal state
̇( ) ( ) ( ) (53)
feedback law of (55) in contrast to the PI controllers of the
where, ( ) ̅( ) ̅( ) state of the art method of [22].The controller gains of (55) are
Since (53) presents a 4th order error dynamics, therefore the detailed in Table II.
controllability matrix is given as:
VII. EXPERIMENTAL VALIDATION OF THE PROPOSED
[̅ ̅ ̅ ̅ ̅ ̅ ̅ ] (54) CONTROL SCHEME
It is easy to show that rank of is 4 which is same as the The proposed novel state space model of linear PMSM and
order of the system of (53). This clearly indicates that the error its optimal linear quadratic regulator based DTFC (‘optimal
dynamics of (53) are always controllable for all values of the DTFC’) are practically validated on a prototype surface-mount
inputs and as well as the disturbance ( ). linear PMSM control system in the laboratory. The main
hardware components of the experimental setup are illustrated
B. Choice of gain matrix for state feedback law
in Fig. 7. The parameters of the surface-mount linear PMSM
From (26) and (52) the stator flux and thrust force dynamics are provided in Table I.
are decoupled from each other, therefore, the linear state
feedback law is:

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 8

Braking Resistor TABLE III

COMPARISON OF RISE TIME, PERCENT OVERSHOOT, AND PERCENT STEADY-
STATE RIPPLE IN THE RESPONSE OF THRUST FORCE PI CONTROLLER UNDER
DTFC1 [22] FOR VARIOUS DAMPING RATIOS

Load LPMSM Damping Rise time Over-shoot

PI Gains
ratio (ms) (%) (%)
Prototype LPMSM 0.0480
0.950 5.0 0.7 3.00
Interface /Measurement 0.0174
Board
0.0606
0.707 3.8 9.1 3.14
DS1104 Controller Board 0.0219
0.0784
0.500 2.6 22.4 3.59
0.0284

A. Dynamic Response with Outer Speed Loop Disabled

3-Phase Inverter
The improvements realized by the optimal DTFC are
demonstrated by comparing its transient performance for
Fig. 7. Experimental setup showing the prototype linear PMSM, voltage thrust force regulation, without the outer speed control loop, to
source inverter and the control circuitry. that of DTFC1. For this purpose a periodic square wave thrust
force reference with amplitude of ±26.5 N and a period of 1.1s
Experimental results indicate improved performance of the is used as reference thrust force command.
optimal DTFC method in terms of steady-state error and The transient and steady-state thrust force control
transient response of both stator flux and thrust force when performances of DTFC1 are shaped by tuning the proportional
compared with the prior PI based DTFC method of [22] gain and the integral gain of the thrust force PI
(referred to as ‘DTFC1’). The controller gains for both
controller shown in Fig. 4. In this research, the tuning criteria
DTFC1 and the proposed optimal DTFC are listed in Table II.
adopted for these gains is to achieve a fast step response of the
All the experimental waveforms for DTFC1 and the optimal
thrust force in terms of rise time with minimum possible or
DTFC are based on the gains of Table II.
near zero precent (%) overshoot (i.e. a deadbeat response with
The proposed optimal DTFC is digitally implemented
a faster rise time). Moreover, a steady-state response of thrust
according to the block diagram of Fig. 6.The optimal state
force with minimal ripple is also of key interest while tuning
feedback control law of (55) as illustrated by Fig. 6 is digitally
the gains of the PI controller for thrust force regulation.
implemented using a dSPACE® DS-1104 controller.
The quantitative performance of the thrust force response in
The gains of thrust force and stator flux regulating PI
terms of rise time, percent (%) overshoot and percent (%)
controllers of DTFC1 were tuned to achieve a damping ratio
steady-state ripple for various values of the proportional
of 0.95 using the root locus method to ensure comparable
conditions for benchmarking. The integrators used in both the gain and the integral gain corresponding to the
optimal DTFC and DTFC1 have an anti-windup scheme with damping ratios of 0.95, 0.707, and 0.5 respectively are
the anti-windup gains for all the integrators set at unity. The summarised in Table III. The percent % steady-state ripple in
proportional and integral gains for the speed PI controller (in thrust force, is defined by (58). It can be observed from
Fig. 4 and 6) are selected as 970 and 6.5 respectively for both Table III that tuning of the gains of thrust force PI controller
DTFC1 and the optimal DTFC to ensure similar conditions for for a higher damping ratio of 0.95 results in the minimum over
the comparison of the two methods. The speed PI controller shoot of 0.7 % and a steady-state ripple of 3.0 % with a rise
also has an anti-windup scheme with the anti-windup gain set time of 5 ms.
at unity. It is important to note that when damping ratio is decreased
to 0.707, the integral gain increases and results in further
TABLE II
reduction of the rise time to 3.8 ms. The PI gains tuned for a
CONTROLLER GAINS USED IN EXPERIMENT FOR THE PERFORMANCE damping ration of 0.5 even further reduce the rise time to 2.6
COMPARISON OF THE OPTIMAL DTFC AND DTFC1 OF [22] ms as observed from Table III, however the percent (%) over
shoot increases to 22.4 % in this case, which is not acceptable.
Proposed optimal DTFC State of the art DTFC1[22] It can be concluded from Table III that a lower damping
ratio results in a significantly larger percent (%) overshoot and
10 Flux PI 8.7
Flux Controller an increase in the steady-state ripple in thrust force. Therefore
20 Controller 17
the gains of PI controllers corresponding to damping ratio of
Thrust force 0.055 Thrust Force 0.0480 0.95 are used for DTFC for benchmarking the performance
controller 0.023 PI Controller 0.0174 and these gains are underlined in Table III and also given in
Table II. The step response of the thrust force corresponding
to the damping ratio of 0.95 ( and
) is shown in Fig. 8.

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 9

50 This clearly validates the superior control performance in

40 Reference
30 terms of thrust force transient response of the optimal DTFC.
20
Force (N)
10 B. Steady-state Regulation with Outer Speed Loop Disabled
0 Estimated
-10 The steady-state regulation of stator flux and thrust force for
-20
-30 both the DTFC1 and optimal DTFC are presented in Fig. 10
5 ms
-40
-50
and 11 respectively. In Fig. 10, from top to bottom the steady-
1.095 1.0975 1.1 1.1025 1.105 1.1075 1.11 1.1125 1.115 state response of thrust force, magnified view of the steady-
Time (s)
state thrust force, steady-state error in thrust force and steady-
Fig. 8. Thrust force step response of DTFC1 with (Experiment). state stator flux are shown and the same order is followed in
Fig. 11. The thrust force reference is set at 26.5 N and the
50 stator flux reference is fixed at 0.0846 Wb according to (14).
40 Reference
30 It is evident from Fig. 10 and 11 that steady-state regulation
20
of thrust force is much improved for the optimal DTFC as it
Force (N)

10
0 Estimated has reduced steady-state ripple when compared to that of
-10
-20 DTFC1. It can be observed from Table III that steady-state
-30
-40
3.88 ms ripple in thrust force under DTFC1is 3 %. However, under
-50
1.095 1.0975 1.1 1.1025 1.105 1.1075 1.11 1.1125 1.115
optimal DTFC the steady-state ripple is reduced to 2.6 %.
Time (s) The magnified view of the thrust force steady-state response
Fig. 9. Thrust force step response of optimal DTFC (Experiment). of Fig. 10 illustrates that under DTFC1 the oscillation in thrust
force is within a band of 4 N. However, from Fig. 11, it is
The step response of thrust force under DTFC1 is compared evident that the thrust force oscillations remain within a band
with that of optimal DTFC and is shown in Fig. 8 and 9 of less than 3 N which clearly validates the effectiveness of
respectively. It can be observed from Fig. 8 that a reasonable the optimal DTFC.
transient response of thrust force is achieved under DTFC1 The steady-state error in thrust force under DTFC1 has a
with a rise time of 5 ms. However, from Fig. 9, a superior peak of 3.5 N whereas in case of the optimal DTFC it
transient response of thrust force under optimal DTFC with a remains within 2.5 N. In addition, it can be observed that the
rise time of 3.88 ms with a negligible overshoot of 3.12 % is steady-state stator flux response is smoother under the optimal
observed which is 28 % faster than that of DTFC1 in terms of DTFC when compared with that of DTFC1.
rise time.
35 35
30 Reference (dashed) 30 Reference (dashed)
25 25
Force (N)
Force (N)

20 20
15 15
10 Estimated (solid) 10 Estimated (solid)
5 5
0 0
1.25 1.275 1.3 1.325 1.35 1.375 1.4 1.25 1.275 1.3 1.325 1.35 1.375 1.4
29 Reference (dashed) 29
Reference (dashed)
28 28
Force (N)
Force (N)

27 27

26 26

25 25
Estimated (solid) Estimated (solid)
24 24
1.3125 1.3175 1.3225 1.3275 1.3325 1.3375 1.3125 1.3175 1.3225 1.3275 1.3325 1.3375
5 5
4 4
Force Error (N)

Force Error (N)

3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
1.25 1.275 1.3 1.325 1.35 1.375 1.4 1.25 1.275 1.3 1.325 1.35 1.375 1.4
0.0855 0.0855
Reference (dashed) Reference (dashed)
0.0853 0.0853
Flux (Wb)
Flux (Wb)

0.085 0.085

0.0848 0.0848
Estimated (solid)
Estimated (solid)
0.0845 0.0845
1.25 1.275 1.3 1.325 1.35 1.375 1.4 1.25 1.275 1.3 1.325 1.35 1.375 1.4
Time (s) Time (s)

Fig. 10. Steady-state response of DTFC1. From top to bottom the thrust force Fig. 11. Steady-state response of optimal DTFC. From top to bottom the thrust
response, magnified thrust force response, steady-state error in the thrust force force response, magnified thrust force response, steady-state error in the thrust
and steady- state stator flux response is shown (Experiment). force and steady- state stator flux response is shown (Experiment).

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 10

C. Start-up Speed Response Outer Speed Loop Enabled DTFC1 Optimal DTFC
The start-up response with the outer speed loop enabled for

Speed (mm/s)

Speed (mm/s)
600 600
the surface-mount linear PMSM from zero speed to 200 mm/s 0 0
Reference Reference
under both the DTFC1 and the optimal DTFC is compared in
-600 Measured -600
Fig. 12 which shows the speed response, the corresponding the Measured

thrust force, stator flux and the stator currents for both the 0.7 0.8 0.9 1 1.1 1.2 1.3 0.7 0.8 0.9 1 1.1 1.2 1.3

DTFC1 and the optimal DTFC respectively. 200 Reference 200 Reference

Force (N)

Force (N)
It can be observed from Fig. 13 that the speed response 100 100
0 0
under optimal DTFC is 5.1 ms (17 %) faster than that of
-100 -100
DTFC1. Estimated Estimated
-200 -200
The thrust force demand of the optimal DTFC is 0.7 0.8 0.9 1 1.1 1.2 1.3 0.7 0.8 0.9 1 1.1 1.2 1.3
comparable to that of DTFC1. However it is important to note 0.087 -
0.087
Reference Reference
that the transient response of thrust force under optimal DTFC

Flux (Wb)

Flux (Wb)
0.086 0.086
is faster than that of DTFC1 which consequently results in the 0.085 0.085
faster speed response during the start-up transient when the 0.084 0.084
Estimated Estimated
linear LPMSM is controlled under the optimal DTFC. 0.083
0.7 0.8 0.9 1 1.1 1.2 1.3
0.083
0.7 0.8 0.9 1 1.1 1.2 1.3
It is observed from Fig. 12, the stator flux response and the
5 5
stator current demand for both the DTFC1 and the optimal

Current(A)
Current(A)
3 3
DTFC do not differ by much. 0 0
It is important to note that the gains for speed PI controller -3 -3
of both DTFC1 and the optimal DTFC are same as mentioned -5 -5
0.7 0.8 0.9 1 1.1 1.2 1.3 0.7 0.8 0.9 1 1.1 1.2 1.3
earlier. However the tuning of these gains is performed to Time (s) Time (s)
achieve damping ratio of 0.95 for DTFC1 and results in
Fig. 14. Speed reversal from -600 mm/s to 600 mm/s and steady-state
response at 600 mm/s with outer speed loop closed. Speed, thrust force, stator
DTFC1 Optimal DTFC flux, duty ratio and stator phase a current responses are shown from top to
300 300 bottom respectively for both the DTFC1 and novel DTFC (Experiment).
Speed (mm/s)

Speed (mm/s)

Reference Reference
200 200
100 100 DTFC1 Optimal DTFC
0 0 Reference Reference
Speed (mm/s)

Speed (mm/s)
Measured Measured 600 600
-100 -100
0.95 1 1.05 1.1 1.15 1.2 0.95 1 1.05 1.1 1.15 1.2 0 Measured 0 Measured
150 Reference 150 Reference
-600 66.8 ms -600 59.6 ms
Force (N)

Force (N)

100 Estimated 100 Estimated

50 50 0.76 0.8 0.84 0.88 0.92 0.96 0.76 0.8 0.84 0.88 0.92 0.96
Time (s) Time (s)
0 0 (a) (b)
0.95 1 1.05 1.1 1.15 1.2 0.95 1 1.05 1.1 1.15 1.2
0.086 0.086 Fig. 15. Magnified view of the speed reversal transient (Experiment).
Estimated Estimated
Flux (Wb)

Flux (Wb)

0.0855 0.0855
negligible overshoot in speed response with a rise time of 35.2
0.085 0.085
0.0845 0.0845
ms as observed from Fig. 13. However, the same set of the
0.084
Reference
0.084
Reference gains, when used for the speed PI controller of the optimal
0.95 1 1.05 1.1 1.15 1.2 0.95 1 1.05 1.1 1.15 1.2
DTFC, results in comparatively larger overshoot in the speed
3 3
Ic (Green) Ib (Blue) response with rise time of 30.1 ms. The larger over shoot in
Current (A)

Current (A)

2 2
1 1
0 0 the speed response under optimal DTFC can be attributed to
-1
Ia (Red) Ib (Blue)
-1
Ic (Green) I (Red)
the faster transient control performance of the optimal inner
-2 -2
-3 -3
a
thrust force controller loop.
0.95 1 1.05 1.1 1.15 1.2 0.95 1 1.05 1.1 1.15 1.2
Time (s) Time (s)
D. Speed Reversal with Outer Speed Loop Enabled
Fig. 12. Startup performance from 0 to 200 mm/s with outer speed loop
closed. Speed, thrust force, stator flux, and stator phase currents responses are
The experimental results comparing the performance of the
shown from top to bottom respectively for both the DTFC1 and optimal DTFC1 and the optimal DTFC during the speed reversal from
DTFC (Experiment). -600 mm/s to +600 mm/s are shown in Fig. 14 which shows
the speed response, the corresponding thrust force, stator flux
DTFC1 Optimal DTFC and the stator currents for both the DTFC1 and the optimal
300 300
Reference Reference DTFC.
Speed (mm/s)

Speed (mm/s)

200 200
100 100 Fig. 15 shows a magnified view of the speed response and
Measured Measured
0
35.2 ms
0 30.1 ms
demonstrates that under the optimal DTFC the speed response
-100 -100 is 7.2 ms (12 %) faster than DTFC1 during the speed reversal
0.95 0.975 1 1.025 1.05 0.95 0.975 1 1.025 1.05 transient. The thrust force response for the optimal DTFC is
Time (s) Time (s)
(a) (b)
4 considerably less oscillatory when compared with DTFC1.
Fig. 13. Magnified view of the speed response during start-up (Experiment).

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 11

The improved stator flux response is evident under the √ ∑ ( () )

optimal DTFC with significantly reduced ripple. The stator ( ) (57)
current demand for both the control schemes is similar during
the speed reversals. √ ∑ ( () )
( ) (58)
E. Steady-state Response with Outer Speed Loop Enabled
The steady-state performance of the prototype linear PMSM √ ∑ ( () )
at 600 mm/s occurs from 0.95s to 1.15s as observed from Fig. ( ) (59)
14 under both DTFC1 and optimal DTFC and is also
compared. where, , and represent the average steady-state
In order to illustrate the comparison of steady-state stator flux, thrust force and mover’s speed respectively and
performances of the two control schemes, the magnified view ( ), ( ) and ( ) are the instantaneous values of stator
of the speed response, stator flux response and thrust force flux, thrust force and mover’s speed.
response, for DTFC1 and optimal DTFC during the steady- It can be observed from Table IV that the optimal DTFC
state are shown in Fig. 16. Fig. 16 demonstrates that under reduces the percentage ripple in steady-state speed, stator flux
optimal DTFC the steady-state low frequency oscillations in and thrust force response compared to DTFC1 which validate
speed, thrust force and stator flux have noticeably reduced the superior steady-state performance of the proposed optimal
compared to DTFC1. DTFC.
The quantitative results for steady-state performance of
DTFC1 and the optimal DTFC at 600 mm/s and 52N (average VIII. CONCLUSIONS
thrust force) in terms of percent stator flux ripple λrip (%), In this research, an optimal linear quadratic regulator-based,
percent force ripple Frip (%), and percent speed direct thrust force control scheme utilizing space vector
ripple ( ) are summarized in Table IV. In this analysis modulation for the linear PMSM is proposed.
λrip (%), Frip (%) and ( ) are given by (57), (58) and (59) A multiple-input multiple-output (MIMO) state space
respectively. model for linear PMSM, comprising the stator flux and thrust
force as states, is formulated which subsequently allows an
optimal linear state feedback control law for direct thrust force
DTFC1 Optimal DTFC
control to be synthesized using the optimal linear quadratic
Reference
Speed (mm/s)

Reference
Speed (mm/s)

700 700 regulator approach. Integral action is incorporated in the

600 600 control scheme by state augmentation of the novel state space
Measured Measured model to reduce the steady-state error.
500 500
Experimental results clearly indicate that the proposed
0.95 1 1.05 1.1 1.15 0.95 1 1.05 1.1 1.15
150 150
control scheme exhibits excellent control of stator flux and
Reference Reference thrust force with faster transient response and reduced steady-
Force (N)

100 100
Force (N)

50 50
state error when compared to the state of the art controller.
0
Estimated
0
The novel state space model is independent of the mover’s
-50
Estimated speed and asymptotically state controllable over the whole
-50
0.95 1 1.05 1.1 1.15 0.95 1 1.05 1.1 1.15 speed range of the linear PMSM. The optimal state feedback
0.087 0.087
Reference Reference control law involves static gains that are independent of
Flux (Wb)

Flux (Wb)

0.086 0.086
mover’s speed. Only one set of gains are sufficient to provide
0.085 0.085
optimal control performance for the whole speed range of the
0.084 Estimated 0.084 Estimated linear PMSM. Excellent steady-state and transient
0.083 0.083
0.95 1 1.05 1.1 1.15 0.95 1 1.05 1.1 1.15 performance including speed/thrust force reversals is
Time (s) Time (s)
achieved.
Fig.16. Steady-state performance at 600 mm/s. From top, Speed, thrust force,
and stator flux responses are shown for both the DTFC1 and optimal DTFC REFERENCES
(Experiment).
[1] L. Zhong, M. F. Rahman, W. Y. Hu, and K. W. Lim, "Analysis of direct
torque control in permanent magnet synchronous motor drives," IEEE
TABLE IV Trans. Power Electron., vol. 12, pp. 528-536, 1997.
COMPARISON OF STEADY-STATE PERFORMANCE OF DTFC1 AND THE [2] M. F. Rahman, L. Zhong, and L. Khiang Wee, "A direct torque-
OPTIMAL DTFC controlled interior permanent magnet synchronous motor drive
incorporating field weakening," IEEE Trans. Ind. Appl., vol. 34, pp.
600 mm/s, 52 N DTFC1[22] Optimal DTFC 1246-1253, 1998.
[3] F. Niu, B. Wang, A. S. Babel, K. Li and E. G. Strangas, "Comparative
λrip (%) 0.34 0.24 Evaluation of Direct Torque Control Strategies for Permanent Magnet
Frip (%) 10.48 5.91 Synchronous Machines," IEEE Trans. Pow. Electron., vol. 31, pp. 1408-
vrip (%) 1.92 1.13 11424, 2016.
[4] K. Yoshida, Z. Dai, and M. Sato, "Sensorless DTC propulsion control of
PM LSM vehicle," in Proc. Power Electronics and Motion Control
Conf. (IPEMC), 2000., pp. 191-196 vol.1.

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2016.2519331, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 12

[5] J. Cui, C. Wang, J. Yang, and L. Liu, "Analysis of direct thrust force International Power Electronics and Motion Control Conference
control for permanent magnet linear synchronous motor," in Proc. (EPE/PEMC), 2012.
Intelligent Control and Automation(WCICA), 2004., pp. 4418-4421 [26] L. M. Grzesiak, "PMSM servo-drive control system with a state
vol.5. feedback and a load torque feedforward compensation " COMPEL: The
[6] J. Cui, C. Wang, J. Yang, and D. Yu, "Research on force and direct International Journal for Computation and Mathematics in Electrical
thrust control for a permanent magnet synchronous linear motor," in and Electronic Engineering, vol. 32, p. 18, 2013
Proc. IEEE Ind. Electron Conference (IECON), 2004, pp. 2269-2272, [27] Ton Duc Do, Sangshin Kwak, Han Ho Choi and Jin-Woo Jung,
vol. 3. "Suboptimal Control Scheme Design for Interior Permanent-Magnet
[7] M. Abroshan, K. Malekian, J. Milimonfared, and B. A. Varmiab, "An Synchronous Motors: An SDRE-Based Approach" IEEE Trans. Power
optimal direct thrust force control for interior Permanent Magnet Linear Electron., vol. 29, pp. 3020-3031, 2014
Synchronous Motors incorporating field weakening," in Proc. [28] Y. S. Huang and C. C. Sung, "Implementation of sliding mode controller
International Symposium on Power Electronics, Electrical Drives, for linear synchronous motors based on direct thrust control theory," IET
Automation and Motion (SPEEDAM), 2008, pp. 130-135. Control Theory & Applications, vol. 4, pp. 326-338, 2010.
[8] S. Cheng-Chung and H. Yi-Sheng, "Based on Direct Thrust Control for [29] J. Faiz and S. H. Mohseni-Zonoozi, "A novel technique for estimation
Linear Synchronous Motor Systems," IEEE Trans. Ind. Electron, vol. and control of stator flux of a salient-pole PMSM in DTC method based
56, pp. 1629-1639, 2009. on MTPF," IEEE Trans. Ind. Electron, vol. 50, pp. 262-271, 2003.
[9] Y. Ren, Z. Q. Zhu, and J. Lue, "Direct Torque Control of Permanent- [30] J. Luukko, O. Pyrhonen, M. Niemela, and J. Pyrhonen, "Limitation of
Magnet Synchronous Machine Drives With a Simple Duty Ratio the load angle in a direct-torque-controlled synchronous machine drive,"
Regulator," IEEE Trans. Ind. Electron., vol. 61, pp. 5249-5258, 2014. IEEE Trans. Ind. Electron, vol. 51, pp. 793-798, 2004.
[10] F. Niu, K. Li and Y. Wang, "Direct Torque Control for Permanent- [31] Z. Jun, X. Zhuang, T. Lixin, and M. F. Rahman, "A Novel Direct Load
Magnet Synchronous Machines Based on Duty Ratio Modulation," Angle Control for Interior Permanent Magnet Synchronous Machine
IEEE Trans. Ind. Electron, vol. 62, pp. 6160-6170, 2015. Drives with Space Vector Modulation," in Proc. International Power
[11] Z. Ma, S. Saeidi, and R. Kennel, "FPGA Implementation of Model Electronics and Drives Systems (PEDS), 2005, pp. 607-611.
Predictive Control With Constant Switching Frequency for PMSM [32] Y. Inoue, S. Morimoto, and M. Sanada, "Examination and Linearization
Drives," IEEE Trans. Ind. Informat., vol. 10, pp. 2055-2063, 2014. of Torque Control System for Direct Torque Controlled IPMSM," IEEE
[12] B. Boazzo and G. Pellegrino, "Model-Based Direct Flux Vector Control Trans. Ind. Appl, vol. 46, pp. 159-166, 2010.
of Permanent-Magnet Synchronous Motor Drives," IEEE Trans. Ind.
Appl., IEEE Transactions on, vol. 51, pp. 3126-3136, 2015. Muhammad Ali Masood Cheema (S’12) received the B. Sc. (first-class
[13] Y. Cho, K. B. Lee, J. H. Song, and Y. I. Lee, "Torque-Ripple honors) with gold medal and M.Sc. degrees both in electrical engineering
Minimization and Fast Dynamic Scheme for Torque Predictive Control from the University of Engineering & Technology, Lahore, Pakistan in 2008
of Permanent-Magnet Synchronous Motors," IEEE Trans. Power and 2012 respectively. He is currently studying towards a PhD degree in
Electron., vol. 30, pp. 2182-2190, 2015. electrical engineering in the University of New South Wales, Sydney,
[14] F. Wang, S. Li, X. Mie, W. Xie, J. Rodriguez, and R. M. Kennel, Australia. His research interests include power electronics, electrical drives,
"Model-Based Predictive Direct Control Strategies for Electrical Drives: nonlinear control and convex optimization.
An Experimental Evaluation of PTC and PCC Methods," IEEE Trans.
Ind. Informat., vol. 11, pp. 671-681, 2015. John E. Fletcher (SM’11) received the B.Eng. (first-class honors) and Ph.D.
[15] M. Preindl and S. Bolognani, "Optimal State Reference Computation degrees from Heriot-Watt University, Edinburgh, U.K., in 1991 and 1995,
With Constrained MTPA Criterion for PM Motor Drives," IEEE Trans. respectively, both in electrical and electronic engineering. He is currently
Power Electron., vol. 30, pp. 4524-4535, 2015. Professor at the University of New South Wales, Sydney. His research
[16] W.Xie, X.Wang, F. Wang, W. Xu, R. M. Kennel, D. Gerling and R. D. interests include power electronics, drives and energy conversion. He is a
Lorenz, "Finite-Control-Set Model Predictive Torque Control With a Fellow of the Institution of Engineering and Technology and a Senior
Deadbeat Solution for PMSM Drives," IEEE Trans. Ind. Electron., vol. Member of the IEEE.
62, pp. 5402-5410, 2015
[17] F. Minghua and X. Longya, "A sensorless direct torque control Dan Xiao obtained his Bachelors and Master’s degree in 2001 and 2004,
technique for permanent magnet synchronous motors," in Proc. IEEE respectively from Shenyang University of Technology, China. He obtained
Ind. Appl. Annual Meeting. Conf. Record (IAS),1999, pp. 159-164, vol.1. the PhD degree in Electrical Engineering the University of New South Wales,
[18] T. Lixin, Z. Limin, M. F. Rahman, and H. Yuwen, "A novel direct Australia. He is currently working as a technical support in Power
torque control for interior permanent-magnet synchronous machine drive Engineering at the University of New South Wales. His research interests are
with low ripple in torque and flux-a speed-sensorless approach," IEEE in sensorless control, matrix converter, and electrical drive systems.
Trans. Ind. Appl., vol. 39, pp. 1748-1756, 2003
[19] T. Lixin, Z. Limin, M. F. Rahman, and H. Yuwen, "A novel direct Muhammad Faz Rahman (F’14) received the Bachelor of Electrical
torque controlled interior permanent magnet synchronous machine drive Engineering degree (first-class honors) from Bangladesh University of
with low ripple in flux and torque and fixed switching frequency," IEEE Engineering and Technology, Dhaka, Bangladesh, in 1972 and the M.Eng.
Trans. Power Electron., vol. 19, pp. 346-354, 2004. and Ph.D. degrees from the University of Manchester Institute of Science and
[20] G. Foo, C. S. Goon, and M. F. Rahman, "Analysis and design of the Technology, Manchester, U.K., in 1975 and 1978, respectively. He is
SVM direct torque and flux controlled IPM synchronous motor drive," currently a Full Professor at the University of New South Wales, Sydney,
in Proc. Australasian Universities Power Engineering Conf. (AUPEC), Australia. His research interests are in electrical machines, drives, and power
2009, pp. 1-6. electronics.
[21] A. Mohammadpour and L. Parsa, "SVM-based direct thrust control of
permanent magnet linear synchronous motor with reduced force ripple,"
in Proc. IEEE International Symposium on Ind. Electron. (ISIE), 2011,
pp. 756-760.
[22] M. A. M. Cheema, J. Fletcher, M. F. Rahman, and D. Xiao, "Modified
direct thrust control of linear permanent magnet motors with sensorless
speed estimation," in Proc. IEEE Ind. Electron Conference (IECON),
2012, pp. 1908-1914.
[23] P. C. Young and J. C. Willems, "An approach to the linear multivariable
servomechanism problem†," International Journal of Control, vol. 15,
pp. 961-979, 1972.
[24] C. Kuan-Teck, L. Teck-Seng, and L. Tong-Heng, "An optimal speed
controller for permanent-magnet synchronous motor drives," IEEE
Trans. Ind. Electron, vol. 41, pp. 503-510, 1994.
[25] T. Tarczewski and L. M. Grzesiak, "State feedback control of the
PMSM servo-drive with sinusoidal voltage source inverter," in Proc.

(c) 2015 Crown Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.