Torque Control of BLDC Motor for Electric Bicycle

Józef Gromba
Department of Power Electronics and Energy Control Systems
AGH UNIVERSITY OF SCIENCE AND TECHNOLOGY
Cracow, Poland
[email protected]

The paper focuses on using the BLDC motor as electrical bike Unfortunately, the construction of a brushless motor is not
drive. As of today on the market there are already different free from imperfections. One of the biggest imperfect is the
methods of controlling electrical motor of a bike. The most ripples of electromagnetic torque. These ripples may cause
common is setting the motor’s speed with a lever mounted on large vibrations during drive operation which, in turn, may lead
handlebars. Controlling motors torque instead of speed makes to instability of rotational speed.
new principle of how the drive works. Proposed method of
control is based on controlling the motor torque in such a way Operation of the BLDC motor is based on the principle of
that allows the user to set a desired torque value keeping the interaction between the magnetic field generated by the
force on bike pedals constant. This article covers different permanent magnets located on the rotor and the magnetic field
methods of controlling the BLDC motor, analysis of electrical generated by the current flowing through the stator windings. A
bike movement, methods of calculating the load torque, transistor bridge, in connection with a controller and Hall
controlling electromagnetic torque of the motor, simulations and sensors, decides which of the motor windings are to be
laboratory research. connected connected to the positive or negative potential of the
supply source at any given time instant.
Keywords— brushless direct current motor; electric bike; motor
torque control; kinematics.
B. Methods fot Controlling the Electric Commutator
The lack of commutator unit and brushes in a BLDC motor
I. INTRODUCTION
makes it necessary to use an electronic commutator, usually
Several electric bicycle drive solutions are currently consisting of six transistors (a voltage source inverter).
available on the market. The most common solution is setting
the electric motor speed using a shifter located on the bicycle Fig.1 presents a diagram for connecting the motor stator
handlebar. In the case of using the electromagnetic torque windings to an electronic commutator, control system, and Hall
control for the motor driving the bike, it is possible to use a sensors. Two groups of transistors are distinguished in such
new principle of operation of the drive. Such control consists in system: upper ones and lower ones.
setting the load torque on the connecting rod with pedals as a
difference between the determined load torque on the driven
wheel of the bicycle and the electric motor torque. The
presented method makes it possible to determine the work of
the bicyclist with a constant torque [1].The paper discusses the
rules for BLDC motor control, analysis of electric bicycle
movement mechanics [2], method for determining the load
torque, control of the motor electromagnetic torque [3], as well
as simulation and laboratory testing of the proposed drive.

II. BLDC MOTOR

A. Description of the electric motor Fig. 1 System for connecting a transistor bridge to a BLDC motor [4]
In a brushless direct current motor (BLDC) the mechanical
Control of the individual transistors is realized with
commutator are replaced with an electronically controlled
feedback using Hall sensors located on the motor stator. There
transistor system connected to the fixed stator winding, while
are several methods for transistor switching
permanent magnets are located on the motor rotor.
control [4],[5],[6],[7] (Fig. 2):
BLDC motors are also characterized by high efficiency,
• C120Q+ method: positive group transistors fulfill the
wide range of rotational speed regulation, higher ratio of the
control function, and negative group transistors fulfill only the
obtained power to the mass and volume in comparison with
commutation function.
other electric motors, low failure rate, and low noise emission
in comparison with other constructions. • C120Q- method: transistor control is executed exactly
opposite to the C120Q+ method, upper transistors fulfill the

978-1-5386-5210-7/18/$31.00 ©2018 European Union

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.
commutation function, and negative group transistors fulfill the Fig. 3 presents the idea of bipolar control. This method
control function. consists in simultaneous generation of PWM signal for
controlling both the upper and the lower transistors at the
The advantage of the above methods is that the control moment of commutation. The advantage of bipolar control is
function is only fulfilled by three keys of the commutator, that all transistors of the electronic commutator are loaded
which simplifies realization of the control system, however, its evenly and have the same switching frequencies.
disadvantage is that transistors are utilized unevenly.
• C60Q+ method: upper and lower transistors for the first III. MEASUREMENT AND DETERMINATION OF ELECTRIC MOTOR
60Û electrical of the period of their operation fulfill the control LOAD
function, and for the next 60Û the commutation function.
In some electrical drive control methods it is necessary to
• C60Q- method: the transistors for the first 60Û of their know the load torque. This problem can be solved by adequate
conducting fulfill the commutation function, and for the next computational method or by hardware.
60Û the control function.
Hardware measurement of torque consists in using several
The advantage of C60Q+ and C60Q- methods is that all strain gauges connected into Wheatstone bridges, which are
transistors are utilized evenly, however, its disadvantage is that located on the shaft between the electric motor and the driven
the transistors switching system is more complex [7]. machine. Another method for load torque measurement is
using a torque sensor located in the same place as strain gauge.
Yet another method for obtaining the information about
load torque value is using a torque observer (computational
algorithm) or determining the values of all torques acting on
the motor shaft. In the case of load torque observer it is
necessary to know the present value of motor current and its
rotational speed. On the basis of these values and an adequate
algorithm it is possible to determine the electric motor load
torque. In the case if another source of driving torque, apart
from electric motor, is also present in the drive (hybrid drive),
which cannot be determined or measured directly, it is
necessary to determine individual values of torques acting on
the drive. These relationships are discussed in the section 5.

IV. ELECTRIC TORQUE CONTROL IN ELECTRIC BICYCLE DRIVE

One of possible applications of the discussed motor torque

control method is constructing an electric bicycle drive, which
extends the conventional bicycle construction with additional
elements: rechargeable batteries, electric motor built into the
Fig. 2 Transistor switching methods front or rear wheel, controller, and speed setting shifter. A
Another method of the commutator switching is bipolar block diagram of the discussed electromagnetic torque control
control [7]. algorithm for a motor driving an electric bicycle is presented
in fig.4.

Fig. 4 Block diagram of the control system

Fig. 3 Bipolar control of electronic commutator

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.
 This torque is described by the formula:
V. ANALYSIS OF BICYCLE MOVEMENT MECHANICS
FR k , (2)
For the purpose of effective and precise control of motor MR =
torque in the proposed electric bicycle drive, it is necessary to i
determine the forces and torques acting on the bicycle. Due to a where:
lot of forces acting on the vehicle in motion, the presented
mathematical torque model is simplified and it only includes FR – force of pressure on the pedal, k – crank arm,
the torques which have the greatest influence on kinematics. i – gear ratio between the front and rear gear wheel.
Such approach made it possible to determine the five major
forces [8] acting on the system in motion (Fig. 5). The values of aerodynamic drag force as well as wind
velocity and direction have the greatest influence on the value
of “fan torque”. This torque can be determined from the
formula:
1 (3)
M W = C x (v − vw ) 2 Aȡr
2
where:
CX –drag coefficient (1 for upright bicyclist) [1], vw – wind
velocity, A – front surface area of the bicycle and bicyclist
system, ȡ – air density (1.226 kg/m3 at 1013hPa and 15ÛC), r –
bicycle wheel radius.
Another torque having a significant influence on the bicycle
movement is the torque coming from the force resulting from
the slope angle of the terrain on which the vehicle is moving.
Fig. 5 Distribution of forces acting on the bicycle

Passive forces: The active torque is presented by the formula:

• FT – Representing the forces of friction in bearings as well M cz = gmr sin(Į ) (4)

as between the tire and the surface (resistance to motion).
where:
Active forces:
m – total mass of bicyclist and bicycle, r – bicycle wheel
• FR – Force created by the pressure exerted on the radius, Į – terrain slope angle.
connecting rod with pedals.
The last discussed torque is the torque of the BLDC motor.
• FW – Force created by aerodynamic drag force and wind
This torque can be determined from the simplified
velocity (resistance to motion)..
formula [3]:
• FCZ – Force resulting from the terrain slope (resistance
to motion). 2 (5)
M E = kt I
• FS – Force coming generated the electric motor. 3

The listed forces are associated with values of respective where:

torques acting on the driven wheel. Due to low influence on the kt – motor torque constant, I – current of motor windings
value of friction torque, the coefficients related to tire pressure,
tread type, tire width, and bearing friction have been neglected.
VI. DESIGN OF DIGITAL ELECTRIC TORQUE CONTROL SYSTEM
After adopting the above assumptions, the friction torque is
described by the formula: The process of designing a digital control system required
the following stages:
M T = gm cos(Į)krsign(v) (1)
• Determination of the BLDC motor mathematical
where: model [6].
g = 9,81 m/s2, m – mass of bicycle and bicyclist,
cos(Į) – terrain slope angle, k – surface type, r – wheel radius, § u SA · § LS − Lȝ 0 0 ·§ diSA / dt ·
¨ ¸ ¨ ¸¨ ¸
v – bicyclist velocity. ¨ u SB ¸ = ¨ 0 LS − Lȝ 0 ¸¨ diSB / dt ¸
The force of the bicyclist acting on the connecting rod with ¨u ¸ ¨ 0 0 LS − Lȝ ¸¹¨© diSC / dt ¸¹ (6a)
© SC ¹ ©
pedals makes it possible to generate the torque directly driving
the bicycle. § RS 0 0 ·§ iSA · § e A ·
¨ ¸¨ ¸ ¨ ¸
+¨ 0 RS 0 ¸¨ iSB ¸ + ¨ eB ¸
¨ 0 0 RS ¸¹¨© iSC ¸¹ ¨© eC ¸¹
©

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.
 where: Execution of the PI control in C language required its form
to be changed from continuous (13a) to digital (13c).
Lμ-linkage inductance, Ls-stator inductance, uSA, uSB, uSC - Discretization has been performed using the Tustin method
voltage of each motor phases, iSA, iSB, iSC - voltage of each motor (13b) with sampling time Tp [9].
phases, RS- stator resistance.
• Determination of the drive model (voltage source Ki
Gr ( s ) = K p + (13a)
inverter, BLDC motor and switching – fig. 2 and 3) s
where:
2 z −1
L = 2( LS − Lȝ ) R = 2 RS (6c) s= (13b)
Tp z + 1
di (6c)
U = RI + L + ȥ eω RiTp z + 1
dt
Gr ( z ) = K p + (13c)
dȦ 2 z −1
J = Me − Mm (6d)
dt In order to determine the output signal value for nth
sampling step, the transfer–function definition (14a) has been
M e = ȥe I (6e) applied.
where: U ( z)
GR ( z ) = (14a)
U –voltage, I –motor 2 phase current, Mm – load torque, s( z )
Me – electromagnetic torque, Ȧ – motor rotational velocity,
ȥe – flux linkage, J – moment of inertia of motor bike and U ( z) K iT p z + 1
bicyclist, R – generalized resistance, L – total inductance, = Kp + (14b)
s( z) 2 z −1
• Parameters selection and settings of the continuous
torque control K iT p
U ( z ) z − U ( z ) = İ ( z ) K p ( z − 1) + İ ( z ) ( z + 1) (14c)
PI control has been applied as the motor torque control, and 2
its parameters have been selected in accordance with the shape Then, after multiplying the expression (14c)
criterion [9], where the transfer–function of the controller is: by z-1, obtained as:
ms + 1 K (7) K iT p
GR ( s ) = = Kp + i U ( z ) = U ( z ) −1 + İ ( z ) K p (1 − z −1 ) + İ ( z ) (1 − z −1 ) (14d)
Vs s 2
Determination of control parameters (m, v) is dependent on Finally, the control algorithm takes the form:
the relationship between the electromechanical constant
IJ and the electromagnetic time constant T. If the condition IJ•4T K iT p K iT p
is true, the parameters are determined from the formulas: U (n) = U (n − 1) + ( + K p )İ ( n ) + ( − K p )İ ( n − 1) (14
2 2
L e)
T= (8)
R The performed calculations made it possible to write
the control equation in samples domain (14e). In order to
JR (9) simplify the equation, the coefficients at ο have been replaced
τ=
ȥ 2N with parameters A0 and A1.
Then the control parameters are determined as: K iT p K iT p
A0 = + Kp A1 = − Kp (15a)
2
ȥ ȥ 2N
4T (10)
N
m = 0.5IJ(1 − 1 − )
τ U (n) = U (n − 1) + A0 İ (n) + A1İ (n − 1) (15c)

K pIJ The final form of the implemented (sec. 8) discrete PI

v=ȕ (11) control is presented by the formula (15c).
( IJ1 − ȕ) R
where: VII. MODEL AND SIMULATIONS OF ELECTRIC BICYCLE
In The analysis of bicycle movement mechanics in the
IJ1 = IJ − m (12) previous chapter made it possible to create an electric bicycle
model in the Matlab-Simulink package. Another torque
ȕ – time constant of the of closed torque control system
component, coming from the electric motor, has been added to
• Discretization of the applied PI control the driving simulation of a classic bicycle (Fig. 6).

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.
 gear ratio coefficient k = 0.5, and surface type coefficient k =
0.005. The value of bicyclist power has been increased in time
from 200 W to 400 W. The mass of the classic bicycle system
has been assumed at the level m = 80 kg (moment of inertia J =
7.2 kg/m2), whereas for the electric bicycle m = 100 kg
(moment of inertia J = 9 kg/m2). Additional value for the
solution with electric motor is the torque value set by the user
WZ = 9 Nm.

Fig. 6 Model of motor with PI coltrol

The motor has been modeled as a second order system

with the following parameters:
R – 0.25Ÿ L – 184.4mH
UN – 48V IN – 23A
MN – 27Nm J – 0.2kg/m2
ȦN – 29.3rad/s ȥe – 1.2875Vs
Control in continuous form has been applied in the
simulation. The gain values of the proportional and integrating
parts have been determined using the shape criterion [9] and
they lead to, respectively: Kp=6.4×10-3, Ki=8.4/fp. The
sampling frequency fp=1 kHz has been obtained by using a
timer clocking the main loop and counting successive cycles of Fig. 7 Diagram of the developed electric bicycle model
the processor clock.
The developed model makes it possible to set various As it results from fig.8, despite the great increase of
external forces acting on the vehicle, and on this basis it mass, caused by using the electric drive, the maximum velocity
calculates the total system torque and the resulting velocity. for the same bicyclist power has been increased twofold. Apply
The value of the electric torque is the difference of the actual the another torque source made it possible to maintain a
load torque and the value set by the user. constant set connecting rod load.

The model of the forces having influence on the electric

bicycle movement in a two-dimensional system, converted to
rear axis torques, is presented in fig.7. The model has been
realized so as to make it possible to observe the influence of
the motor torque generated by the BLDC motor on the total
torque and the velocity developed by the drive system.
The developed model made it possible to compare the
operation of the system with and without electric motor, for
identical external conditions. Simulations have been performed
for various values of the weather conditions (wind velocity and
terrain slope angle) in which the bicycle is moving. Adding the
electric drive with controller and rechargeable battery
influences the bicycle mass. The total mass of the bicyclist and
the electric bicycle has been adequately increased in
comparison with the classic solution. The values of bicycle
chain gear ratio, surface type, and set torque (in the case of
electric bicycle) have been set at a constant level. Fig. 8 Simulation results for the system with and without motor
Fig.8 presents the simulation results for common set values
of terrain slope angle Į = 2°, wind velocity vs = 5 m/s, chain

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.
 VIII. DESIGN OF CONTROL SYSTEM
The presented electric torque control method and the
bicycle model presented in section 6 have been used for
executing an electric bicycle drive prototype, where the digital
control algorithm (14e) has been used.
The BLDC motor model Magic Pie 3 manufactured by
Golden Motor company has been used as the drive unit of the
bicycle, due to its high efficiency. Its main advantages are:
wide range of power supply voltage from 24 V to 60 V, output
power from 48 W to 1000 W, and maximum torque within the
limit of 26.66 Nm. The use of a motor of such construction
makes it necessary to apply an external electronic commutator
constructed on the basis of MOSFET transistors IRF4905 and
IRFZ48N [10].
Fig. 9 Measurement of motor’s current and speed, terrain slope angle
The microcontroller STM32F407VGT6 based on ARM varying from
Cortex-M4 core with 168 MHz clock has been applied as the The performed measurements made it possible to
motor control data processing element. An important element conclude that the constructed control system is operating in
of the chosen microcontroller is the built-in FPU (Floating- accordance with the assumptions made beforehand, because
Point Unit) coprocessor, used for performing operations on increasing the terrain slope angle resulted in increasing the
single-precision floating-point numbers. Terrain slope current drawn by the motor, as well as the electromagnetic
measurements have been performed using LIS302DL torque generated by the drive. The next stage in the
accelerometer, included in the STM32F407VG6 Discovery investigation of proposed electric bicycle conception will be
development kit. LEM current transducer LA-55-P-SP1 has supplementing the control system with the required
been applied in order to perform current measurement. Wind measurement systems. And finally, implementing the proposed
velocity can be determined using a system of two pressure electromagnetic motor torque control system (Fig. 7), maybe
sensors located at the front and back of the bicycle. The with additional cruise control (cascade control of electric motor
proposed location of sensors would make it possible to e.g. [11]) and load torque observer implementation [12].
determine the pressure difference, and thus the wind velocity.
This solution has not been realized due to the high cost and low
accuracy of the commercially available sensors. Another REFERENCES
possible method for wind velocity measurement is using a
spherical thermo-anemometer but, like in the case of pressure [1] J. Gromba, P. Gawron, Master’s Thesis: Microprocessors controlled
sensors, this solution has not been realized due to excessively electromagnetic torque of BLDC Motor, AGH Kraków 2016.
high cost of the device. [2] K. J. Åström, Control System Design Lecture Notes for ME 155A,
Department of Mechanical and Environmental Engineering University
of California Santa Barbara 2002.
IX. CONCLUSIONS [3] G. Sieklucki, “Analysis of the transfer-function models of electric drives
with voltage controlled source,” Przeglad Elektrotechniczny 88(7A),
In order to check the correctness of operation of the 250–255, 2012
developed control system, its measurements have been
[4] Z. Goryca, “Controlling methods of BLDC motors,” Power Electronics
performed. One of them consisted in observing the changes of and Drives, Nr.66/2012
current and velocity of the motor for constant power supply [5] I. Boldea, D. Nasar, The Induction Machine Handbook, Boca Raton,
voltage, set load torque value equal to 20 Nm, and constant CRC Press, 2002.
load torque. The measurement consisted in increasing the [6] R. Krishnan, Electric Motor Drives. Modelling, Analysis and Control,
accelerometer inclination angle from 0° to 3° (Fig. 9). Upper Saddle River, Prentice Hall, 2001.
[7] A. Domoracki,K, Krykowski, BLDC drives - the classical control
strategies. Electrical Machines - Transaction Journal Nr.72/2005
[8] R. P. Feynman,R.B. Leignton, M. Sanbs, The Feynman Lectures on
Physics, Volume I, The Perseus Books Group 2011r.
[9] G. Sieklucki, Automation of the drive, AGH Kraków 2009
[10] J. Gromba, M. Waákowicz, Engeering Thesis: Construction of electric
bicycle drive, PWSZ Tarnów 2015r.
[11] R. Klempka, B. Filipowicz, “Comparison of using the genetic algorithm
and cuckoo search for multicriteria optimisation with limitation,”
Turkish Journal of Electrical Engineering and Computer Sciences, 25
(2), pp. 1300-1310, 2017.
[12] G. Sieklucki, T. Orzechowski, M. Tondos, R. Sykulski, “Optimization of
load torque observer at quadratic performance index,” Przeglad
Elektrotechniczny, 84 , nr7, pp. 29-35,2008

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on July 27,2020 at 21:21:23 UTC from IEEE Xplore. Restrictions apply.