Static and Dynamic Ergonomic Corrects of Torque Controlled in Bicycle Ergometer

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Static and Dynamic Ergonomic Corrects of Torque

Controlled in Bicycle Ergometer

Cezary Bielak Krzysztof Nowopolski, Bartłmiej Wicher


Agregaty Polska Institute of Control and Information Engineering
Poznan, Poland Poznan University of Technology
cbielak@agregatypolska.pl Poznan, Poland
krzysztof.nowopolski@put.poznan.pl

Abstract — In this paper various methods of ergonomics ergonomic improvements of ergometer operation, that do not
improvement in electric bicycle pedaling are discussed. The imply heavy flywheel application are discussed.
main goal is to control an average amperage of generator
powered by a cyclist, but considering the unsteadiness of human II. OVERVIEW OF THE ERGOMETRY SYSTEM
legs torque vs. angular position of the support. In the paper,
four different approaches for generating the current reference A. Mechanic and Electric Layout
trajectory are compared – two of them use only position signal, Figure 1 presents overall flowchart of the test setup. A
and the next two use also velocity signal in the ergonomic three-phase brushless DC (BLDC) motor with embedded hall
improvements calculations. sensors was used. Power section consists of a 3-phase rectifier,
a boost converter, voltage and current meters. Power section is
I. INTRODUCTION integrated with a control section, containing STM32F4
Each ergometric system designed for human body microcontroller, optocouplers and MOSFET driver. The
measurements needs to satisfy not only high accuracy control section communicates with the PC through UART.
requirements, but also ergonomic factors. Proper sport The measurement software is running on PC and enables to
machine design has to consider all relationships between save measured data and change parameters of the control
human muscle force and velocity of the moving joints [1]. The algorithm, implemented in the microcontroller. The energy
angular position of limb needs to be taken into consideration produced during the tests was feeding a gel battery.
as well, because the force produced by a muscle changes
within area of the possible movement.
For training purpose it is very important to control an
average load torque of the exercise machine. If the machine is
powered electrically, accurate and dynamic control of the
sportsman load is possible. This control is usually combined
with continuous physiometric measurements. Unfortunately,
this torque load often may be only active. The most common
way is to increase the moment of inertia by equipping the
exercise system with a flywheel. This approach leads to
increased weight of the sport machine. Other way is electronic
imitation of all passive torques and inertia. To keep the
sportsman movement natural, it is necessary to consider
ergonomic aspects of machine operation and implement
methods of torque correction, making the pedaling more
similar to the one observed during riding on a real bicycle.
The sportsman load should not be kept at the constant value, Figure 1. Mechanical and electric flowchart
but needs to be fluctuated. These required fluctuations may be
derived from the measurements of human movement B. Measurement Methodology
characteristics. Only this way the ergometric measurements Motor velocity and angular pedal position synchronization
may be treated as realistic. In this paper, four methods of cycle are the only two mechanical measurements that were done
during ergometric tests. The measured electric quantities were:
Cezary Bielak, Agregaty Polska, Poznan, Poland, voltage and current values after rectification of the motor
cbielak@agregatypolska.pl

978-1-4673-5508-7/13/$31.00 ©2013 IEEE 161


phases currents. The generator torque may be easily estimated, PWM duty cycle 50%
since the mechanical characteristics of the BLDC had been PWM duty cycle 30% 90 40
tested before the ergometric tests. The gained characteristics 120 60
30
were: rectified voltage vs. velocity and current vs. mechanical
torque. 150 20 30

The cyclist was asked to remain an even pedaling speed 10


(cadency). In the human pedaling characteristics
measurements, presented and discussed in the chapter III, 180 0
a constant pulse width of the boost converter voltage signal
was adjusted. The rectified motor current measurements
allowed to estimate the torque. It was assumed that the gear 210 330
and pedal efficiency may be neglected.
Tests of ergometric corrects were conducted in a closed 240 300
current control loop. Since in the operation range of the cycle 270
ergometer, motor current is proportional to the torque, it is
assumed that the current control loop directly controls the
Figure 3. Pedaling characteristic with two PWM voltage values
cyclist load torque. The reference current was modified in
order to relieve the cyclist in these positions, where his
pedaling force decreases. Full control loop is shown in the
figure 2. The values measured to compute the reference PWM duty cycle 50%
90
PWM duty cycle 30% 25
current correction are: cadency and position of pedals. 120 60
Cadency is measured by multiplication of generator velocity 20

by support gear factor. Hall sensors located in the stator of 15


BLDC motor were used to derive the motor rotational speed 150
10
30

by counting the time between each Hall sensor signal 5


transition. Position of pedals is derived by counting motor
Hall sensor ignals. The position is synchronized once per 180 0
rotation using an additional Hall sensor located near support
axis bush. Alternative measurement system has been
presented in [2, 3]. 210 330

240 300
270

Figure 4. Complement of the human pedaling characteristic to maximum


measured torque

To allow natural pedaling, but with controlled average


load torque, it is necessary to take into consideration described
angular fluctuations. Therefore the average torque shall be
Figure 2. Control loop overview. corrected with TC value, to keep the natural cyclist’s
characteristic. This corrective torque may be approximated by
III. HUMAN PEDALLING CHARACTERISTICS equation (1), when the fluctuation amplitude K and minimum-
Following the method presented in section II.B, the torque angle ϕ are known.
characteristic of human pedaling were researched. In the
figure 3 two cyclist torque vs. angular position of pedals Tc (θ ) = K ⋅ sin(θ − ϕ ) (1)
characteristics (polar coordinates) are shown. These
characteristics was collected under two values of PWM duty
cycle. The duty cycle was considered as a load value in this
part of research. Two shown representative loads can clearly IV. ERGOMETRICS CORRECTS FOR REFERENCE CURRENT
determine the main trend, which can be described in the The goal of this part was to keep the average current in
following two points: generator circuit at the reference value, but taking into
1. For larger loads, fluctuations of the force generated by consideration momentary or local corrects related to
human legs become more significant. In the result also the fluctuations described in section III. The subjective feelings
measured torque fluctuates. are difficult to be documented and compared, it is difficult to
find a numerical indicator corresponding with it. Variation of
2. There are two points in each pedaling cycle, where the angular velocity ω divided by average velocity was found as
torque is minimal – when pedal position is perpendicular to the best criterion and was used as an indicator of ergonomics
the ground (these positions are marked as 90 and 270 deg in quality (2).
the figure 3).

162
∑ (ω )
N
2
i −ω 8

J= i =1
(2) 7
ω
Both dynamic and static corrective methods were taken 6
into account. Dynamic methods focus on angular velocity and
acceleration measurements, static methods base on

Iref [A]
5
characteristics shown in figure 5. Comparison of all methods
is presented in table 1.
4

25
PWM fill = 50% 3
PWM fill = 30%

20 2
0 1 2 3 4 5 6
Required Torque Correct [Nm]

Support angle [rad]

15 Figure 6. Exemplary asymmetric sine function.

10
⎧ ⎛ ⎛θ ⋅π π ⎞⎞
⎪ I ref 0 ⋅ ⎜⎜1 − A ⋅ sin⎜⎜ − ⎟⎟ ⎟⎟ θ < θ s
⎪ ⎝ ⎝ θs 2 ⎠⎠ (3)
5
I ref =⎨
⎪ I ⋅ ⎛⎜1 − A ⋅ sin⎛⎜ θ ⋅ π + π ⎞⎟ ⎞⎟ θ ≥ θ
⎪ ref 0 ⎜ ⎜ π −θ 2 ⎟⎠ ⎟⎠
s
0
0 1 2 3 4 5 6 ⎩ ⎝ ⎝ s
Pedal angle [deg]
This approximation is much closer to the real pedaling
Figure 5. Complement of human pedaling torque to constant torque characteristic complement. Its results are shown in the table 1.
vs. angular position – Carthesian coordinates Authors have not found any more reliable method of
ergonomic correction utilizing only the position of pedals.
A. Sinusoidal Corrective Methods
The simplest approximation of characteristic presented in B. Velocity Derivative Feedback
the figure 3 is just a sinusoidal function of support angle. The two following methods are based on the dynamic state
Unfortunately, this representation does not allow to get of the mechanical system. In particular, the pedals and motor
satisfactory results – this major disadvantage is documented in velocities measurements are essential to compute proper
table 1. The second negative feature is high computational correction value using this methods.
demands, since nonlinear function values need to be
determinate in real time. Although, contemporary First dynamic method is focused on a current velocity
microprocessors often incorporate floating point unit, so the derivative measurement. Simulation of inertia of real bicycle
second disadvantage does not constitute any significant is essential in this correction method. The derivative dVel is
difficulty in the implementation. On the other hand, for not only computed as a difference factor of current and
implementation in legacy devices, the sine function values previous velocities: a digital implementation of the first order
may be preliminary computed and tabulated in an array. inertial element, given by formula 4, where k1 is the inertia
constant and Ts is the velocity sampling time.
Both pure sine calculations and an absolute value of sine
of current pedals angle gave similar unsatisfactory results. ⎛ ω[k ] − ω[k − 1] ⎞ (4)
dVel[k ] = dVel[k − 1] + k1 ⋅ ⎜⎜ − dVel[ k − 1] ⎟⎟
These two approximations were evaluated as unsatisfactory ⎝ Ts ⎠
and not worthy of implementation in ergometric
measurements and test. The velocity derivative signal is used to form a feedback
loop signal. The dVel values is gained and then added to the
Better results were obtained using asymmetric sinusoidal reference current value as an ergonomic correction. The whole
approximation of complement function shown in the figure 5. control system diagram is shown in the figure 7.
As it may be easily noticed, for higher cadencies, the shape of
the complement function is asymmetric: one slope is steeper This way, an external control loop was formed, in order to
than the second one. To approximate this relationship, formula keep a constant pedaling speed. If the cyclist increases the
3 may be used. A is the maximal correction value (0 ≤ A ≤ 1), speed in these positions, where his force applied to pedals is
Iref0 is the original reference current value and θs is the angle of higher, the controller increases the generator current to limit
switch between two slopes of compensation characteristic. An the velocity increase. On the other hand, if the cyclist's force is
exemplary function is shown in figure 6. smaller in the current pedal position, the controller reduces the
generator current (and cyclist load) to limit the derivative to
low level, and keep an even cadency.

163
The Virtual Flywheel Method gave the best results of four
tested methods. In the figure 8 stages of acceleration, constant
speed may be easily recognized. Even at high load values, the
pedaling may be keep smooth. Operation with and without
corrections is compared in figure 9 and 10. Virtual Fly Wheel
method allowed to avoid velocity unsteadiness caused by the
characteristic of human muscles force at different angle
position of leg joints.
Figure 7. Control loop with Velocity Derivate Feedback Method.

It is worth to indicate, that in the presented approach,


cyclist is considered as a disturbance, that forces the changes
of generator velocity, that need to be reduced by the controller
operation. In this method, the current controller operation is
extended by the addition of ergonomic corrections, to a very
soft, proportional controller of motor acceleration (with
constant reference value equal to 0) at one time.
The results of the Velocity Derivative Feedback Method
are slightly better than those in static methods (see table 1),
but in subjective impression, that is difficult to document in
mathematical manner, the quality of the pedaling fluency is
still not satisfactory and differs from traditional bicycle
pedaling.

C. Flywheel
The last tested method of the cyclist load correction is
based on the model of flywheel equipped with a ratchet. The
basic model of the flywheel, that have been implemented in
the tested system may be described by the equation 4, where J Figure 8. Acceleration with corrective method
is the moment of inertia of the mechanism transformed to the
motor shaft, I, ωm – real motor current and virtual wheel
rotational speed, k3 – motor constant, R – viscous friction
torque coefficient.
dω m
J⋅ = k3 ⋅ I − R ⋅ ω m (4)
dt
The model of ratchet was implemented considering the
following two states.
1. If velocity of the virtual flywheel is lower than the
velocity of the real mechanism, the virtual flywheel ωm
velocity approaches the real velocity ω; the process is given
by the equation 5. This way a virtual ratchet is smoothly
locked. In this state the ergonomic correction is proportional
to the difference between real and model velocities.
ω m [k ] = k 4 ⋅ (ω[k ] − ω m [k − 1]) (5)

2. If the velocity of the flywheel is higher than the real


mechanism velocity, the virtual flywheel slows with the Figure 9. Motor velocity vs. sample number (proportional to time) - pedaling
passive friction torque (see equation 4), until the velocity of without ergonomic correction
real mechanism will be approached. In this state the correction
is equal to the current corresponding the torque required to
maintain the momentary dynamic state (equation 6) as if the
virtual inertia was really connected to the motor.
J dω (6)
I ref = I ref 0 + ⋅
k 3 dt

164
TABLE I. CORRECTIVE METHODS COMPARISON
1.05
Corrective Method (Method of Quality
1
reference current calculation) Indicator J
No correction 0,0653
Motor rotation frequency [Hz]

0.95 Sinusoidal Corrective Methods


0,0272
(IV.A)
0.9 Velocity Derivative (IV.B) 0,0226

Virtual Flywheel (IV.C) 0,0061


0.85

0.8

0.75
REFERENCES

0.7
0 100 200 300 400 500 600 700 800 900 1000
Sample number [1] D. R. Wilkie, “The Relation Between Force and Velocity in Human
Muscles”, J. Physiol (1950) 110, pp. 249-280
Figure 10. Pedaling with Virtual Flywheel corrective method [2] L. Jiong, “Diagnosis and Feedback System for Bicycle Pedaling
Technique and its Application”, International Conference on Future
V. CONCLUSIONS Computer Science and Education (2011), pp. 205-209
[3] J. Y. Chen, “A new automatic measurement system for a bicycle”,
Tests and analysis presented in this article allows to infer IEEE International Conference on Mechatronics (2005), pp. 514-519
that dynamic methods presented in sections IV.B and IV.C
provide much more natural pedaling than static ones.
Dynamic aspects of human force application cannot be
omitted in discussion of ergonomics of stationary exercise
machines equipped with pedals.

165

You might also like