RATIO CONTROL OF AN ELECTROMECHANICAL DUAL ACTING PULLEY CONTINUOUSLY VARIABLE TRANSMISSION (EMDAP- CVT) SYSTEM USING PD-FUZZY LOGIC

C CONTROLLER
Bambang Supriyo1, Kamarul Baharin Tawi, Hishamuddin Jamaluddin and Sugeng Ariyono2
Faculty of Mechanical Engineering, Universiti Technologi Malaysia 81310 UTM Skudai, Johore, Malaysia
1

Department of Electronic Engineering, Politeknik Negeri Semarang

Jl. Prof.H. Sudarto, S.H. Tembalang, Semarang 50329, Indonesia E-MAIL : bsupriyo7763@yahoo.com

Department of Mechanical Engineering, Politeknik Negeri Semarang

Jl. Prof.H. Sudarto, S.H. Tembalang, Semarang 50329, Indonesia

Abstract
Currently, a hydraulic system has been widely used in automotive application for changing transmission ratio of metal pushing V-belt Continuously Variable Transmission (CVT) utilizing a single acting pulley system. This system needs continuous power to supply force to both movable pulleys to maintain the desired transmission ratio and clamping force. The application of continuous hydraulic force introduces power loss. This paper introduces an alternative way of changing transmission ratio using an electromechanical system applied to dual acting pulley CVT system. This system has a viable solution for reducing the power consumption, since it operates only during changing the transmission ratio. The dual acting pulley system incorporates two movable pulley sheaves in both primary and secondary shafts to keep the belt position always aligned in the center of shafts, even during the ratio change, hence eliminating the belt misalignment effects. The electromechanical system introduces two DC motors as actuators and power screw mechanisms as mechanical linkages to adjust and to brake the axial position of movable pulleys. The primary motor is used for changing the transmission ratio, while the secondary one is used for preventing the belt from slipping. This paper proposes proportional differential (PD) fuzzy logic controller to control both primary and secondary DC motors such that both primary and secondary pulley radii satisfy the desired transmission ratio. The initial value of the PD gain is derived using Astrom and Hagglund tuning method. Then, this gain is fine tuned using fuzzy logic to improve the performance of the transmission ratio control automatically until it converges to a predetermined optimal global criterion. Computer simulation results are presented to demonstrate the effectiveness of the proposed controller.

Keywords :
Electromechanical CVT, Dual Acting Pulley CVT, Ratio Control, Fuzzy Logic Controller.

1. Introduction
Due to its wide range coverage, continuously variable transmission (CVT) enables the engine to run in its fuelefficient operating point for any vehicle load, hence lowering the engine fuel consumption. Most current metal pushing V-belt CVTs are hydraulically actuated. This hydraulic system requires continuous power to supply clamping force. This continuous energy consumption becomes the major loss hence reducing CVT efficiency [1].

Although some efforts on improving the performance of hydraulic systems have been done using a new servo hydraulic actuation system [2], a new electromechanically actuated CVT has been introduced in [3] to improve CVT efficiency. Slightly different with [3], this paper introduces an electromechanical CVT with dual acting pulley system. This system adopts two movable pulley sheaves in each shaft to eliminate belt misalignment effects. Long term application of this misalignment may damage the belt. The metal belt misalignment has been studied intensively in [4]. Two DC

servomotors are used as actuators. These motors will be activated when transmission ratio is changed, hence cutting the actuator operation time and reducing the energy loss. PID (Proportional, Integral and Derivative) controller has been the basis in the simple linear control systems. The PID controller is a well-known technique for various industrial control applications, mainly due to its simple design, straightforward parameters tuning, and robust performance. PID controllers are usually used to control servomotors. Position controls utilizing PID can be seen in [5],[6],[7],[8]. The conventional approach to determine the PID parameters is to use a mathematical model of the process and a simple tuning law with a fixed set of gain parameters. One famous example of such approach is the Ziegler-Nichols method [9]. However, an accurate model of a real industrial process is difficult to obtain due to its complex characteristics and parameter variations. Therefore, tuning laws based on these inaccurate models are no longer adequate. In order to increase the capabilities of the PID controller in improving its performance, a fuzzy logic controller (FLC) will be used. The FLC presents an algorithm for translating an expert knowledge-based linguistic control strategy into an automatic control strategy [10]. Fuzzy-logic-based PID controllers offer knowledge representation and learning capabilities to tackle a wide range of complex dynamical systems, which may be ill defined or subjected to numerously varying parameters. An auto tuning of PD controller using Takagi-Sugeno type fuzzy logic controller is proposed for performing the ratio control of the EMDAP CVT system.

Figure.1. Variator Geometry Based on this variator geometry, the relationship between speed and pulley radii can be given as follows:

s Rs = p R p

(1) (2)

s p R rg = p Rs
rs =

(3)

Where Rp, Rs , p , s ,rs , rg are belt contact radius on the secondary pulley, belt contact radius on the primary pulley, angular speed of primary pulley, angular speed of secondary pulley, speed ratio, and geometrical ratio respectively. The following implicit relationship between belt length (L) and the pulley radii (Rp, Rs) can be derived as:
L = ( + 2 ) R p + ( 2 ) Rs + 2c Cos

(4) (5)

= arcCos (

R p Rs c

where c is the pulley center distance, and is the half of the belt wrapped angle on the primary pulley.

2.1. Metal V-Belt CVT

A metal V-belt developed by Van Doornes Transmissie of Holland consists of thin, flat tension bands (sometime called as rings) and metal V- segments ( sometime called as blocks), which filled the entire length of the band. Each band is approximately 0.2 mm thick, and is laminated. Thin segments are strung together by two sets of thin bands positioned in slots at each side of the segments. One set of the band consists of about nine bands, and there are about 300 segments for each belt [11]. The metal belt structure [12] can be seen in figure 2.

2. Background of CVT
The idea behind a continuously variable transmission is based on the use of a variator instead of gears. The variator consists of a primary and a secondary pulleys and a metal belt. The variator geometry is given in figure 1. By assuming that no gross belt slip occurs, the belt is inextensible and it runs at perfect circles with radii Rp (on the primary pulley) and Rs (on the secondary pulley), the tangential velocities of primary pulley, secondary pulley, and the belt will be the same.

Figure 2. The structure of a Metal V-belt CVT (Srivastava and Haque ,2004)

2.2. Electromechanical Dual Acting Pulley CVT

The electromechanical dual acting pulley CVT (EMDAP CVT) system utilizes two DC servomotors as actuators. The system consists of two sets of pulleys, namely primary pulley placed on input fixed shaft, and secondary pulley placed on secondary fixed shaft. Each set of pulley has two movable sheaves that can be shifted axially along the shaft. The servomotor shaft is directly connected to a series of gear reducers and power screw mechanism to axially move the pulley sheaves. A Van Doornes metal pushing V-belt is placed between pulley sheaves, and runs on the surface of the sheaves. This metal belt connects the primary and secondary pulleys to transmit the power and torque from the input to the output shaft by means of friction between the belt and the pulley contact [13],[14]. The contact radius of the belt with primary pulley (Rp) and the contact radius of the belt with secondary pulley (Rs) determine the transmission ratio. The primary and secondary parts of the CVT system have the same components. The components of each part (primary or secondary) consist of a dc servo motor, a gear reducer with ratio of (30:1), a gear reducer with gear ratio of (60:14), power screw mechanism, and two movable metal pulley sheaves for clamping the metal belt. A spring disc is inserted in the back of each secondary pulley sheave to provide continuous clamping force to the belt, and to reduce excessive slip during transmission ratio change. The block diagram of the EMDAP-CVT system can be seen in figure 3.

Tm - TL = Jm .s. +Bm.s2. = /s Where : Va : Motor Voltage [V] La : Motor Inductance [H] ia : Motor Current [A] Ra : Motor Resistance [] Ka : Back emf constant [mV/(rad/sec)] : Motor shaft angular velocity [rad/sec] : angular displacement [rad] Tm : Motor Torque [Nm] Km : Torque Contant [Nm/A] TL : Load Torque [Nm] Jm : Motor Inertia [Nm.sec2] Bm : Viscous friction coefficient [Nm/rad/sec]

(8) (9)

3.2. Gear Reducer and Power screw Mechanism A gear reducer serves as a speed reducer and torque multiplier. The gear reducers are coupled with the servomotor shaft. The gear reducer output is connected to power screw mechanism for shifting the pulley sheaves. There are two gear reducers employed in the EMDAP-CVT system. The first one has ratio of (30:1), while the second one has the ratio of (60:14). The total gear ratio of these reducers is (128.57:1). The power screw mechanism is used to move the pulley sheaves axially. It converts every 360 of rotation into 2 millimeters axial movement. The combination of gear reducers and power screw mechanism is used to help the motor in providing significant torque to turn the power screw mechanism.

4. Proposed Controller
This paper proposes a Fuzzy-PD controller for each DC servomotor system to control the transmission ratio of the CVT. The primary DC servomotor system will tracks the desired value of the primary pulley radius, and the secondary DC servomotor system will take the actual value of the primary pulley radius as its input and set the secondary pulley radius accordingly to prevent belt slip. The common problem of tuning PID is dealing with specifying three gain parameters to meet a certain requirement of control specifications, such as set-point following, load disturbance attenuation, robustness due to model uncertainties, and noise rejection [15]. Many different techniques have been proposed in the literature to cope with these tuning problems. The use of Ziegler-Nichols formula for tuning the PID parameters usually gives a good result in load disturbance attenuation, but poor in both lowering the overshoot and reducing the settling time. These two aspects are mainly triggered by the selection of high proportional gain. To cope with this problem, a fuzzy logic controller is proposed to obtain the suitable weight value to adjust the

Figure 3. The block diagram of the EMDAP-CVT

3. System Modeling
3.1. Modeling of DC Servomotor The dynamic model of the DC servomotor is represented as follow: (6) Va = sLa ia + Ra ia + K a
Va K a . = s.La .i a + Ra .i a

(7)

proportional gain with respect to the current values of the system error (e) and its change of error rate (de). The block diagram of the proposed controller scheme is given in the figure 4.
TR_set-toRp_set + + Kp X DC Servomotor Gear Reducer Rp_act

+ -

Process Output PROCESS

TR_set

d/dt -

Kd

d/dt

FUZZY Controller

Rp_act-toRs_set

d/dt

Kd

DC Servomotor

Gear Reducer

Rs_act

Kp

Figure. 5 The relay feedback controller

d/dt

FUZZY Controller

4.2. Fuzzy Auto Tuning Fuzzy logic is used to perform online tuning for PID parameters. A tuning algorithm proposed in this paper only tunes the proportional gain (Kp ) of the PID controller. The fuzzy auto-tuner has two inputs: the error (e) and the change error rate (de) of the output response. To normalize these inputs into the range of [-1,1] , these inputs are scaled by two coefficients, Ke and Kde respectively. The fuzzy system can be started to operate at a certain error value by specifying the value of Ke. This paper selects Ke=10, meaning that the system will effectively operate when the output error is in the range of [-0.1,0.1]. The Kde can be set initially to 1. Three triangular membership functions (see figure 6) are defined for each input (N = Negative, Z = Zero, P = Positive), while five constant values of Takagi-Sugeno type in the range of [-1,1] are defined for the output (NB = Negative Big, NS = Negative Small, Z = Zero, PS=Positive Small, PB=Positive Big).

Figure 4. Block diagram of the Fuzzy-PD controller

4.1.

Parameter Tuning for PID Controller

The transfer function of a PID controller has the following form: (10) G PID = K p + K i / s + K d s

G PID = K p (1 + 1 /(Ti s ) + Td s

(11 )

where Kp, Ki, and Kd are the proportional, integral, and derivative gains, respectively, Ti=Kp/Ki and Td=Kd/Kp. Ti and Td are the integral time constant and the derivative time constant, respectively. In order to obtain initial parameters of PID controller, the Astrom-Hagglund method [16] will be used to determine both critical period of waveform oscillation (Tc) and critical gain (Kc). These two values can be obtained by running the closed loop control of DC servomotor system with relay feedback controller. The oscillation period of the output waveform is considered as the critical period attained from a proportional feedback. The critical gain is given as follows:

Kc =

4d a

( 12)

(a) error

Where d is the amplitude of the relay output, and a is the amplitude of the waveform oscillation. Based on the values of Tc and Kc, the PID parameters (Kp, Ti , and Td) can be specified using Ziegler-Nichols formula [9] (see Table 1.). Table 1. Ziegler-Nichols parameter tuning. Kp Ti Td 0.5 Kc 0.45 Kc 0.85 Tc 0.6 Kc 0.5 Tc 0.125 Tc
(b) the change of the error

P PI PID

wi = Aj (e(k )) x Bj (de(k ))
j=1,2,3, i=1,2,3

(18 )

(c) Output Figure 6. Membership function of input and output

This type of defuzzification method is selected due to its simple and fast computation [8].
a 1

The error (e (k)) and the error change rate (de(k)) are defined as follow: (13) e (k) = xr(k) xa(k) de(k) = e(k) e(k-1) (14 ) The fuzzy output (Yf) is the singleton value in the range of [ -1, 1] of the Takagi-Sugeno Type. This output value is then multiplied by the proportional gain (Kp) and then summed with the derivative gain to give the overall controller gain (Gc) : Gc= Yf * Kp + Kd (15)

e b 0 d f g c 1 2 3 4 5 6

Figure 7. Error curve of output Table 2. The prototype of the fuzzy control rules

The fuzzy rules to determine the output fuzzy (Yf) is the following equation: Ri: if e(k) is Aj and de(k) is Bj then Yfji Where i (i=1,,9) is the number of rules. (16)

Rule No. 1 2 3 4 5

e P Z N Z Z

de Z N Z P Z

Yf PB NS NB PS Z

Reference Point a,e b c d f

Table 3. the justification of a fuzzy control rules

The rule justification is based on the scale mapping method developed by King and Mamdani [10]. This justification is done by referring to a closed system trajectory in a phase plane. The control rules of the consequence of (Yf) are determined from the error curve of the system step response [9] (see figure 7). This justification requires both knowledge of controller parameter adjustment based on the phase plane analysis, such as overshoot and rise time, and intuitive feeling of the behavior of the closed loop system. The prototype of the fuzzy control rules is given in table 2. While the justification of a fuzzy control rules is presented in table 3. Defuzzifcation is needed for the fuzzy logic controller to convert its internal fuzzy output variables into crisp values usable for the controlled system. This paper will adopt the weighted average method represented by:

Rule No. 6 7 8 9

e P N N P

de N N P P

Yf PB NB NB PB

Reference range 1,5 2 3,6 4

5. Simulation Studies
Simulation studies of the proposed Fuzzy-PD controller are carried out in order to investigate its effectiveness in this ratio control application. In these studies, the DC servomotor has the following important parameters as shown in Table 4. Table 4. DC servomotor parameters Parameters Motor Voltage Armature Resistance Armature Inductance Back emf Constant Torque Constant Nominal Torque Nominal Speed Rotor Inertia Values 24 V 0.03 0.1 mH 7 mV/rpm 0.0674 N/A 1.6 Nm 3000 rpm 0.1555e-4 N.sec2

yf =

w c
i =1 i

j ,i

(17 )

w
i =1

Where : n is the number rules, Cj,I is the vale of the center of gravity of the fitness of i-th rule in the premise, and the fitness (wi) is defined as:

The simulation of the controller was performed using MATLAB/FUZZY-SIMULINK packages. Control performance is determined based on trajectory errors. An input excitation of sinusoidal waveform is used to examine the performance of the conventional PD and the Fuzzy-PD controllers. The initial proportional gain will be set to the same value to ensure reasonable comparisons between these two controllers. In order to obtain initial parameters of PID, the AstromHagglund method based on a relay feedback controller is carried out to attain the critical period of waveform oscillation (Tc) and critical gain (Kc). The amplitude of the relay controller is set to 10 since the input voltage in the range of [-5,+5 volts] is needed to drive the servo system. From simulation results, these following parameters are found: Tc= 0.4 s, a = 0.1, and d = 10 . By using equation (12), the critical gain (Kc) is 127.4. Finally, by using these values, together with Ziegler-Nichols formula and both equations (10) and (11), the three parameters of PID can be specified as follows: Kp=76.4, Ki =636.9, and Kd=3.8. From these data, it was found that the integral gain ( Ki) is too big and it is not suitable for this application, so the PID controller will be reduced to PD controller. Figure 8 and figure 9 show the output responses of the sinusoidal wave excitation. From figure 9, it can be seen that the Fuzzy-PD controller has smaller position error (about -0.003 to +0.003) compared to the conventional PD controller (about -0.02 to + 0.02). Thus, the Fuzzy controller has significantly improved the performance of the conventional PD controller, especially in terms of minimizing the position error of trajectory tracking.

6. Conclusion
The conventional PID is initially tuned using AstromHagglund method and Ziegler-Nichols formula. However, only both proportional and derivative gains of the PID are used to control the servomotor system. The integral gain is not used, since it is too big and unsuitable for this application. The proposed controller utilizes a fuzzy logic controller to tune the proportional gain of the PD controller and to improve the PD performance. The simulation results have shown that the Fuzzy controller has significantly improved the performance of the conventional PD controller, especially in terms of minimizing the position error of trajectory tracking for ratio control of the EMDAPCVT.

Reference
[1] Akerhurst, S. and N.D. Vaughan.1999. An Investigation into the Loss Mechanisms Associated with an Automotive Metal V-belt CVT . In European Automotive Congress Vehicle Systems Technology for The Next Century, STA991407, Barcelona. Veenhuizen, P.A. and S. Cools. 2004. Servo Hydraulic Control of a Continuously Variable Transmission. In 7th International Symposium on Advanced Vehicle Control (AVEC04), Arhem, Netherland. Van de Meerakker, K., P. Rosielle, B.Bonsen and T. Klaassen. 2004. Design of An Electromechanical Ratio and Clamping Force Actuator for a Metal VBelt Type CVT. In 7th International Symposium on Advanced Vehicle Control (AVEC04), Arhem, Netherland. Tawi, K.B. Investigation of Belt Misalignment Effects on Metal Pushing V-Belt Continuously Variable Transmission, Ph.D. Thesis, Cranfield University. 1997. Shi Jingzhuo, Xu Dianguo, and Wang Zongpei, 2000. Hybrid Stepping Motor Position Servo System Based On DSP. In IECON 2000, 20th Annual Conference of the IEEE Industrial Electronics Society IEEE, Vol.2, 22-28 Oct 2000, pp.1475-1479. Habib, R. Maki, 2001. Designing Fuzzy Logic Controllers fro DC Servomotors Supported by Fuzzy Logic Control Development Environment. In IECON01: The 27th Annual; Conference of the IEEE Industrial Electronics Society. Yu Yang, Wen-Ge Wang, De-Jie Yu, nd Ge Ding, 2002. A Fuzzy Parameters Adaptive PID Controller Design of Digital Positional Servo System. In Proc. Of The First International Conference on Machine Learning and Cybernetics. Beijing. Yusuf I. Bulale, M.J.E. Salami, Marizan Sulaiman, 2003. Design and Development of DSP-based hybrid

[2]

[3]

[4]

[5]

Figure 8. Set-point follower

[6]

[7]

Figure 9. Position error of set-point follower

[8]

Controller For Servo Driver Applications. In Proceeding of 2003 IEEE Conference on Control Applications. Vol.1, 23-25,pp.773-778. [9] Hyung-Soo Hwang, Jeng Nae Choi, Won Hyok Lee, Jin-Kwon Kim, 1999. In A Tuning for The PID Controller Utilizing Fuzzy Theory,IJCNN99 International Joint Conference on Neural Network. Vol.4,10-16,pp. 2210-2215. [10] Lee, C.C, 1990. Fuzzy Logic in Control Systems: Fuzzy Logic Controller- Part I. IEEE on Trans. On System, Man, and Cybernetics. Vol.20. no.2. [11] Fujii, T. , T. Kurokawa and S. Kanehara. 1993. A Study of a Metal Pushing V-Belt CVT Type Part 1: Relation Between Transmitted Torque and Pulley Thrust, SAE Paper No. 930666. [12] Srivastava, N. and I. Haque. 2004. On the Operating Regime of a metal pushing V-belt CVT under steady state microslip conditions. In 2004 International

[13]

[14]

[15]

[16]

Continuously Variable Transmission and Hybrid Transmission Congress, September 23-25. Kanehara, S. , T. Fujii and S. Oono.1996. A Study of a Metal Pushing V-Belt Type CVT : Macroscopic Consideration for Coefficien of Friction between Belt and Pulley, SAE Paper No. 9636277. Fushimi, Y. , T. Fujii and S. Kanehara. 1996. A Numerical Approach to Analyse the Power Transmitting Mechanisms of a Metal Pushing V-Belt Type CVT. SAE Technical Paper Series 960720. Visioli, A.,1999. Fuzzy Logic Based Set-Point Weight Tuning of PID Controllers. IEEE Trans. On Systems, Mans, and Cybernetics-Part A: System and Humans, Vol.29, No.6. K.J. Astrom and T. Hagglund, 1984. Automatic Tuning of Simple regulators with Specifications on phase and amplitude margins. Automatica, 20,pp.645651.