International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Sliding Mode Control of Active Suspensions for a Full Vehicle Model

N.YAGIZ*

and I.YUKSEK**

*Department of Mechanical Engineering, Faculty of Engineering , University of Istanbul, Avcilar Istanbul,Turkey, e-mail: nurkany@istanbul.edu.tr, Fax:+90 212 591 19 97 **Department of Mechanical Engineering, Faculty of Engineering , Yildiz Technical University,Yildiz, Istanbul,Turkey, e-mail: ismail@mel.go.jp

Abstract: In this study, a linear Seven Degrees of freedom Vehicle Model is used in order to design and
check the performance of Sliding Mode Controlled Active Suspensions. Force Actuators are mounted as parallel to the four suspensions and a non-chattering control is realized. Sliding Mode Control is preferred because of its robust character since any change in vehicle parameters should not affect the performance of the active suspensions. Improvement in ride comfort is aimed by decreasing the amplitudes of motions of vehicle body. Body bounce, pitch and yaw motions of the vehicle are simulated both in time domain in case of travelling on a limited ramp type of road profile and frequency domain. The robustness of the controller has been proved by using different vehicle parameters such as vehicle mass and damper ratios. Also phase plane plots of them are checked. Simulation results are compared with the ones of passive suspensions. Keywords : sliding mode control, active suspensions, full vehicle model, simulation.

1 Introduction

The main functions of a vehicle suspension system are to provide effective isolation from road surface unevenness, stability and directional control during handling manoeuvres and to provide vehicle support. Traditional vehicle suspension systems are composed of two parallel components which are springs and viscous dampers. The passive suspension system designers are faced with the problem of determining the suspension spring and damper coefficients. They have to compromise two important factors which are conflicting each other. These are ride comfort and road holding. Good ride comfort needs soft springs. But this means poor road holding. Furthermore when talking about passive suspensions, there is no way to get rid of the resonance frequencies such as the most important one around 1 Hz. which is the result

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

of vehicle body dynamics. Therefore the improvement of the vehicle suspension systems has gained more interest and been the subject of the research and development in recent years. This activity has two reasons which are commercial and scientific. The main reason of the commercial activity is the desire of the automotive manufacturers to improve the performance and quality of their products. On the other hand, researchers and control system designers have claimed that the automatic control of the vehicle suspension systems is possible when the developments in actuators, sensors and electronics have been considered. When the performance characteristics of a desired suspension system has been taken into consideration, the suspension control has become more attractive . The last fifteen years, many studies have been published on active and semi active suspension systems. Most of the investigators used quarter car model. Procop and Sharp (1995) studied active Automotive Suspensions by road preview on a quarter model. Hrovat (1993) surveys applications of optimal control techniques for the design of active suspensions in one of his study starting with a quarter model. Nonlinear control of a quarter car active suspension is reviewed by Alleyne and Hedrick(1995). Burton et al (1995) have brought together analysis of active and passive quarter-car systems and a full-scale test rig in their paper. Redfield and Karnopp (1988) examined the optimal performance comparisons of variable component suspensions on a quarter car model. Yu and Crolla (1993) presented an optimal self-tuning control algorithm using a quarter model considering both external and internal disturbances. Although the quarter car model has been proved to be useful for designing control strategies, it does not reflect the terms such as pitch and roll. But some of investigations based on three dimensional vehicle model have been published. Chalasani (1986) studied ride isolation performance and road holding qualities of the active suspension on a seven degrees of freedom vehicle model. Esmailzadeh and Fahimi (1997) presented a method for designing active suspensions where the dampers of a passive system are replaced by actuators which are controlled with optimal full state vector feedback on a three dimensional vehicle model. Abdel Hady and Crolla (1989) outlined techniques for

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

obtaining control laws for an active suspension mounted on four wheels where ground input information can be omitted. Yuksek and Kaya (1995) discusses vibration optimisation of a vehicle by calculating optimum values of control forces such that overall maximum vehicle response amplitude is minimised on a full car model.

The aim of this study is to apply the non-chattering sliding Mode control to automotive suspension systems. If not prevented, the chattering causes damages to the mechanical components. Sabanovic proposed an effective method for chattering free sliding Mode applications (Sabanovic,1994). The improvements in electromagnetic force sources and sensors make it possible anymore . Dan Cho (1993) presented the application of sliding mode control to stabilise an electromagnetic suspension system with experimental results. The sliding Mode control proposed first time in Soviet Union by Emelyanov and Utkin. A survey paper by Utkin referances many of the early contributions available in translation (Utkin,1993). Because of the language and reference problems, it has taken a long time to enter the western literature (Hung,1993). Utkin (1981) published an excellent book on Sliding Mode Control. Young (1978) showed that the method is successfully applicable to robot manipulators. Yagiz et al. (1997) proposed the application of Sliding Mode Control on a relatively simple vehicle model. The superiorities of this method are the applicability on nonlinear systems, simplicity, high performance and its robust character . On these days, this method has been applied to robot control, flight control, motor control and power systems successfully.

2 The full vehicle model

In the full vehicle model shown in Figure 1, V is the vehicle speed, a and b are the longitudinal distances of the front and rear wheel centres from the body centre of gravity. The

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

road input to the four wheels are denoted as z1(t), z2(t), z3(t) and z4(t). There will be t = (a+b)/V delay time between the front and rear wheel inputs. The controlling action will be realized by force actuators which work in parallel with suspension elements. u1, u2, u3 and u4 are the control forces. Referring to Figure 1, mi, ki, ci, kti are the unsprung mass, suspension spring rate, suspension damper rate and tyre spring rate of the corresponding elements respectively, where c and d are the horizontal distances of the left and right wheel centres from the body centre of gravity, M represents body sprung mass and Iyy , Ixx are the body pitch and roll moments of inertias. Sprung mass is allowed to have bounce, pitch and roll motions x1, x2 and x3 while unsprung mass can move freely with respect to ground and the sprung mass. x4 to x7 represent the axle displacements vertically from equilibrium. The angles of rotational motion are assumed to be small. The vehicle model has seven degrees of freedom.

Figure 1 The full vehicle model.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

The state vector x = [ x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14 ]T comprises the displacements x1, x2, x3, x4, x5, x6, x7 and their corresponding velocities respectively. Using the states above, the non-linear set of the state equations can be written in the form:

. x = f ( x ) + [B]* u

(1)

where f (x) is the vector of the non-linear equations of the system excluding the control inputs, [B] is the input matrix having the dimension of (14 x 4) and u = [ u1 u2 u3 u4 ]T is the vector of the four control forces for the current model. The complete set of fourteen state equations and [B] are given in Appendix.

The sliding mode controller design

Sliding Mode Control Theory has been applied in numbers of non-linear systems. The main idea is to bring the error on sliding surface such that system is on sliding surface and insensitive to the disturbances and parameter changes. If the system is defined as in equation (1) where dim[B] = n* m, dim(f(x)) = n*1 and dim(u) = m*1; f(x) is continuos, but u(t) may be discontinuous. The aim is to hold the system motion on a sliding surface S. The surface can be expressed as: S = { x : ( x, t ) = 0 } (2)

In order to obtain a stable solution of the system, it must stay on this surface, i.e. ( x, t ) = 0 as shown in Figure 2.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Figure 2 Phase plane diagram of the state variables.

The sliding surface equation for the control of the system can be selected as follows:

(x,t) = [G] * ( xref - x ) = [G] * e

(3)

In this equation xref represents the state vector of the reference, and the constant [G] matrix represents the slope of the sliding surface. The same equation also can be written as:

(x,t) = (t) - a( x) where,

(4)

(t) = [G] * xref (t)

and a( x) = [G] * x

(5)

The first step in design is to select Lyapunov function . According to Lyapunov Stability Criteria, Lyapunov function must have a value greater than zero whereas its derivative is smaller than zero. Selecting the function as in Equation (6) makes its value greater than zero:

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

= T(x,t) * (x,t)/2 > 0

(6)

In order to have the value of the derivative of Lyapunov Function smaller than zero:

d = - T (x,t)* D* (x,t) < 0 dt

(7)

Thus Lyapunovs Stability Criteria has been satisfied. If we equate (7) to the derivative of (6):

d (x,t)/dt = - D* (x,t) and d (x,t)/dt + D* (x,t) = 0

(8)

As it is seen in equation (8), sliding function goes to zero at infinity. But our goal is to send it very close to zero. If equation (4) is differentiated and (2) used, the derivative of the sliding surface is obtained as:

d (x,t)/dt = d(t)/dt - {a (x)/x } * dx/dt = d(t)/dt - [G]*(f(x) + [B]*u(t))

(9)

and ( [G]*[B] )-1 must exist. The controller is designed as below by inserting (9) in (8):

u(t) = ueq(t) + ( [G]* [B] )-1 * D * (x,t)

(10)

ueq(t) = - ( [G]* [B] )-1 * ( [G]* f(x) - d(t)/dt)

(11)

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

If the knowledge of f(x) and [B] matrices are very poor, then the equivalent control calculated will be too far off from the actual equivalent control. In the literature a number of approaches are proposed for the estimation of ueq , rather that calculating it. In this study, the approach suggested uses the fact that the equivalent control is the average of the total control. Let us design an averaging filter for calculation of the equivalent control as below.

u eq =

1 u s + 1

(12)

This is actually a low-pass filter. The value of

1/ gives the cut-off frequency. The logic

behind the designing a low pass filter is that low frequencies determine the characteristics of the signal and high frequencies come from unmodeled dynamics. Then:

u(t) = u eq + ( [G].[B] )-1 * D* (x,t)

(13)

At this type of non-switching control, the system trajectories will stay only in the vicinity of the sliding surface.

4 Simulation

First, the vehicle model with passive suspensions was considered. Then by adding force actuators, active system was formed. After Sliding Mode Controller design, the control force inputs were calculated. The state equations of the passive system and active system were solved numerically both in time and frequency domain. The road profile is shown in Figures 3 a,b. These figures show that the vehicle rises 0.01 m. for the travel of first 0.01 m. Afterwards

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

vehicle travels on a straight road. The road input has been started at the end of the first second of simulation. Simulation period is five seconds. The vehicle travels at a constant speed of 40 km/h. Road surface input effects the rear wheels after a certain time delay. Because of the asymmetry of the vehicle mass centre, this road surface input is enough to cause bounce, pitch and roll motion. Since the inertia forces acting on passengers are very important, the acceleration of the sprung mass has been plotted. Vehicle Body acceleration for the road surface input is shown in Figures 3.c,d. Figure 3.c is the response of the passive system and figure 3.d is that of the active one. It could be seen that maximum vertical acceleration of the vehicle body in the active system is the half of the that of the passive system and the active system returns rest faster. This means when the vehicle body is following zero reference successfully, the inertia forces are also decreased by two times when the maximum values are considered.

Figure 3 Road surface input and Body Accelerations.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Sprung mass pitch and roll angular displacements for controlled and uncontrolled cases are shown in Figures 4.a and 4.b respectively. The success of the controller is obvious when both body acceleration and angular displacements are considered. Phase plane plots of the error in body bounce and roll are shown in Figures 4.c and 4.d, respectively. These figures demonstrate the action of the sliding surfaces.

Figure 4 Angular Motions and Phase Plane Plots for Active and Passive Systems.

Since the vehicle mass can be changed depending on the amount of luggage and number of passengers, the robustness of the controller is very important. As seen in Figure 5, the increased value of the vehicle mass does not effect the success of controller significantly.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Also in Figure 6, for different values of damping rates, the robust character of the controller is presented.

Figure 5 Vehicle Body Acceleration for Different Values of the Vehicle Body Mass.

Figure 6 Vehicle Body Acceleration for Different Values of the Damper Rates.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

When the ride comfort is considered, the frequency response of the vehicle body must be examined too. The frequency response of the passive and active systems are given in Figures 7.a,b and 7.c,d for body bounce and pitch motions. There are practically two effective resonance frequencies belonging to sprung and unsprung masses in case of passive suspensions. When we check the frequency response of the system without controllers, two resonance frequencies are observed around 1.1 and 10 Hz. On the other hand, when the controllers are active, resonance peak of the sprung mass vanishes and the amplitude of

motion, almost throughout the frequency range, considerably gets smaller additionally.

Figure 7 Frequency response of the passive and active systems

The control forces u1, u2, u3 and u4 are shown in Figure 8. It is observed that there is no sharp change in values known as chattering which can harm vehicle components. The value of the forces changes smoothly.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Figure 8. Control force inputs

5 Conclusion

In this study, a Sliding Mode Controller for a vehicle has been designed and simulation results have been presented. The main idea behind proposing this controller is its robustness and the ability of using these type of controllers on vehicles with developing technology. Since vehicle dynamics changes with load and road conditions, this method gains more importance. The results of this study prove that the performance of active suspension of this type is highly superior then the one of passive one. Against the disturbances coming from the road, the passengers are almost insensitive and it is foreseen that they feel ride as if on an excellent road surface.

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

Nomenclature

x u
xi

Vector of the state variables Vector of the control forces ith state variable

x ref Vector of the state references

zi(t) ith road surface disturbance kti ki ci Iyy a b c d Stiffness coefficient of ith tire Stiffness coefficient of ith suspension springM Damping coefficient of ith suspension damper Sprung mass pitch moment of inertia

Mass of the vehicle body Ixx Sprung mass roll moment of inertia

Vector of selected Lyapunov function

Distance between sprung mass center and front of the vehicle Distance between sprung mass center and rear of the vehicle Distance between sprung mass center and right of the vehicle Distance between spring mass center and left of the vehicle

f ( x ) Vector of the state equations without control inputs

Vector of the road surface disturbances

Appendix
State equations excluding control inputs:

8, f1 ( x ) = x

9, f 2 (x) = x

10 , f 3 (x) = x

11 , f 4 (x ) = x

12 , f5 (x) = x

13 , f 6 (x) = x

14 . f 7 (x) = x

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

f 8 (x ) =

1 { (c1 + c 2 + c 3 + c 4 ) x 8 + (cc1 + dc 2 cc 3 + dc 4 ) x 9 + (ac1 + ac 2 bc 3 bc 4 ) x 10 M c1 x 11 c 2 x 12 c 3 x 13 c 4 x 14 + (k 1 + k 2 + k 3 + k 4 ) x 1 +(ck 1 + dk 2 ck 3 + dk 4 ) x 2 + (ak 1 + ak 2 bk 3 bk 4 ) x 3 k 1 x 4 k 2 x 5 k 3 x 6 k 4 x 7 }

f 9 (x ) = I 1 { (cc1 + dc 2 cc3 + dc 4 )x 8 + (c 2 c1 + d 2 c 2 + c 2 c 3 + d 2 c 4 )x 9 + (cac1 + dac 2 + cbc3 dbc 4 )x 10

+ (cc1 )x 11 (dc 2 )x 12 + (cc 3 )x 13 (dc 4 )x 14 + (ck1 + dk 2 ck 3 + dk 4 )x 1 + (c 2 k1 + d 2 k 2 + c 2 k 3 )x 2 + d 2 k 4 x 2 + ( cak1 + dak 2 + bck 3 dbk 4 )x 3 + (ck1 )x 4 (dk 2 )x 5 + (ck 3 )x 6 (dk 4 )x 7 } f10 (x) = 1 { (ac1 + ac2 bc3 bc4 ) x 8 + (acc1 + adc2 + bcc3 dbc4 )x 9 + (a 2 c1 + a 2 c 2 + b 2 c3 + b 2 c 4 )x10 Iyy (ac1 )x11 (ac2 )x12 + (bc3 )x13 + (bc4 )x14 + (ak1 + ak 2 bk3 bk 4 )x1 + (cak1 + adk2 + bck3 )x 2 + (bdk4 )x 2 + (a 2 k1 + a 2 k 2 + b 2 k 3 + b 2 k 4 )x 3 (ak1 )x 4 (ak 2 )x 5 + (bk3 )x 6 + (bk 4 )x 7 }

f11 ( x ) =

1 {c1 x 8 + (cc1 )x 9 (ac1 )x10 + c1x11 k1x1 + ck1 x 2 ak1x 3 + k1 x 4 + k t1 x 4 k t 1z1 m1

f 12 ( x ) =

1 m2

{ c 2 x 8 (dc 2 )x 9 (ac 2 )x 10 + c 2 x 12 k 2 x 1 dk 2 x 2 ak 2 x 3 + k 2 x 5

+k t2x5 k t2z2}

f13 ( x ) =

1 + (cc 3 )x 9 + (bc 3 )x 10 + c 3 x 13 k 3 x 1 + ck 3 x 2 + bk 3 x 3 + k 3 x 6 + k t 3 x 6 k t 3 z 3 { c3y m3

f14 ( x ) =

1 { c 4 x 8 (dc 4 )x 9 + (bc 4 )x 10 + c 4 x 14 k 4 x 1 dk 4 x 2 + bk 4 x 3 + k 4 x 7 + k t 4 x 7 k t 4 z 4 } m4

Control force matrix :

0 0 [B ] = 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

1 M 1 M 1 M 1 M

a I xx a I xx b I xx b I xx

c I yy d I yy c I yy d I yy

1 m1 0 0 0

0 1 m2 0 0

0 0 1 m3 0

0 0 1 m4 0

International Journal of Vehicle Design, Vol. 26, Nos. 2/3, 2001, pp. 264-276.

References
Abdel Hady,M.B.A, and Crolla,D.A. (1989) "Theoretical Analysis of Active Suspension Performance Using a Four-Wheel Vehicle Model" Imeche,Vol.203.pp.125-136. Alleyne, A., Hedrick, J.K.,(1995) Non linear Adaptive Control of Active Suspensions, IEEE Transactions on Control Systems Technology, Vol.3, No.1, pp. 95-101. Burton, A.W., Truscott, A.J., Wellstead, P.E.,(1995) Analysis, Modelling and Control of an Advanced Automotive Self-Levelling Suspension System, IEE Proc.-Control Theory App., vol.142, No.2, pp. 129-139. Chalasani, R.M.,(1986) "Ride Performance Potential of Active Suspension Systems, Part-II", Symposium on Simulation and Control of Ground Vehicles and Transportation Systems , ASME Monograph, AMD, vol.80,DSE,No.2,pp.205-234. Dan Cho, D.,(1993) Experimental Results on Sliding Mode Control of an Electromagnetic Suspension, Mechanical Systems and Signal Processing, vol. 7(4), pp. 283-292. Esmailzadeh, E,Fahimi, F. (1997) "Optimal Adaptive Active Suspension for a Full Car Model ", Vehicle System Dynamics, vol 27, pp.89-107. Hrovat, D.,(1993) Applications of Optimal Control to Advanced Automotive Suspension Design, Transections of ASME, Vol. 115, pp.328-342. Hung, J., Y., (1993) "Variable Structure Control: A Survey" , IEEE Transactions on Industrial Electronics, vol. 40, No:1, pp. 2-22. Prokop, G., Sharp, R.S.,(1995) Performance Enhancement of Limited-Bandwidth Active Automotive Suspensions by Road Preview, IEE Proc.-Control Theory App., vol.142, No.2, pp.140-148. Redfield,R.C., and Karnopp, D.C.,(1988) "Optimal performance of Variable Component Suspensions" Vehicle System Dynamics, vol. 17, pp.231-253. Sabanovic, A.,(1994) Chattering Free Sliding Mode 1 Turkey. Utkin, V., I.,(1977) Variable Structure Systems with Sliding Mode , IEEE Transactions on Automatic Control, vol.AC-22, pp. 212-222. Utkin, V., I., (1991) "Sliding Mode in Control and Optimisation," Springer - Verlag. Yagiz, N., Ozbulur, V., Inan, N., Derdiyok, A., (1997) Sliding Mode Control of Active Suspensions, 12th IEEE International Symposium on Intelligent Control, Istanbul. Young, K., D.,(1978) "Controller Design for a Manipulator Using Theory of Variable Structure Systems" , IEEE Transactions on Systems Man and Cybernetics, vol.SMC-8, pp. 210- 218. Yu, F., Crolla, D.A.,(1998) An Optimal Self-Tuning Controller for an Active Suspension, Vehicle System Dynamics, vol. 29, pp. 51-65. Yuksek, I., and Kaya, F.,(1995) '' Vibration Optimisation of Vehicle Systems'' ASME Structural Dynamics and Vibration , PD-Vol.70, January 1995, pp. 217-221. St Turkey Automatic Control Symposium , Istanbul-