Chapter 3 Model Formulation

In this chapter, the pertinent vehicle representation and control schemes are modeled and formulated for use within the simulation program. The first section of the chapter will explain and formulate the roll-plane vehicle model used for this research. This is a planar model with four degrees of freedom that represent the heave and roll of the vehicle body, as well as the wheel hop of the left and right tires. In the second and final section of the chapter, the equations are developed for the semiactive damping control schemes as they pertain to the roll-plane model. The semiactive control schemes formulated for this research include on-off skyhook, continuous skyhook, on-off groundhook, and fuzzy logic damping control.

3.1 Roll-Plane Vehicle Model The purpose of this research is to examine the effect of different control methods on the transient response of the heave and roll motion of a class 8 truck. The front end of the truck is represented by a four-degree-of-freedom, roll-plane model, as shown in Figure 3.1.

b , Tb
d b, l

z b , Fb d b, r

mb , I xx k s, l zt , l kt, l mt , l cs, l k s, r mt , r kt,r c s, r zt ,r

z in , l

z in , r

Figure 3.1. Roll-Plane Vehicle Model

25

The model consists of three masses. The top mass, mb , represents the truck body, while the two lower masses, mt ,l and mt ,r , represent the left and right tires of the front axle, respectively. The parallel spring and damper combinations located between the truck body and each tire ( k s ,l , cs ,l and k s ,r , cs ,r ) represent the stiffness and damping of the vehicle suspension system. The respective stiffness of the left and right tires are

represented by the lower springs k t ,l and k t ,r . On the truck body, a heave force, Fb , and a roll force (torque), Tb , can be exerted. The model parameters and their respective units are summarized in Table 3.1.

Table 3.1. Roll-Plane Vehicle Model Parameter Definition Symbol cs ,l cs ,r d b ,l d b ,r Fb I xx k s ,l k s ,r k t ,l k t ,r mb mt ,l mt ,r Tb Description suspension damper, left side suspension damper, right side dimension from suspension to CG, left side dimension from suspension to CG, right side body heave force body roll moment of inertia suspension stiffness, left side suspension stiffness, right side tire stiffness, left side tire stiffness, right side vehicle body mass tire mass, left side tire mass, right side body roll force (torque) Units lb-sec/in lb-sec/in inch inch lb slug-in2 lb/in lb/in lb/in lb/in slug slug slug in-lb

The states of the model are shown by zb , b , z t ,l , z t ,r , zin ,l , and zin ,r , where zb and b represent the heave and roll motions of the truck body, z t ,l and z t ,r represent the heave motions of the left and right vehicle tires, and zin ,l and zin ,r represent the base (or road) inputs into the left and right tires of the model. The displacement, velocity, and

acceleration components of the vehicle states are described in Table 3.2.

26

Table 3.2. Roll-Plane Vehicle Model States Symbol zb &b z && zb zin ,l zin ,r z t ,l &t ,l z &&t ,l z z t ,r &t ,r z &&t ,r z b & Description body heave displacement body heave velocity body heave acceleration road input displacement, left side road input displacement, right side tire heave displacement, left side tire heave velocity, left side tire heave acceleration, left side tire heave displacement, right side tire heave velocity, right side tire heave acceleration, right side body roll displacement, left side body roll velocity, left side body roll acceleration, left side Units inch in/sec in/sec2 inch inch inch in/sec in/sec2 inch in/sec in/sec2 rad rad/sec rad/sec2

&& b

The dynamics of the roll-plane model in Figure 3.1 are described by: & ) + c (z & ) + k (z z d ) &b z &t ,l d b ,l &t , r + d b , r mb && zb + c s , l ( z b s ,r &b z b s ,l b t ,l b ,l b + k s ,r ( zb z t ,r + d b ,r b ) = Fb

(3.1)

& ) k (z z d ) + k z = k z &b z &t ,l d b ,l mt ,l && z t ,l cs ,l ( z b s ,l b t ,l b ,l b t ,l t ,l t ,l in ,l

(3.2)

& ) k (z z + d ) + k z = k z &b z &t ,r d b ,r mt ,r && z t ,r c s ,r ( z b s ,r b t ,r b ,r b t ,r t ,r t ,r in ,r

(3.3)

&& + c d ( z & ) + c d (z & ) k d (z &t , l d b , l &t , r + d b , r I xx b s ,l b ,l &b z b s ,r b ,r &b z b s ,l b ,l b z t ,l d b,l b ) + k s ,r d b ,r ( zb z t ,r + d b ,r b ) = Tb

(3.4)

where all parameters and variable states are defined in Tables 3.1 and 3.2. Equations (3.1) through (3.4) can be represented in matrix form as: 27

v v v && & + Kz Mz + Cz = f

(3.5)

where M , C , and K represent the mass, damping, and stiffness matrices described by:

M = diagonal mb , mt ,l , mt ,r , I xx ,

(cs , l + cs , r ), cs ,l , C= cs ,r , ( cs , l db, l + cs , r db , r ),

cs , l , cs , l , 0, cs , l db , l ,

cs , r , 0, cs ,r , cs , r db, r ,

( cs , l db, l + cs , r db , r ) cs ,l db , l , cs , r d b , r (cs , l db2, l + cs ,r db2, r )

( k s , l + k s , r ), ks,l , ( k s , l + k t , l ), k s,l , K = 0, k s,r , k s ,l d b , l , ( k s , l db ,l + k s ,r db , r ),

( k s , l d b , l + k s , r db , r ) 0, k s , l db , l , ( k s , r + k t , r ), k s ,r d b , r ( k s , l db2, l + k s , r db2, r ) k s ,r db , r ,

k s, r ,

and the force vector, f , is defined as:

Fb k z t ,l in ,l f = . k t ,r z in ,r Tb v The displacement vector z is described by:

zb v zt ,l z= . zt ,r b 28

v v v & and && The velocity and acceleration vectors, z z , are similar to z , and are defined as:

&b z z &t ,l v , &= z z &t ,r & b

and

zb && && zt ,l v . && = z && zt ,r && b

3.2 Semiactive Damping Control Models In Chapter 2, a quarter-car model was used to explain the semiactive damping schemes of on-off skyhook, continuous skyhook, on-off groundhook, and fuzzy logic control. In the following sections, each of these control schemes will be formulated for the roll-plane model of Figure 3.1.

3.2.1 On-Off Skyhook Control In Section 2.3.1, the on-off skyhook control policy was explained for a quarter-car model. In Equations (2.2a) and (2.2b), the basis for on-off skyhook control was presented. When the product of the absolute velocity of the vehicle body and the relative velocity across the damper is greater than or equal to zero, the damper is adjusted to its high state. If this product is negative, then the low state of the damper is applied.

For the roll-plane model of Figure 3.1, the relative velocity across each of the suspension dampers is computed as:

29

& &b z &t , l d b , l v rel ,l = z b & &b z &t ,r + d b ,r v rel ,r = z b

(3.6a) (3.6b)

where v rel ,l and v rel ,r , respectively represent the relative velocity across the left and right dampers of the roll-plane model. The absolute velocity of the vehicle body mass attached to the left and right side of each damper is thus calculated as: & &b d b ,l v abs,l = z b & &b + d b , r v abs,r = z b

(3.7a) (3.7b)

Therefore, the on-off skyhook control policy as it applies to the roll-plane model of Figure 3.1 can be formulated by:

Left Damper v abs ,l v rel ,l 0; v abs ,l v rel ,l < 0; cs ,l = high state cs ,l = low state (3.8a) (3.8b)

Right Damper v abs,r v rel ,r 0; v abs ,r v rel ,r < 0; cs ,r = high state cs ,r = low state (3.9a) (3.9b)

3.2.2 Continuous Skyhook Control An extension of the on-off skyhook control policy was used as one method of continuous control. As explained in Section 2.3.2, when the product of the absolute velocity of the vehicle body and the relative velocity across the damper is less than zero, low state damping is applied. When this product is positive or equal to zero, the damping value applied by the damper is equal to a gain times the absolute velocity of the vehicle body,

30

maintained within the high and low state limits of the damper. This concept, as it pertains to the left and right dampers of the roll-plane model, is shown by:

Left Damper v abs ,l v rel ,l 0; v abs ,l v rel ,l < 0; cs ,l = max{low state, min[( v abs,l ) , high state]} (3.10a) cs ,l = low state (3.10b)

Right Damper v abs ,r v rel ,r 0; v abs ,r v rel ,r < 0; cs ,r = max{low state, min[ ( v abs,r ), high state]} (3.11a) cs ,r = low state (3.11b)

The symbol represents the gain of the damper. Due to the symmetry of the model, the same gain is used for each damper.

3.2.3 On-Off Groundhook Control For the roll-plane model of Figure 3.1, the relative velocity across each damper was previously determined in Equations (3.6a) and (3.6b), and the absolute velocities of the & t,l and z & t,r, vehicle tire masses attached to the left and right side suspension dampers are z respectively. Thus, the on-off groundhook control policy as it applies to the roll-plane model can be summarized by:

Left Damper &t ,l v rel ,l 0; z &t ,l v rel ,l > 0; z cs ,l = high state cs ,l = low state (3.12a) (3.12b)

Right Damper &t ,r v rel ,r 0; z &t ,r v rel ,r > 0; z c s ,r = high state c s ,r = low state (3.13a) (3.13b)

31

As explained in Section 2.3.3, in on-off groundhook control, the product of the tire heave velocity and the relative velocity across the damper is used to determine the damper state. If this product is negative or zero, the high state of the damper is applied. Otherwise, the damper is adjusted to its low state.

3.3 Fuzzy Logic Control In Chapter 2, Section 2.4, the three basic steps of a fuzzy logic controller were presented and explained. These steps were referred to as fuzzification, execution of rules, and defuzzification. In the following paragraphs, these three steps are presented as they pertain to the fuzzy logic controller design used for this research.

3.3.1 Step One: Fuzzification As explained previously, the first step of a fuzzy logic controller is the fuzzification of the controller inputs, based on the membership function design for each input. For the rollplane model of Figure 3.1, there exist two controllable dampers and, therefore, two controllers. Each fuzzy logic controller was designed with four inputs. For each damper &t ,l and z &t ,r , the controller, these inputs include the heave velocity of each suspension tire, z relative velocity across the damper and the absolute velocity of the vehicle body, defined by Equations (3.6) and (3.7), and the absolute acceleration of the vehicle body, defined as: && a abs,l = && zb d b ,l b && a abs ,r = && zb + d b , r b

(3.14a) (3.14b)

where a abs ,l and a abs ,r respectively represent the absolute acceleration of the left and right sides of the vehicle body.

For each input, a trapezoidal membership function was used. Figure 3.2 shows the shape of the trapezoidal membership function as it was used for the fuzzification of the controller inputs.

32

Base / 50

Base

Figure 3.2. Trapezoidal-Shaped Membership Function Definition

For the trapezoidal-shaped membership functions used for this research, the width of the top of the trapezoid was defined as one-fiftieth of the width of the trapezoid base. The reason for using this flattened top triangle instead of a completely triangular-shaped function is that the trapezoid has been shown to smooth the response of the controller [21].

In Figure 3.3, each input and its membership function are defined for the left damper controller.

33

vabs,l 1 N Z P

rel , l

(1)

v abs pos

0 (a)

v abs pos

v abs,l

v rel pos (1)

0 (b)

v rel pos ( 1)

v rel ,l

&t ,l

rel ,l

(2 )

z & t pos

0 (c)

z & t pos

z & t ,l

v rel pos ( 2 )

v rel ,l

0 (d)

v rel pos ( 2 )

a abs,l 1 N Z P

rel ,l

( 3)

a abs pos

0 (e)

a abs pos

a abs,l

v rel pos ( 3)

0 (f)

v rel pos ( 3)

v rel ,l

Figure 3.3. Left Damper - Controller Input Membership Functions Each membership function is defined by three linguistic variables, Negative(N), Zero(Z), and Positive(P), and is symmetric about zero. There is a total of six membership functions used for four inputs because the relative velocity across the damper is fuzzified three times, using three separate membership functions.

In Figure 3.4, each input and its membership function are defined for the right damper controller.

34

 vabs , r

v rel , r ( 1)

v abs pos

v abs pos

v abs ,r

v rel p o s ( 1)

v rel pos (1)

v rel ,r

(a)
z&t ,r
v rel , r ( 2 )

(b)

&t pos z

&t pos z

& t ,r z

v rel pos ( 2 )

v rel pos ( 2 )

v rel ,r

(c)
a abs , r
v rel ,r ( 3 )

(d)

a abs pos

a abs pos

a abs , r

v rel p o s ( 3 )

v rel pos ( 3)

v rel ,r

(e)

(f)

Figure 3.4. Right Damper - Controller Input Membership Functions

Because of the symmetry of the roll-plane model, the shapes of the membership functions for the left and right dampers of the model are the same, and the values used to describe the maximum values of each linguistic variable will be identical.

3.3.2 Step Two: Execution of Rules As detailed in Section 2.4.2, prior to establishing the rules of the rule-base, the output membership function must be defined. For this research, the output of the controller is the value of damping to be exerted by the left and right dampers of the roll-plane model, cs ,l and cs ,r . For the outputs, a triangular-shaped membership function is used. These

functions are shown in Figure 3.5.

35

 cs , l

c s ,r

cs small cs medium cs l arge

cs,l

cs small cs medium cs l arge

cs,r

(a) Left Damper Controller Output

(b) Right Damper Controller Output

Figure 3.5. Controller Output Membership Functions

The same membership function is used for the controller output to the left and right damper. Each is described by three linguistic variables, Small(S), Medium(M), and

Large(L), and the maximum value of each linguistic variable is the same for each damper. The output membership function will always be symmetric about cs medium , which is defined as: (c s l arg e c s small ) cs medium = c s small + 2

(3.15)

The rules of the system can now be developed. The fuzzy logic controller rule-base for the left damper of the half-car model is detailed in Figure 3.6.

36

vabs ,l

v rel ,l ( 1)

N N Z P

&t ,l z

v rel , l ( 2 )

N N Z P

L M S

M M M

S M L

S M L

M M M

L M S

(a) Rule Table 1L

(b) Rule Table 2L

N N Z P

L M S

M M M

S M L

(c) Rule Table 3L

Figure 3.6. Left Damper - Controller Rule-Base

The controller for this damper consists of three rule tables in the rule-base. For future reference, the rule tables of Figure 3.6 (a), (b), and (c), will be referred to as 1L, 2L, and 3L, respectively, as indicated on the figure. Each table is a function of either the absolute body velocity, body acceleration, or tire heave velocity and the relative velocity across the damper. As discussed previously, an individual membership function is used to describe the relative velocity of each rule table.

The logic of rule table 1L follows what may be called a fuzzy skyhook policy.

By

examining this rule table, it can be seen that when the v abs ,l and v rel ,l are both positive or both negative, corresponding to a positive product, the output is large, and when their signs are opposite, corresponding to a negative product, the output is small. When cs l arg e is defined as the high-state damping and cs small is defined as the low-state damping and when each input is fully Negative or fully Positive, a skyhook policy is observed. However, when each input is not fully Negative or fully Positive, it is fuzzified according to the membership function, and a fuzzy skyhook algorithm is used.

&t ,l and In a similar manner, rule table 2L is deemed a fuzzy groundhook policy. When z

37

v abs ,l are the same sign, the output is small, and when they have opposite signs, the output is large. When each input is either fully Negative or fully Positive, a groundhook policy is observed; otherwise, the input is fuzzified according to its membership function and a fuzzy groundhook result is obtained.

Lastly, rule table 3L is identical to table 1L, with the exception that the control is based on the absolute body acceleration, not its velocity. During tuning of the controller, it was found that rule tables 1L and 2L were successful at reducing the body and tire displacements, but not the body accelerations. Thus, rule table 3L was added to assist in reducing the body accelerations to the desired level. This table may be considered a fuzzy, acceleration-based skyhook policy.

In Figure 3.7, the rule-base for the right damper of the roll-plane model is shown.

vabs ,r

v rel , r (1)

N N Z P

&t , r z

v rel , r ( 2 )

N N Z P

L M S

M M M

S M L

S M L

M M M

L M S

(a) Rule Table 1R

(b) Rule Table 2R

N N Z P

L M S

M M M

S M L

(c) Rule Table 3R

Figure 3.7. Right Damper - Controller Rule-Base

As with the left damper, the controller for the right damper consists of three rule tables in the rule base. Because of the symmetry of the model, the rule-base for the right tire is identical to the left tire rule-base of Figure 3.6. For future reference, the rule tables of

38

Figures 3.7 (a), (b), and (c) will be referred to as 1R, 2R, and 3R, as indicated on the figure.

3.3.3 Step Three: Defuzzification As discussed previously, defuzzification converts the fuzzy values obtained from execution of the rule tables into a single, crisp value. The method of defuzzification used for this research is the weighted average method of Equation (2.5). However, the method has been expanded and revised to account for the existence of the three rule tables in the rulebase.

The weighted average method (explained in Section 2.4.3), as it pertains to the left damper of the vehicle model, is given by:

* s ,l

1 L cs , l c s , l

1 L cs ,l

[ ]

1L

+ 2 L cs , l c s ,l + 2 L cs , l

1L

[ ]

2L

+ 3 L c s , l c s ,l

2L

+ 3 L cs , l

[ ]

3L

(3.16)

3L

where cs , l represents the weighting function of the output, c s ,l represents the maximum (median) linguistic variable value based on the output membership function, and 1L , 2 L , and 3 L represent the ranking functions of rule tables 1L, 2L, and 3L, respectively, from Figure 3.6. In the defuzzification method of Equation (3.16), in the numerator, the products of the weighting function and corresponding maximum linguistic variable value are summed for each rule. In turn, these summations are each multiplied by the ranking function of their corresponding rule table before they are added. In the denominator, the summation of the weighting functions from each rule are also multiplied by their corresponding ranking function before their products are added. Thus, the ranking

functions are used to rank the importance of each rule table during defuzzification.

In Equation (3.17), the weighted average defuzzification method is shown as it applies to the right damper of the roll-plane model.

39

* s ,r

1 R c , r c s , r

1R cs ,r

[ ]

1R

+ 2 R c s , r c s , r
1R

+ 2 R cs ,r

[ ]

2R

+ 3 R cs , r c s ,r

2R

+ 3 R c s , r

[ ]

3R

(3.17)

3R

The defuzzification method for the right damper is identical to the defuzzification method used for the left damper. The symbols cs , r and cs , r respectively represent the weighting function of the output and the maximum (median) linguistic variable value based on the output membership function, while 1R , 2 R , and 3R represent the ranking functions of rule tables 1R, 2R, and 3R, respectively, from Figure 3.7. Due to the symmetry of the model, the ranking values for the right damper will be identical to those used for the left damper.

Because the semiactive damper is limited by its high and low state, the crisp output values from Equations (3.16) and (3.17) must be maintained within the allowable damping range. This is accomplished for the left controller output by: cs ,l = max{low state, min[ cs*,l , high state]}

(3.18)

Equation (3.18) implies that the crisp output value of the left controller, cs*,l , is limited to within the high-state and low-state damping, and that the actual value applied by the damper is determined by the computation result, cs ,l .

The crisp output of the right controller, cs*,r , is limited to within its high-state and lowstate damping through: cs ,r = max{low state, min[ cs*,r , high state]}

(3.19)

40