Vous êtes sur la page 1sur 24

NEES at CU Boulder

CU-NEES-08-15
01000110 01001000 01010100
The George E Brown, Jr. Network for Earthquake Engineering Simulation

Dynamic Force Control with Hydraulic Actuators Using Added Compliance and Displacement Compensation
By

Mettupalayam V. Sivaselvan
University of Colorado, Boulder

Andrei M. Reinhorn, Xiaoyun Shao, and Scot Weinreber


University at Buffalo

October 2008

Center for Fast Hybrid Testing Department of Civil Environmental and Architectural Engineering University of Colorado UCB 428 Boulder, Colorado 80309-0428

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS Earthquake Engng Struct. Dyn. 2007; 00:110 Prepared using eqeauth.cls [Version: 2002/11/11 v1.00]

Dynamic force control with hydraulic actuators using added compliance and displacement compensation
Mettupalayam V. Sivaselvan1 , Andrei M. Reinhorn2 , Xiaoyun Shao2 and Scot Weinreber2
1

Department of Civil, Environmental and Architectural Engineering, University of Colorado at Boulder, Boulder, CO 80309

Department of Civil, Structural and Environmental Engineering, University at Bualo, Bualo, NY 14260

SUMMARY A new approach to dynamic force control of mechanical systems, applicable in particular to frame structures, over frequency ranges spanning their resonant frequencies is presented. This approach is implemented using added compliance and displacement compensation. Hydraulic actuators are inherently velocity sources, that is, an electrical signal regulates their velocity response. Such systems are therefore by nature high-impedance (mechanically sti) systems. In contrast for force control, a force source is required. Such a system logically would have to be a low-impedance (mechanically compliant) system. This is achieved by intentionally introducing a exible mechanism between the

Correspondence

to: 428 UCB, University of Colorado at Boulder, Boulder, CO 80309

Email: siva@colorado.edu, Phone: (303)735-0925, FAX: (303)492-7317

Contract/grant sponsor: George E. Brown Network for Earthquake Engineering Simulation, National Science Foundation; contract/grant number: #CMS-0086611 and #CMS-0086612

Received Copyright c 2007 John Wiley & Sons, Ltd. Revised

M. V. SIVASELVAN ET. AL.

actuator, and the structure to be excited. In addition, in order to obtain force control over frequencies spanning the structures resonant frequency, a displacement compensation feedback loop is needed. The actuator itself operates in closed-loop displacement control. The theoretical motivation as well as the laboratory implementation of the above approach is discussed along with experimental results. Having achieved a means of dynamic force control, it can be applied to various experimental seismic simulation techniques such as the Eective Force Method and the Real-time Dynamic Hybrid Testing Method. Copyright c 2007 John Wiley & Sons, Ltd.

key words:

Dynamics Force Control, Hydraulic Actuators, Natural Velocity Feedback, Smith

Predictor, Advanced Seismic Testing

1. INTRODUCTION Advanced seismic testing techniques such as the eective force method [3] and forms of realtime dynamic hybrid testing [15] require the implementation of dynamic force control in hydraulic actuators. Dynamic force control with hydraulic actuators is however a challenging problem. By its physical nature, a hydraulic actuator is a velocity source, i.e., a given controlled ow rate into the actuator results in a certain velocity. Moreover, hydraulic actuators are typically designed for good position control, i.e., to move heavy loads quickly and accurately. They are therefore by construction, high impedance (mechanically sti) systems [12]. In contrast a force source is required for force control. Such a system logically would have to be a low-impedance (mechanically compliant) system. Force control with hydraulic actuators is associated with many problems. Actuators designed for position control have sti oil columns, making force control very sensitive to control
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

parameters often leading to instability. Moreover friction, stick-slip, breakaway forces on seals, backlash etc. cause noise in the force measurement, making force a dicult quantity to control. Several strategies have been introduced to work around this problem. For instance, a dual compensation scheme [11] uses a primary displacement feedback loop with force as a secondary tracking feedback. This scheme also supports features such as acceleration compensation to overcome some of the eects that distort the force measurements. In robotics, the impedance control strategy has been employed wherein the force-displacement relationship is controlled at the actuator interface [7, 19]. Pratt et. al. [14] have used the idea of series elastic actuators where a exible mechanism is intentionally introduced between the actuator and the point of application of force, along with force feedback. They applied this to non-resonant systems. Furthermore, in force control the dynamics of the structure on which force is applied, is coupled in a feedback system with the dynamics of the actuator, resulting in a natural velocity feedback. When the structure is resonant, this results in a set of complex conjugate zeros of the open loop transfer function. Shield et. al. [3, 17] in their work on the eective force method, compensate for this eect by using velocity feedforward. It was also recognized by Conrad and Jensen [1], that closed-loop control with force feedback is ineective without velocity feedforward, or full state feedback. In this paper, a new approach to dynamic force control is presented, in which a compliance in the form of a spring is intentionally introduced between the actuator and the structure, and a displacement feedforward compensation is used. The method does not use direct force feedback. It also allows for an added physical design parameter in the control system, namely the stiness of the added compliance. In the following, a standard linearized dynamic model of a hydraulic actuator is rst presented. The natural velocity feedback problem and the solution of Shield et.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

M. V. SIVASELVAN ET. AL.

al. [3] are then discussed side by side to emphasize the dierences and commonalities with the approach presented in this paper. The motivation behind the proposed solution using added compliance and displacement compensation is then discussed. The analysis of the proposed solution and some experimental results are then presented.

2. LINEAR MODEL OF A SERVO-HYDRAULIC ACTUATOR A hydraulic actuator driving a single degree of freedom structure is shown in Figure 1. The analysis is this paper is based on linear models of the actuator and of the structure. For this, the dynamics of the actuator are linearized about the equilibrium point at the mid-stroke of the actuator. The linearized equations are standard (see for example [9, 2, 5, 18]) and are given by xp = vp Ap 2 P st xp 2st st vp M P = 2 (Ap vp 1 P + 2 xv ) Ap L vp = xv = Kv 1 + u v v

(1)

Here, xp and vp are respectively the displacement of the SDOF structure, P is the dierential pressure between the actuator chambers, xv is the valve spool displacement, M is the combined mass of the actuator piston and the SDOF system, L is half the stroke of the actuator, Ap is the area of the actuator piston, v is the servovalve time constant, Kv is the servovalve gain, is the bulk modulus of the oil, 1 is a dissipative constant that depends on the chamber and valve leakage ows, 2 is a gain coecient, st and st are the natural frequency and damping ratio of the SDOF structure and u is the control input to the servovalve. A block diagram
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

Supply PS

Figure 1. Model of a hydraulic actuator driving a SDOF structure

Valve Command, u

K vs +1
Servovalve

h Figure 2. Block diagram representation of the linear model of equation (1). C12 =

model of this linear system is shown in Figure 2. The quantity

is referred to as the oil column frequency. This is the imaginary part of a complex conjugate eigenvalue pair of the linearization, in the absence of a structure stiness.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

Return PR ~ 0 2 1 P1

xv
3 4

xp
P2

kst

M cst

Flow

Ap C12 s + 1
Actuator

Applied Force, f

Natural Velocity Feedback

Ap
1 s 2 2 M s + 2 st st s + st
Structure

Ap L ,K 2

= Kv 2

oil =

2Ap LM

(2)

M. V. SIVASELVAN ET. AL.

3. NATURAL VELOCITY FEEDBACK AND THE ASSOCIATED CONTROL PROBLEM It can be seen from the third part of equations (1), that the velocity of the mass aects the rate of change of the dierential pressure. This feedback can also be seen in the block diagram of Figure 2. This has been termed natural velocity feedback [3]. If the dissipation related to leakage ows, 1 is assumed to be zero, then the resulting transfer function Huf from the control input u to the applied force f is given by

Huf =

2 K s2 + 2st st s + st 2 + 2 s (v s + 1) s2 + 2st st s + st oil

(3)

It can be seen that this transfer function has a complex conjugate pair of zeros corresponding to the natural frequency and damping ratio of the SDOF structure. This implies that the force applied on the structure at this frequency becomes small. Feedback control using for example a PID controller does not improve the performance because these zeros persist in the closed-loop transfer function also. Therefore, additional control strategies are necessary.

3.1. Strategy of Velocity Feedback Compensation (Shield et. al. [3, 17]) It can be seen that there is a negative feedback of velocity at the summing junction in Figure 1. If we can add a positive feedback at this junction of the same amount, then the eect of the natural velocity feedback can be nullied. But since this is a physical junction that in inaccessible, the strategy of Shield et. al. [3, 17] is to add this positive feedback to the valve command. However, this signal has to now be preconditioned by the pseudo-inverse of the servovalve transfer function. This is done using a lead-lag compensator. In addition force feedback is also used. Figure 1 shows the resulting control strategy [3].
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

Force Feedback

Desired Force

Valve Command

Flow

Ap

Ap C12 s
Actuator

Achieved Force

Servovalve

Natural Velocity Feedback

Velocity Compensation

Compensator

1 s 2 M s 2 + 2 st st s + st
Structure

Figure 3. Block diagram showing velocity feedforward correction loop

4. MOTIVATION FOR SOLUTION BASED ON ADDED COMPLIANCE AND DISPLACEMENT COMPENSATION

It is known from experience that hydraulic actuators are more conveniently tuned in closed-loop position control, than in force control. It is therefore suggested to indirectly control force by controlling position. To do this, a compliance, a spring of stiness kLC , is introduced between the actuator and the structure. In this section, for simplicity of illustration, the eect of the reaction force from the spring on the actuator is ignored, i.e., perfect disturbance rejection is assumed. Perfect tracking is also assumed over all frequencies of interest. The full linear dynamics of the actuator is however considered in the analysis in section 5. First, the scenario the scenario of applying a force f on a rigid structure is considered as shown in Figure 4. It is easily seen that to apply a force f , the actuator piston needs to move an amount f /kLC . Thus the actuator can be operated in closed-loop position control, and be commanded to the position f /kLC . If the structure were not rigid but exible, then the applied force would cause it to displace by an amount xst . Thus the actuator needs to be commanded to the position f /kLC + xst . This leads to the need for displacement compensation. The structure displacement xst may be obtained by from a model of the structure, or by measurement. These
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

M. V. SIVASELVAN ET. AL.

Measured Force, f

Desired Force, f

1 / kLC

Position Command Actuator in Closed-loop Position Control Added Compliance, kLC

Rigid Structure

Figure 4. Applying a desired force on a rigid structure by controlling the position of the actuator

possibilities are shown in Figure 5. It will be seen later that a mix of the two approaches leads to the Smith Predictor approach. In addition, since the assumptions of perfect tracking and perfect disturbance rejection in the above discussion are not realistic, additional compensation is needed for the dynamics of the actuator. This is presented in section 5 below. However, rst the relationship of this approach to that of Shield et. al. is shown.

4.1. Comparison of the Proposed Approach to Velocity Compensation The relationship of the proposed approach to the velocity feedback compensation strategy of Shield et. al. [3, 17] can be shown by rearranging terms in the block diagram in Figure 3. Factoring Ap s suitably in Figure 3, the block diagram in Figure 6 is obtained. Comparing the block diagrams in Figures 6 and 5(c), it is seen that the in the absence of added compliance, the oil column behaves as a spring providing the compliance required for force control. Relative deformation occurs across this spring and force is applied through it. However, the compliance of the oil column spring is xed for a given actuator. In the approach proposed here, this compliance becomes an additional physical design parameter for the control system.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

Measured Force, f 1 / kLC Desired Force, f Structure Model


+ +

Position Command Actuator in Closed-loop Position Control Added Compliance, kLC Flexible Structure

(a) Using a model to obtain the structure displacement


Measured Force, f Desired Force, f

1 / kLC

+ +

Position Command Actuator in Closed-loop Position Control Added Compliance, kLC Flexible Structure

Structure Displacement, xst

(b) Using measured structure displacement


Desired Force

1/kLC

Displacement Command

Actuator in Position Control

Actuator + Displacement

kLC
Added Compilance

Achieved Force

Displacement Compensation

Compensator

Structure Displacement

2 M ( s 2 + 2 st st s + st )

Structure

(c) Block diagram representation of (b)

Figure 5. Applying a desired force on a exible structure

5. ANALYSIS OF THE PROPOSED CONTROL SOLUTION

The analysis of the proposed strategy for dynamic force control with added compliance and displacement compensation is based on a linearized model, which is rst presented.

Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls

Earthquake Engng Struct. Dyn. 2007; 00:110

10

M. V. SIVASELVAN ET. AL.

Force Feedback

Desired + Force

Valve Command

K Ap s
Servovalve + Actuator

Actuator + Displacement

2 Ap L
Oil spring

Achieved Force

Compensator Displacement Compensation

Structure Displacement

2 M ( s 2 + 2 st st s + st )

Structure

Figure 6. Refactoring of block diagram in Figure 3

5.1. Linear Modeling Modifying the model in equation (1) suitably, the linearized model of the actuator and the structure with the added compliance is obtained as xp = vp vp = Ap P kLC (xp xst ) mp

(Ap vp 1 P + 2 xv ) P = 2 Ap L Kv 1 u xv = + v v xst = vst


2 vst = st xp 2st st vp kLC (xst xp )

(4)

Here, mp is the mass of the piston, xxt and vst are the displacement and velocity of the structure (which are now dierent from those of the actuator piston because of the added compliance), kLC is the stiness of the added compliance and the other symbols are as dened before. The block diagram representation of this system along with the position controller C1 and the displacement feedforward compensator C2 are shown in Figure 7. In the gure, A1 and A2 are actuator transfer functions respectively from the valve command to the actuator
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

11

A2

Desired Force, f

1/kLC

C2

Position + Command

C1

A1

xp +

kLC

Applied Force

xst

Figure 7. Block diagram of the linear model of the actuator and structure with added compliance and the inner and outer loop controllers
Desired Force, f

1/kLC

C2

Position + Command

C1

A1 (1 + k LC S ) 1 + k LC ( A2 + S )

xp +

kLC

Applied Force

xst

Figure 8. Block diagram of Figure 7 rearranged

displacement, and from the force on the piston to the actuator displacement. These are given by A1 = 4K 2 mp Ls (v s + 1) (s2 + 2a oil s + oil ) (5)

s + 2a oil A2 = 2 + 2 s + 2 ) mp s (s a oil oil where oil =


2Ap mp L

is the oil column frequency, a = 1

mp a3 L p

is the actuator damping ratio,

and S is the transfer function of the SDOF structure, S= 1 2 mst (s2 + 2st st s + st ) (6)

The block diagram in Figure 7 can be rearranged as shown in Figure 8. The block diagram consists of an inner loop with controller C1 whose role is to track the position command, and an outer loop which provides displacement feedforward compensation. The role of the C2 is to compensate for the dynamics of the inner loop. The inner loop dynamics and the controller C1
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

12

M. V. SIVASELVAN ET. AL.

Position + Command

C1

A1 (1 + k LC S ) 1 + k LC ( A2 + S )

xp

Figure 9. Inner Loop

are presented in section 5.2. The outer loop and the compensator C2 are discussed in section 6.

5.2. Inner Loop The inner loop is shown in Figure 9. It can be seen that a more compliant spring, i.e. a lower kLC relative to the structure stiness and the oil column stiness, results in reducing the inuence of the structure dynamics S, and of the eect of the reaction force A2 on the dynamics of the actuator A1 . Physically, this can be interpreted as the compliant spring isolating the dynamics of the actuator from that of the structure for displacement tracking. The role of the controller C1 is to track the position command. For this purpose, a proprietary control system, typically implementing a PID control can be used. The control system can be tuned with the structure connected to the actuator through the spring. Experience shows that the controller C1 can be tuned in most cases so that the inner loop dynamics has a nearly at frequency response magnitude with a linearly increasing phase lag over the bandwidth of interest. The inner loop dynamics can therefore be modeled reasonably as a pure time delay. This approach is used in modeling the inner loop dynamics.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

13

5.3. Response without Outer-Loop Compensation If an explicit outer loop compensator is not used, i.e. C2 is set equal to 1, then the transfer function from the desired force to the measured force is given by
2 s2 + 2st st s + st fachieved = 2 kLC 2 fdesired s + 2st st s + st + mst (1 IL)

(7a)

where IL is the inner loop transfer function. If the inner loop dynamics is modeled by a pure time-delay, and a rst order Taylor series approximation of the delay is used (i.e., (1IL) s), then this transfer function reduces to
2 s2 + 2st st s + st fachieved = fdesired 2 s2 + 2st st + kLC d s + st mst

(7b)

where d is the time-delay of the inner loop dynamics. It is seen that the delay, to a rst order approximation, has eect of increasing the damping of the poles of the transfer function. For a lightly damped structure, lightly damped zeros still exist in the transfer function. These zeros manifest as a drop in the frequency response magnitude at the resonant frequency of the SDOF structure as shown in Figure 15(a). This necessitates the design of the outer loop compensator C2 .

6. OUTER LOOP COMPENSATOR DESIGN

Motivated by the fact that the inner loop dynamics can be reasonably modeled as a pure timedelay, we consider the Smith predictor is considered as an approach to design the compensator C2 of Figure 8. The Smith predictor was developed as a time-delay compensation algorithm in chemical process control [6]. It is however applicable to compensate for other types of dynamics as well. In the following, the basic idea of the Smith predictor is rst reviewed, followed by a description of how it can be used to compensate for the inner loop dynamics.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

14 6.1. The Smith Predictor

M. V. SIVASELVAN ET. AL.

The basic idea of the Smith predictor is described by constructing it based on motivation. Figure 10(a) shows a standard feedback control system where a controller C has been designed for the plant P in such a way that the closed loop system has certain desired characteristics. The closed-loop transfer function is
PC 1+P C .

However, the control input cannot be applied directly

to the plant, but has to be applied through an actuator A. The dynamics of the actuator may be thought of as undesirable dynamics in the feedback path. In order to regain the original closed loop structure, feedback is obtained from a model of the plant, P instead of from the plant itself as shown in Figure 10(b). However due to modeling error, the feedback obtained from the model of the plant P will not be the same as what would have been obtained from the actual plant P in the absence of the undesirable dynamics. Therefore, an additional error feedback is used as shown in Figure 10(c). Here, A is the transfer function model of the actuator dynamics. This leads to the Smith Predictor architecture. It can be seen that if the models were exact, i.e. A = A and P = P , then the transfer function from reference to output is
PC 1P C A,

and the Smith Predictor has the eect of moving the undesirable dynamics out

of the feedback loop. The Smith Predictor is also intimately related to the Internal Model Control idea (see for example, [10]).

6.2. Smith Predictor for Compensation of Inner Loop Dynamics As discussed in section 4, a desired force f is applied on the SDOF structure by imposing a displacement of f /kLC +xst to the end of the added compliance. Thus the feedback structure is as shown in Figure 11, corresponding to the idea depicted in Figure 5(b). This is the desired feedback structure corresponding to Figure 10(a). However in reality, also present in this
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

15

reference

output

(a) Standard feedback control system


reference

output

(b) Using feedback from the model to avoid undesirable dynamics A in the feedback path
reference

P +

output

error

(c) The Smith Predictor

Figure 10. The concept of the Smith Predictor

Desired Force, f

1/kLC

f + xst k LC +
-

kLC
Added Compilance

Achieved Force

xst

2 M ( s 2 + 2 st st s + st )

Structure

Figure 11. Desired feedback structure for displacement compensation

Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls

Earthquake Engng Struct. Dyn. 2007; 00:110

16

M. V. SIVASELVAN ET. AL.

feedback loop is the undesirable dynamics of the inner loop as shown in Figure 8. The corresponding Smith Predictor architecture in line with Figure 10(c) is therefore as shown in Figure 12. In this gure, IL is the inner loop transfer function and quantities with hats are the modeled values of the actual physical parameters. The part shown in the dotted box is the controller C2 dened in Figure 8. In can be seen that this part as a whole has two inputs (the reference and the feedback) and a single output, the actuator command. For digital implementation, the blocks in this part can be therefore composed into two transfer functions, to avoid algebraic loops. These are then transformed to a discrete time transfer functions using the bilinear transform, s =
2 z1 T z+1

[4]. As described above, if the models were exact, then the

Smith Predictor has the eect of moving the undesirable inner loop dynamics out of the outer loop. If the model were not exact, it can be veried that the transfer function becomes fachieved = fdesired 1 1+
[mst (1+IL)mst (1+IL)]s2 +[cst (1+IL)cst (1+IL)]s+[kst (1+IL)kst (1+IL)] kLC mst s2 +cst s+kst

IL

As the stiness of the added compliance decreases relative to the structure stiness and the oil column stiness, the sensitivity of the performance of the Smith Predictor to modeling error decreases. This is a further benet of the added compliance.

7. EXPERIMENTAL RESULTS Experiments were performed using a small-scale pilot test setup shown in Figure 13 to study the performance of the proposed force control strategy, before it was applied to large-scale actuators. A hydraulic actuator with 1 kN (2.2 kip) force capacity and 100 mm (4 in) stroke was used. The actuator was tted with a MTS 252.22 two-stage servo-valve with a 19 liters/minute (5 gpm) ow capacity. The MTS FlexTest GT system was used for the inner loop controller.
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

17
Achieved Force

Desired Force, f

1/kLC

IL

+ -

kLC
Added Compilance

2 mst ( s 2 + 2 st st s + st )

Structure

k LC mst s 2 + cst s + k st + k LC

IL

error

Figure 12. Smith Predictor structure for displacement compensation

The outer loop controller was implemented using using Simulink and xPC Target [8]. For the SDOF structure, a simple one story shear building model was used. A 305mm x 203mm x 25mm (12in 8in x 1in) steel plate served as the oor, while four 12.7mm (0.5in) diameter aluminum threaded rods served as columns. Braces were installed in the transverse direction on both sides of the structure to limit out-of-plane motion. Lead blocks are used to provide additional mass. Two dierent damping scenarios were considered for the structure one with merely the inherent damping in the structure, and another with model dashpots installed as shown in Figure 13(a). Helical springs were used for the added compliance as show in Figure 13(c). Four compression-only springs were used. They were pre-compressed so that they could act in both tension and in compression. The properties of the structure and the added compliance are summarized in Table I. The inner loop controller C1 was tuned for position tracking, and the resulting frequency
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

18

M. V. SIVASELVAN ET. AL.

mst Case 1 Case 2 77.27kg (170lb)

kst 16.67N/m (156lb/in)

st 3 Hz

st 0.01 0.17

kLC 12.57N/m (111lb/in)

Table I. Structure Properties

Mass

Dashpot

Braces

(a) The SDOF structure

Load Cell Stroke

Servovalve

Hydraulic Supply

Structure Displacement Transducer

Reaction Frame

(b) The hydraulic actuator

(c) Spring used for added compliance

Figure 13. Experimental Setup

Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls

Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

19

2 1.8 1.6 1.4


achieved desired

1.2 1 0.8 0.6 0.4 0.2 0 0 1 2 3 4 5 6 Frequency (Hz) 7 8 9 10

/x

(a) Magnitude
0 2 4 6 Phase (degress) 8 10 12 14 16 18 20 0 1 2 3 4 5 6 Frequency (Hz) 7 8 9 10

(b) Phase

Figure 14. Inner loop Frequency Response Function

response function of the inner loop is shown in Figure 14. For the frequency range of interest, it is seen that the inner loop dynamics can be modeled as a pure time-delay, , of 5.6 ms. The force control performance was studied by measuring the frequency response of the ratio of the applied force to the desired force. This was done using a crest factor-minimized multisine input [13] for the desired force. Figure 15(a) shows the results for the structure with low
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

20

M. V. SIVASELVAN ET. AL.

damping (st = 0.01). For the case with C2 = 1, the analytically obtained FRF considering the actuator as a pure time-delay of 5.6 ms agrees well with the experimentally measured FRF. This implies that that it is in fact reasonable to model the inner loop dynamics as a pure delay. It is also seen that using a Smith predictor for C2 improves the force control performance. However, the frequency response function still exhibits some drop (about 20 %) at the resonant frequency of the structure and a bump at the resonant frequency of the structure (about 10 %) with the added compliance. This is because the damping, being very small is not known accurately and hence is modeled imprecisely. Figure 15(b) shows the results for the structure with the added dashpots, and hence higher damping (st = 0.17). The frequency response with C2 = 1 still shows a drop a the resonant frequency of the structure, but the drop is smaller (about 20 %) because the zeros of the transfer function of equation (7) are more highly damped. Since damping is modeled more accurately in the Smith Predictor, the frequency response with compensation is almost ideally at one.

8. SUMMARY AND CONCLUDING REMARKS From both the analytical and the experimental studies, the strategy of adding compliance and providing displacement feedforward compensation appears adequate for dynamic force control using hydraulic actuators over frequencies spanning resonances. The strategy does not use force feedback, the measurement of which generally is noisier and is corrupted by stick-slip, breakaway forces on seals, backlash etc. in the hydraulic actuator. The strategy results in two controllers an inner loop controller, a typical PID controller, whose role is to track a position command, and outer loop controller whose role is to compensated for the inner loop dynamics. The inner loop dynamics can be reasonable modeled as a pure time-delay. In this work, the
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

21

2 1.8 1.6 1.4


achieved desired

C2 = 1 Analytical with C = 1
2

C = Smith Predictor
2

1.2 1 0.8 0.6 0.4 0.2 0 0 1 2 3 4 5 6 Frequency (Hz) 7 8 9 10

/f

(a) Case 1: st = 0.01


2 1.8 1.6 1.4
achieved desired

Open Outer Loop C2 = 1 C = Smith Predictor


2

1.2 1 0.8 0.6 0.4 0.2 0 0 1 2 3 4 5 6 Frequency (Hz) 7 8 9 10

/f

(b) Case 2: st = 0.17

Figure 15. Force transfer function

outer loop controller has been designed using the Smith predictor approach. This requires approximate models of the SDOF structure as well as of the inner loop dynamics. It is seen that the performance of the system is less sensitive to the accuracy of these models when the added compliance is made more exible. The tuning of the inner loop controller also becomes less sensitive to the dynamics of the SDOF structure with increase in this exibility. The added
Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

22

M. V. SIVASELVAN ET. AL.

compliance thus provides an additional physical parameter in the dynamic force control system design. The exibility cannot be arbitrarily reduced, for this comes at the expense of increased stroke requirement on the actuator. This method of force control has been successfully used in a unied approach to real-time dynamic hybrid simulations [16]

ACKNOWLEDGEMENT

The authors gratefully acknowledge the nancial support from the National Science Foundation through the George E. Brown Network for Earthquake Engineering Simulation (NEES) development program, grants #CMS-0086611 and #CMS-0086612.

REFERENCES

1. F. Conrad and C. J. D. Jensen. Design of hydraulic force control systems with state estimate feedback. In IFAC 10th Triennial World Congress, Munich, Germany, 1987. 2. J. P. Conte and T. L. Trombetti. Linear dynamic modeling of a uni-axial servo-hydraulic shaking table system. Earthquake Engineering and Structural Dynamics, 29(9):1375, 2000. 3. J. Dimig, C. Shield, C. French, F. Bailey, and A. Clark. Eective force testing: A method of seismic simulation for structural testing. Journal of Structural Engineering-Asce, 125(9):10281037, 1999. 4. G. F. Franklin, J. D. Powell, and M. L. Workman. Digital control of dynamic systems. Addison-Wesley, Menlo Park, Calif., 3rd edition, 1998. 5. J. Kuehn, D. Epp, and W. N. Patten. High-delity control of a seismic shake table. Earthquake

Engineering and Structural Dynamics, 28(11 Nov):12351254, 1999. 6. J. E. Marshall. Control of time-delay systems. P. Peregrinus, Stevenage [Eng.]; New York, 1979. 7. M. Mason. Compliant motion. In Michael Brady, editor, Robot motion: planning and control, MIT Press series in articial intelligence, pages xv, 585 p. MIT Press, Cambridge, Mass., 1982. 8. Mathworks. Simulink and xpc target, 2003. 9. H. E. Merritt. Hydraulic control systems. Wiley, New York,, 1967. 10. M. Morari and E. Zariou. Robust process control. Prentice Hall, Englewood Clis, N.J., 1989. Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls Earthquake Engng Struct. Dyn. 2007; 00:110

DYNAMIC FORCE CONTROL

23

11. MTS. 793.xx software system, 2003. 12. C. L. Nachtigal. Instrumentation and control: fundamentals and applications. Wiley series in mechanical engineering practice. Wiley, New York, 1990. 13. R. Pintelon and J. Schoukens. System identication: a frequency domain approach. IEEE Press, New York, 2001. 14. J. Pratt, B. Krupp, and C. Morse. Series elastic actuators for high delity force control. Industrial Robot, 29(3):234241, 2002. 15. A. M. Reinhorn, M. V. Sivaselvan, S. Weinreber, and X. Shao. Real-time dynamic hybrid testing of structural systems. In Third European conference on structural control, Vienna, Austria, 2004. 16. X. Shao. Unied control platform for realtime dynamic hybrid simulations. PhD thesis, University at Bualo - State University of New York (SUNY), 2006. 17. C. K. Shield, C. W. French, and J. Timm. Development and implementation of the eective force testing method for seismic simulation of large-scale structures. Philosophical Transactions of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences, 359(1786):19111929, 2001. 18. M. V. Sivaselvan and J. Hauser. Modeling of a hydraulic shaking table for optimal control. page (In Preparation), 2007. 19. D. E. Whitney. Historical-perspective and state-of-the-art in robot force control. International Journal of Robotics Research, 6(1):314, 1987.

Copyright c 2007 John Wiley & Sons, Ltd. Prepared using eqeauth.cls

Earthquake Engng Struct. Dyn. 2007; 00:110

Vous aimerez peut-être aussi