+ All Categories
Home > Documents > UPB Sci Bull ShortV

UPB Sci Bull ShortV

Date post: 08-Apr-2018
Category:
Upload: bogdan176
View: 219 times
Download: 0 times
Share this document with a friend

of 16

Transcript
  • 8/6/2019 UPB Sci Bull ShortV

    1/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    BEHAVIOR MODEL CONTROL FOR CASCADED

    PROCESSES: APPLICATION TO AN ELECTRICAL DRIVE

    B. VULTURESCU1,2

    , A. BOUSCAYROL1, F. IONESCU

    2, JP. HAUTIER

    1

    Behavior model control (BMC) is known to increase the robustness of a

    process control. It has already been applied in single-loop electrical drives control.This paper introduces the BMC among the others model control strategies andextend the single-loop utilization to more complex controls, which needs severalloops. Simulations and experimental results point out that the proposed topologyproduces better results than a reference classical control, increasing the robustnessagainst parameter variations and external disturbances.

    Keywords: Behavior model control, electrical drives, robustness.

    1. Introduction

    Electrical drives are widely spread in industry application thanks to their

    growing performances and their flexibility [1]. But electrical machines are

    sensitive to some physical phenomenon which robustness problem in their

    control: non-linear character of their relation as the magnetic saturation, time-

    varying parameters as the resistance evolution according to the temperature,unknown perturbation as the load torque for some applications...

    A lot of control techniques have been used in the last decade in order to

    improve the robustness performances of these drives [2]. Most of them are based

    on the use of a machine model, in order to take into account the difference

    between the actual process and the model one. They can be differentiated by the

    use of the error between the model and the process. Closed loop observers use the

    error to modify the model to converge to the process [6]. Model reference

    adaptive controls use this error in order to adapt the control parameters [4].

    Internal model controls use the error as input of the control, in order to be

    compensate as equivalent perturbation [5].

    Another model control has recently been defined in order to solve therobustness problem of electrical drive: the behavior model control (BMC) [3]. It

    uses the model error to modify the control input value in order to impose the

    process convergence to the model. This original control strategy has already been

    validated for simple electrical drives [19, 22]. In this paper, the BMC is extended

    for more complex drive with internal and external control loops. A DC machine

    application validates this study with experimental results.

  • 8/6/2019 UPB Sci Bull ShortV

    2/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    In the section 2, the BMC and the other model control strategies are

    depicted with the same representation in order to point out their differences. The

    section 3 is devoted to the analysis of the classical BMC with a single control

    loop. This principle is extended for process which need cascaded control loop in

    the section 4. Finally, the theoretical methods are validated by simulation end

    experimental result for a DC machine.

    2. The BMC among other robust strategies

    The objective of this section is to present the BMC among other model

    control techniques. In order to point out its characteristics, a simplified four-block

    representation is proposed for all described control strategies. These simplified

    structures allow presenting the principle of each control, but they do not attempt

    to make an exhaustive description or classification of them.

    The process block corresponds to the real plant (Figure 1). It can be

    characterized by its input vector u and its output vectory.

    zest

    ProcessControlyref

    ymes

    u

    Model

    y

    Adaptation?

    ymod

    Figure 1: Four-block description

    The "control" block has to define an appropriated control variable u, in

    order to obtain the wished reference vector yref on the process. This objective isgenerally realized through the measure of the process output vector ymes and/or

    other estimated variableszest.

    The "model" block is a process simulation, which yields to estimate

    variableszestneeded by the control. The control variable u can often be its inputvector. In most of cases, this block can be a simplified model of the process to

    avoid a long calculation time for real-time implementation.

    The difference between the process output y and the model one ymod istaken into account by the "adaptation" block. The output of this block can be used

    as input for other blocs; this information acts on different blocs according to the

    described robust strategies.

  • 8/6/2019 UPB Sci Bull ShortV

    3/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    This four-block description allows distinguishing the following

    techniques: control with observers (closed-loop model estimation), model

    reference adaptive control (MRAC), internal model control (IMC) and behavior

    model control (BMC). You can notice that some control strategies are a

    combination of the presented one's [21] [7].

    An observer imposes the model to follow the process. So, it yields

    accurate estimated variables [6].

    The output of the adaptation block is imposed as a supplementary input for

    the model (closed-loop estimation). The adaptation mechanism is a simple gain

    (or a linear controller) in most of the cases [8], but it could be even a non-linear

    mechanism [9].In drives applications, a lot of observer structures have been used for flux

    estimation of ac machines: reduced-order observer [10], stochastic and/or

    extended one [11]

    ProcessControlyref u

    Adaptation

    Model

    ymes

    y

    zest

    Model

    Processyref u

    Adaptation

    ymes

    yControl

    Figure 2: Example of a control structure with observer Figure 3: Example of MRAC structure

    The adaptive control [1] allows to adapt the parameters of controllers or of

    the model in real-time. So it yields an increasing of the control performances for

    processes with time-varying parameters. The most popular in drive applications is

    the model reference adaptive control (Figure 3).

    The adaptation output acts directly on the control [12] and/or the model

    [13] but by modifying their parameters.A lot of structures have been developed: yref as model input (the most

    classical), u as model input [12, 20], a double model [14]...

    The MRAC allows increasing the robustness of the induction machine

    control [15] and can lead to speed estimation [13, 14].

    The internal model control [2] rejects an equivalent perturbation of the

    process. The adaptation output defines this perturbation, which is compensated in

  • 8/6/2019 UPB Sci Bull ShortV

    4/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    the control (Figure 4). The adaptation mechanism can be a simple unitary gain or

    a noise filter.

    This control strategy has been applied to the induction machine [16, 24]

    and to DC/DC converters [23] in order to solve the disturbances of a load

    changing.

    Model

    ProcessControlyref u

    Adaptation

    y

    ymes

    Model

    ProcessControlyref u

    Adaptation

    ymes

    y

    Figure 4: Example of IMC structure Figure 5: Example of BMC structure

    3. Behavior Model Control (BMC)

    The behavior model control [3] imposes the process to follow the model.

    At the opposite of other structures, its adaptation output acts directly on

    the process by a supplementary input. The adaptation mechanism can be a simple

    gain [19] or a classical controller [25].Before applying BMC to a double closed-loop system, a general

    presentation of the single closed-loop BMC is given.

    The BMC single closed-loop structure is depicted in Figure 6. The

    structure uses a behavior model (M(s) and its modeled perturbation dmod) in

    parallel with the controlled processP(s) (and its perturbation d). About the chosenmodel, the authors of [3] propose dynamics close to the real process ones.

    The adaptation mechanism is the behavior controller CB(s). The difference

    between process and model outputs is the behavior controller input. Its output

    uregis added to the main controller output, ureg.

    IfCB(s) is well tuned,y andymodare identical. Both can be used as input ofthe main controller, Cm(s). In most of the cases ymod is used to lessen the noise;

    usingy, disturbances are much more attenuated but the noise level increases.From the BMC structure (see Figure 6), y andymodcan be expressed by:

    [ ]{ }[ ]{ }

    +=

    ++=

    modmodmodmod

    modmodmod

    )()()()()(

    )()()()()()()()()()(

    ddsysysCsMsy

    sdsdsysysCsysysCsPsy

    refm

    Brefm(1)

  • 8/6/2019 UPB Sci Bull ShortV

    5/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    P

    M

    CB

    +

    +

    -

    -

    -ureg

    d

    y

    ureg

    ymod

    +

    dmod

    Cm+yref

    -+

    ++

    +

    process

    modeldmod

    ucontrol

    adaptation

    Figure 6: Behavior model control structure

    To make easy the writing script, Laplace operator,s, will be omitted in thispaper.

    In order to point out two closed-loops, two different transfer functions can

    be obtained:

    ( )( )

    ( ddPC

    Py

    PCM

    MCPy mod

    B

    mod

    B

    B +

    ++

    +=

    11

    1) (2)

    ( modmodm

    ref

    m

    m

    mod ddMC1

    My

    MC1

    MCy

    ++

    += ) (3)

    Equation (3) yields to the main loop, a closed-loop composed by themodel and the main controller (Figure 7). Obviously, the second term, the

    perturbation, is null: it is indicated to point out the difference with the behavior

    loop. IfCm is well tuned with the model,ymodfollows the referenceyref.

    The condition of well-tuning for the main controller is:

    1MCm >> (4)

    yref

    +-

    -

    ymod

    dmod

    +

    dmod

    + +Cm M

    Figure 7: Equivalent main loop of BMC structure

    The model has well-known non-varying parameters and the perturbation is

    perfectly defined. So, the main controller tuning leads to a great robustness of this

    closed-loop (the perturbation is completely compensated).

    In order to simplify (2) let assume a big gain of the behavior controller:

    1>>BMC (5)

    Using (5), the simplified expression of (2) is:

  • 8/6/2019 UPB Sci Bull ShortV

    6/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    ( ddPCP

    yPC

    PCy mod

    B

    mod

    B

    B+++= 11 ) (6)

    Expression (6) yields to a closed-loop, called behavior loop. It consists

    of the behavior controller, the process, and the equivalent perturbation d* (afunction ofd-dmod).

    ymod

    +-

    -

    y

    d*

    +CB P

    Figure 8: Equivalent behavior loop of BMC structure

    If CB is well tuned with the process, y follows its reference ymod. Thiscondition can be written as:

    1>>BPC (7)

    Using (7), the equivalent perturbation d* becomes:

    ( modB

    * ddC

    d =1

    ) (8)

    It can be attenuated, if the perturbation model dmod is close to the process

    perturbation d.

    As it can be noticed, the output of the main loop is the reference of the

    behavior loop. So, the slowest loop imposes the dynamics of the global BMC.

    BMC has been already applied to solve robustness problems of electrical

    drives. In [19, 25] the process, a DC machine and its mechanical load, has varying

    parameters. [17] employs the strategy to drive a synchronous machine. Applying

    the BMC, the parameter variations have a reduced impact on the control. In [22]

    an induction machine is perturbed by a non-linear load torque. This disturbance is

    rejected using a model without perturbation (dmod=0).

    BMC has been applied to the control of a non-linear process, a wheel-rail

    contact law of a traction system [26]. A linear behavior of the process is obtained

    due to the linear model used and the BMC.

    But in such process, the BMC is only applied to a part of the process,

    which is of the first order.

    4. Double closed-loop structures

    In this section, the BMC is used in a double closed-loop, cascaded system.

    This approach is a new one and it is validated in numerical simulations and

    experiments in section 5.

  • 8/6/2019 UPB Sci Bull ShortV

    7/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    The classical control structure of a cascaded system is depicted in Figure 9

    (dashed rectangle). A cascaded control allows the independent management of

    each controlled variable.

    P1 and P2 are the transfer functions of the process; Cm1 and Cm2 are the

    corresponding controllers. The inner loop is made up by the main controller Cm1and the first processP1. The outer loop is made up by the main controller Cm2 and

    the processP2.

    In order to simplify the control it is often made the assumption of the

    separation in dynamics if the settling time of the inner loop is shorter than the

    settling time of the outer loop.

    These kind of cascaded structures are often use in motor drives: internal

    current loop and external speed loop in DC machine control, as an example.

    BMC structures

    A BMC structure brings M1 and M2 as behavior models ofP1 andP2, and

    CB1 and CB2 as behavior controllers.The behavior controllers outputs could act on the same spot. This kind of

    structure is called structure with global action (of the behavior controllers). A

    distributed action is the opposite of the global one: each behavior controller acts

    on its associated loop.

    When the output of the M1 is the input ofM2, both models lead to a global

    one (BMC with global model). The distributed models are totally independents,its do not put in common any input or output.

    There are two types of controllers and two types of models. Therefore,

    there are four BMC double closed-loops structures to be analyzed.

    We make the assumption of a perfect compensation of external

    disturbances, to underline the differences between these structures, not their

    specific proprieties. So, perturbation will not be taken into account in the next

    paragraphs.

    In the first structure (Figure 9), a global model is considered. It is makes

    up of both M1 and M2 models, directly connected.

    Both behavior controllers lead to a distributed action. The behaviorcontroller CB1 defines a supplementary output, u1reg, which acts in the inner loop.

    The other behavior controller CB2 acts in the outer loop through its output y1reg.

    Assuming the well tuning of inner loop controllers, Cm1 and CB1, this

    structure lead to a new process-dependent structure, a non-robust one.

  • 8/6/2019 UPB Sci Bull ShortV

    8/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    y2ref y1ref y1 y2

    +-

    +-

    ++

    u1reg

    y1reg

    u1ref

    y1mod y2mod

    u1regP2

    M2

    CB2

    +

    +

    -

    M1

    P1

    CB1

    +

    Cm1Cm2

    +-

    y1reg

    Figure 9: BMC structure with distributed action and global model.

    In the next structure (Figure 10), a global model is considered, where both

    M1 and M2 models are directly connected.

    The behavior controllers lead to a global action. The behavior controller

    CB1 defines a supplementary output, u1reg, which acts in the inner loop. The other

    behavior controller CB2 acts in the same spot, as CB1, through its output u2reg.

    y2ref y1ref y1 y2

    +- -

    +

    y1mod y2mod

    u2reg

    u1reg

    u1reg

    Cm2 Cm1 P1 P2

    CB1

    CB2

    M1 M2

    -

    +

    -

    +

    +++

    Figure 10: BMC structure with global action and global model.

    In order to makey2/y2modindependent of the ratioP2/M2,y1/y1mod must tendtowards M2/P2 ratio. Supplementary simplification assumption is necessary, other

    than (4), (5) or (7). This new simplification assumption is more complex and

    seems hard to accomplish.

  • 8/6/2019 UPB Sci Bull ShortV

    9/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    y2ref y1ref y1 y2

    +-

    +-

    y1reg

    + +

    u1reg

    u1reg

    y1reg

    y1mod

    y2mod

    Cm2 Cm1

    M2

    CB2

    P2P1

    M1

    CB1

    +

    -

    -

    +

    ++

    Figure 11: BMC structure with distributed action and distributed model.

    In Figure 11 the third structure is presented, the BMC structure with

    distributed action and distributed model. It means there is no direct connection

    between the models M1 and M2. The behavior controllers lead to a distributed

    action. The behavior controller CB1 defines a supplementary output, u1reg, which

    acts in the inner loop. The other behavior controller CB2 acts in the outer loop, as

    CB1, through its output y1reg.

    Using the same assumptions as a classical-single loop BMC, this structure

    leads to a robust cascaded process. The separation in dynamics is a supplementary

    assumption specific to a cascaded system. So, this structure could be generalized

    for more than two cascaded processes.

    The last structure (Figure 12) is a distributed model-global action

    structure. It means there is no direct connection between the models M1(s) and M2.The behavior controllers lead to a global action. The behavior controller CB1

    defines a supplementary output, u1reg, which acts in the inner loop. The other

    behavior controller CB2 acts in the same spot, as CB1, through its output u2reg.

    This structure is very similar with the second one and the study leads tothe same conclusion: a new complicated simplification assumption is necessary.

    This structure is not retained as a robust one.

    We have seen four possibilities to apply BMC to a cascaded system. The

    distributed action-global model structure (Figure 9) does not comply with the

    BMC principle.

  • 8/6/2019 UPB Sci Bull ShortV

    10/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    Two other structures, global action-global model (Figure 10) and global

    action-distributed model (Figure 12), need new and hard to accomplish

    assumptions.

    y2ref y1ref y1 y2

    +- -

    +

    y1mod

    y2mod

    u2reg

    u1reg

    u1reg

    M2

    M1

    CB1

    CB2

    P1 P2Cm1Cm2

    +

    -

    +

    -

    ++

    Figure 12: BMC structure with global action and distributed model.

    The distributed model-distributed action (Figure 11), comply with the

    BMC principle and it has a useful expression of the output/input transfer function.

    It can be generalized for more than two processes. This structure is tested in

    simulation and experimentally in the next section.

    5. Application to a DC machine

    In this section, the distributed action, distributed model BMC structure is

    applied to an experimental system, a classical electrical DC machine.

    Let assume that the process is a DC machine with permanent magnets. It

    can be described by a cascaded process: an electrical part and a mechanical one:

    =

    +=

    fTTdtdJ

    ueRidi

    L

    L

    dt(9)

    WhereR andL are the resistance and the inductance of the rotor windings,

    J and fare the total inertia and friction coefficients, i the current, u the supplyvoltage and e the back EMF.

    The coupling equations link electrical variables with the mechanical one

    through the flux coefficientK:

  • 8/6/2019 UPB Sci Bull ShortV

    11/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    =

    =

    KeiKT (10)

    A classical control of the DC machine is depicted in Figure 13 (the dashed

    rectangle).

    The behavior model (distributed action-distributed model structure) is

    applied to both electrical and mechanical parts (M1 and M2 are the models of the

    electrical and mechanical process and CBI and CBthe behavior controllers). Theelectrical and mechanical part of the DC machine are represented by theirs

    corresponding transfer functions, P1 and P2. The back EMF and the load torquecan be compensated by their estimations, eestand TL_est, if it is possible.

    The DC machine parameters are:

    ==

    ==

    msR

    L

    RK

    E

    E

    67.16

    416.01 1

    ==

    ==

    sf

    J

    Nmsradf

    K

    M

    M

    34.8

    //202.01

    ANmKKC /139.0==

    The main controllers, CmI and Cm, are the same in BMC as in a classical

    control. They are Integral Proportional controllers [18]. Its set the current settling

    time at (trM)I= 12 ms and the speed settling time at (trM)= 400 ms. The damping

    factor is imposed to 1, in both main loops. The assumption of decoupling modes is

    accomplished by the choice of the settling times.The behavior controllers, CBi and CB, are classical Proportional Integral

    controllers:

    +

    s

    s1K

    . Their gains are chosen to ensure (5). Increasing theses

    gains, the robustness of the control increases, but the noise too.

    The constants are chosen in order to have a good dynamics of the

    external perturbation rejections:

    Robustness test

    A test has been performed in order to compare the robustness of a classicalcontrol and BMC. At t= 0 we simulate the response to a set point step change; the

    speed goes from 0 to 100 rad/s (a third of the rated value) and back again at t= 2s.The robustness test has associated an ideal trajectory. This one is defined as the

    speed response of the studied control, when all parameters are well known and all

    perturbations are well compensated. A real trajectory is the test response of thestudied control with parameter variations or bad compensations. These trajectories

    are depicted in Figure 14.

  • 8/6/2019 UPB Sci Bull ShortV

    12/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    E

    E

    s

    K

    +1

    mod_E

    mod_E

    s

    K

    +1

    KC

    mod_E

    mod_E

    s

    K

    +1

    TL

    TL_mod

    -

    -

    KC

    CK

    1

    -

    -

    e

    emod

    emod

    CmICmM

    M

    s

    K

    +1

    ref Treg ir

    - -

    CBI

    CB

    Tref

    + +

    ureg

    ureg

    Treg

    imod

    mod

    +

    TL_mod

    +

    +

    +++

    Figure 13: BMC applied to DC machine (cascaded system)

    An error is used in order to compare the transient performances of the

    classical control and BMC. This error is defined as follow:

    100input_Step

    trajectory_Realtrajectory_Ideal(%)

    = (11)

    Assuming a parameter variation, the defined error is drawn in Figure 14.

    Some assumptions have been made in order to simplify simulations: ideal

    power converter, ideals sensors, and no limitation of the control values.

    0 1 2 3 4

    20

    0

    20

    40

    60

    80

    100

    120

    Time (s)

    Speed(rad/s)

    Speed ReferenceIdeal TrajectoryReal Trajectory

    0 1 2 3 4

    20

    0

    20

    40

    60

    80

    100

    120

    Time (s)

    Speed(rad/s)

    Speed ReferenceIdeal TrajectoryReal Trajectory

    0 1 2 3 4

    15

    10

    5

    0

    5

    10

    15

    Time (s)

    SpeedError(%)

    Figure 14: Speed trajectories and the error

    Firstly, the current closed-loop is affected by a parameter variation

    (R=1.5Rmod), as imposed by a temperature increasing. The back EMF iscompensated.

  • 8/6/2019 UPB Sci Bull ShortV

    13/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    Both classical control and BMC work well (Figure 15). Using (11), the

    errors between the ideal speed trajectories and the real ones are less than 1%,

    (Figure 15). The same result is obtained for inductance saturation:L=1.4Lmod.Even the error is small, BMC works better.

    0 1 2 3 4

    1

    0.5

    0

    0.5

    1

    Time (s)

    pee

    rror

    Classical ControlBMC

    0 1 2 3 4

    1

    0.5

    0

    0.5

    1

    Time (s)

    SpeedError(%)

    Classical ControlBMC

    Figure 15: Electrical parameter variation (R=1.5Rmod) effect Figure 16: Back EMF effect

    The second comparison concerns the back EMF, which is a slow

    perturbation. There are no parameter variations and the back EMF is compensated

    neither in the classical control nor in the BMC. Both controls work well

    (Figure 16) and the BMC error is smaller.

    It has been seen that the current closed-loop is the fastest. The influence of

    a parameter variation on the slow closed-loop (speed loop) is studied in thisparagraph.

    The affect of -50% parameter variation on the inertia J (J=0.5Jmod), a

    mechanical parameter, is shown in Figure 17. The BMC works better than the

    classical control. Using (11), the error decreases from 11,85% to 4,5%.

    0 1 2 3 4

    15

    10

    5

    0

    5

    10

    15

    Time (s)

    SpeedE

    rror(%)

    Classical ControlBMC

    0 1 2 3 4

    5

    0

    5

    10

    15

    20

    25

    Time (s)

    SpeedE

    rror(%)

    Classical ControlBMC

    Figure 17: Mechanical parameter variation (J=0.5Jmod) effect Figure 18: Load torque effect

  • 8/6/2019 UPB Sci Bull ShortV

    14/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    A physical friction variation does not affect the classical control, nor the

    BMC significantly.

    Finally, the machine was subjected to a step load change of 0.32 Nm (25%

    of rated load torque) to allow load rejection to be analyzed. The load torque step

    in at t = 1 s and it is not compensated in any controls. The BMC reduces the

    transient error from 22% to 4%. The error level can be decrease increasing the

    gain of the speed behavior controller. The transient of the error is controlled by

    the B constant of the speed behavior controller.

    Experimental results confirm the theoretical ones. The experimental

    platform consists of a 0.4 kW DC machine. A 1 kW IGBT chopper is used to

    supply the machine. All drive control is implemented on a DSpace DS1102 DSPcontroller board. 16 kHz was chosen as switching frequency and 10 kHz assampling frequency. The experiments are made without load.

    Comparative performances of the classical control and BMC are shown in

    the Figure 19.

    A parameter variation (J=0.5Jmod) affects the speed control. Without

    BMC, two effects are noticed: the 12% overshot and the effect of a non-linear

    torque, due to the Coulombus frictions (Figure 19). BMC drives better than the

    classical control because the error decrease to 4,5%, as simulations have

    predicted.

    The effect of a non-linear torque can be seen inA area of the Figure 19.

    We have chosen to show only the effect of that parameter variation(J=0.5Jmod) because of its important error. Others parameter variations are toosmall to be relevant, for this process.

    0 1 2 3 4

    20

    0

    20

    40

    60

    80

    100

    120

    Time (s)

    Speed(rad/s)

    Classical ControlBMC

    0 1 2 3 4

    15

    10

    5

    0

    5

    10

    15

    Time (s)

    SpeedError(%)

    Classical ControlBMC

    Figure 19: Mechanical parameter variation

  • 8/6/2019 UPB Sci Bull ShortV

    15/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    Conclusion

    The behavior model control (BMC) can be an alternative of other robust

    control strategies in the field of electrical machines. It is based on a

    supplementary input of the process in order to impose it to follow the model. If

    the behavior loop is well tuned, the principal control loop is so more robust

    because it is defined with the well-known model.

    The BMC is extended to cascaded systems in this paper. Four possibilities

    have been shown with global or distributed models, and with global or distributed

    actions. The proposed studies point out that the BMC with local models and

    distributed actions can lead to better performances. This specific structure has

    been validated with an experimental DC machine system.

    The double BMC for the DC machine does not increase the drive

    performances than a single BMC one, based on the mechanical part. But, these

    cascaded BMC can be now applied to AC machines, where the control has

    robustness problems (phase of transformations, flux estimations) and where

    several closed-loops with close dynamics are used (currents and flux loops) [2].

    REFERENCES

    [1] B.K. Bose, Power electronics and motion control - Technology and recent trends, IEEE Trans.

    on Industry Applications, vol. 29, 1993, pp 902-909.

    [2] R.D. Lorenz, T.A. Lipo, D. Nowotny, Motion control with induction motors, Proceeding of theIEEE, vol. 82, August 1994, pp 1215-1240.

    [3] J.P. Hautier, J.P. Caron, Systmes automatiques. Tome 2: Commande des processus, Edition

    Ellipses, Paris, 1997.

    [4] I.D. Landau, Adaptive Control: The Model Reference Approach, Marcel Dekker, New York,

    1979.

    [5] M. Morari, E. Zafiriou, Robust Process Control, Prentice Hall, Englewood Cliffs, New Jersey,

    1989.

    [6] G. Verghese, S. Sanders, Observers for flux estimation in induction machines, IEEE

    Transactions on Industrial Electronics, vol. 35, no 1, February 1988.

    [7] M. Elbuluk, N. Langovsky, D. Kankam, Design and implementation of a closed-loop observer

    and adaptive controller for induction motor drives, IEEE Transactions on Industry Applications,

    Vol. 34, no 3, May/June 1998, pp 435-443.

    [8] P. Jansen, R. Lorenz, A physically insightful approach to the design and accuracy assessmentof flux observers for field oriented induction machine drives, IEEE Transactions on Industry

    Application, vol. 30, no 1, January/February 1994, pp 101-109.

    [9] R. Nielsen, M. Kazmierkowski, Reduced-order observer with parameter adaptation for fast

    rotor flux estimation in induction machines, IEE Proceeding D, vol. 136, no 1, January 1989, pp

    35-43.

    [10] T. Orlowska-Kowalska, Application of extended Luenberger observer for flux and rotor time-

    constant estimation in induction motor drives, IEE Proceeding D, vol. 136, no 6, November 1989,

    pp 324-330.

  • 8/6/2019 UPB Sci Bull ShortV

    16/16

    U.P.B. Sci. Bull., Series D, Vol. 65, No. 1, 2003

    [11] T. Du, P. Vas, F. Stronach, Design and application of extended observers for joint state and

    parameter estimation in high-performance AC drives, IEE Electronics Power Applications, vol.142, no 2, March 1995, pp 71-78.

    [12] W. Wang, C. Wang, A rotor-flux-observer-based composite adaptive speed controller for an

    induction machine, IEEE Transactions on Energy Conversion, vol. 12, no. 4, December 1997, pp

    323-329.

    [13] C. Schauder, Adaptive speed identification for vector control of induction motors without

    rotational transducers, IEEE Transactions on Industry Applications, vol. 28, no 5,

    September/October 1992, pp 1054-1061.

    [14] L. Zhen, L. Xu, Sensorless field orientation control of induction machines based on a mutual

    MRAS scheme, IEEE Transactions on Industrial Electronics, vol. 45, no. 5, October 1998, pp 824-

    830.

    [15] T. Rowan, R. Kerkman, D. Leggate, A simple on-line adaptation for indirect filed orientation

    of an induction machine, IEEE Transactions on Industry Applications, vol. 27, no.4, July/August

    1991, pp 720-727.[16] L. Harnefors, H.P. Nee, Model-based current control of AC machines using the internal

    model control method, IEEE Transactions on Industry Applications, Vol. 34, no 1,

    January/February 1998, pp 133-141.

    [17] B. Robyns, Y. Fu, F. Labrique, H. Buyse, Commande numrique de moteurs synchrones

    aimants permanents de faible puissance (in French), Journal de Physique III, vol. 5, no 3, August

    1995, pp 1255-1268.

    [18] P.K. Nandam, P.C. Sen, Analogue and digital speed control of DC drives using proportional-

    integral and integral-proportional control techniques, IEEE Ton Transactions on Industrial

    Electronics, vol. IE-34, May 1987, pp 227-233.[19] P.J. Barre, J.P. Hautier, X. Guillaud, B. Lemaire-Semail, Modelling and axis control of

    machine tool for high speed machining, Proceeding of IFAC'97, Belfort,1997, pp. 63-68.

    [20] H. Naitoh, M. Hirano, S. Tadakuma, Microprocessor-based adjustable speed dc motor drives

    using model reference adaptive control, Proceeding of IAS Annual Meeting, 1985, pp 524-528.

    [21] F. Palis, A. Buch, U. Ladra, R. Kurrich, Q. Ila, M. Negnevitsky, Fuzzy and neuronal control

    of drive systems with changing parameters and load, Proceeding of EPE-EDDA Conference,

    Nancy 1996, pp 183-185.

    [22] I. Stefan, C. Forgez, B. Lemaire-Semail, X. Guillaud, Comparison between neural

    compensation and internal model control for induction machine drive, ICEM'97 Conference,

    Istanbul September 1997, pp 1330-1334.

    [23] I. Gadoura, T. Suntio, K. Zenger, P. Vallitu, Internal model control for DC/DC converters,

    EPE'99 Conference, Lausanne (Switzerland), September 1999.

    [24] J.L. Thomas, M. Boidin, An internal model control structure in field oriented controlled VSI

    induction motors, Proceeding of EPE'91, Firenze (Italy), 1991, vol. 2, pp 202-207.

    [25] B. Vulturescu, A. Bouscayrol, J.P. Hautier, X. Guillaud, F. Ionescu, Behaviour model control

    of a DC machine, ICEM'2000, Conference Espoo (Finland), August 2000.

    [26] J. Pierquin, P. Escan, A. Bouscayrol, M. Pietrzak-David, J.P. Hautier, B. de Fornel,

    Behaviour model control of a high speed traction system, EPE-PEMC'2000 Conference, Kocise

    (Slovak Republic), September 2000.


Recommended