JP6174063B2 - System identification apparatus and system identification method - Google Patents

Description

  The present invention relates to a system identification device and a system identification method.

  For example, in feed-forward control, a mathematical model taking into account the actual control object is created in advance, and an input command to the control system is determined by using this mathematical model. In order to improve the accuracy of the feedforward control, it is important to make the response of the mathematical model as close as possible to the response of the actual controlled object. Therefore, system identification is performed in which the parameters of the mathematical model are determined based on the response of the actual control target (see, for example, Patent Documents 1 and 2).

  The system identification as described above is not limited to feedback control, and can be used for abnormality diagnosis of a control target, for example. More specifically, by monitoring changes over time in the parameters of the mathematical model determined by system identification, it is possible to grasp abnormalities and aging degradation that occur in the controlled object.

Japanese Patent Laid-Open No. 7-114531 Japanese Patent Laid-Open No. 2004-30065

  In order to improve the accuracy of the feedforward control and abnormality diagnosis described above, it is necessary to increase the accuracy of system identification. However, when the mathematical model includes nonlinear elements, it is difficult to perform system identification with high accuracy, and it is difficult to improve the accuracy of feedforward control and the accuracy of abnormality diagnosis.

  An example of the nonlinear element is backlash. For example, as shown in FIG. 14, the motor shaft 100 and the load shaft 102 mesh with each other by gears 104 and 106, and a multi-inertia torsional mechanical device that transmits the rotational force of the motor to the load is provided between the motor and the load. There is a backlash between each gear 104, 106 coupled between them. This backlash is “play” between the tooth surfaces when a pair of gears are engaged with each other, and is expressed as a dead zone characteristic or a hysteresis characteristic in the mathematical model.

  The dead zone characteristic that defines the backlash is expressed as a characteristic having the center of the dead zone width as a zero point, as shown in FIG. This characteristic presupposes that the zero point of the sensor provided on each of the motor side and the load side is located at the center of the backlash. However, in an actual mechanical apparatus, the zero point of the sensor provided on each of the motor side and the load side does not always coincide with the center of the backlash. Therefore, in order to further improve the accuracy of system identification, it is necessary to perform system identification in consideration of such a shift amount of the dead zone center position.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a system identification device and a system identification method capable of improving the accuracy of system identification of a mathematical model including nonlinear parameters. .

  A first aspect of the present invention is a two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. A system identification device for identifying a motor-side residual equation including a motor-side residual term for a motor-side residual equation, wherein the motor current or motor torque, the motor A motor-side error calculation unit that calculates a plurality of motor-side internal signals and the motor-side residual using the rotation position of the load and the rotation position of the load as input information, and a plurality of included in the motor-side residual calculation formula The calculation result of the motor side error calculation unit is used for the parameter estimation formula in which the linear parameter and the plurality of nonlinear parameters are each represented by an error component. Accordingly, error components of the plurality of linear parameters and the plurality of nonlinear parameters are respectively estimated, and the linear parameters and the nonlinear parameters are identified from the estimated error components of the linear parameters and the error components of the nonlinear parameters. A motor-side parameter identification unit, wherein the parameter estimation formula is expressed by a linear combination of a plurality of linear terms and a plurality of nonlinear terms, and each linear term includes a linear internal signal and a single linear parameter. Each nonlinear term is represented by a product of a nonlinear internal signal and a single nonlinear parameter error component, and one of the plurality of nonlinear parameters is a center of the deadband function. It is a system identification device that is a position.

According to the system identification apparatus, the motor-side error calculation unit calculates the internal signal and the motor-side residual included in the motor-side residual calculation formula, and uses the internal signal and the motor-side residual as a parameter estimation formula. By using the motor-side parameter and the nonlinear parameter on the motor side included in the two-inertia system model, the motor-side parameter identification unit identifies the motor-side linear parameter and the nonlinear parameter. Here, the parameter estimation formula is a calculation formula in which the linear parameter and the nonlinear parameter are expressed by error components in the motor side residual calculation formula, and a function in which a plurality of linear terms and a plurality of nonlinear terms are linearly combined. It is represented. Furthermore, each linear term contains only a single linear parameter, and each nonlinear term contains only a single nonlinear parameter. Therefore, by using such a parameter estimation formula, it is possible to handle the motor-side model that is nonlinear as a linear model. Thereby, the identification accuracy of the nonlinear parameter can be improved, and the system identification accuracy of the two-inertia system model including the nonlinear element can be improved.
Furthermore, since the system identification is performed in consideration of the center position shift of the dead band function, the accuracy of the system identification can be further improved.

  A second aspect of the present invention is a two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. A system identification device for identifying a load-side residual equation including a load-side residual term for a load-side residual equation and a rotational position of the motor and the load A load-side error calculation unit that calculates a plurality of load-side internal signals and the load-side residual, a plurality of linear parameters and a plurality of nonlinearities included in the load-side residual calculation formula By using the calculation result of the load-side error calculation unit for the parameter estimation formula in which each parameter is represented by an error component, a plurality of the linear parameters and a plurality of the non-linear parameters are used. A load-side parameter identification unit that estimates each linear parameter and each nonlinear parameter from the estimated error component of each linear parameter and each estimated nonlinear error component, The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms, and each of the linear terms is represented by a product of a linear internal signal and a single error component of the linear parameter. The nonlinear term is expressed by a product of a nonlinear internal signal and an error component of a single nonlinear parameter, and one of the plurality of nonlinear parameters is a system identification device that is a center position of the dead band function.

According to the system identification apparatus, the load side error calculation unit calculates the internal signal and the load side residual included in the load side residual calculation formula, and uses the internal signal and the load side residual as the parameter estimation formula. By using it, the load-side linear parameter and the nonlinear parameter included in the two-inertia system model are identified by the load-side parameter identification unit. Here, the parameter estimation formula is a calculation formula in which the linear parameter and the nonlinear parameter are respectively expressed by error components in the load side residual calculation formula, and a function in which a plurality of linear terms and a plurality of nonlinear terms are linearly combined. It is represented. Furthermore, each linear term contains only a single linear parameter, and each nonlinear term contains only a single nonlinear parameter. Therefore, by using such a parameter estimation formula, it is possible to handle a load-side model that is nonlinear as a linear model. Thereby, the identification accuracy of the nonlinear parameter can be improved, and the system identification accuracy of the two-inertia system model including the nonlinear element can be improved.
Furthermore, since the system identification is performed in consideration of the center position shift of the dead band function, the accuracy of the system identification can be further improved.

  In the system identification apparatus, the plurality of nonlinear parameters may include a dead band width of the dead band function.

  According to the system identification device, not only the center position shift of the dead zone function but also the dead zone width of the dead zone function can be made closer to the actual mechanical device. As a result, the system identification accuracy can be further improved.

  According to a third aspect of the present invention, a control signal is generated using the system identification device described above and the inverse model of the two-inertia system model reflecting the linear parameter and the nonlinear parameter identified by the system identification device. A motor control system including a feedforward control unit.

  According to a fourth aspect of the present invention, the system identification device described above, and the amount of change of each linear parameter and each nonlinear parameter identified by the system identification device are monitored, and each amount of change is set according to the parameter. An abnormality diagnosis system comprising an abnormality detection unit that detects an abnormality when the predetermined threshold value is exceeded.

  A fifth aspect of the present invention is a two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. A system identification method for identifying a motor-side residual equation that expresses a motor-side equation of motion including a motor-side residual term with respect to a motor-side residual calculation formula, wherein the motor current or the motor torque, the motor Using the rotational position of the load and the rotational position of the load as input information, calculating a plurality of motor-side internal signals and the motor-side residual, a plurality of linear parameters included in the motor-side residual calculation formula, and A plurality of the motor-side internal signals and the motor-side residuals are used for a parameter estimation formula in which a plurality of nonlinear parameters are represented by error components, respectively. A step of estimating error components of the plurality of linear parameters and the plurality of nonlinear parameters, respectively, and each linear parameter and each nonlinear parameter from the estimated error component of each linear parameter and the error component of each nonlinear parameter. The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms, each linear term being a linear internal signal and a single error component of the linear parameter. Each nonlinear term is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter, and one of the plurality of nonlinear parameters is a center position of the deadband function. It is a system identification method.

  A sixth aspect of the present invention is a two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. A system identification method for identifying a load-side residual equation including a load-side residual term including a load-side residual term with respect to a load-side residual calculation formula And calculating a plurality of load-side internal signals and the load-side residual, and a plurality of linear parameters and a plurality of non-linear parameters included in the load-side residual calculation formula, respectively. By using a plurality of the load side internal signals and the load side residuals for the parameter estimation formula represented by the component, a plurality of the linear parameters and a plurality of the nonlinear Estimating each parameter error component, and identifying each linear parameter and each nonlinear parameter from the estimated error component of each linear parameter and each nonlinear parameter error component, and the parameter estimation formula Is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms, wherein each said linear term is represented by the product of a linear internal signal and a single error component of said linear parameter, and each said nonlinear term is The system identification method is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter, and one of the plurality of nonlinear parameters is a center position of the dead band function.

  According to the present invention, it is possible to improve the identification accuracy of nonlinear parameters.

It is the figure which showed one structural example of the control object of the motor control system which concerns on one Embodiment of this invention. It is the figure which showed the block diagram when the mechanical apparatus shown in FIG. 1 was shown with the mathematical model of a two-inertia system. It is a figure showing functional composition of a motor control system concerning one embodiment of the present invention. It is the figure which showed an example of the dead zone function model which prescribed | regulated the backlash. It is the figure which represented the motor side error calculation part and the load side error calculation part as a block diagram. It is a figure for demonstrating a dead zone width change function. It is a figure for demonstrating a dead zone width change function. It is a figure for demonstrating a dead zone center position change function. It is a figure for demonstrating a dead zone center position change function. It is the figure which represented the parameter estimation formula which represented the model parameter contained in the motor side residual calculation formula as an error component with the block diagram. It is the figure which represented the parameter estimation formula which represented the model parameter contained in the load side residual calculation formula as an error component with the block diagram. It is the figure which showed an example of the simulation result of the estimation precision of a dead zone width and a dead zone center position by the system identification apparatus which concerns on one Embodiment of this invention. It is the figure which showed one structural example of the abnormality diagnosis system which concerns on one Embodiment of this invention. It is a figure for demonstrating backlash. It is a figure for demonstrating the dead zone characteristic which prescribes | regulates a backlash.

[Motor control system]
Hereinafter, an embodiment in the case of applying a system identification device and a system identification method of the present invention to a motor control system will be described with reference to the drawings.

FIG. 1 shows a configuration example of a control target of the motor control system 1 according to the present embodiment. As shown in FIG. 1, a control target is a multi-inertia torsional mechanical device 2 including a motor 20, a load 21, and a transmission mechanism 22 that transmits power from the motor 20 to the load 21. The transmission mechanism 22 includes, for example, a gear 24 connected to the motor shaft 23 and a gear 26 connected to the load shaft 25, and the force is transmitted when these gears mesh with each other. When the gear 24 and the gear 26 are engaged with each other, backlash occurs between the tooth surfaces, and this is defined as a nonlinear element in a mathematical model described later.
In the present embodiment, a gear is described as an example of the transmission mechanism 22, but the transmission mechanism 22 is not limited to a gear, and may be, for example, a screw.

The motor control system 1 receives information on the target position θ _t of the load 21 from a host controller (not shown), and makes the rotational position (hereinafter referred to as “load position”) θ _L of the load 21 coincide with the target position θ _t. The motor torque is calculated and a torque command signal τ is output to the motor 20. Thus, the motor 20 is rotationally driven with a torque based on the torque command signal τ, and the torque is transmitted from the motor 20 side to the load 21 side through the transmission mechanism 22. As a result, the load 21 rotates according to the transmission torque τ ′. In this way, the load position theta _L is controlled to a desired target position theta _t.

When the mechanical device 2 is regarded as a two-inertia system, a plurality of parameters representing characteristics unique to the mechanical device 2 (hereinafter referred to as “actual machine parameters”) indicate parameters indicating characteristics on the motor 20 side and characteristics on the load 21 side. The parameters are roughly divided into parameters indicating the characteristics of the transmission mechanism 22. For example, as parameters indicating characteristics on the motor 20 side, a torque constant (torque efficiency) K _T , a motor side moment of inertia J _M , a motor side viscosity coefficient D _M , and a motor side coulomb friction coefficient τ _fM represent the characteristics on the load 21 side. The parameters shown include the load side moment of inertia J _L , the load side viscosity coefficient D _L , and the load side Coulomb friction coefficient τ _fL . Parameters indicating the characteristics of the transmission mechanism 22 include gear efficiency η, torsional stiffness coefficient Kg, and torsional viscosity coefficient Dg. Further, in the present embodiment, the deadband width BL and the deadband center position SH are given as parameters related to backlash. It is done.

FIG. 2 shows a block diagram when the mechanical device 2 shown in FIG. 1 is represented by a mathematical model of a two-inertia system.
As shown in FIG. 2, the motor 20, the motor-side Coulomb friction coefficient τ _{torque fM} or the like is subtracted from the torque τ obtained by multiplying the torque constant K _T of the motor current i _r as an input, transfer element 1 / (J _Ms + D _M ) and a transmission element 1 / s, which is expressed as a transmission system that outputs a rotational position (hereinafter referred to as “motor position”) θ _M of the motor.

The load 21 receives a torque obtained by subtracting the load-side Coulomb friction coefficient τ _fL from the transmission torque τ ′ transmitted from the motor 20 through the motor shaft 23, and transmits the transmission element 1 / (J _Ls + D _L ) and the transmission element 1 /. through s are expressed as a transmission system for outputting a load position theta _L.
In the middle of the transmission system shown in FIG. 2, the motor angular velocity ω _M and the load angular velocity ω _L are shown.

The transmission mechanism 22 receives the deviation between the motor position θ _M and the load position θ _L, and passes through the dead band function F1 that defines the characteristics of the backlash, the transmission element K _g and the transmission element D _g s to the load 21. It is expressed as a transmission system that outputs an applied transmission torque τ ′. Here, the dead band function F1 is a non-linear function BKLS (θ _M −θ _L ) whose output changes according to the deviation (θ _M −θ _L ) between the motor position and the load position, and is a dead band width BL, which is an actual machine parameter. This is an element including the dead zone center position SH.

Torque constant (torque efficiency) K _T , motor side inertia moment J _M , motor side viscosity coefficient D _M , motor side coulomb friction coefficient τ _fM , load side inertia moment J _L , load side viscosity coefficient D _L , load side coulomb friction The coefficient τf _L , the gear efficiency η, the dead zone width BL, and the dead zone center position SH are actual machine parameters that represent actual characteristics of the mechanical device 2, and are unknown parameters that cannot be individually observed. On the other hand, the motor current i _{r as} an input, the motor position θ _{M as} an output, and the load position θ _L are observable parameters.

FIG. 3 is a diagram illustrating a functional configuration of the motor control system 1 according to the present embodiment. As shown in FIG. 3, the motor control system 1 includes, for example, a feedback control unit 10, a feedforward control unit 11, and a system identification device 12.
The feedback control unit 10 calculates a torque such that the deviation (θ _t −θ _L ) between the target position θ _t of the load 21 and the load position θ _L becomes zero, and outputs a feedback torque command signal τ _FB . At this time, the feedback control unit 10 uses the motor angular velocity omega _{M (time} differential of the motor position theta _M) calculated based on the motor position theta _M, calculates the torque as appropriate and rapid feedback control is performed May be.

The feedforward control unit 11 calculates and outputs a feedforward torque command signal τ _FF using a two-inertia model MOD that is a mathematical model representing the mechanical device 2 as a two-inertia system. The two-inertia model MOD is an inverse model of the mechanical device 2 shown in FIG. 2 and includes model parameters corresponding to the actual machine parameters of the mechanical device 2 described above. Specifically, the torque parameters (torque efficiency) K _T0 , motor-side inertia moment J _M0 , motor-side viscosity coefficient D _M0 , motor-side coulomb friction coefficient τ _fM0 , load-side inertia moment J _L0 , load-side viscosity are model parameters. It includes a coefficient D _L0 , a load-side Coulomb friction coefficient τ _fL0 , a gear efficiency η ₀ , a dead band width BL ₀ , and a dead band center position SH ₀ .
These model parameters (K _T0 , J _M0 , D _M0 , τ _fM0 , J _L0 , D _L0 , τ _fL0 , η ₀ , BL ₀ , SH ₀ ) are estimated and updated by the system identification device 12 described later. .

Thus, in the present embodiment, the dead zone width BL ₀ and the dead zone center position SH ₀ that are parameters related to backlash are also considered as model parameters, and these model parameters are also identified by the system identification device 12 described later. As a result, the two-inertia system model MOD can be brought closer to the characteristics of the actual machine device 2, and the system identification accuracy can be further improved.

The feedback torque command signal τ _FB output from the feedback control unit 10 and the feed forward torque command signal τ _FF output from the feed forward control unit 11 are added by the adding unit 13, and the torque command signal τ of the mechanical device 2 is added. It is output to the motor 20.

The system identification device 12 includes model parameters (K _T0 , J _M0 , D _M0 , τ _fM0 , J _L0 , D _L0 , τ _fL0 , η ₀ , BL ₀ , SH of the two-inertia model MOD of the feedforward control unit 11. ₀ ) is identified. The system identification device 12 includes, for example, a motor side error calculation unit 120, a motor side parameter identification unit 121, a load side error calculation unit 122, and a load side parameter identification unit 123. Each part will be described below.

[Motor side error calculator]
Motor-side error calculator 120, the motor-side residual calculation formula represents the motor side motion equation for sections of the motor side residuals e _M includes a term of the motor side residuals e _M, the motor current i _r, the motor using location theta _M, and the load position theta _L as input information, calculates the motor-side internal signals q1~q5 and the motor side residuals e _M.

For example, if the response error of the motor-side model of the two-inertia model MOD feedforward control unit 11 is provided is defined as a motor-side residuals e _M, the motor-side equation of motion includes a term of the motor side residuals e _M is the following It is represented by the formula (1).

Here, sign (sθ _M ) takes a value of “+1” when sθ _M is positive (sθ _M > 0), and a value of “−1” when sθ _M is negative (sθ _M <0). Is a nonlinear function.
Further, as shown in FIG. 4, BKLS ₀ (θ _M −θ _L ) is a dead zone function model F1 _MOD that defines backlash, has a dead zone width (single width) BL ₀ , and a dead zone center position SH _0. Take.

(1) reacting a becomes the following formula (2) represent the motor side residuals e _M, or less, the (2) equation of the motor side residual equation.

In equation (2), the backlash characteristic is regarded as a fixed characteristic represented by the dead band function model F1 _MOD defined in the current two-inertia model MOD, and the known portion is regarded as the motor side internal signal qi (i = 1, 2). ..)), The following equation (3) is obtained.

  Here, q1 to q5 are as in the following formula (4).

Motor-side error calculator 120 calculates the motor-side internal signal q1~q5 shown in the above (4) equation, to calculate the motor side residuals e _M, and outputs them to the motor-side parameter identification unit 121.
FIG. 5 shows a block diagram of the motor-side error calculation unit 120.

[Motor side parameter identification section]
Motor-side parameter identification unit 121 using the motor side residuals e _M and the motor-side internal signal q1~q5 output from the motor-side error calculating unit 120, to identify the motor side model parameters.

  First, the motor side residual calculation formula is represented by the following formula (2) as described above.

Here, assuming that the motor-side residual e _M is caused by a modeling error of each model parameter, each model parameter is expressed by an error component as defined in the following equation (5). Can be expressed by the following equation (6).

  Further, in the above equation (6), when the motor side internal signal qi (i = 1, 2,...) Is defined and arranged as in the following equation (7), equation (8) is obtained.

In this way, by partially differentiating each term of the motor-side residual calculation formula with the corresponding model parameter and expressing the model parameter included in the motor-side residual calculation formula as an error component, the motor-side residual calculation formula is It can be expressed by a linear combination of a linear term and a nonlinear term. Here, the linear term is expressed as the product of the linear internal signal and the error component of a single linear parameter, and the nonlinear term is expressed as the product of the nonlinear internal signal and an error component of a single nonlinear parameter.
By estimating each model parameter on the motor side using such a parameter estimation formula, it becomes possible to handle the motor-side model that has been nonlinear as a linear model, so that the identification accuracy of the nonlinear parameter can be improved. Therefore, it is possible to improve the system identification accuracy of the two-inertia system model including the nonlinear element.
The parameter estimation formula represented by the above formula (8) can be represented by a block diagram as shown in FIG.

In equation (7), {∂BKLS (θ _M −θ _L ) / ∂BL}, which is an element of the nonlinear internal signal q6, is a nonlinear function indicating the amount of change in the characteristics of the dead band function F1 due to the dead band width error component δBL. Yes, {∂BKLS (θ _M −θ _L ) / ∂SH} is a nonlinear function indicating the amount of change in the characteristic of the dead zone function F1 due to the error component δSH of the dead zone center position.

Here, as shown in FIG. 6, assuming a dead band function model F1 _MOD having an error component of δBL (δBL> 0) in the dead band width with respect to the dead band function F1 defining the characteristics of the mechanical device 2, the dead band function F1. A characteristic indicating the amount of change in the dead band function model F1 _MOD with respect to, in other words, a function obtained by subtracting the dead band function model F1 _MOD from the dead band function F1 is a non-linear function as shown in FIG. 7 (hereinafter referred to as “dead band width change function”). F2 = ∂BKLS (θ _M −θ _L ) / ∂BL. Thus, the dead band width change function F2 is a non-linear function indicating the amount of change in the dead band function F1 (= BKLS (θ _M −θ _L )) due to the dead band width error component δBL.

Similarly, as shown in FIG. 8, assuming a dead zone function model F1 _MOD having an error component of δSH (δSH> 0) at the dead zone center position with respect to the dead zone function F1 defining the characteristics of the mechanical device 2, the dead zone function The characteristic indicating the amount of change of the dead zone function model F1 _MOD with respect to F1, in other words, the function obtained by subtracting the dead zone function model F1 _MOD from the dead zone function F1 is a non-linear function as shown in FIG. 9 (hereinafter referred to as “dead zone center position change function”). .) F3 = ∂BKLS (θ _M -θ _L ) / ∂SH. Thus, the dead zone center position change function F3 is a non-linear function indicating the amount of change in the dead zone function F1 (= BKLS (θ _M −θ _L )) due to the error component δSH of the dead zone center position.

Next, each model parameter included in the above equation (8) is estimated. For example, it can be estimated using the following method.
First, internal signals q1 to q7 are multiplied on both sides of equation (8), and both sides are integrated over time T, for example, over a period of one cycle when the mechanical device 2 is driven with a sine wave, (9) Get the formula.

  Further, the formula (9) is transformed into the following formula (10).

Subsequently, the motor side parameter identification unit 121 calculates error components (δK _T , δJ _M , δD _M , δτ _fM , δη, δBL, δSH) of the motor side model parameters according to the following equation (11).

  Then, the motor-side parameter identification unit 121 calculates each new model parameter (each motor-side linear parameter and each motor-side nonlinear parameter) by adding the error component of the motor-side model parameter to the current model parameter. To do.

[Load side error calculation section]
Load error calculation unit 122, the load side residual calculation expression indicating the section of the load-side residuals e _L the load side equation of motion includes a term of the load side residuals e _L, motor position theta _M and the load by using the position theta _L, and calculates the load side internal signal and the load side residual.

For example, if you define the response error of the load-side model of the two-inertia model MOD feedforward control unit 11 is provided with a load side residual e _L, the load-side equation of motion includes a term of the load side residuals e _L is the following It is expressed by equation (12).

Equation (12) the result with the following equation (13) and representing the load side residuals e _L, or less, the (13) equation that the load side residual equation.

In the equation (13), when the backlash characteristic is regarded as a fixed characteristic represented by the dead band function model F1 _MOD and the known part is arranged as the load side internal signal qi (i = 8, 9,...), (14)

  Here, q8 to q11 are as in the following formula (15).

Load error calculation unit 122 calculates the load side internal signal q8~q11 shown in the above (15), calculates the load side residuals e _L, and outputs them to the load side parameter identification unit 123.
FIG. 5 shows a diagram representing the load-side error calculation unit 122 as a block diagram.

[Load side parameter identification section]
Load parameter identification unit 123, using the load-side error calculation unit load residual output from 122 e _L and load side internal signal Q8～q11, identifying load model parameters.

  First, the load side residual calculation formula is expressed by the following formula (13) as described above.

Here, when the load side residuals e _L is assumed to be caused by modeling error of each model parameter, expressed in error component so as to define a respective model parameters in the following equation (16), (13) Can be expressed by the following equation (17).

  Furthermore, in the above equation (17), when the load side internal signal qi (i = 8, 9...) Is defined and arranged as in equation (18), equation (19) is obtained.

In this way, by partially differentiating each term of the load side residual calculation formula with each model parameter and expressing the model parameter included in the load side residual calculation formula with an error component, Similarly, the load side residual calculation formula can be expressed by a linear combination of a linear term and a nonlinear term. By using such a parameter estimation formula, it is possible to handle a load-side model that is nonlinear as a linear model. Thereby, the identification accuracy of the nonlinear parameter can be improved, and the system identification accuracy of the two-inertia system model including the nonlinear element can be improved.
Here, the non-linear internal signal relating to each backlash is as described in the motor side.
The parameter estimation equation represented by the above equation (19) can be represented by a block diagram as shown in FIG.

  Next, each model parameter included in the equation (19) is estimated using, for example, the same method as that on the motor side described above. That is, both sides of the above equation (19) are multiplied by internal signals q8 to q13, and both sides are time-integrated over time T, for example, a period of one cycle when the mechanical device 2 is driven with a sine wave (20 ) Get the formula.

  Furthermore, the equation (20) is transformed into the following equation (21).

Thereby, the load side parameter identification unit 123 calculates error components (δJ _L , δD _L , δτ _fL , δη, δBL, δSH) of the load side model parameters according to the following equation (22).

  Then, the load-side parameter identification unit 123 calculates each new model parameter (each load-side linear parameter and each load-side nonlinear parameter) by adding the error component of the load-side model parameter to the current model parameter. To do.

As described above, the non-linear parameters η ₀ , BL ₀ , and SH ₀ overlap between the motor side model and the load side model. Therefore, for these nonlinear parameters, values determined by either the motor-side parameter identification unit 121 or the load-side parameter identification unit 123 may be used for the other. Thereby, reduction of a processing burden etc. can be aimed at. In this case, it is preferable to use the nonlinear parameter identified by the motor side parameter identification unit 121 for the load side parameter identification unit 123. This is because the motor side is the input side of the mechanical device 2, and therefore there is less noise than the load side and the identification accuracy is relatively high.
Note that the calculation method for obtaining each model parameter from the parameter estimation formulas represented by the above-described formulas (8) and (14) is not limited to the above example, and a known method can be appropriately employed. .

Next, the operation of the motor control system 1 according to this embodiment will be described.
First, an error between the target position θ _t and the load position θ _L is input to the feedback control unit 10, and a feedback torque command signal τ _FB that makes the load position θ _L coincide with the target position θ _t is calculated and output. . On the other hand, the feedforward control unit 11, the feed-forward torque command signal tau _FF is calculated using the target position theta _t and two-inertia model MOD, are output. The feedback torque command signal τ _FB and the feedforward torque command signal τ _FF are added by the adding unit 13 and are given to the mechanical device 2 that is the control target as the torque command signal τ. Thus, the motor is rotated by the torque based on the torque command signal τ applied to the motor, the rotational power is transmitted to the load side by the transmission mechanism 22, the control so that the load position theta _L becomes the target position theta _t Is done.

The motor position θ _M and the load position θ _L are detected at a predetermined timing by a sensor (not shown), and these detected values are fed back to the feedback control unit 10. In addition, the system identification device 12, the motor position theta _M is detected by the _sensor, the load position theta _L and the motor current i _r is input.
In system identification device 12, the motor-side error calculator 120, the motor position theta _M, using a load position theta _L and the motor current _{i r,} calculates the motor-side internal signals q1~q5 and the motor side residuals _{e M.} Motor-side parameter identification unit 121 using a motor-side internal signals q1~q5 and the motor side residuals _{e M} calculated by the motor-side error calculator 120, the error component .delta.K _T linear parameters of the motor side, .delta.j _M, δD _M , δτ _fM and nonlinear parameter error components δη, δBL, δSH are calculated, and these error components are respectively calculated as current linear parameters K _TM0 , J _M0 , D _M0 , τ _fM0 and nonlinear parameters η ₀ , BL _0. , SH ₀ are added to calculate new model parameters K _TM0 , τ _fM0 , D _M0 , J _M0 , η ₀ , BL ₀ , SH ₀ .

Similarly, load-side error calculation unit 122, using the motor position theta _M and the load position theta _L, and calculates the load side internal signals q8~q11 and load side residuals _{e L.} Load parameter identification unit 123, using the load-side internal signals q8~q11 and load side residuals _{e L} calculated by the load-side error calculating unit 122, the error component .delta.j _L of linear parameters of the load, [delta] D _L, δτ _fL and nonlinear parameter error components δη, δBL, δSH are calculated, and these error components are added to the respective current linear parameters J _L0 , D _L0 , τ _fL0 and nonlinear parameters η ₀ , BL ₀ , SH ₀ . Thus, each model parameter J _L0 , D _L0 , τ _fL0 , η ₀ , BL ₀ , SH ₀ is identified. At this time, for the model parameters η ₀ , BL ₀ , SH ₀ , the number of model parameters determined by the load-side parameter identifying unit 123 is reduced by using the model parameters determined by the motor-side parameter identifying unit 121 as known. It is possible. As a result, the processing burden can be reduced.

Motor-side model parameters K _TM0 , τ _fM0 , D _M0 , J _M0 , η ₀ , BL ₀ , SH ₀ and load-side model parameters J _L0 , D _L0 , τ _fL0 , η ₀ , BL determined in the system identification device 12 ₀ and SH ₀ are reflected in the two-inertia system model MOD of the feedforward control unit 11. As a result, it is possible to reduce an error between the model parameter of the two-inertia system model MOD and the actual machine parameter of the mechanical device 2, and it is possible to improve the accuracy of the feedforward control.

As described above, according to the system identification apparatus and the method thereof according to the present embodiment, the parameter estimation formula expressing the model parameter included in the motor side residual calculation formula or the load side residual calculation formula as an error component is used. Thus, the identification accuracy of the non-linear parameter can be improved. Thus, the system identification accuracy of the two-inertia system model including the non-linear element can be improved. Then, by reflecting the model parameters identified with such high accuracy in the feedforward control unit 11, it is possible to improve the control accuracy of the feedforward control unit 11.
Furthermore, since the system identification is performed in consideration of the center position shift of the dead band function, the accuracy of the system identification can be further improved.

FIG. 12 is a diagram illustrating an example of a simulation result of estimation accuracy of the dead zone width and the dead zone center position by the system identification device 12 according to the present embodiment. As shown in FIG. 12A, in this simulation, the actual parameter BL = 0.3 [rad] of the dead zone width (single width) has an error component δBL = + 0.2 [rad] and the dead zone. A dead band function model F1 _MOD having an error component δSH = 0.1 [rad] with respect to the actual machine parameter SH = 0 at the center position is assumed, and these nonlinear parameters BL ₀ and SH ₀ are identified using the system identification device 12. We assumed the case. Here, other model parameters are set to the same values as actual machine parameters.

FIG. 12B shows a simulation result of the dead zone width BL ₀ , and FIG. 12C shows a simulation result of the dead zone center position SH ₀ . As shown in FIGS. 12B and 12C, the dead zone width BL ₀ and the dead zone center position SH ₀ are estimated to have almost the same values as the actual machine parameters, and the dead zone width estimation error is 0.52 [%]. The estimation error of the dead zone center position was −2.27 [%].
Thus, it was confirmed that the system identification device 12 according to the present embodiment can achieve sufficient estimation accuracy.

[Abnormality diagnosis system]
Next, an embodiment when the system identification apparatus and method of the present invention are applied to an abnormality diagnosis system will be described with reference to the drawings.
In the motor control system 1 described above, the error component between the actual machine parameter and the model parameter is regarded as a modeling error, and the purpose is to bring the two-inertia system model MOD in the feedforward control unit 11 closer to the characteristics of the mechanical device 2.
On the other hand, for example, when the reliability of the model parameter identified by the system identification device 12 is already secured, that is, when the control error by the motor control system 1 satisfies a predetermined error range, the model parameter The error component generated in can be regarded as a change on the mechanical device 2 side due to deterioration over time or abnormality.

  The abnormality diagnosis system according to the present embodiment grasps a change in the characteristics of the machine device 2 by monitoring the amount of change in each model parameter identified by the system identification device 12, and based on the monitoring result, It is a system that detects or predicts abnormalities.

FIG. 13 is a diagram showing a configuration example of an abnormality diagnosis system according to an embodiment of the present invention. As shown in FIG. 13, the abnormality diagnosis system 3 includes a system identification device 12 and an abnormality detection unit 15.
Since the system identification device 12 is the same as the system identification device 12 included in the motor control system 1 described above, description thereof is omitted.

The anomaly detection unit 15 _calculates the amount of change of each model parameter K _T0 , J _M0 , D _M0 , τ _fM0 , J _L0 , D _L0 , τ _fL0 , η ₀ , BL ₀ , SH ₀ identified by the system identification device 12. Monitor each one. For example, the abnormality detection unit 15 holds a reference value (for example, initial value) of each model parameter, and also holds a change amount threshold value as a reference for determining that each model parameter is abnormal. And the abnormality detection part 15 monitors the variation | change_quantity with respect to a reference value about each model parameter, when the variation | change_quantity exceeds the threshold value, it detects abnormality and alert | reports that.

According to such an abnormality diagnosis system 3, each model parameter K _T0 , J _M0 , D _M0 , τ _fM0 , J _L0 , D _L0 , τ _fL0 , η ₀ , BL ₀ , SH identified by the system identification device 12 is used. ₀ is input to the abnormality detection unit 15. The abnormality detection unit 15 calculates the amount of change with respect to the reference value for each model parameter, and determines whether or not the amount of change is greater than or equal to a threshold set for each model parameter. When there is a parameter whose change amount is equal to or greater than the threshold, it is determined that there is an abnormality.

As described above, according to the abnormality diagnosis system 3 according to the present embodiment, the dead zone width of the dead zone function that defines the backlash of the mechanical device 2 and the dead zone center position are identified by the system identification device 12 that handles them as model parameters. Since abnormality diagnosis is performed based on the model parameters, it is possible to improve the accuracy of abnormality diagnosis of the mechanical device 2.
Note that the abnormality determination method by the abnormality detection unit 15 described above is not limited to the above example, and a known method can be employed. For example, instead of or in addition to the change amount, the change rate of the model parameter may be monitored, and it may be determined that an abnormality occurs when the change rate of the model parameter exceeds a predetermined threshold.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the invention.

For example, in the above-described embodiment, the model parameters are the torque constant (torque efficiency) K _T0 , the motor side moment of inertia J _M0 , the motor side viscosity coefficient D _M0 , the motor side coulomb friction coefficient τ _fM0 , the load side moment of inertia J _L0 , the load The side viscosity coefficient D _L0 , the load side Coulomb friction coefficient τ _fL0 , the gear efficiency η ₀ , the dead zone width BL ₀ , and the dead zone center position SH ₀ are specified. It is possible to select appropriately according to the accuracy.

In the embodiment described above, it is used as input information the motor current i _r, instead of this, the input and the motor torque command signal tau, the motor torque multiplied by the torque constant K _T of the specified motor current i _r It may be used as information.

  In the above-described embodiment, the integration is performed in the equations (9) and (20). However, the integration may be performed in real time without performing the integration.

DESCRIPTION OF SYMBOLS 1 Motor control system 2 Mechanical apparatus 3 Abnormality diagnosis system 10 Feedback control part 11 Feedforward control part 12 System identification apparatus 15 Abnormality detection part 20 Motor 21 Load 22 Transmission mechanism 24, 26 Gear 120 Motor side error calculation part 121 Motor side parameter identification Unit 122 load side error calculation unit 123 load side parameter identification unit

Claims (7)

A two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. ,
A motor current or a motor torque, a rotational position of the motor, and a rotation of the load with respect to a motor side residual calculation formula representing a motor side equation of motion including a term of a motor side residual with respect to the term of the motor side residual A motor-side error calculation unit that calculates a plurality of motor-side internal signals and the motor-side residual using the position as input information;
A plurality of linear parameters and a plurality of nonlinear parameters included in the motor-side residual calculation formula are used as parameter estimation formulas each represented by an error component, by using the calculation result of the motor-side error calculation unit, Motor side parameters for estimating error components of the linear parameters and the plurality of nonlinear parameters, respectively, and identifying the linear parameters and the nonlinear parameters from the estimated error components of the linear parameters and error components of the nonlinear parameters An identification unit,
The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms,
Each linear term is represented by the product of a linear internal signal and a single linear parameter error component;
Each nonlinear term is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter;
The system identification apparatus, wherein one of the plurality of nonlinear parameters is a center position of the dead band function. A two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. ,
Using the rotational position of the motor and the rotational position of the load as input information with respect to the load-side residual calculation formula expressing the load-side equation of motion including the load-side residual term for the load-side residual term A load side error calculation unit for calculating a plurality of load side internal signals and the load side residual;
A plurality of linear parameters and a plurality of nonlinear parameters included in the load-side residual calculation formula are respectively calculated by using the calculation result of the load-side error calculation unit for the parameter estimation formulas represented by error components. Load-side parameters for estimating error components of the linear parameters and the plurality of nonlinear parameters, respectively, and identifying the linear parameters and the nonlinear parameters from the estimated error components of the linear parameters and error components of the nonlinear parameters An identification unit,
The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms,
Each linear term is represented by the product of a linear internal signal and a single linear parameter error component;
Each nonlinear term is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter;
The system identification apparatus, wherein one of the plurality of nonlinear parameters is a center position of the dead band function.   The system identification apparatus according to claim 1, wherein the plurality of nonlinear parameters include a dead band width of the dead band function. The system identification device according to any one of claims 1 to 3,
A motor control system comprising: a feedforward control unit that generates a control signal using an inverse model of the two-inertia system model in which the linear parameter identified by the system identification device and the nonlinear parameter are reflected. The system identification device according to any one of claims 1 to 3,
The amount of change in each of the linear parameter and the nonlinear parameter identified by the system identification device is monitored, and an abnormality is detected when the amount of change is equal to or greater than a predetermined threshold set according to the parameter. An abnormality diagnosis system comprising an abnormality detection unit. A two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. ,
A motor current or a motor torque, a rotational position of the motor, and a rotation of the load with respect to a motor side residual calculation formula representing a motor side equation of motion including a term of a motor side residual with respect to the term of the motor side residual Calculating a plurality of motor side internal signals and the motor side residual using the position as input information;
A plurality of the motor-side internal signals and the motor-side residuals are used for a parameter estimation formula in which a plurality of linear parameters and a plurality of nonlinear parameters included in the motor-side residual calculation formula are represented by error components, respectively. Respectively, estimating respective error components of the plurality of linear parameters and the plurality of nonlinear parameters;
Identifying each linear parameter and each nonlinear parameter from the estimated error component of each linear parameter and the error component of each nonlinear parameter,
The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms,
Each linear term is represented by the product of a linear internal signal and a single linear parameter error component;
Each nonlinear term is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter;
The system identification method, wherein one of the plurality of nonlinear parameters is a center position of the dead band function. A two-inertia system model representing a mechanical device having a motor and a load, wherein the two-inertia system model includes a dead band function as a transfer characteristic for transmitting power from the motor to the load. ,
Using the rotational position of the motor and the rotational position of the load as input information with respect to the load-side residual calculation formula expressing the load-side equation of motion including the load-side residual term for the load-side residual term Calculating a plurality of load side internal signals and the load side residual;
A plurality of load side internal signals and a plurality of load side residuals are used for a parameter estimation formula in which a plurality of linear parameters and a plurality of nonlinear parameters included in the load side residual calculation formula are represented by error components, respectively. Respectively, estimating respective error components of the plurality of linear parameters and the plurality of nonlinear parameters;
Identifying each linear parameter and each nonlinear parameter from the estimated error component of each linear parameter and the error component of each nonlinear parameter,
The parameter estimation formula is represented by a linear combination of a plurality of linear terms and a plurality of nonlinear terms,
Each linear term is represented by the product of a linear internal signal and a single linear parameter error component;
Each nonlinear term is represented by a product of a nonlinear internal signal and a single error component of the nonlinear parameter;
The system identification method, wherein one of the plurality of nonlinear parameters is a center position of the dead band function.

Images