JP2020194375A - Servo system parameter identification method - Google Patents

Abstract

To provide a method for identifying various parameters in a short time without using a special operation or a large capacity memory.SOLUTION: In the method for identifying an inertia moment or the like based on time-series data of four signals (motor acceleration, motor speed, sign of motor speed, and torque) when predetermined conditions are met, the predetermined conditions are made to be a velocity absolute value greater than a first threshold value and an acceleration absolute value greater than a second threshold value, the four signals at a time ti are denoted as Xk (ti) (k=1 to 4), variables are defined as Skl=Σ[Xk (ti)×Xl (ti)] (k, l=1 to 4), which are updated when the predetermined conditions are met for |X1| and |X2| at the time ti, Skl is set to zero during a period that |X2| is equal to or less than a first threshold while the update of Skl is started when |X2| exceeds the first threshold, the update of Skl is completed when |X2| becomes the first threshold or less due to deceleration, and an inertia moment J, a viscous resistance coefficient D, and a Coulomb friction Tf are simultaneously identified as solutions of a given determinant.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for identifying a mechanical parameter including a moment of inertia, a viscous drag coefficient, and Coulomb friction of a controlled object in a servo system that controls the motion of the controlled object including a motor and its load.

In servo control, it is necessary to know the mechanical parameters such as the moment of inertia and load torque of the controlled object in order to make appropriate control settings. In particular, the moment of inertia is an important parameter, and if it is unknown, an appropriate control gain cannot be set, and good followability to position commands and speed commands cannot be obtained, or vibrational movements occur. It may become.

Here, various techniques for estimating the moment of inertia of the controlled object have been conventionally provided.
For example, in Patent Document 1, as shown in FIG. 10, a signal ΔT _{b in} which the torque command signal is calculated by the high-pass filter 11, a signal Δb in which the acceleration feedback signal of the motor is calculated by the high-pass filter 12, and an estimated value storage means 13 are described. Using the output, the error en of the difference between the current value and the previous value of the torque command signal is obtained, and when the vibration in the acceleration feedback signal is _equal to or greater than a predetermined threshold, the estimated gain of the moment of inertia is 0 or more and less than 1. the value obtained by multiplying a coefficient α is multiplied to the error e _n, by the time the vibration is less than the threshold for multiplying the estimated gain as the error e _n, calculating the moment of inertia without being influenced by steady disturbance or vibration A method for sequentially identifying the value J _n is described.

Further, in the servo control device described in Patent Document 2 that simultaneously estimates the moment of inertia and friction of the electric motor drive shaft, as shown in FIG. 11, a sinusoidal command is superimposed on the torque command output from the speed control unit 21. Then, the torque constant K _t is multiplied via the current control unit 22 and the amplifier 23 and given to the control target 30. Then, the inertia / friction estimation unit 40 takes in the current feedback value i (n) _FB and the velocity feedback value ω (n) _FB via the sampling units 41 and 42, and the inverse function estimation unit 44 takes in the estimation error e (n). ) Is calculated for each sampling period so that the coefficients (estimated inertia J _m and estimated friction C _f (estimated viscous friction and estimated Coulomb friction)) are calculated. Further, the inverse function model 43 updates the calculation formula based on the above coefficient to generate an estimated current value x (n), and estimates the deviation between this x (n) and the i (n) _FB as an estimation error e (n). Is input to the inverse function estimation unit 44.

Further, Patent Document 3 and Non-Patent Document 1 are given a periodic velocity command of positive and negative symmetry, and the value estimated as the load torque by the disturbance observer is multiplied by the velocity, acceleration, etc. and integrated to obtain the moment of inertia and the like. A servo control device that calculates the estimation error and updates the estimated values of moment of inertia, viscous resistance coefficient, and Coulomb friction is described.

FIG. 12 is a block diagram of this servo control device, in which positive and negative symmetric periodic speed commands ω _r and acceleration commands ω _r'are input to and output from the feedback control unit 52 and the feed forward control unit 51, respectively. feedback signal U _FB and torque corresponding to the torque command tau _r and the disturbance compensation signal (estimated disturbance) d _a torque command u obtained by subtracting a is applied to the load machine 53 that. Disturbance compensation module 54 outputs from the angular velocity detection value ω of the torque command u and the load machine 53 to the disturbance compensation signal d _a and the internal signal vector [phi] a and calculates parameter identification unit 55, parameter identification unit 55, inertia moment, the viscous resistance coefficient, steady disturbance and the identified values of the Coulomb friction ^{_{J *, D *, d c}} *, the feedforward control unit 51 obtains a _T ^{c *,} the feedback control unit 52, and outputs the disturbance compensation module 54.
Figure 13 shows the parameter identification unit 55, an estimation error δJ performing multiplication and integration with internal signals _{q 1, q 1 ', q} 2 and the disturbance compensation signal d _a that constitute the internal signal vector [phi] ^* , δD ^* and steady disturbance T _c ^* are calculated, and the moment of inertia J _a and the viscous resistance coefficient D _a of the observer model are used to obtain the moment of inertia identification value J ^* and the viscous resistance coefficient identification value D ^* , and T _c. ^* Is output as it is as the Coulomb friction identification value, and the steady disturbance identification value d _c ^* is obtained based on the internal signal q ₀ and the disturbance compensation signal da _a .

Japanese Patent No. 3796261 ([0008], [0024], FIG. 6, etc.) Japanese Patent No. 4813318 ([0033] to [0057], FIGS. 2 to 7, etc.) Japanese Unexamined Patent Publication No. 6-309008 ([0007] to [0040], FIGS. 1 to 4, etc.)

Ichiro Awaya et al., "Methods for identifying parameters of mechanical motion systems on which Coulomb friction acts", JSME Proceedings (C), Vol. 59, No. 567 (1993-11), No. 93-0155

In many conventional techniques such as Patent Documents 1 and 2, an online identification method using a sequential least squares method or the like that can save the memory capacity in the control device is adopted. In principle, these online identification methods have the advantage that data from an infinitely long past can be reflected and sequentially identified, which not only saves memory but also enables robust identification against noise. is there.

On the other hand, in the servo system, sometimes the mechanical parameters may change suddenly. For example, when various loads are loaded on a uniaxial stage and transported, the moment of inertia of the servo system changes depending on the load, but the influence of the past estimated value is affected until the estimated value of the moment of inertia converges to a new true value. It takes time to receive it. For such problems, the influence of past estimates can be reduced by using, for example, the sequential least squares method with a forgetting coefficient, but since it is a trade-off with robustness against noise, inertia Choosing the optimum value for the moment is not easy.

Further, according to the methods described in Patent Document 3 and Non-Patent Document 1, each parameter can be identified in a short time by appropriately selecting the cycle of the speed command which is a periodic signal. However, for identification, it is essential to give a periodic speed command to the system, and in the case of the above-mentioned uniaxial stage example, it is not possible to identify the parameters during normal operation. ..
Further, when the same machine is operated under the same conditions, the friction torque and the viscous drag coefficient usually do not change significantly, but when an abnormality occurs in the machine, the friction torque or the like is large as a sign that a significant abnormality occurs. It may look different. If this kind of change can be accurately captured, it can be applied to accident prevention, etc., but it is necessary to add special operation for the purpose of monitoring various parameters on a daily basis to the device for which the same machine is operated under the same conditions. There is a request to avoid it if possible.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to apply a current corresponding to a torque command to a motor to control the motion of a controlled object including the motor and its load machine. In a servo system, even in a system where mechanical parameters such as the moment of inertia change, from the moment of inertia, viscous resistance coefficient, and Coulomb friction in a short time without the need for special operation for identification or a large amount of memory. It is an object of the present invention to provide a method for identifying a servo system parameter, which enables the identification of a mechanical system parameter in a short time.

In order to solve the above problems, the invention according to claim 1 is intended for a servo system that controls the motion of a controlled object including the motor and its load machine by energizing a motor with a current corresponding to a torque command.
The moment of inertia J and the viscous resistance coefficient of the controlled object are based on time-series data consisting of four signals of the motor acceleration, the motor speed, the code of the motor speed, and the applied torque when the predetermined conditions for updating are satisfied. In the method of identifying the parameters of the servo system for identifying D and the Coulomb friction T _f ,
The predetermined condition is that the absolute value of the motor speed is larger than the first threshold value and the absolute value of the motor acceleration is larger than the second threshold value.
Wherein at time t _i 4 signal when expressed as _{X k (t i) (k} = 1~4) each variable motor speed and motor acceleration at time t _i is updated only if it meets a predetermined condition _{S kl = Σ [X k (} t i) × X l (t i)] (k, l = 1~4) defines, while the absolute value of the motor speed is below the first threshold value the variable S Zero is given to _{kl, and} when the absolute value of the motor speed becomes larger than the first threshold value, the update of the variable _Skl is started, and then the motor decelerates and the absolute value of the motor speed is equal to or less than the first threshold value. At the time when the variable _Skl is updated, the moment of inertia J, the viscous resistance coefficient D, and the Coulomb friction _Tf are simultaneously identified as the solution of the mathematical formula 1 described later.

The invention according to claim 2 is the method for identifying parameters of a servo system according to claim 1.
Immediately after solving the mathematical formula A to identify the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f , the identification result and the cumulative count number SUM = Σ1 which is the number of updates of the S _kl and the S _kl are used. Therefore, when the estimation error ε obtained by the equation 5 described later exceeds a predetermined threshold value ε _max , the identification result at that time is invalidated.

The invention according to claim 3 is the method for identifying parameters of a servo system according to claim 1 or 2.
The identification results of the moment of inertia J, the viscous resistance coefficient D, and the Coulomb friction T _f obtained as the solution of Equation 1 are defined as the first identification results, and the moving average value of the first identification results is defined as the second identification result. At the same time, every time the first identification result is updated as a valid value, a moving average calculation is executed to update the second identification result, and the second identification result is the moment of inertia J, the viscous resistance coefficient D, and Coulomb. It is characterized in that the friction T _f is used for control.

The invention according to claim 4 is the method for identifying parameters of the servo system according to claim 3.
The moving average score and the reset timing of the moving average value for obtaining the second identification result are individually given for each of the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f .

The invention according to claim 5 is the method for identifying parameters of a servo system according to claim 1 or 2.
The identification results of the inertial moment J, the viscous resistance coefficient D, and the Coulomb friction T _f obtained as the solution of Equation 1 are defined as the first identification results, and the viscous resistance coefficient D and the Coulomb friction T _f are transferred from the first identification result. The second identification result obtained as an average value is defined, and every time the first identification result is updated as a valid value, a moving average calculation is executed to update the second identification result, and the control of the servo system is performed. Using the second identification result,
For the Coulomb friction T _f , the third identification result and the fourth identification result obtained as moving average values for each velocity polarity are defined with respect to the first identification result, and the first identification result is updated as a valid value. Each time, the moving average calculation on the side corresponding to the velocity polarity is executed to update the third identification result or the fourth identification result, and the control of the servo system is performed by the third identification result or the third identification result according to the velocity polarity. It is characterized by using the fourth identification result.

The invention according to claim 6 is the method for identifying parameters of the servo system according to claim 5.
The moving average score and the reset timing of the moving average value for obtaining the third identification result or the fourth identification result are individually given for each of the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f. To do.

The invention according to claim 7 is the method for identifying parameters of a servo system according to any one of claims 1 to 6.
The servo system moment of inertia changes every performing acceleration and deceleration operation, applies the viscous resistance coefficient D and Coulomb friction T _f estimated by the operation up to the previous, the viscous resistance coefficient from applying the torque of the motor D and Coulomb friction T _f A provisional estimated value of the moment of inertia is obtained during motor acceleration based on the value obtained by subtracting the value and the motor acceleration, and the absolute speed of the motor becomes equal to or less than the first threshold value during the deceleration process of the motor. It is characterized in that the provisional estimated value is used for controlling the servo system until the true values of D and the Coulomb friction T _f are obtained.

The invention according to claim 8 is characterized in that, in the method for identifying a parameter of a servo system according to any one of claims 1 to 7, a floating point arithmetic operation is used for the operation of the variable _Skl .

According to the present invention, in a servo system that controls the motion of a controlled object composed of a motor and its load machine by energizing a motor with a current corresponding to a torque command, even in a system in which the moment of inertia changes, it is possible to identify the motor. Mechanical parameters consisting of moment of inertia, viscous resistance coefficient and Coulomb friction can be identified in a short time without the need for special operation or large capacity memory.

It is a block diagram of the servo system to which embodiment of this invention is applied. It is a flowchart which shows 1st Example of the parameter identification method which concerns on this invention. It is a flowchart which shows the 2nd Example of the parameter identification method which concerns on this invention. It is a flowchart which shows the 3rd Example of the parameter identification method which concerns on this invention. It is a figure which shows the specific example of the moving average processing by step S410 of FIG. It is a flowchart which shows 4th Example of the parameter identification method which concerns on this invention. It is a flowchart which shows 5th Example of the parameter identification method which concerns on this invention. It is a flowchart which shows 6th Example of the parameter identification method which concerns on this invention. It is a timing chart of the motor speed and torque for demonstrating the 7th Example of the parameter identification method which concerns on this invention. It is a block diagram which shows the main part of the moment of inertia estimation apparatus described in Patent Document 1. It is a block diagram which shows the main part of the electric motor control device described in Patent Document 2. It is a block diagram of the servo control device described in Patent Document 3. It is a block diagram of the parameter identification part in FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram of a servo system to which the mechanical parameter identification method according to this embodiment is applied. This servo system is a system that controls the motion of a controlled object 100 composed of a motor and a load machine, and in order to drive the motor, an inverter or the like is supplied from the speed controller 101 based on a speed command v ^* and a motor speed v. current source 102 provides a torque command tau ^* to a system for supplying a current I in accordance with the torque command tau ^* to the motor, the motor speed v and the acceleration a with applied by the motor torque τ ₍₌ K τ I) It shall be an acquireable system. In FIG. 1, 103 is a subtraction means, 104 is a model of a controlled object 100 represented by a transfer function (1 / Js), and 105 is an approximate friction characteristic.

Here, the acceleration a of the motor may be obtained by the numerical differentiation of the speed v, or the motor speed v may be obtained by the numerical differentiation of the rotor position of the motor. Further, the torque applied by the motor τ may be obtained from the torque command τ ^{* or} may be obtained from the detected value of the current I energizing the motor.
Further, it is assumed that the load torque can be approximated to the motor speed v in the form of Dv + T _f sign (v) except near the zero speed of the motor. In addition, sign (v) is a code of motor speed v.
In the above load torque (Dv + T _f sign (v)), the viscous drag coefficient D and the Coulomb friction T _f are unknown parameters together with the total moment of inertia (hereinafter, simply referred to as the moment of inertia) J of the controlled object including the motor and the load machine. It is assumed that it is a mechanical parameter to be identified according to the present invention.

Next, FIG. 2 is a flowchart showing a first embodiment of the parameter identification method according to the present invention.
The flowcharts shown in FIGS. 2 and 2 may be executed each time each data such as the motor speed v and the motor acceleration a is updated, or may be executed in an integer multiple of the data update cycle. Further, in each flowchart, "measuring" indicates that measurement related to parameter identification is being performed.
Furthermore, the respective flowcharts, the motor acceleration a at the sampling time _{t i,} the motor speed v, the motor speed code sign (v), 4 two signals _X k _(t i) consisting of applying torque τ (k = 1~4), _They will be referred to as X ₁ , X ₂ , X ₃ , and X ₄ , respectively. As described above, the motor acceleration X ₁ may be a value obtained by numerical differentiation of motor speed X _2, also the motor speed X ₂ may be a value obtained by numerical differentiation of the rotor position of the motor.

In FIG. 2, first, when the motor speed is being measured (S201Yes) and the absolute motor speed | X ₂ | is larger than v _min as the first threshold value (S202Yes), or when the motor speed is not being measured (S201No), the motor speed When the absolute value | X ₂ | is larger than the first threshold value of v _{min and} the flag being measured is turned on (S205Yes, S206), the process proceeds to step S203. When | X ₂ | ≦ v _min in step S202 (S202No), it is determined that the measurement is completed and the flag being measured is turned off (S208), and then the process proceeds to step S209, and in step S205. | _{X 2} | a ≦ _v if a _min (S205No) is a variable _{S kl} to be described later to 0 (S207), and ends the identifying periodic process.

In step S203, the absolute value of motor acceleration | X ₁ | is compared with a _min as the second threshold value, and only in the case of | X ₁ |> a _min (S203 Yes), the variable S _kl related to identification is S _kl ← S. _{It is} updated by _kl + X _k X _l (S204), and when | X ₁ | ≦ a _min , the periodic identification process is terminated (S203 No). Here, the subscripts k = 1 to 4, l = 1 to 4.

Further, when the motor decelerates during the measurement and the measurement is completed when | X ₂ | ≤ v _min (S202No, S208), the calculation of S _kl ← S _kl + X _k X _l at that time, determines that calculates completion of _{S kl} = _{ΣX k} X _l in other words, using the above variables (matrix _{elements) S kl,} equation moment of inertia J, the viscosity resistance coefficient D and Coulomb friction _{T f} below It is obtained as the solution of 1 (same as formula A in the scope of patent claim).

従って、数式２によりモデル化した場合の二乗誤差ｆは、数式３となる。

Hereinafter, the calculation of the above-mentioned formula 1 will be described.
If J, D, and T _f are all true values and there is no modeling error, the torque τ is expressed as in Equation 2 using the acceleration a and velocity v of the motor.

Therefore, the square error f when modeled by the mathematical formula 2 is the mathematical formula 3.

If the squared error f in Equation 3 is partially differentiated by J, D, and T _f , and the values are set to 0, Equation 4 is obtained.

Here, when the mathematical formula 4 is transformed by using S ₁₁ = Σa ^2, etc., the mathematical formula 1 is obtained. That is, the calculation of Equation 1 is in principle the same as the offline least squares method. However, since the matrix element calculation required at the final stage of the offline least squares method is performed during operation, it is not necessary to secure the memory of the time series data. Further, when calculating the matrix element of the variable _Skl , in addition to excluding the extremely low speed region in which the load torque is clearly non-linear with respect to the velocity v by step S202, the inertia estimation is performed by step S203. Since the low acceleration that hinders the above is also excluded, the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f can be identified with high accuracy.
If the operation pattern at the time of identification is not specified, it is not possible to predict in advance the score of the data and the order of the value of the variable _Skl when the measurement is completed, but by adopting the floating point operation for the variable _Skl , overflow etc. Risk can be avoided.

なお、図３におけるステップＳ３０１〜Ｓ３０３，Ｓ３０５〜Ｓ３０９は、図２に示した第１実施例のステップＳ２０１〜Ｓ２０３，Ｓ２０５〜Ｓ２０９にそれぞれ相当している。 Next, FIG. 3 is a flowchart showing a second embodiment of the parameter identification method according to the present invention.
In the second embodiment, in addition to the steps in the first embodiment, in step S304, previously obtained the _S SUM ← SUM + 1 as the integrated count number is a number of updates of _kl, i.e. SUM = .SIGMA.1 with matrix elements _{S kl} .. Then, when the update of _Skl is completed, J, D, and T _f are calculated by the formula 1 (S308, S309), and then the estimation expressed by the formula 5 (the same as the formula B in the scope of the patent claim). The error ε is calculated (S310), and if this estimated error ε is larger than the threshold ε _max , the J, D, T _f obtained at that time is discarded (S311Yes, S312), and the estimated error ε is equal to or less than the threshold ε _max . Only in certain cases, J, D, T _f are adopted as the identification result (S311No, S313).

Note that steps S301 to S303 and S305 to S309 in FIG. 3 correspond to steps S201 to S203 and S205 to S209 of the first embodiment shown in FIG. 2, respectively.

In this second embodiment, the significance of adding the processing according to steps S311 to S313 is as follows.
For the actual torque τ, the square error f'per point when modeled by the mathematical formula 2 is expressed by the mathematical formula 6.

That is, the estimation error ε shown in the above equation 5 represents the squared error per point of the data used in the identification calculation. For example, if the motor speed v slightly exceeds the first threshold value v _min and then immediately decelerates to the first threshold value v _min or less, the parameters cannot be correctly identified even if the formula 1 is formally calculated. However, in that case, the square error f'(= ε) per point becomes a large value. Therefore, when the estimation error ε calculated by Equation 5 exceeds the predetermined threshold value ε _max , the identification result of each parameter at that time is invalidated and discarded to prevent the wrong identification result from being adopted. It is the one that was made.

Next, FIG. 4 is a flowchart showing a third embodiment of the parameter identification method according to the present invention. In addition, steps S401 to S409 of FIG. 4 correspond to steps S201 to S209 of the first embodiment shown in FIG. 2, respectively.
The part where this third embodiment is different from the first embodiment is that the value calculated by the mathematical formula 1 is used as the first identification result (S409) and the moving average value thereof is used as the second identification result (S410). Each parameter according to is adopted for control.

A specific example of the moving average processing according to step S410 will be described with reference to FIG.
As shown in FIG. 5A, in the moving average processing means 501 in which the first identification result in step S409 is input, the moving average score N can be set for each of the parameters J, D, and T _f . Further, the reset for the moving average value of each parameter J, D, T _f can be individually executed according to the reset command. For example, when driving a certain device for the first time, the moving average values of all parameters are reset and then started, and when only the moment of inertia J increases or decreases for the same device, the moving average value for the moment of inertia J You only have to reset it. The moving average score N for each parameter may be set based on the variation in the identification results.

FIG. 5B is a specific example of the moving average processing means 501, in which the storage position of the first identification result is sequentially moved and input to the ring buffer 501a, and the data of n points is input to the arithmetic averaging means 501b. Obtain the second identification result. The arithmetic mean means 501b outputs a result each time a new input is obtained, and then increases n by 1 until n = N (n = 1 at reset).

Next, FIG. 6 is a flowchart showing a fourth embodiment of the parameter identification method according to the present invention.
Steps S601 to S612 in FIG. 6 correspond to steps S301 to S312 of the second embodiment shown in FIG. 3, and steps S613 in FIG. 6 correspond to steps S410 of the third embodiment shown in FIG. There is.
That is, in this fourth embodiment, when the estimation error ε obtained by Equation 5 is larger than the threshold value ε _max , the identification result at that time is discarded as invalid (S611Yes, S612), and the estimation error ε is equal to or less than the threshold value ε _max. The identification result of the case is subjected to moving average processing as the first identification result, and the moving average value is output as the second identification result (S611No, S613), and as in the second embodiment, due to an erroneous identification result. It is possible to prevent the control from being disturbed.

Next, FIG. 7 is a flowchart showing a fifth embodiment of the parameter identification method according to the present invention.
Steps S701 to S709 of FIG. 7 correspond to steps S401 to S409 of the third embodiment shown in FIG. 4, respectively.
This fifth embodiment differs from the third embodiment in that the Coulomb friction T _f is individually subjected to moving average processing according to the polarity of the motor speed v, that is, the rotation direction of the motor, and the rotation direction is also performed in actual control. The point is to use the Coulomb friction T _f (third identification result, fourth identification result according to the forward and reverse of the rotation direction) identified by obtaining the moving average value according to the above. On the other hand, for the moment of inertia J and the viscous resistance D, the moving average process is performed without considering the polarity of the motor speed v (step S710 in FIG. 7).

In this fifth embodiment, when a term caused by gravity is added to the load torque, it appears as a constant torque regardless of the rotation direction, and when Coulomb friction is further added, the Coulomb apparently differs depending on the rotation direction. This is in case the friction seems to work.
The servo control performance can be improved by identifying the Coulomb friction T _f for each rotation direction of the motor as in this embodiment and performing friction compensation according to the rotation direction using the result.

Next, FIG. 8 is a flowchart showing a sixth embodiment of the parameter identification method according to the present invention.
Steps S801 to S812 of FIG. 8 correspond to steps S401 to S412 of the fourth embodiment shown in FIG. 6, and step S813 of FIG. 8 corresponds to step S710 of the fifth embodiment shown in FIG. ing.
According to the sixth embodiment, it is possible to prevent being disturbed by an erroneous identification result as in the fourth embodiment, and as in the fifth embodiment, friction compensation according to the rotation direction is performed. It is possible to improve the servo control performance.

Next, a seventh embodiment of the parameter identification method according to the present invention will be described with reference to FIG.
In the seventh embodiment, in a servo system in which the moment of inertia J can change each time the acceleration / deceleration operation is performed, the viscous resistance is obtained from the applied torque by using the viscous resistance coefficient D estimated by the previous operation and the Coulomb friction T _f. A provisional estimated value of the moment of inertia J is obtained during motor acceleration based on the value obtained by subtracting the coefficient D and the Coulomb friction T _f and the motor acceleration.

FIG. 9 is a timing chart of the torque τ and the motor speed v in this embodiment.
Here, the case where the pattern of the motor speed v is trapezoidal (trapezoidal acceleration / deceleration pattern) is illustrated, but the speed pattern is not limited to this shape.
In this case, apart from obtaining the provisional estimated value of the moment of inertia J, J, D, T are used using the period ΔT ₀ from when the motor speed v> v _min to when v ≦ v _min by the method described above. Identification of _f is also performed in parallel. Then, the provisional estimated value is obtained before the moment of inertia J is determined.

As the initial value of the provisional estimated value of the moment of inertia J, a typical value of the moment of inertia considered in actual operation may be given. When this typical value is unknown, for example, it may be the moment of inertia of a single motor, and when the operation and stop are repeated without changing the moment of inertia, the moment of inertia J identified last time is used as the provisional estimated value. It may be the initial value.

Ｄ，Ｔ_ｆが真値であれば、τ_ａｃｃは慣性×加速度に一致する。加速度ａの絶対値がａ_ｍｉｎを超える期間において、慣性モーメントＪの暫定推定値Ｊ_ｔｅｍｐを、例えば数式８のように更新する(αは１未満の正数)。

When acceleration is started at time t _{0 in} FIG. 9, the acceleration torque _τac is obtained by the formula 7. In this mathematical formula 7, D and T _f are the viscous drag coefficient and the Coulomb friction identified by the operation up to the previous time, respectively.

If D and T _f are true values, τ _acc corresponds to inertia × acceleration. During the period when the absolute value of the acceleration a exceeds a _min , the provisional estimated value _Jtemp of the moment of inertia J is updated as in Equation 8, for example (α is a positive number less than 1).

As long as the machine is operating normally, the viscosity resistance coefficient D and Coulomb friction T _f does not change rapidly, J deterministic D, with T _f becomes the v <v _min as the time t ₄ after Until this is done, if the J _emp of Equation 8 is applied to the control, the moment of inertia set to a value close to the true value can be used in a short time even if the moment of inertia changes significantly.
In FIG. 9, the period ΔT _{0 at} times t _{1 to} t ₄ is the “measuring” period during which the measurement related to parameter identification is performed as described above, and the periods ΔT _{1 at} times t _{1 to} t ₂ and The period ΔT _{2 at} times t _{3 to} t ₄ corresponds to, for example, the period in which steps S202 and S203 in FIG. 2 branch to the Yes side and the variable _Skl is updated.

100: Controlled object 101: Speed controller 102: Current source 103: Subtracting means 104: Model 105: Approximate friction characteristic 501: Moving average processing means 501a: Ring buffer 501b: Arithmetic mean means

Claims (8)

Targeting a servo system that controls the motion of a controlled object consisting of the motor and its load machine by energizing the motor with a current corresponding to a torque command.
The moment of inertia J and the viscous resistance coefficient of the controlled object are based on time-series data consisting of four signals of the motor acceleration, the motor speed, the code of the motor speed, and the applied torque when the predetermined conditions for updating are satisfied. In the method of identifying the parameters of the servo system for identifying D and the Coulomb friction T _f ,
The predetermined condition is that the absolute value of the motor speed is larger than the first threshold value and the absolute value of the motor acceleration is larger than the second threshold value.
Wherein at time t _i 4 signal when expressed as _{X k (t i) (k} = 1~4) each variable motor speed and motor acceleration at time t _i is updated only if it meets a predetermined condition _{S kl = Σ [X k (} t i) × X l (t i)] (k, l = 1~4) defines, while the absolute value of the motor speed is below the first threshold value the variable S Zero is given to _{kl, and} when the absolute value of the motor speed becomes larger than the first threshold value, the update of the variable _Skl is started, and then the motor decelerates and the absolute value of the motor speed is equal to or less than the first threshold value. Parameter identification of the servo system, which is characterized in that the update of the variable _Skl is completed at the time when becomes, and the inertial moment J, the viscous resistance coefficient D and the Coulomb friction T _f are simultaneously identified as the solution of the following mathematical formula A. Method.

In the servo system parameter identification method according to claim 1,
Immediately after solving the mathematical formula A to identify the moment of inertia J, the viscous resistance coefficient D, and the Coulomb friction T _f , the identification result and the cumulative count number SUM = Σ1 which is the number of updates of the S _kl and the S _kl are used. A method for identifying parameters of a servo system, which invalidates the identification result at that time when the estimation error ε obtained by the following mathematical formula B exceeds a predetermined threshold value ε _max .

In the servo system parameter identification method according to claim 1 or 2.
The identification results of the inertial moment J, the viscous resistance coefficient D, and the Coulomb friction T _f obtained as the solution of the mathematical formula A are defined as the first identification results, and the moving average value of the first identification results is defined as the second identification result. Along with the definition, every time the first identification result is updated as a valid value, a moving average calculation is executed to update the second identification result, and the inertial moment J, the viscous resistance coefficient D, and the viscosity resistance coefficient D, which are the second identification results, are updated. A method for identifying parameters of a servo system, which comprises using Coulomb friction T _f for control. In the servo system parameter identification method according to claim 3,
The moving average score and the reset timing of the moving average value for obtaining the second identification result are individually given for each of the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f . Parameter identification method. In the servo system parameter identification method according to claim 1 or 2.
The identification results of the moment of inertia J, the viscous resistance coefficient D, and the Coulomb friction T _f obtained as the solution of the mathematical formula A are defined as the first identification results, and the viscous resistance coefficient D and the Coulomb friction T _f are the first identification results. A second identification result obtained as a moving average value is defined, and every time the first identification result is updated as a valid value, a moving average calculation is executed to update the second identification result to control the servo system. Uses the second identification result and
For the Coulomb friction T _f , the third identification result and the fourth identification result obtained as moving average values for each velocity polarity are defined with respect to the first identification result, and the first identification result is updated as a valid value. Each time, the moving average calculation on the side corresponding to the velocity polarity is executed to update the third identification result or the fourth identification result, and the control of the servo system is performed by the third identification result or the third identification result according to the velocity polarity. A method for identifying parameters of a servo system, which comprises using the fourth identification result. In the servo system parameter identification method according to claim 5,
The moving average score and the reset timing of the moving average value for obtaining the third identification result or the fourth identification result are individually given for each of the moment of inertia J, the viscous drag coefficient D, and the Coulomb friction T _f. How to identify the parameters of the servo system. In the servo system parameter identification method according to any one of claims 1 to 6.
The servo system moment of inertia changes every performing acceleration and deceleration operation, applies the viscous resistance coefficient D and Coulomb friction T _f estimated by the operation up to the previous, the viscous resistance coefficient from applying the torque of the motor D and Coulomb friction T _f A provisional estimated value of the moment of inertia is obtained during motor acceleration based on the value obtained by subtracting the value obtained by subtracting the above and the motor acceleration. A method for identifying parameters of a servo system, which comprises using the provisional estimated values for controlling the servo system until the true values of D and Coulomb friction T _f are obtained. In the servo system parameter identification method according to any one of claims 1 to 7.
A method for identifying parameters of a servo system, which comprises using a floating-point arithmetic for the arithmetic of the variable _Skl .

Images