JP7585970B2 - Mechanical constant estimation device and motor control device - Google Patents

Description

The present invention relates to a mechanical constant estimation device that estimates the mechanical constants of a motor and its mechanical load, and a motor control device equipped with this mechanical constant estimation device.

In order to perform good speed control of a motor that drives a mechanical load, it is necessary to acquire the inertia of the motor and its mechanical load, the viscosity coefficient (viscous friction coefficient), Coulomb friction, and other load torques, and to reflect these in the speed control conditions.
Furthermore, even if a motor continues to drive the same mechanical load, the load torque may change. In such cases, it is convenient to be able to estimate mechanical constants such as inertia regardless of the operating conditions.

Here, Non-Patent Document 1 discloses a method for estimating inertia together with Coulomb friction and viscosity coefficient, on the premise that the load torque is represented by the sum of Coulomb friction of a sign function type independent of speed and a viscosity coefficient proportional to speed.
In this document, a positive-negative symmetrical periodic signal is given as the speed command, and as shown in FIG. 8 (FIG. 4 of Non-Patent Document 1), signals τ _e , q ₀ , q ₀ ^' , q ₁ based on the torque command u and angular velocity ω are multiplied together to obtain a value, which is integrated at periodic intervals using equations (37) to (44) described in the same document to determine elements φ ₁₁ , φ ₁₃ , φ ₂₂ , φ ₂₃ , φ ₃₃ of matrix Φ and elements v ₁ , v ₂ , v ₃ of vector V, and then a calculation Φ ^-1 V is performed to identify the machine constants including inertia.

Patent document 1 describes a motor control device that performs adaptive identification calculations on torque command differential values, motor acceleration, motor jerk, etc. through a low-pass filter with the same characteristics to identify inertia and viscosity coefficients. Furthermore, patent document 2 describes a mechanical constant identification device for a driving machine that inputs weighted speed, acceleration, and torque using a weighting signal that is updated according to the time after the driving machine is started into a least squares calculation unit to estimate inertia and viscosity coefficients.

JP 2006-217729 A ([0036] to [0059], FIG. 1, etc.) Japanese Patent No. 3683121 ([0092] to [0103], Figs. 7 and 8, etc.)

Ichiro Awaya et al., "Parameter Identification Method for Mechanical Motion Systems with Coulomb Friction," Transactions of the Japan Society of Mechanical Engineers (Series C), Vol. 59, No. 567 (1993), pp. 108-114

According to the method for identifying mechanical constants described in Non-Patent Document 1, for example, even when the motor is driven at a small acceleration at which it is difficult to obtain a desired estimation accuracy of the inertia, the estimated inertia value is updated regardless of the estimation accuracy. Alternatively, even when the motor is driven at a small maximum speed at which it is difficult to obtain a desired estimation accuracy of the viscosity coefficient, the estimated inertia value is updated regardless of the estimation accuracy.
In other words, even if a more reliable estimation result was obtained in the past, the estimated value may be updated to the latest less reliable estimation result. Furthermore, while it is desirable to estimate the machine constants as accurately as possible, there are also cases where it is desired to estimate the machine constants within a short period of time. However, with the method described in Non-Patent Document 1, it is difficult to achieve both improved estimation accuracy and reduced estimation time.

In addition, in the conventional technology disclosed in Patent Document 1, the inertia and viscosity coefficients are identified quickly and with high accuracy even when the motor rotation speed is low, while in the conventional technology disclosed in Patent Document 2, the weights are reduced during the period from when the drive machine is started until a predetermined time has elapsed, thereby reducing the estimation error caused by Coulomb friction.
However, Patent Documents 1 and 2 do not specifically disclose a means for estimating the Coulomb friction.

The problem to be solved by the present invention is to provide a mechanical constant estimation device that can estimate inertia, viscosity coefficient, and Coulomb friction with high accuracy and in a short time in a manner that is unlikely to lead to deterioration of the estimated values even when the motor speed or acceleration is inappropriate when estimating the mechanical constants, and a motor control device that uses the mechanical constants estimated by this estimation device to control the motor speed, etc.

In order to solve the above problem, a mechanical constant estimation device according to claim 1 is a mechanical constant estimation device that estimates mechanical constants consisting of inertia, viscosity coefficient, and Coulomb friction of a motor and a mechanical load driven by the motor, the mechanical constant estimation device comprising:
an integrating means for starting a numerical integration using the product of any two of the torque, acceleration, speed, and speed sign of the motor as an integrand function when the absolute value of the speed of the motor exceeds a predetermined estimation start speed after the motor is started, and for ending the numerical integration when the absolute value of the speed of the motor becomes equal to or less than the estimation start speed as a result of the deceleration of the motor;
an estimated value calculation means for calculating an estimated inertia value _J1 , an estimated viscosity coefficient value _D1 , and an estimated Coulomb friction value _Tf1 during an acceleration/deceleration period of the motor based on the integrated value by the integrating means;
a means for calculating weights W _J1 , W _D1 , and W _Tf1 corresponding to the estimated inertia value J ₁ , the estimated viscosity coefficient value D ₁ , and the estimated Coulomb friction value T _f1, respectively;
an inertia estimation value averaging means for updating an inertia estimation average value _J2 and an integrated weight _WJ2 calculated before the acceleration/deceleration period using the inertia estimation value _J1 and the weight _WJ1 during the acceleration/deceleration period, and outputting the updated inertia estimation average value _J2 as an inertia estimation result;
a viscosity coefficient estimated value averaging means for updating a viscosity coefficient estimated average value _D2 and an integrated weighting _W2 calculated before the acceleration/deceleration period using the viscosity coefficient estimated value _D1 and the weighting W _D1 during the acceleration/deceleration period, and outputting the updated viscosity coefficient estimated average value _D2 as a viscosity coefficient estimated result;
a Coulomb friction estimated value averaging means for updating a Coulomb friction estimated average value _Tf2 and an integrated weight _WTf2 calculated before the acceleration/deceleration period using the Coulomb friction estimated value _Tf1 and the weight _WTf1 during the acceleration/deceleration period, and outputting the updated Coulomb friction estimated average value _Tf2 as a Coulomb friction estimation result;
The present invention is characterized by comprising:

A mechanical constant estimation device according to claim 2 is the mechanical constant estimation device according to claim 1,
The estimation value calculation means
When the acceleration a(t), velocity v(t), velocity sign(v(t)), and torque T(t) during an acceleration/deceleration period t of the motor are denoted by _x1 (t), _x2 ( _t ), _x3 (t), and y(t), respectively, the _inertia estimated value J1, _viscosity coefficient estimated value D1, and Coulomb friction estimated value Tf1 are calculated according to Equation 1 described in the claims using m _ij =∫{x _i (t) x _j (t)}dt and q _i =∫{x _i (t) y(t)}dt (where i, j = 1 to 3) where the product of any two of x _{1 (t)} , x 2 (t ₎ , x 3 (t), and y(t ₎ is used as an integrand function.

The mechanical constant estimation device according to claim 3 is the mechanical constant estimation device according to claim 1 or 2, characterized in that the torque, acceleration, and speed used to calculate the integrand function are values obtained after calculation through a low-pass filter with the same time constant.

The mechanical constant estimation device of claim 4 is the mechanical constant estimation device according to claim 1 or 2, characterized in that the weight _WJ1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the acceleration of the motor, the weight _WD1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the speed of the motor, and the weight _WTf1 is a value obtained by numerically integrating a value that is independent of either the speed or the acceleration of the motor.

The mechanical constant estimation device of claim 5 is the mechanical constant estimation device described in claim 3, characterized in that the weight _WJ1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the acceleration of the motor after calculation by the low-pass filter, the weight _WD1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the speed of the motor after calculation by the low-pass filter, and the weight _WTf1 is a value obtained by numerically integrating a value that is independent of either the speed or the acceleration of the motor.

According to a sixth aspect of the present invention, there is provided a mechanical constant estimation device according to the fourth or fifth aspect, wherein, each time the inertia estimated value _J1 and the weight _WJ1 are newly obtained, the inertia estimated value averaging means sets a current value of the integrated weight _WJ2 to a sum of a previous value of the integrated weight _WJ2 and a current value of the weight _WJ1 limited by a predetermined upper limit value _WJmax , and then:
Current value of inertia estimated average value _J2 =Previous value of inertia estimated average value _J2 +(current value of weight _WJ1 /current value of integrated weight _WJ2 )×(current value of inertia estimated value _J1 −previous value of inertia estimated average value _J2 )
The present invention is characterized in that it calculates the following:

The mechanical constant estimation device according to claim 7 is the mechanical constant estimation device according to any one of claims 4 to 6, wherein, each time the viscosity coefficient estimated value _D1 and the weight W _D1 are newly obtained, the viscosity coefficient estimated value averaging means sets a value obtained by limiting the sum of the previous value of the integrated weight W _D2 and the current value of the weight W _D1 by a predetermined upper limit value W _Dmax as the current value of the integrated weight W _D2 , and then:
Current value of the estimated viscosity coefficient average value _D2 = Previous value of the estimated viscosity coefficient average value _D2 + (Current value of the weighting W _D1 / Current value of the integrated weighting W _D2 ) × (Current value of the estimated viscosity coefficient value _D1 - Previous value of the estimated viscosity coefficient average value _D2 )
The present invention is characterized in that it calculates the following:

The mechanical constant estimation device according to an eighth aspect of the present invention is the mechanical constant estimation device according to any one of the fourth to seventh aspects, wherein, each time the Coulomb friction estimated value _Tf1 and the weight W _Tf1 are newly obtained, the Coulomb friction estimated value averaging means sets a value obtained by limiting the sum of the previous value of the integrated weight W _Tf2 and the current value of the weight W _Tf1 by a predetermined upper limit value W _Tfmax as the current value of the integrated weight W _Tf2 , and then:
Current value of Coulomb friction estimated average value _Tf2 =Previous value of Coulomb friction estimated average value _Tf2 +(Current value of weight W _Tf1 /Current value of integrated weight W _Tf2 )×(Current value of Coulomb friction estimated value _Tf1 −Previous value of Coulomb friction estimated average value _Tf2 )
The present invention is characterized in that it calculates the following:

The mechanical constant estimation device of claim 9 is the mechanical constant estimation device according to any one of claims 1 to 8, characterized in that the inertia estimation average value _J2 , the viscosity coefficient estimation average value _D2 , the Coulomb friction estimation average value _Tf2 , and the integrated weights _WJ2 , _WD2 , and _WTf2 are stored in a non-volatile memory and can be initialized externally.

The motor control device according to claim 10 is characterized in that it controls the motor using the mechanical constants estimated by the mechanical constant estimation device according to any one of claims 1 to 9.

According to the present invention, even when it is difficult to properly estimate the mechanical constants, for example because the motor speed or acceleration is small, it is possible to estimate the mechanical constants with high accuracy and in a short time in a manner that is unlikely to lead to deterioration of the estimated value.

1 is a block diagram showing an overview of a mechanical constant estimation device according to an embodiment of the present invention. FIG. 2 is a configuration diagram showing a first embodiment of a mechanical constant estimating unit in FIG. 1 . FIG. 4 is a configuration diagram of an integration start/end determination unit in each embodiment of the machine constant estimation unit. FIG. 2 is a configuration diagram showing a second embodiment of the mechanical constant estimating unit in FIG. 1 . FIG. 4 is a configuration diagram of an estimated inertia value averaging means in each embodiment of the mechanical constant estimating unit. FIG. 4 is a configuration diagram of a viscosity coefficient estimated value averaging means in each embodiment of the mechanical constant estimating unit. FIG. 13 is a configuration diagram of a Coulomb friction estimated value averaging means in each embodiment of the mechanical constant estimating unit. FIG. 1 is an explanatory diagram of the conventional technology described in Non-Patent Document 1.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
1 is a block diagram showing an outline of a mechanical constant estimation device according to the present embodiment. This mechanical constant estimation device estimates the inertia, viscosity coefficient, and Coulomb friction of a motor/mechanical load 200 connected to the motor (collectively referred to as a motor/mechanical load 200) based on the speed and torque of the motor.

1, a speed control unit 100 outputs torque so that the motor speed obtained from a motor/mechanical load 200 coincides with a speed command. Here, the torque may be a torque command or a torque estimate obtained by detecting the current flowing through the motor.
The motor speed and torque are input to a mechanical constant estimation unit 300. The mechanical constant estimation unit 300 estimates the inertia, viscosity coefficient and Coulomb friction by an operation described later, and performs weighting processing to calculate average values _J2 , _D2 and _Tf2 thereof.
The speed control unit 100 and the mechanical constant estimation unit 300 are realized by an arithmetic processing device such as a microcomputer and a program therefor.

FIG. 2 is a configuration diagram showing a first embodiment of the mechanical constant estimating section 300 (mechanical constant estimating section 300A).
2, the motor speed obtained from the motor/mechanical load 200 is provided to integration start/end determination means 320, differentiation means 301, and sign determination means 302 in a mechanical constant estimation section 300A, and the motor torque is provided to multiplication means 303. In addition, an external reset command is input to inertia estimated value averaging means 340, viscosity coefficient estimated value averaging means 360, and Coulomb friction estimated value averaging means 380, which will be described later.

Now, let the motor acceleration be a(t), the motor speed be v(t), the sign of the motor speed be sign(v(t)), and the torque be T(t), and let these be _x1 (t), _x2 (t), _x3 (t), and y(t), respectively. The motor acceleration _x1 (t), motor speed _x2 (t), which are the outputs of the differentiation means 301, the speed sign _x3 (t), which is the output of the sign determination means 302, and the torque y(t) are input to the multiplication means 303.

The multiplication means 303 calculates x _i (t) x _j (t) and x i (t) y (t) (i, _j = 1 to 3) using two parameters from among x ₁ (t), x ₂ (t), x ₃ (t), and y (t). The calculations in the multiplication means 303 also include the calculations of x ₁ (t) ² , x ₂ (t) ² , and x ₃ (t) ² (i.e., i = j = 1, i = j = 2, i = j = 3).
The results of the multiplication by the multiplication means 303 are input to the integration means 304 at the next stage.

The integration means 304 performs numerical integration of m _ij = ∫{x _i (t) x _j (t)} dt and q _i = ∫{x _i (t) y(t)} dt over a certain period of time and sends these calculation results to the estimate calculation means 311 .
In actual numerical integration, for example,
m _ij =Σ{x _i (n)x _j (n)}, q _i =Σ{x _i (n)y(n)}
The above calculation is performed to define m _ij and q _i , and when the n-th data point is obtained,
m _ij ←m _ij +x _i (n)x _j (n),
q _i ←q _i +x _i (n)y(n)
Then, m _ij and q _i are updated as follows.

FIG. 3 is a block diagram of the integration start/end determination means 320 in FIG.
In FIG. 3, absolute value calculation means 321 calculates the absolute value of the motor speed and sends it to comparison means 322, which compares the speed absolute value with a preset estimated start speed.
When the absolute velocity value exceeds the estimation start velocity, the comparison means 322 sends an integration start instruction to the integration means 304 in Fig. 2 via the instruction generation means 323, and when the absolute velocity value becomes equal to or less than the estimation start velocity, the comparison means 322 sends an integration end instruction to the integration means 304 via the instruction generation means 324. The period from the integration start instruction to the integration end instruction is the integration period. The instruction generation means 324 also outputs instructions to update the inertia estimated value _J1 , viscosity coefficient estimated value _D1 , and Coulomb friction estimated value _Tf1 to the estimate value calculation means 311 in Fig. 2, and instructions to update the weights _WJ1 , _WD1 , and _WTf1 to the inertia estimated value averaging means 340, viscosity coefficient estimated value averaging means 360, and Coulomb friction estimated value averaging means 380.

The reason why the integration start/end determination means 320 issues the integration start instruction, integration end instruction, and update instruction at the above-mentioned times is as follows.
When a motor drives a mechanical load, a load torque accompanied by hysteresis often occurs near zero speed. In order for the estimated value calculation means 311 to correctly estimate the mechanical constant, it is necessary to exclude the above-mentioned hysteresis region from the period during which the integrating means 304 performs integration. Therefore, the integration period is set so that integration is started when the absolute speed value exceeds a predetermined estimation start speed after the motor starts, and integration is stopped when the absolute speed value becomes equal to or lower than the estimation start speed as the motor decelerates, and an instruction to update each parameter is issued to determine the estimated inertia value _J1 , estimated viscosity coefficient value _D1 , estimated Coulomb friction value _Tf1 , and weights _WJ1 , _WD1 , and _WTf1 during the integration period.
In this case, by making the absolute motor speed values at the integration end instruction time point and the integration start instruction time point equal (i.e., using the same estimation start speed), it is possible to reduce errors in each estimated value when the load torque is nonlinear.

If there is a problem with using the speed signal, its differential acceleration signal, and torque signal as they are for calculation due to detection errors in the motor speed or current flow, it is better to use the results of calculations performed by low-pass filters 312 and 313 on the motor speed and torque, as in the mechanical constant estimation unit 300B of the second embodiment shown in Figure 4. In this case, it is desirable to make the time constants of low-pass filters 312 and 313 equal.

Returning to FIG. 2, the estimate calculation means 311 uses m _ij , q _i (i, j = 1 to 3) calculated by the integration means 304 to calculate the inertia estimate value J ₁ , viscosity coefficient estimate value D ₁ , and Coulomb friction estimate value T _f1 during one acceleration/deceleration period of the motor according to Equation 1.

This formula 1 corresponds to finding the parameters _J1 , D1, and Tf1 for one acceleration/deceleration period by the least squares method for a system whose equation of motion is expressed as T( _t ) = J(dv/dt ₎ + Dv + _Tf sign(v).

The above will be specifically explained. In the least squares method,
Integral value I=∫{T(t)-J(dv/dt)-Dv-T _f sign(v)} ² dt
To this end, the integral value I is partially differentiated with respect to J, D, and _{Tf ,} respectively, and then rearranged to obtain Equation 2.

数式３における積分演算∫ｄｔを積算演算Σに置き換えると、数式４が得られる。

By substituting the above-mentioned x ₁ (t), x ₂ (t), x ₃ (t), and y(t) into the above-mentioned equation 2, equation 3 is obtained.

By replacing the integral operation ∫dt in Equation 3 with the multiplication operation Σ, Equation 4 is obtained.

この数式５を変形すると数式６が得られ、数式６におけるＪをＪ_１、ＤをＤ_１、Ｔ_ｆをＴ_ｆ１とおけば、前述の数式１によって慣性推定値Ｊ_１、粘性係数推定値Ｄ_１、及びクーロン摩擦推定値Ｔ_ｆ１を求めることができる。

Furthermore, by replacing the n-th data x _i (n), x _j (n), and y(n) of the above-mentioned m _ij = Σ{x _i (n) x j (n)}, q _i = Σ{x _i (n ₎ y (n)} with the data x _i (t), x _j (t), and y(t) at time t, respectively, we obtain m _ij = Σ{x _i (t) x _j (t)}, q _i = Σ{x _i (t) y (t)}, and therefore Equation 4 can be expressed as Equation 5.

By transforming this equation 5, we obtain equation 6. If J in equation 6 is replaced by _J1 , D by _D1 , and _Tf by _Tf1 , the estimated inertia value _J1 , the estimated viscosity coefficient value _D1 , and the estimated Coulomb friction value _Tf1 can be calculated using equation 1 described above.

2, the motor acceleration _x1 (t) is input to a first integrand generating means 305, which generates an integrand _f1 that increases monotonically with the absolute value of _x1 (t), and the motor speed _x2 (t) is input to a second integrand generating means 307, which generates an integrand _f2 that increases monotonically with the absolute value of _x2 (t). Also, a third integrand generating means 309 generates an integrand _f3 that does not depend on the motor acceleration or speed. Here, for example, _f1 may be the square of _x1 (t), _f2 may be the square of _x2 (t), and _f3 may simply be 1.

The integrands _f1 , _f2 , and _f3 are integrated by integrating means 306, 308, and 310, respectively, and the results are input as weights _WJ1 , _WD1 , and _WTf1 to inertia estimated value averaging means 340, viscosity coefficient estimated value averaging means 360, and Coulomb friction estimated value averaging means 380, respectively.
The inertia estimation value averaging means 340, the viscosity coefficient estimation value averaging means 360, and the Coulomb friction estimation value averaging means 380 perform weighting calculations using weights W _J1 , W _D1 , and W _Tf1 as follows on the inertia estimation value J ₁ , the viscosity coefficient estimation value D ₁ , and the Coulomb friction estimation value T _f1 sent from the estimation value calculation means 311 to calculate the respective estimated average values J ₂ , D ₂ , and T _f2 .

The configurations and operations of the inertia estimated value averaging means 340, the viscosity coefficient estimated value averaging means 360, and the Coulomb friction estimated value averaging means 380 will be described below.
For example, when the inertia is small and the acceleration is also small, it is considered that the inertia cannot be estimated with sufficient accuracy by only one acceleration/deceleration operation, even if the estimation calculation is performed by the above-mentioned estimated value calculation means 311. Also, when an acceleration/deceleration operation with a small maximum speed is performed, it is considered that the viscosity coefficient cannot be estimated with sufficient accuracy by only one acceleration/deceleration operation, and further, when the acceleration/deceleration operation period is very short, it is considered that the Coulomb friction cannot be estimated with sufficient accuracy by only that one acceleration/deceleration operation.

When the same pattern of acceleration/deceleration operation is repeated, the estimation accuracy can be improved by a method such as applying a simple moving average to each estimated value obtained for each acceleration/deceleration operation. However, when it is desired to improve the estimation accuracy of each machine constant in a situation where the operation is not repeated, the reliability of each estimation differs for each acceleration/deceleration operation.
In this embodiment, therefore, the estimation accuracy is improved by the following operations performed by the inertia estimated value averaging means 340, the viscosity coefficient estimated value averaging means 360, and the Coulomb friction estimated value averaging means 380.

First, the inertia estimate averaging means 340 will be described with reference to FIG.
It is assumed that the switching means 344, 348 are in the state shown in the figure until a reset command is input. This also applies to switching means 364, 368 in Fig. 6 and switching means 384, 388 in Fig. 7, which will be described later.

5, the inertia estimation value averaging means 340 operates to update the inertia estimation average value _J2 and the weight _WJ2 every time a new inertia estimation value _J1 and weight _WJ1 are obtained during one acceleration/deceleration period. Note that the weight _WJ2 is always calculated as the sum of the current value (latest value) and the previous value of the weight _WJ1 , and therefore will be referred to as the "accumulated weight _WJ2 " below.
5, the previous value of the integrated weight _WJ2 held by the previous value holding means 343 and the current value (latest value) of the weight _WJ1 are added by the addition/subtraction means 341, and the result is limited by a preset upper limit value _WJmax to obtain the current value of the integrated weight _WJ2 . Then, the current value of the weight _WJ1 is divided by the current value of the integrated weight _WJ2 by the division means 342, and the result is input to the multiplication means 346.

Also, the deviation between the previous value of the inertia estimated average value _J2 held by the previous value holding means 349 and the current value of the inertia estimated average value _J1 is found by the addition/subtraction means 345, and this deviation is input to the multiplication means 346 where it is multiplied by the output of the division means 342. Furthermore, the output of the multiplication means 346 and the previous value of the inertia estimated average value _J2 are added by the addition/subtraction means 347, and the result is output as the current value of the inertia estimated average value _J2 .

The above operations can be expressed by the following equations.
_J2 current value = _J2 previous value + ( _WJ1 current value / _WJ2 current value) × ( _J1 current value - _J2 previous value)
Incidentally, the integrated weight _WJ2 and the estimated inertia average value _J2 can be initialized by switching the switching means 344, 348 to the "0" side in response to an external reset command, thereby essentially removing the loop including the previous value holding means 343, 349, etc. Specifically, for example, when the mechanical load is changed, the switching means 344, 348 is switched to the "0" side to initialize _WJ2 , _J2 , and thereafter _WJ2 , _J2 are updated while being stored in storage means such as a non-volatile memory. Also, when the system is restarted, the values of _WJ2 , _J2 are read from the storage means prior to operation after the restart.

The method of updating the estimated inertia average value _J2 will now be described in more detail.
First, if the value obtained by adding the previous value _{of W_J2} and the current value of _{W_J1} by the addition/subtraction means 341 remains below the upper limit value _{W_Jmax} ,
_J2 current value = _J2 previous value + {W _J1 current value / (W _J2 previous value + W _J1 current value)} × ( _J1 current value - _J2 previous value)
= ( _J2 previous value × _WJ2 previous value + _J1 current value × _WJ1 current value) / ( _WJ2 previous value + _WJ1 current value)
It becomes.
This is equivalent to finding a weighted average value using data from all past acceleration/deceleration driving. In the case of an acceleration/deceleration driving pattern in which it is difficult to obtain high inertia estimation accuracy (for example, in the case of driving with low acceleration), _x1 (t) in FIG. 2 becomes smaller and the weight _WJ1 is suppressed to a small value, so that a highly reliable inertia estimation average value _J2 can be obtained.

However, if operation is continued without placing any limit on the value obtained by adding the previous value of _WJ2 and the current value of _WJ1 , _WJ2 will eventually reach infinity and the output of the division means 342 will approach zero, so that the inertia estimated average value _J2 will no longer be updated even if a new acceleration/deceleration operation is performed.
Therefore, as shown in FIG. 5, the sum of the previous value of _WJ2 and the current value of _WJ1 is limited to an upper limit value _WJmax or less, so that the latest driving result is reflected in the inertia estimated average value _J2 as the acceleration/deceleration driving continues.

Here, without defining weights,
_J2 current value = _J2 previous value + constant x ( _J1 current value - _J2 previous value) (0 < constant < 1)
Let us consider the case where the _J2 current value is calculated by the following equation.
Even when the calculation is performed in this way, the accuracy of the inertia estimation increases as the number of acceleration and deceleration operations increases. However, in this case, the _J2 current value will be smaller than the true value until the number of acceleration and deceleration operations reaches about (3/constant). Also, when low acceleration or short-term acceleration and deceleration operation that is not suitable for inertia estimation is performed, the inertia estimation value will be updated with the result as with other driving patterns.

In contrast, in this embodiment, for example, by initializing the values of _WJ2 and _J2 by a reset command when the mechanical load is changed, an inertia estimation value close to the true value can be obtained even if the number of acceleration/deceleration operations is small, and a highly accurate inertia estimation average value _J2 can be obtained in a short time.
In addition, the weight _WJ1 is calculated by numerically integrating _f1, which increases monotonically with respect to the absolute value of the acceleration _x1 (t), and the estimated inertia value _J1 is updated using this weight _WJ1 and an integrated weight _WJ2 . Therefore, the estimated inertia average value _J2 is less likely to be disturbed even when the vehicle is driven in a manner that is not suitable for inertia estimation.

Next, FIG. 6 shows the configuration of the viscosity coefficient estimated value averaging means 360.
This viscosity coefficient estimated value averaging means 360 operates to update the viscosity coefficient estimated average value _D2 and the integrated weight W _D2 every time a new viscosity coefficient estimated value _D1 and weight W _D1 are obtained during one acceleration/deceleration period.

6, the viscosity coefficient estimated value averaging means 360 includes addition/subtraction means 361, 365, and 367, division means 362, previous value holding means 363 and 369, switching means 364 and 368, multiplication means 366, and W _Dmax , which is an upper limit set for the sum of the previous value W _D2 and the current value W _D1 . The overall operation of the viscosity coefficient estimated value averaging means 360 is similar to that of the inertia estimated value averaging means 340, except for the input and output signals.
In estimating the viscosity coefficient, the viscosity coefficient estimated value _D1 is updated using the weight W _D1 and the integrated weight W _D2 obtained by numerically integrating f ₂ , which monotonically increases with respect to the absolute value of the speed x ₂ (t), to obtain the viscosity coefficient estimated average value D _2. Therefore, the viscosity coefficient estimated average value D ₂ is less likely to be disturbed even for driving patterns that are not suitable for estimating the viscosity coefficient, such as driving with a small maximum speed.

FIG. 7 also shows the configuration of the Coulomb friction estimated value averaging means 380 .
The Coulomb friction estimated value averaging means 380 operates to update the Coulomb friction estimated average value T _f2 and the integrated weight W _Tf2 every time a new Coulomb friction estimated value T _f1 and weight W _Tf1 are obtained during one acceleration/deceleration period.

7, Coulomb friction estimated value averaging means 380 includes addition/subtraction means 381, 385, and 387, division means 382, previous value holding means 383 and 389, switching means 384 and 388, multiplication means 386, and W _Tfmax is an upper limit set for the addition result of the previous value W _Tf2 and the current value W _Tf1 . The overall operation of this Coulomb friction estimated value averaging means 380 is similar to the inertia estimated value averaging means 340 and viscosity coefficient estimated value averaging means 360 described above, except for input and output signals.
In estimating the Coulomb friction, the Coulomb friction estimated value Tf1 is updated using the weight W _Tf1 and the integrated weight W _Tf2 obtained by numerically integrating the value _f3 that is independent of either the acceleration x ₁ (t) or the speed x ₂ ( _t) , and the Coulomb friction estimated average value _Tf2 is calculated in this way. Therefore, even if driving that is not suitable for estimating the Coulomb friction is performed, the Coulomb friction estimated average value _Tf2 is less likely to be disturbed.

As described above, according to the mechanical constant estimation device of this embodiment, even if the operating pattern is such that the motor speed, acceleration, or acceleration/deceleration operating period is inappropriate for estimating the mechanical constants, the mechanical constants can be estimated with high accuracy and in a short time in a manner that is unlikely to lead to deterioration of the estimated values.
Furthermore, by implementing this mechanical constant estimation device in a motor control device and reflecting the mechanical constants estimated for each acceleration/deceleration period in accordance with the mechanical load in the speed control and torque control of the motor, it is possible to reduce tracking errors and vibrations in the mechanical drive system.

100: Speed control section 200: Motor/machine load 300, 300A, 300B: Machine constant estimation section 301: Differentiation means 302: Sign determination means 303: Multiplication means 304, 306, 308, 310: Integration means 305, 307, 309: Integral value generation means 311: Estimated value calculation means 312, 313: Low-pass filter 320: Integration start/end determination means 321: Absolute value calculation means 322: Comparison means 323, 324: Instruction generation means 340: Inertia estimated value averaging means 341, 345, 347: Addition/subtraction means 342: Division means 343, 349: Previous value holding means 346: Multiplication means 344, 348: Switching means 360: Viscosity coefficient estimated value averaging means 361, 365, 367: Addition/subtraction means 362: Division means 363, 369: Previous value holding means 366: Multiplication means 364, 368: Switching means 380: Coulomb friction estimated value averaging means 381, 385, 387: Addition/subtraction means 382: Division means 383, 389: Previous value holding means 386: Multiplication means 384, 388: Switching means

Claims (10)

A mechanical constant estimation device that estimates mechanical constants, which are composed of inertia, viscosity coefficient, and Coulomb friction, of a motor and a mechanical load driven by the motor, comprising:
an integrating means for starting a numerical integration using the product of any two of the torque, acceleration, speed, and speed sign of the motor as an integrand function when the absolute value of the speed of the motor exceeds a predetermined estimation start speed after the motor is started, and for ending the numerical integration when the absolute value of the speed of the motor becomes equal to or less than the estimation start speed as a result of the deceleration of the motor;
an estimated value calculation means for calculating an estimated inertia value _J1 , an estimated viscosity coefficient value _D1 , and an estimated Coulomb friction value _Tf1 during an acceleration/deceleration period of the motor based on the integrated value by the integrating means;
a means for calculating weights W _J1 , W _D1 , and W _Tf1 corresponding to the estimated inertia value J ₁ , the estimated viscosity coefficient value D ₁ , and the estimated Coulomb friction value T _f1, respectively;
an inertia estimation value averaging means for updating an inertia estimation average value _J2 and an integrated weight _WJ2 calculated before the acceleration/deceleration period using the inertia estimation value _J1 and the weight _WJ1 during the acceleration/deceleration period, and outputting the updated inertia estimation average value _J2 as an inertia estimation result;
a viscosity coefficient estimated value averaging means for updating a viscosity coefficient estimated average value _D2 and an integrated weighting _W2 calculated before the acceleration/deceleration period using the viscosity coefficient estimated value _D1 and the weighting W _D1 during the acceleration/deceleration period, and outputting the updated viscosity coefficient estimated average value _D2 as a viscosity coefficient estimated result;
a Coulomb friction estimated value averaging means for updating a Coulomb friction estimated average value _Tf2 and an integrated weight _WTf2 calculated before the acceleration/deceleration period using the Coulomb friction estimated value _Tf1 and the weight _WTf1 during the acceleration/deceleration period, and outputting the updated Coulomb friction estimated average value _Tf2 as a Coulomb friction estimation result;
A mechanical constant estimation device comprising: 2. The mechanical constant estimation device according to claim 1,
The estimated value calculation means
A mechanical constant estimation device characterized in that, when the acceleration a(t), velocity v(t), _velocity sign(v(t)), and torque T(t) during an acceleration/deceleration period t of the motor are denoted as _x1 (t), _x2 (t), _x3 (t), and y(t), respectively, an _inertia estimated value J1, a viscosity coefficient estimated value D1, and a Coulomb friction estimated value Tf1 _are calculated by the following Formula 1 using m _ij = ∫{x _i (t) x _j (t)} dt and q _i = ∫{x _i (t) y(t)} dt (where i, j = 1 to 3) where the product of any _two of x _{1 (} t), x 2 (t _{), x 3 (t)} , and y(t) is an integrand function:

3. The mechanical constant estimation device according to claim 1,
A mechanical constant estimation device, characterized in that the torque, acceleration, and speed used to find the integrand function are values obtained after being calculated through a low-pass filter having the same time constant. 3. The mechanical constant estimation device according to claim 1,
A mechanical constant estimation device characterized in that the weight _WJ1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the acceleration of the motor, the weight _WD1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the speed of the motor, and the weight _WTf1 is a value obtained by numerically integrating a value that is independent of either the speed or the acceleration of the motor. 4. The mechanical constant estimation device according to claim 3,
a mechanical constant estimation device characterized in that the weight _WJ1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the acceleration of the motor after calculation by the low-pass filter, the weight _WD1 is a value obtained by numerically integrating a value that monotonically increases with respect to the absolute value of the speed of the motor after calculation by the low-pass filter, and the weight _WTf1 is a value obtained by numerically integrating a value that is independent of either the speed or the acceleration of the motor. 6. The mechanical constant estimation device according to claim 4,
The inertia estimate averaging means
Every time the inertia estimated value _J1 and the weight _WJ1 are newly obtained, the sum of the previous value of the integrated weight _WJ2 and the current value of the weight _WJ1 is limited by a predetermined upper limit value _WJmax, and the sum is set as the current value of the integrated weight _WJ2 .
Current value of inertia estimated average value _J2 =Previous value of inertia estimated average value _J2 +(Current value of weight _WJ1 /Current value of integrated weight _WJ2 )×(Current value of inertia estimated value _J1 −Previous value of inertia estimated average value _J2 )
A mechanical constant estimation device comprising: In the machine constant estimation device according to any one of claims 4 to 6,
The viscosity coefficient estimated value averaging means
Every time the viscosity coefficient estimated value _D1 and the weight W _D1 are newly obtained, the sum of the previous value of the integrated weight W _D2 and the current value of the weight W _D1 is limited by a predetermined upper limit value W _Dmax , and the sum is set as the current value of the integrated weight W _D2 .
Current value of the estimated viscosity coefficient average value _D2 = Previous value of the estimated viscosity coefficient average value _D2 + (Current value of the weighting W _D1 / Current value of the integrated weighting W _D2 ) × (Current value of the estimated viscosity coefficient value _D1 - Previous value of the estimated viscosity coefficient average value _D2 )
A mechanical constant estimation device comprising: In the mechanical constant estimation device according to any one of claims 4 to 7,
The Coulomb friction estimate averaging means
Every time the Coulomb friction estimated value _Tf1 and the weight W _Tf1 are newly obtained, the sum of the previous value of the integrated weight W _Tf2 and the current value of the weight W _Tf1 is limited by a predetermined upper limit value W _Tfmax , and the sum is set as the current value of the integrated weight W _Tf2 .
Current value of Coulomb friction estimated average value _Tf2 =Previous value of Coulomb friction estimated average value _Tf2 +(Current value of weight W _Tf1 /Current value of integrated weight W _Tf2 )×(Current value of Coulomb friction estimated value _Tf1 −Previous value of Coulomb friction estimated average value _Tf2 )
A mechanical constant estimation device comprising: In the machine constant estimation device according to any one of claims 1 to 8,
A mechanical constant estimation device characterized in that the inertia estimation average value _J2 , the viscosity coefficient estimation average value _D2 , the Coulomb friction estimation average value _Tf2 , and the integrated weights _WJ2 , _WD2 , and _WTf2 are stored in a non-volatile memory and can be initialized externally. A motor control device that controls the motor using the mechanical constants estimated by the mechanical constant estimation device according to any one of claims 1 to 9.

Images