JP3019359B2 - Detection method of instantaneous rotation speed of motor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a digital using a microprocessor which supplies electric power to a load such as an electric motor via a static electric power converter and controls the torque and the rotation speed of the electric motor. The present invention relates to a suitable method for detecting an instantaneous rotational speed of a motor used in obtaining an instantaneous speed of the electric motor in a control device.

[Conventional technology]

One of the encoders as a typical example of a rotary position detector of a servomotor is an incremental encoder.
In the case of an optical type, for example, a rotating disk having a slit through which light passes through a plate surface is arranged between a light emitting element and a light receiving element, and the rotating disk rotates the motor. As it rotates with, light passes through where there is a slit,
Light is cut off at places where there is no light, and the number of slits passing by rotation is simply counted based on the number of times the light passes and is cut off.

As a method for detecting the rotation speed of the electric motor using such an incremental encoder, there is the following conventional method.

(B) A method of integrating and counting input pulses representing the result of counting the number of slits coming from an encoder within a certain measurement time. (Ii) A reference pulse is generated at a constant frequency of a high frequency generated while a predetermined number of input pulses are counted. Counting the number and calculating the reciprocal of the counted number of reference pulses (c) The number (Q) of input pulses from the encoder arriving within a certain time (T) and the arrival of input pulses immediately before the lapse of a certain time Counting the number (P) of high-frequency fixed-cycle reference pulses generated between the time and a certain time;
At regular time intervals, the number of reference pulses (P) from the previous count value Pn- ₁ and the current count value Pn, and the number of input pulses (Q) from the current count value Qn A method of detecting the speed by an arithmetic operation (see Japanese Patent Application Laid-Open No.
(D) While a pulse from the encoder arrives within a certain time (T), the speed is detected by a method based on the same principle as the method (c), and within a certain time (T) Since the speed cannot be detected by the method (c) at a low speed at which the pulse from the encoder does not arrive, the speed is estimated and detected by the following method.

That is, as a speed detection method at a low speed, the load torque is estimated every time an encoder pulse is generated. That is, at the time when the (i + 1) th encoder pulse is generated, n1 (i): Average value of motor speed between the generation points of the i-th (i + 1) th encoder pulse 1 (i): Estimated value of n1 (i) kp, ki: Calculate proportional and integral gain (T), load torque estimated value Is fixed, the speed estimation value f is obtained as follows (generally, Those marked with indicate estimated values). Ie J: Inertia of the motor τr (i, j): Torque command value Tn: Encoder pulse interval f (i, j): Speed estimation value at the time after the elapse of jTs from the time when the speed detection value was obtained (“speed estimation Digital Servo Using Observer ”, IEEJ Transactions on Electronics, Vol. 107, No. 12) In the conventional methods described above, all of the methods (a) to (c) detect the average speed value within the measurement time. A dead time due to the averaging (a difference between the time when the average value is obtained and the time when the speed actually takes the average value) occurs in the loop, and the speed decreases in the conventional method (d). As the fixed time (T) approaches the period of the pulse generated by the encoder, the range of change in the dead time of detection increases,
These cause the control performance to deteriorate.

The dead time due to the averaging means the following.

In FIG. 6, assuming that the speed changes linearly with respect to time, the speed is A at a certain timing t1, and the speed is B at the next timing t2, the average value is ( A + B) / 2. However, the time when the average value is obtained is the timing t2, and the time when the speed is actually (A + B) / 2 is tM.

In other words, the point in time when the average value (A + B) / 2 is obtained by averaging (timing t2) means that the speed is actually the average value (A
+ B) / 2 (at timing tM) is Δt
(= T2-tM).
This time delay Δt is called dead time. The larger the dead time Δt is, the larger the deviation from the actual speed value is, which causes a decrease in control performance for control performed by detecting the speed.

The reason why the change width of the detection dead time increases as the fixed time (T) approaches the period of the pulse generated by the encoder is as follows.

In FIG. 7 (a), the average value between a certain timing t1 and the next timing t2 (constant time T) means that the timing t2 starts from the pulse (encoder generated pulse) that arrives immediately before the timing t1. Since the time average value of the number of pulses up to the arriving pulse, that is, the average value during the period L, is referred to, the detection timing of the average value should be tL.
Since detection is performed at t2, a time delay of Δt (dead time) occurs during the detection.

FIG. 7 (b) shows a case where the period of the generated pulse of the encoder is approaching the fixed time (T), and as a result, the dead time Δt similarly generated is the same fixed time T.
It will be noted that the expansion is relatively large with respect to (between timing t1 and the next timing t2). As described above, the width of the dead time Δt is changed as described above.

In the above method (d), the speed is detected by a method based on the same principle as the conventional method (c) while the pulse from the encoder arrives within the fixed time (T). In the method of estimating the speed at a lower speed, the speed estimation value obtained at every fixed time (T) within the encoder pulse interval (Tn) is integrated as shown in the above equation (4). Value obtained is the encoder pulse interval (Tn)
Is used as the estimated value of the average speed, a calculation error occurs due to the asynchronous between the point of generation of the encoder pulse and the point of time of the speed estimation calculation, and there is a problem that the speed cannot be estimated accurately.

Therefore, as described above, (a) the problem that occurs when a pulse from the encoder arrives within a certain time (T), that is, the dead time and the speed that occur due to detection of the average speed value are: The control performance decreases due to an increase in the variation width of the detection dead time which occurs as the period of the generated pulse decreases and the period of the pulse generated by the encoder approaches. (B) From the encoder within the predetermined time (T) In order to solve the problem that occurs at a low speed at which no pulse arrives, that is, the problem that the speed cannot be accurately estimated and detected at regular time intervals (T), Japanese Patent Application No. 63-156697 discloses a motor as described below. Has been proposed.

This detecting method includes an integrating means for counting and integrating pulses generated by an increment type encoder attached to a rotating shaft of a motor for a predetermined period Ts, and an input pulse (encoder immediately before a lapse of a certain sample timing). Pulse), a time length measuring means for measuring a time length from the arrival time (rising edge of the pulse) to the elapse of the timing pulse, an integrated value of the pulse by the integrating means, and a time measured by the time length measuring means. Calculating means for calculating the average rotation speed of the motor using the length, load current detection means for the motor, and the instantaneous rotation speed of the motor at a certain sample timing from the detected load current value and the average rotation speed obtained by the calculation means. And an estimating operation means for estimating the operation.

In the method for detecting the instantaneous rotational speed of the electric motor according to the already proposed method, it is necessary to first calculate the average rotational speed of the electric motor. FIG. 12 is a diagram illustrating the principle of the operation method.

Referring to FIG. An integrated value Pi is obtained by an integrating means for counting and integrating the encoder pulses 1 to Pi generated by the increment type encoder during a predetermined period Ts. Further, time length measuring means for measuring the time length from the arrival time (rising edge of the pulse) of the input pulse (encoder pulse) immediately before the lapse of a certain sample timing to the lapse of the timing pulse,
A calculating means for calculating a time length t _i when (i) is selected as the timing pulse and a time length t _{i + 1} when selecting (i + 1) as the timing pulse, and calculating the average rotation speed of the electric motor, The average rotation speed Vi during the time T _{di at} the sample timing (i + 1) is obtained by the following equation (5).

Vi = K (Pi / T _di ) (5) where K: operation constant T _di : time length (= Ts + t _i −t _{i + 1} ) (6) In this way, first, the average rotation of the motor is performed. Calculate the speed.

Next, the operation of the means for estimating and calculating the instantaneous rotation speed at the sample timing from the detected load current value and the average rotation speed obtained by the calculation means will be described in the following procedure.

(1) Prerequisites for estimating and calculating the instantaneous rotational speed (2) State equation in continuous time system of the mechanical system of the motor,
Output equation (3) Derivation of state equation of discrete time system and output equation from (1) and (2) above (4) Method of estimating and calculating instantaneous rotational speed from above (3) (5) Within fixed time (T) (6) Method for estimating and calculating instantaneous rotation speed at a speed at which pulses from the encoder no longer arrive (6) Method for estimating and calculating instantaneous rotation speed to compensate for the effect of dead time due to speed calculation time (7) Speed calculation A method for estimating and calculating the instantaneous rotation speed at a low speed at which no pulse from the encoder arrives within a certain time (T) in order to compensate for the effect of the dead time due to time.

(1) Preconditions for estimating and calculating the instantaneous rotational speed The actual speed n (t) between the sample points, the speed sample points (i), (i + 1) _,. FIG. 8 shows the relationship with the detected speed average value. Generally, since the control sample period (Ts) is set sufficiently small with respect to the response frequency of the control system, the following assumption may be satisfied.

(A) substantially constant change in speed within the aforementioned time length (T _di) That is, (put a [i]) mean value Vi of time length (T _di) is the sample point from (i) m _i Ts time It may be regarded as an instantaneous value at a point where the time has passed. (Equal to the instantaneous value at half the time length T _di ) (b) Adjacent time lengths are almost constant. That is, T _di-1 T _di T _{di + 1} T _d where T _d (1-m _i + m _{i + 1} ) Ts (7)

Further, since the current control system of the motor can be approximated by a first-order lag, and the relationship between the mechanical time constant (TM) and the time constant (Ta) of the current control system is TM≫Ta, the following linearization is performed. Also has little effect on speed. That is, the torque current i _T (t) between the sample points (i) and (i + 1) is converted to the detected current values i [i], i at the sample points (i) and (i + 1).
Using [i + 1], it can be expressed as i _T (t) = i [i] (1-t / Ts) + i [i + 1] / Ts (8) Here, the time t is the elapsed time from the sample point (i).

(2) State equation of the mechanical system of the motor in a continuous time system,
Output equation The block diagram of the mechanical system of the motor is shown in FIG. 9, the input variable is the torque current i _T (t) of the motor, and the output variable is the speed n.
(T), the state variable is velocity, and the disturbance torque τ _d (t) is the continuous-time state equation and the output equation are as follows:
a, 9b). Where u (t) is an input vector, y
(T) represents an output vector, and x (t) represents a state vector.

_{(T) = A c x (} t) + b c u (t) ...... (9a) y (t) = c c x (t) ...... (9b) _{However, dτ d (t) / dt} = 0 ...... (10) x (t) = [n (t), τ _d (t)] ^T (T means transposed matrix) (11) u (t) = i _T (t) (12) y (t) = n (t) (13) b _c = [φ / TM 0] (15) c _c = [10] (16) τ _d is torque, n is rotation speed, i _T is torque current (load current), φ is magnetic flux, TM machine time constant (3) above (1), (2) the state equation of the discrete-time system from claim, the solution of the derived continuous-time state equation of output equation (9a), the state variable at the initial time t ₀ x (t0) and continuous input u (t) Given by Here, when t ₀ = iT _d and t = (i + 1) T _d are substituted, Expression (17) becomes Expression (18).

Here, [i + 1] and [i] represent the state variables at the virtual sample points (i) ′ and (i + 1) ′ in FIG. 8, respectively. As can be seen from FIG. 8, the virtual sample point is a timing (time point) that is located between the actual sample point and the sample point and exactly coincides with the average speed therebetween. I called.

By transforming equation (8) by substituting equation (18), Then, a state equation and an output equation represented by equations (21a) and (21b) are obtained.

[I + 1] = A (i) + bU [i] (21a) [i] = c [i] (21b) where [i] = [[i], _d [i]] ^T () 22) [i] = [i] ... (23) b = [φTd / TM 0] ... (25) c = [1 0] ... (26) [i] :( i + m _i) the instantaneous value of the velocity at Ts time clogging virtual sampling point (= sample point (i + 1 _D ) [i]: (i + m _i ) instantaneous value of disturbance torque at Ts, that is, a virtual sample point (= average value of disturbance torque detectable at sample point (i + 1)). Method for estimating and calculating the instantaneous rotation speed from the term (3) Since it is clear that (c, A) is observable in the above equations (21a) and (21b), an observer of the disturbance torque can be configured. Here, the literature (B. Gopinath, “On the
Control of Linear Multiple Input-Output Syst
ems '' The Bell System Technical Journal Vol.50, N
When a minimum-dimensional observer is designed based on o.3 (1971), the estimated value of the disturbance torque is obtained by equation (27b), with the virtual state variables set as in equation (27a).

Therefore, the estimated value [i at the sample point (i + 1)
+1] is given by equation (17). Can be represented by

From equation (28), the estimated disturbance torque at the sample point (i + 1) The instantaneous speed estimated value ([i + 1]) is calculated by the following (29) and (30).
It is found like an expression.

(5) Method for estimating and calculating the instantaneous rotation speed at a speed at which no pulse from the encoder arrives within a fixed time (T), the actual speed n (t) between the sample points, the speed sample point (i, 0), (I, 1), ... (i, j), ... (i + 1,0), ..., the time length (in this case, the time length from the virtual sample point to the next virtual sample point), and the detected speed average value Relationship No. 14
Shown in the figure. The algorithm of equations (20), (21), and (30) is changed as follows.

(A) The time length (time length from the virtual sample point to the next virtual sample point) Td is set as follows.

Td (n _{ai + 1} + n _bi −m _i + m _{i + 1} ) Ts (31) (b) U [i] in Expression (20) is set as follows.

(C) [i + 1] (= [i + 1, j]) in equation (30)
As follows.

Here, [i + 1, j] represents an estimated value of the instantaneous speed after the sample point at which the pulse is present and the average speed is obtained.

(6) for the time required for speed calculation to compensate for the effects exerted as dead time, m _i ≧ mc the time required for the method operations for estimating the instantaneous rotational speed as mc (known), m _i <mc
The relationship between the actual speed n (t) between sample points, the speed sample points (i), (i + 1),..., The time length, and the detected speed average value in the case of FIG. Are shown below. The algorithm of equations (20) and (30) is changed as follows.

[(6) -1] When m _i ≥ mc (a) Put U [i] in equation (20) as follows.

However, the torque current i _T (t) between the sample values (i) ″ and (i + 1) ″ is changed to the sample points (i) ″, (i +
Using the current detection values i [i] ″, i [i + 1] ″ in 1), i _T (t) = i [i] ″ [1-t / ((1-mc) Ts)] + i [i + 1] t / ((1-mc) Ts) Expression (36) was obtained by the expression (36) (time t is the elapsed time from the sampling point (i) ″) (b) Expression (30) ) Of [i + 1] (= [i + 1] ″)
As follows.

Here, [i + 1] ″ represents the estimated value of the velocity at the sample point (i + 1) ″.

[(6) -2] When m _i <mc (a) U [i] in equation (20) is set as follows.

However, the equation (37) was obtained from the equation (35).

(B) [i + 1] (= [i + 1] ″ in Equation (30) is set as follows.

Here, [i + 1] ″ represents the estimated value of the velocity at the sample point (i + 1) ″.

(7) A method for estimating and calculating the instantaneous rotational speed at a low speed at which no pulse from the encoder arrives within a certain time (T) for compensating the effect of the time required for the speed calculation as a dead time m _i ≧ mc , m _i <mc, actual speed n between sample points
(T), the relationship between the speed sample points (i), (i + 1),..., The time length, and the detected speed average value are shown in 16 (a) and (b).
Each is shown in the figure. The algorithm of equations (20) and (30) is changed as follows.

[(7) -1] When m _i ≥ mc (a) The time length Td is as follows.

Same as equation (31) [Td (n _{ai + 1} + n _bi −m _i + m
_{i + 1} ) Ts] (b) U [i] in equation (20) is set as follows.

However, it was determined by using equation (36).

(C) [i + 1] in equation (30) (= [i + 1,
j] ″) as follows.

Here, [i + 1, j] ″ represents an estimated value of the instantaneous speed after the sample point at which the pulse exists and the average speed is obtained.

[(7) -2] When mi <mc (a) The time length Td is as follows.

Same as equation (31) [Td (n _{ai + 1} + n _bi −m _i + m
_{i + 1} ) Ts] (b) U [i] in equation (20) is set as follows.

However, it was determined by using equation (36).

(C) [i + 1] in equation (30) (= [i + 1,
0] ″) as follows.

Here, [i + 1, j] represents the estimated value of the instantaneous speed at the sample point where the pulse exists and the average speed is obtained.

FIG. 11 is a block diagram showing an example of a device for implementing the detection method according to the above-mentioned proposal, and FIG. 12 is a waveform diagram showing operation waveforms of the devices.

In these figures, reference numeral 1 denotes an increment type encoder attached to a motor shaft, which generates a pulse having a frequency proportional to a rotation speed. Reference numeral 2 denotes a waveform shaping circuit which shapes a pulse from the encoder 1 into a rectangular pulse having a fixed time width. An input pulse counting circuit 3 counts output pulses of the waveform shaping circuit 2. Reference numeral 4 denotes a reference pulse counting circuit, which counts reference pulses A (timing clock shown in FIG. 12). Latch circuits 5 and 6 hold the count contents of the count circuits 3 and 4 in response to a hold command by the sample timing pulse B.

Reference numeral 7 denotes a timer circuit, which outputs a sample timing pulse B each time the reference pulse A is counted by a number corresponding to a predetermined time (Ts), and the latch circuits 5, 6 are counted by the sample timing pulse B. The auxiliary timing pulse C is generated with a slight time delay sufficient to hold the count contents of the circuits 3 and 4 normally.
The auxiliary timing pulse C is used as a zero clear signal of the count circuit 3. The counter circuit 4 is cleared to zero by a change in the input pulse.

8 is a CPU and 9 is a memory. In order to control the rotation speed or the torque of the electric motor, other hardware is required. However, here, the minimum hardware necessary to detect the instantaneous speed is shown.

13 shows a flowchart of a program of the instantaneous speed detection method according to the proposed method of FIG. This shows a program for detecting the instantaneous speed, and excludes a control program necessary for controlling the speed of the electric motor. The detailed operation will be described according to this program.

Now, the interrupt signal at the sample point (i + 1) is
, Before executing the speed control program, the CPU 8 executes the program of FIG. 13 to calculate the instantaneous speed. First, a torque current i [i + 1] is detected. Then, the input pulse integrated value (P
i) to determine whether this value is 1 or more,
That is, it is determined whether or not there is an output pulse from the encoder 1 (hereinafter, sometimes referred to as an encoder pulse) between sample points. If there is an encoder pulse between the sample points, the integrated time (t
_{i + 1} ) is read to check whether there is a pending flag prepared in the memory, that is, whether there was an encoder pulse between sample points last time.

If there is no pending flag, the average speed is obtained by equation (5). Thereafter, the number of sample points from the sample point when there was an encoder pulse between the sample points to the next same state, ie, the sample point when there was an encoder pulse between the sample points (n _{Ti in} FIG. 14), the sample point If the average speed is regarded as an instantaneous value from the sample point when there is an encoder pulse between them, the number of sample points from the sample point where the instantaneous value exists to the next sample point (n _{ai in} Fig. 14), and others (N _bi (= n _Ti −n _ai ) in FIG. 14) is obtained. In this case, since there is no pending flag, n _Ti = 1, n _ai = 1, n _bi =
Set to 0.

After the above operation, _mi is obtained and U [i-1] is obtained from the arithmetic expression (20), and this U [i-1] and the previous torque estimated value are obtained. And the average speed [i] obtained by the arithmetic expression (5) and the torque estimated value by the arithmetic expressions (27a) and (27b). Is calculated, the instantaneous speed [i + 1] is obtained by Expression (30).

Next, if there is a pending flag, the pending flag is reset, and the average speed is calculated by the equation (5). Thereafter, the number of sample points from the sample point when there was an encoder pulse between the sample points to the next same state, ie, the sample point when there was an encoder pulse between the sample points (n _{Ti in} FIG. 14), the sample point If the average speed is regarded as an instantaneous value from the sample point when there is an encoder pulse in between, the number of sample points from the sample point where the instantaneous value exists to the next sample point (Fig. 14
n _ai ) and other sample points (n _bi (=
n _Ti −n _ai )). Further, _mi is calculated and U [i-1]
Is calculated from equation (32), and this U [i-1] and the previous estimated torque value And the average speed [i] obtained by the arithmetic expression (5) and the torque estimated value by the arithmetic expressions (27a) and (27b). Is calculated, the instantaneous speed [i + 1] is obtained by Expression (33).

If there is no encoder pulse between the sample points, the pending flag is set, and the instantaneous speed [i + 1] is obtained from the arithmetic expression (42).

By repeating the above operation for each sample point, the instantaneous speed at the sample point can be estimated and calculated.

FIG. 15 shows a flowchart of a program of another specific example (when the dead time (mc Ts) due to the calculation time cannot be ignored) for implementing the method according to the proposal. This also shows a program for detecting the instantaneous speed, similarly to FIG. 13, except for a control program necessary for controlling the speed of the electric motor. According to this program, only operations different from the specific example of FIG. 13 will be described in detail.

The difference from the embodiment of Figure 13, the magnitude relationship between the absence may have pen computing flags dead time according to the value and operation time of the calculated m _i (mc Ts), the input variable U
The difference is that the formula for calculating the instantaneous speed is different from [i-1]. Specifically, it is as follows.

When there is no penning flag (Fig. 15, part 1), when m _i ≧ mc, the input variable U [i-1] is obtained by equation (35), and the instantaneous speed is obtained by equation (37). When m _i <mc input variable U [i-1] of the formula (38), the instantaneous velocity is obtained by equation (39), also when there is a pen computing flag (Fig. 15 Part 2), the input variables U when m _i ≧ mc [ i-1] of the formula (40), the instantaneous velocity is obtained by equation (41), the input variables U when m _i <mc [i-1] is obtained by equation (43), the instantaneous speed of the formula (44), The other operations are the same as those in the specific example of FIG. 13. By repeating these operations for each sample point, the instantaneous speed at the sample point can be estimated and calculated.

[Problems to be solved by the invention]

The present invention aims to correctly detect the instantaneous rotational speed of an electric motor in a wide speed range from an extremely low speed to a high speed, and an object of the present invention is to solve the above-mentioned problem in the low-speed region. That is, the problems to be solved by the present invention are the following four points.

1) In the prior art, as shown in the above equation (7), adjacent time lengths with a time length (time length from a virtual sample point to the next virtual sample point) _Td are regarded as substantially constant, At low speeds, this assumption does not always hold, which causes an error in the estimation of the disturbance torque.

2) At low speeds where the pulse interval of the encoder pulse is not longer than the sample period of the control device, the convergence of the disturbance torque estimator is also reduced, and the error in the instantaneous speed estimated based on this is also large. To solve this point.

In the prior art, the midpoint of _Td is defined as a virtual sample point using the time length _Td defined by the means for calculating the average rotation speed of the motor, and the time length T between the virtual sample points is defined.
_a is defined as a variable as T _ai = T _di / 2 + T _{di +2} / 2, and an estimator (observer) of a disturbance torque is configured using T _ai . Since the poles of the observer are arranged according to this T _ai , the operation of the estimator does not become unstable even during low-speed rotation where the encoder pulses are sparse. That is, the operation of the estimator can be guaranteed regardless of the rotation speed of the electric motor.

However, in this method, at a speed at which the interval between encoder pulses is shorter than the sample period length Ts, an error occurs in the estimated value of the disturbance torque estimator for the following reason. Hereinafter, for the sake of simplicity, the “speed at which the interval between encoder pulses is equal to the sample period length Ts” is defined and used as “synchronous speed”.

Below the synchronous speed, the number of encoder pulses detected within the sample period is 0 or ± 1 (+ or-depending on the rotation direction).
The average speed Vi is obtained by the reciprocal of the time length T _di , that is, the following formula. When the speed is constant (the encoder pulse interval is also constant), the T _di also becomes a constant value T _d constantly.

Vi = K · (1 / T _di ) = K · (1 / T _d ) where K is a constant. However, above the synchronous speed, the number of encoder pulses Pi arriving within a certain sample period becomes smaller than the encoder within an adjacent sample period. It is possible that the number of pulses is increased or decreased by ± 1 from the number of pulses, and T _{di at} this time is ± 1 more than the adjacent T _d.
Only the encoder pulses become longer (or shorter), also increases or decreases the length of the T _a which is calculated using this T _di.

 This relationship is shown in FIG.

In the figure, although the sampling period lengths Ts, Ts, and Ts each include only one encoder pulse (the number of encoder pulses Pi is 1), the sampling period length Ts happens to include two encoder pulses. It will be noted that it is included (the encoder pulse number Pi is 2). Above the sync speed, depending on that speed, this can happen by chance.

Then, in FIG. 19, T _di becomes longer by one encoder pulse than the adjacent T _{d + 1} or T _d−1 , and as a result, T _di
T _ai and T _ai-1 calculated using are calculated longer than T _{ai + 1} , T _ai-2 , and T _ai-3 calculated without using other than T _di You will see that.

That is, as described above, _Ta may fluctuate even when the rotation speed of the electric motor is constant above the synchronous speed.
Variation of inline not T _a on the speed fluctuation causes an error in the estimated value of the disturbance torque estimator. Disturbance torque estimate τ
The operation expression of _d [i] is shown below from the expressions (27a) and (27b).

Z [i + 1] = Z [i] + _i · (T _di / TM) · (τ _d [i] −U [i]) τ _d [i] = Z [i] + _i · y [i] _i = (Λ-1) T _M / T _ai (where y [i] = n [i], λ; pole of the estimator (set value)) When T _{ai in} this equation fluctuates, _i fluctuates. The disturbance torque estimated value will fluctuate. Therefore,
In this way, an error may be generated in the disturbance torque estimation value.

3) At a low speed similar to 2), the number of data and the number of calculation steps required for estimating the disturbance torque and the instantaneous speed increase as the speed decreases. However, in the control device, the number of data steps that can be held is limited in the number of calculation steps that can be allowed.Below a certain speed, such a large amount of data cannot be held and the number of calculation steps cannot be tolerated. This method is no longer practical, so propose a practical alternative.

4) The torque current value of the motor is required for the estimation calculation of the disturbance torque and the calculation of the instantaneous speed. However, when the torque current value is obtained from the line current of the motor, the calculation for converting the line current into the torque current is required. Become. Further, the current control system is often constituted by a discrete circuit, and in this case, an interface circuit to the CPU for performing calculations is required, which generally increases the cost of the device. Therefore, the present method should be realized without such cost increase.

[Means for solving the problem]

1) Rather than keeping the time length between one sample timing and the next sample timing constant as in the prior art,
A disturbance torque estimator (CPU) is provided, which is regarded as a variable and is based on a state equation of a discrete value system with a variable sampling time. This is an improvement over the use of a disturbance torque estimator based on the state equation of a discrete value system with a fixed sampling time.

2) The pole arrangement set in the disturbance torque estimator (observer) is varied according to the speed of the motor. That is, when the speed of the motor decreases, means (CPU) for changing the pole arrangement is provided so as to increase the response in the estimator.

When pole arrangement is performed by the T _ai value, above the synchronous speed defined above, even if the rotation speed of the motor is constant, T _a may fluctuate, thereby causing an error in the torque estimation value with an error. I will. Therefore, in order to suppress the occurrence of such an error, the pole arrangement is arbitrarily performed using the average speed value of the electric motor instead of the _Tai value. As described above, if the pole arrangement is performed using the average speed of the motor, in the pole arrangement using T _ai , it is possible to suppress the error of the torque estimation value by the disturbance torque estimator that may occur even when the motor speed is constant. . However T _ai value, whereas a substantially constant value even accompanied by variation (= Ts) is a synchronous speed or higher, the average speed continues to increase, to ensure the stability of the estimator of the operation in the speed or An algorithm must be provided.

3) When the motor speed becomes lower than a certain constant value, the method for detecting the instantaneous rotation speed of the motor according to the present invention is stopped and the encoder pulse from the encoder used for detecting the rotation speed of the motor is stopped. At each arrival, an output signal consisting of a combination of a constant speed value output over a certain period and a speed zero and an output over a period until the next encoder pulse arrives is output as an indication of the rotation speed of the motor. Then, a method of detecting the rotation speed of the motor, which detects the rotation speed as the rotation speed of the motor, is adopted. Provide means (CUP) for returning to the instantaneous rotation speed detection method.

4) When a torque current command value of the motor is given and torque current value control for matching the actual value of the torque current to the command value is performed, the estimation of the disturbance torque and the estimation of the instantaneous rotation speed of the motor are performed. As the motor torque current value at the current sample timing required to perform, the torque current command value is delayed and used for a period required for the actual torque current value to approach the torque current command value by control, and the delay for that is used. Means are provided.

[Action]

1) When the pulse interval from the encoder used for detecting the rotation speed of the motor and the sampling interval of the control device are close to each other or are longer than the sampling interval, the time length (from the virtual sampling point to the next virtual The length of the time length adjacent to each other changes greatly with T _d ), but the estimation error can be estimated by using a disturbance torque estimator based on the state equation of a discrete-valued system with variable sampling time. And the accuracy of detecting the instantaneous rotational speed of the motor is improved.

2) At low speeds where the pulse interval from the encoder is longer than the sampling interval of the control device, the sampling period of the discrete-value state equation extends, so that the sampling dead time increases. The pole arrangement set in the disturbance torque estimator is changed so that the response is increased so that the influence of the sampling dead time at low speed is small.

By determining _i using the average rotation speed of the motor, the estimation error of the disturbance torque estimator is suppressed, and the accuracy of detecting the instantaneous rotation speed of the motor is improved.

3) At low speeds where the pulse interval from the encoder is longer than the sampling interval of the control device,
The amount of data and the amount of calculation required for designating the disturbance torque and the instantaneous speed increase as the speed decreases. Due to the physical limitations of the control device, the amount that can be held is naturally limited, and it is not possible to hold all of the necessary data amount and calculation amount because it can be suppressed to a certain appropriate amount. Cannot be effectively implemented.

Therefore, instead of the method according to the present invention, a detection method based on a method similar to a so-called F / V conversion method (a method of converting a frequency F into a voltage V and outputting the same) is employed at a low speed. That is,
Each time an encoder pulse from the encoder arrives, a combination of a constant speed value output over a certain period of time and a zero speed output over a period until the next encoder pulse arrives is used to represent the rotation speed of the electric motor. And outputs it as a rotation speed of the electric motor.

Here, a combination of a constant speed value output over a certain period and a zero speed output after that until the next encoder pulse arrives corresponds to the frequency F, and corresponds to the frequency F. The average of the constant speed value output over the constant period over the entire period up to the arrival of the signal V corresponds to the voltage V.

4) If the torque current command value is delayed so as to be regarded as equivalent to the actual value, the estimation calculation of the disturbance torque and the calculation of the instantaneous speed can be executed with the delayed torque current command value.
The calculation for detecting the line current of the electric motor and converting the line current into the torque current is not required.

[Embodiment] FIG. 2 is an explanatory diagram showing the relationship between various temporal quantities when the interval between encoder pulses and the sampling interval of the control device are close to each other.

In the figure, in the time length _Td as a calculation target of the average speed, the length between the i-th one and the one before and after it is about twice different. Further, at a low speed in which the interval between encoder pulses is larger than the sampling interval of the control device, a slight change in speed greatly changes the time length.

As described above, when detecting a motor speed that is smaller than the motor speed such that the interval between the encoder pulses and the sampling interval of the control device are close to each other, a time length close to each other (for example, T _di and T _{di + Since} the condition assumed in the detection method according to the proposed method that ₁ and T _di-1 ) are almost equal is not satisfied, the time length T _d needs to be dealt with as a variable of a discrete system. Therefore new, the time length T _a from the virtual sample point to the next virtual sample points define as follows.

T _ai = T _di / 2 + T _di +1/2 (45) Since the input variable U [i] in this case is the average value of the torque current value i (t) within the time length T _ai , the following equation must be expressed. Can be.

Here, the torque current value i (t) is obtained by calculating the equation (2) which is a linear approximation based on the detection value at the sampling point of the control device. Is the same.

Next, the estimated value of the disturbance torque Is obtained by the following equation.

(If [i + 1] is the current value, [i] indicates the previous value.) Here, _i = (λ−1) TM / T _ai (47c) where λ is a pole set in the estimator and | λ |
<1.

The flowchart of the program as one embodiment of the present invention for solving the first problem based on this is shown in the first embodiment.
FIG. Time length T _ai
By using as a variable, it is possible to perform an accurate calculation.

An embodiment of the present invention for solving the second problem will be described with reference to FIG. In this embodiment, the pole value set in the disturbance torque estimator is set to a constant pole value λ _{0 up} to a certain motor speed n ₀ , and below that speed, the pole value is gradually increased from λ _0. Approach 0.

n ₀ is the motor speed when the pulse interval of the encoder pulse matches the sampling interval, and can be expressed by the following equation.

The absolute value λ of the pole at a speed of n ₀ or less is given by

In the above equation, λ ₁ is the lower limit of the absolute value of the pole.

One embodiment of the present invention for solving the third problem is shown in FIG.

The figure shows that at a certain motor speed or lower, a speed detection value of a certain value is output for a certain period every time an encoder pulse is input, and thereafter, immediately before the next encoder pulse is input. Until then, the speed is regarded as zero, and as a result, the average value of the entire period is used as the speed detection value, so-called F / V conversion type speed detection method described above is adopted. Fig. 4 shows an algorithm that employs a motor instantaneous speed detection method based on a disturbance torque estimator.

Details will be described below (provided that the current speed is a low speed in which the interval between encoder pulses is longer than the sample period). When an encoder pulse is input during the previous sampling period, the routine proceeds to the routine shown in FIG. Since the pending flag has been set according to the above-described premise, the routine proceeds to the routine.

Then the pulse interval of the encoder calculates an average speed of between, it is equal to or less than the speed n _L a predetermined, proceeds to the routine. At these speeds Pi
= From it is 1, the average velocity is determined by the pulse interval T _d, T _d is the time length T _d max average speed is n _L when exceeded
It is determined that: Thus T _d is exceeds T _d max because the size is meaningless, if n _Ti is beyond the value commensurate with the T _d max, prepared flags proceeds to routine, the count of n _Ti By canceling, even if _Td becomes infinitely large, for example, near a stop, processing can be performed.

In the routine, the first routine for performing a so-called F / V conversion method specific speed detection described above, the n _L predetermined period to output (polarity at this time is consistent with the previous value) as first speed detection value. Further, a counter K for managing a period for outputting n _L is initialized to one. Also set the F / V flag.

In the next sample period, since no pulse is input, the routine proceeds to the routine. Since the F / V flag has been set in the previous sampling period, the routine proceeds to the routine. After setting the pending flag the pulse is not input here, as the speed detection value by the value of the counter K, whether to output the n _L, decide whether to zero.

The speed detection operation in the F / V mode will be described with reference to FIG. Average speed of motor from pulse interval _{Td of} encoder
When you ask for Vi Next, the average value Vi ′ of the detected speed values during T _d ′ in FIG.
Is (It corresponds to Kmax = 2 in FIG. 5.) Now, under the assumption that T _d ≫Ts (where Ts is a sampling period), T _d ≒ T _d ′. Then, when n _L is determined so that the equations (50) and (51) match, Thus, also in the F / V mode speed detection method, it is possible to correctly detect the average speed. By the way, in the routine of FIG. 4, once the speed detection method of the F / V mode is used, the estimation calculation of the disturbance torque is not performed.
When returning from the mode speed detection method to the original speed estimation method, the initial value of the disturbance torque estimator is required.

Therefore, the routine in the F / V mode speed detection method,
In, the calculation of the initial value of the estimated value is performed. Here, based on the above equation (47b), Is the output of the speed controller, [i] is n _L and Z [i]
Is obtained by calculation.

Next, an embodiment of the present invention for solving the second problem will be described.

FIG. 20 is a flowchart showing such an embodiment. In the figure, a portion surrounded by a dotted line is an additional portion according to the present invention. In this example, first, the average speed (referred to as Vi) is detected, and if the average speed Vi is equal to or less than the synchronous speed defined above,
The _{i i = (λ-1)} Vi · TM / K ...... (60) ( although, K is the average constant for velocity calculation) Vi = K · Pi / T d (Pi; is detected within T _d If the average speed Vi is equal to or higher than the synchronous speed, _i
I = (λ−1) TM / Ts (= constant value) (61) Below the synchronous speed, when the rotation speed is constant, the average speed detection value Vi
In this speed range, at a constant speed, T _di is a constant value T _{d as} described above, so that T _ai = T is given by: Vi = K · (1 / T _d ) = K · (1 / T _ai ) Since it is obtained by _{d / 2} + _{Td / 2} = _Td , the equation (60) eventually becomes _i = (λ-1) TM / _Tai .

That is, in this example, as shown in FIG.
As can be seen from equation (61), the pole of the estimator is set to a constant value λ in the entire speed range. However, since the average speed value is used, the estimated torque value does not involve an error.

Still another embodiment for solving the second problem will be described with reference to FIG. FIG. 21 is a flowchart showing such an embodiment.

In the embodiment shown in FIG. 21, the stability of the estimator is ensured even at a low speed at which rotation unevenness is likely to occur, even if the encoder pulse interval becomes longer and the pole of the estimator changes, so that _i does not change. keep giving that _i. In this example, the poles of the estimator are gradually changed so that the response of the estimator becomes faster as the speed increases, and the response of the estimator becomes practical at high speed.

In the figure, V1 and V2 are designated speeds of the low speed, medium speed and high speed breaks, and _i is a constant value _i when the average speed value is equal to or less than V1.
(The pole is assumed to be λ1 close to 1 above the synchronous speed.) At the medium speed, the pole λi is gradually set to λ2 (a value close to zero) so that the response of the estimator becomes faster in accordance with the average speed value Vi according to the following equation. At high speeds above V2 where the estimator has a sufficiently fast response, _i is again a constant value of ₂
(The poles are located at λ2).

However, if V1 and V2 are specified to be equal to or higher than the synchronization speed, and if the encoder pulse interval becomes longer and the poles of the estimator change below the synchronization speed as described above, the estimator remains within the operating range of the estimator. To ensure the stability of See FIG. 23 for the arrangement of the poles.

Vi <V1; _i = (λ1-1) TM / Ts = _i (71) V1 <Vi <V2; _i = (λi-1) TM / Ts (72) where λi = λ1 + (λ2- λ1) ・ (Vi-V1) /
(V2−V1) V2 <Vi; i = (λ2-1) TM / Ts = 2 (73) In both of the above two embodiments, a speed range in which the speed detection value can be regarded as almost zero at a very low speed is set in advance. In this speed range, the operation of the estimator is stopped and the detected speed value is set to zero. (V0 in the figure corresponds to this.) One embodiment of the present invention for solving the fourth problem is shown in FIG.
This will be described with reference to FIG.

FIG. 17 shows that the torque current command value of the motor is given,
5 is a timing chart showing a relationship between the torque current command value and the actual torque current value when the CPU performs torque current value control for matching the actual value of the torque current to the command value.

In the figure, since the current command value output by the CPU is a constant value during the sampling period Ts, the initial value (indicated by a circle) of the sampling period is held. The actual current value follows this and is approximated by a first-order lag, as shown by the solid line graph in FIG. The solid line obtained by linearly approximating the solid line in the sampling period is the dotted line graph in the same figure, which coincides with the connection of the current command values delayed by one sampling period. Therefore, when the current command value and the actual value are approximated by the relationship as shown in the figure, the current actual value detected at each sampling timing can be approximated by the current command value delayed by one sampling period.

FIG. 18 is a flowchart for executing the disturbance torque current calculation and the instantaneous speed calculation shown in FIGS. 1 and 4 using the torque current command value delayed by one sampling period (that is, the flowchart for solving the fifth problem). (Example of the invention) is shown.

Here, the CPU that executes torque current value control and the like includes:
Usually, register A is used.
The torque current command value is temporarily stored in this register A (step S2 in FIG. 18), and during the next sampling period,
Read it as the detected torque current value (step S
1) Perform the operation.

Here, the embodiment in which the torque current command value delayed by one sampling period is used as the torque current detection value has been described. However, depending on the configuration of the current control system, the relationship between the current command value and the actual value may be determined. Therefore, it is necessary to change the delay time accordingly.

〔The invention's effect〕

1) According to the invention for solving the first problem, in the state equation of the discrete value system based on the encoder pulse, the sampling time is made variable so that the interval between the encoder pulses becomes the sampling period of the control device. The disturbance torque can be more correctly estimated even at a low speed where the length becomes longer. As a result, it is possible to realize a speed detector with high accuracy up to extremely low speed and small detection delay.

2) According to the invention for solving the second problem, since the pole set in the disturbance torque estimator is made variable according to the speed, the pole is determined from the stability of the system above the medium speed, At a low speed when the sampling period of the discrete state equation increases and the response becomes a problem, the pole can be made to have a high response, and there is an advantage that a great deal of flexibility can be given to the system design.

Furthermore, even at a speed at which the encoder pulse interval becomes longer at a low speed and the encoder pulse does not arrive within the sample period, the disturbance torque estimator is operated stably,
The effect that the instantaneous speed can be detected can be obtained without causing an error in the disturbance torque estimated value and the instantaneous speed detected value.

3) According to the present invention for solving the third problem, in order to switch to the speed detection method in the F / V mode at the time of extremely low speed, the control device requires too many data at the time of extremely low speed. There is an advantage that a correct and practical speed detection value can be obtained over the entire speed range, even if there is a hardware restriction such that the speed cannot be held.

4) According to the invention for solving the fourth problem, since the estimation calculation of the disturbance torque and the calculation of the instantaneous speed are performed using the torque current command value which is the data inside the CPU, the actual current value of the motor is calculated. Since an interface circuit for loading into the CPU is not required, and an operation for converting the line current of the motor into a torque current is not required, not only high performance but also an inexpensive speed detector can be realized. it can.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an embodiment of the present invention for solving the first problem. FIG. 2 is a flowchart showing the actual speed between sample points when the pulse interval from the encoder and the sample period of the control device are close to each other. FIG. 3 is an explanatory diagram showing a relationship between a value, a time length, a detected average speed, etc. FIG. 3 is a characteristic diagram showing an embodiment of the present invention for solving the second problem, and FIG. 4 is a present invention for solving the third problem. FIG. 5 is a flowchart showing one embodiment of the present invention.
Explanatory drawing showing the motor speed detection method in F / V mode, FIG.
FIG. 7 is an explanatory diagram for explaining a problem in the prior art, and FIG. 8 is an actual speed value between sample points.
FIG. 9 is a block diagram of a mechanical system of an electric motor, and FIGS. 10A and 10B are actual speed values between sample points, a time length, and a detected average. FIG. 11 is a hardware block diagram of a speed detection circuit for implementing a detection method according to the present invention.
FIG. 12 is a timing diagram of hardware, FIG. 13 is a flowchart showing a speed detection method according to the proposed method, FIG. 14 is an explanatory diagram showing a relationship between an actual speed between sample points, a time length, a detected average speed, and the like. FIG. 15 is a flowchart showing another speed detection method according to the proposed method. FIGS. 16 (a) and 16 (b) are explanatory diagrams showing the relationship between the actual value of speed between sample points, time length, detected average speed, and the like. The figure is a timing chart showing an example of the relationship between the torque current command value and the actual torque current value,
FIG. 18 is a flowchart showing an embodiment of the present invention for solving the fourth problem, and FIG. 19 is a time length Td, Ta used for the instantaneous rotation speed detection calculation when the speed is constant (the interval between encoder pulses is constant). However, FIG. 20 is an explanatory diagram showing that the encoder pulse interval may fluctuate at a speed higher than the speed at which the encoder pulse interval becomes shorter than the sample period length Ts. FIG. FIG. 4 is a flowchart showing another embodiment of the present invention for solving the second problem,
FIG. 22 is a characteristic diagram showing the relationship between speed and pole in one embodiment for solving the second problem, and FIG. 23 is a characteristic diagram showing the relationship between speed and pole in another embodiment for solving the second problem. FIG. DESCRIPTION OF SYMBOLS 1 ... Incremental encoder, 2 ... Waveform shaping circuit, 3 ... Input pulse count circuit, 4 ... Reference pulse count circuit, 5,6 ... Latch circuit, 7 ... Timer circuit, 8 ... CPU, 9… ROM, RAM

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-59-117479 (JP, A) JP-A-59-6782 (JP, A) JP-A-60-171464 (JP, A) 17588 (JP, B2) (58) Fields investigated (Int. Cl. ⁷ , DB name) H02P 5/00

Claims (4)

(57) [Claims] A first sample time is defined as a current sample timing, and a period average rotational speed of the motor during a period between the current sample timing and the previous sample timing is obtained at the current sample timing. The period, the period between a certain sample timing and the next sample timing is not always constant, and the variable period includes at least a first period which is a period between the current sample timing and the previous sample timing. , A second period which is a period between the previous sample timing and the last two sample timings is stored, a time point at the center of the first period is set as a current virtual sample timing, and a time point at the center of the second period is stored. When the time point is the previous virtual sample timing, the average torque current value at the current virtual sample timing is A second step of calculating from the torque current detection value of the sample timing and the torque current detection value of the previous sample timing; an average rotation speed obtained in the first step; and a current virtual sample timing obtained in the second step. A third step of estimating the disturbance torque by a disturbance torque estimator using the average torque current value of the above, an estimated value of the disturbance torque estimated in the third step, and a value obtained in the second step. A fourth step of obtaining the average torque current value of the current virtual sample timing, the inertia of the known motor-mechanical system, and the time rate of change of the motor rotation speed at the current virtual sample timing; and And the time change rate of the motor rotation speed obtained in the fourth step, and A fifth step of obtaining a change in the rotation speed of the motor, and adding the change in the rotation speed of the motor obtained in the fifth step to a period-average average rotation speed of the motor obtained in the first step. A sixth step of estimating and calculating the instantaneous rotational speed of the motor at the current sample timing. 2. The method for detecting an instantaneous rotational speed of an electric motor according to claim 1, wherein the pole arrangement of the disturbance torque estimating means used for estimating the disturbance torque of the electric motor is such that the average rotational speed of the electric motor is relatively low. And changing the disturbance torque estimating means to increase the response speed. 3. A method for detecting an instantaneous rotational speed of an electric motor according to claim 1 or 2, wherein an arrival of an encoder pulse from an encoder directly connected to an electric rotating shaft used for detecting an average rotational speed of the electric motor. Each time, as a detected speed value of the motor, a signal consisting of a combination of a fixed speed detection value output over a certain period and a zero speed output over a period until the next encoder pulse arrives is used as an average of the motor. The rotation speed detection method of the motor to be detected as the rotation speed is a second speed detection method, when the average rotation speed of the motor has become a low speed below a certain limit,
The method for detecting the rotation speed of the motor is the second method from the first method.
A method for detecting an instantaneous rotational speed of an electric motor, characterized by switching to the method of (1). 4. The method for detecting an instantaneous rotational speed of a motor according to claim 1, wherein the detected torque current value of the motor at a current sample timing is a torque at a previous sample timing or at a previous sample timing. A method for detecting an instantaneous rotational speed of a motor, wherein a current command value is used.