Technical field [0001] 　The present invention relates to a motor control device that performs vector control of a drive of a three-phase brushless motor in a dq-axis rotating coordinate system and compensates dead time of an inverter, and an electric power steering device equipped with the same, and particularly to dq-axis dead time compensation. At the same time, the dq-axis disturbance estimation observer estimates the disturbance that cannot be compensated by the dead time compensation, and further performs space vector modulation to further improve the steering response, the abnormal noise and vibration of the motor, and the performance of the torque ripple. The present invention relates to a motor control device having improved responsiveness and noise resistance in accordance with the number, improved motor model accuracy in response to dq axis current command values, and improved compensation accuracy, and an electric power steering device equipped with the same. Background technology [0002] 　A motor control device that controls an object by driving a motor is mounted on an electric power steering device (EPS: Electric Power Steering System), an electrically assisted bicycle, a train, an electric vehicle, or the like. BACKGROUND ART An electric power steering device, which is equipped with a motor control device and applies a steering assist force (assist force) to a steering mechanism of a vehicle by a rotational force of a motor, drives the driving force of a motor as an actuator through a reduction gear to a gear or a belt. A steering assist force is applied to the steering shaft or the rack shaft by a transmission mechanism such as the above. Such a conventional electric power steering apparatus performs motor current feedback (FB) control in order to accurately generate the torque of the steering assist force. The feedback control adjusts the motor applied voltage so that the difference between the steering assist command value (current command value) and the detected motor current value becomes small, and the motor applied voltage is generally adjusted by PWM (Pulse Width Width). Modulation) The duty of the control is adjusted. [0003] 　A general configuration of the electric power steering apparatus will be described with reference to FIG. 1. The column shaft (steering shaft, handle shaft) 2 of the steering wheel 1 includes a reduction gear 3, universal joints 4a and 4b, a pinion rack mechanism 5, a tie rod 6a, 6b, and further connected to steering wheels 8L, 8R via hub units 7a, 7b. A steering angle sensor 14 that detects the steering angle θh of the steering wheel 1 and a torque sensor 10 that detects the steering torque Th of the steering wheel 1 are provided on the column shaft 2 to assist the steering force of the steering wheel 1. The motor 20 is connected to the column shaft 2 via the reduction gear 3. Electric power is supplied from the battery 13 to the control unit (ECU) 30 that controls the electric power steering device, and an ignition key signal is input via the ignition key 11. The control unit 30 calculates the current command value of the assist (steering assistance) command based on the steering torque Th detected by the torque sensor 10 and the vehicle speed Vs detected by the vehicle speed sensor 12, and the calculated current command value. The current supplied to the motor 20 is controlled by the voltage control command value Vref that has been compensated for. The steering angle sensor 14 is not essential and may not be provided, and the steering angle (motor rotation angle) θ can be obtained from a rotation sensor such as a resolver connected to the motor 20. [0004] 　The control unit 30 is connected to a CAN (Controller Area Network) 40 that sends and receives various kinds of vehicle information, and the vehicle speed Vs can also be received from the CAN 40. The control unit 30 can also be connected to a non-CAN 41 other than the CAN 40 that exchanges communication, analog/digital signals, radio waves, and the like. [0005] 　In such an electric power steering apparatus, the control unit 30 is mainly composed of a CPU (Central Processing Unit) (including an MPU (Micro Processor Unit) and an MCU (Micro Controller Unit)). The general function to be executed is shown in FIG. 2, for example. [0006] 　The function and operation of the control unit 30 will be described with reference to FIG. 2. The steering torque Th from the torque sensor 10 and the vehicle speed Vs from the vehicle speed sensor 12 are input to the steering assist command value calculation unit 31 to calculate the steering assist command value. The unit 31 calculates the steering assist command value Iref1 based on the steering torque Th and the vehicle speed Vs using an assist map or the like. The calculated steering assist command value Iref1 is added by the adder 32A to the compensation signal CM from the compensator 34 for improving the characteristic, and the added steering assist command value Iref2 is limited by the current limiter 33 to the maximum value. Then, the current command value Irefm whose maximum value is limited is input to the subtraction unit 32B and subtracted from the motor current detection value Im. [0007] 　The deviation current ΔI (=Irefm-Im), which is the subtraction result of the subtraction unit 32B, is subjected to current control such as PI (proportional integration) by the PI control unit 35, and the current-controlled voltage control command value Vref is a modulation signal (triangular wave). (Carrier) CF is input to the PWM control unit 36 ​​to calculate the duty command value, and the motor 20 is PWM-driven via the inverter 37 by the PWM signal for which the duty command value is calculated. The motor current value Im of the motor 20 is detected by the motor current detector 38, input to the subtractor 32B and fed back. [0008] 　The compensator 34 adds the detected or estimated SAT (Self-Aligning Torque) to the inertia compensation value 342 in the adder 344, further adds the convergence control value 341 in the adder 345, and adds the result. The result is input to the adder 32A as the compensation signal CM to improve the characteristics. [0009] 　In recent years, three-phase brushless motors have become the mainstream actuator for electric power steering devices, and since electric power steering devices are in-vehicle products, the operating temperature range is wide and inverters that drive motors are household appliances from a fail-safe perspective. It is necessary to increase the dead time (industrial equipment VR1), and is equal to or more than the predetermined voltage VR2. Is a characteristic that holds a constant limit value DTCa2. The compensation amount limit value DTCa is input to the contact point a1 of the switching unit 252 and the comparing unit 255, and also to the inverting unit 254. Further, the three-phase loss voltage PLB (Vloss_u, Vloss_v, Vloss_w) is input to the comparison units 255 and 256, and is also input to the contact b1 of the switching unit 252. The output “−DTCa” of the inverting unit 254 is input to the contact a2 of the switching unit 253. The contacts a1 and b1 of the switching unit 252 are switched based on the comparison result CP1 of the comparison unit 255, and the contacts a2 and b2 of the switching unit 253 are switched based on the comparison result CP2 of the comparison unit 256. [0065] 　The comparison unit 255 compares the compensation amount limit value DTCa with the three-phase loss voltage PLB, and switches the contacts a1 and b1 of the switching unit 252 according to the following Expression 14. Further, the comparison unit 256 compares the compensation amount limit value −DTCa with the three-phase loss voltage PLB, and switches the contacts a2 and b2 of the switching unit 253 according to the following Expression 15. ( 　　Equation 14) When the three-phase loss voltage PLB≧compensation amount upper limit value: (DTCa), the contact a1 of the switching unit 252 is ON (contact point b2 of the switching unit 253=DTCa) 　　Three-phase loss voltage PLBcompensation amount lower limit value: (-DTCa), contact b2 of switching unit 253 is ON (dead time compensation value DTC =Output of Switching 　Unit 252) Next, FIG. 16 which is the second embodiment of the dead time compensation unit 200 will be described with reference to FIG. The second embodiment (dead time compensating unit 200B) of FIG. 16 has a feature that the compensation accuracy is high even in the low speed/medium speed steering region where the angle and the phase of the phase current are in phase. [0066] 　In the dead time compensating unit 200B of this embodiment, the q-axis steering assist command value i qref corresponding to the steering assist command value Iref2 in FIG. 2 is input, and the motor rotation angle θ and the motor angular velocity ω are input. There is. The dead time compensation unit 200B includes a current control delay model 208, a compensation code estimation unit 209, multiplication units 232, 233d and 233q, an addition unit 221, a phase adjustment unit 207, an inverter applied voltage sensitive gain unit 220, and an angle-dead time (DT). ) Compensation value function units 230U, 230V and 230W, multiplication units 231U, 231V and 231W, a three-phase AC/dq axis conversion unit 203, and a current command value sensitive gain unit 240. [0067] 　The detailed configuration of the dead time compensation unit 200B is shown in FIG. 17, and the q-axis steering assist command value i qref is input to the current control delay model 208. A delay occurs due to a noise filter or the like of the ECU until the current command values ​​i d * and i q * of the dq axes are reflected in the actual current. Therefore, when trying to determine the sign directly from the current command value i q * , timing deviation may occur. In order to solve this problem, the delay of the entire current control is approximated as a first-order filter model to improve the phase difference. The current control delay model 208 of the present embodiment uses T as a filter time constant and is a primary filter of the following Expression 16. [0068] [Equation 16] The 　current command value Icm output from the current control delay model 208 is input to the current command value sensitive gain unit 240 and the compensation code estimation unit 209. The dead time compensation amount may be overcompensated in the low current region, and the current command value sensitive gain unit 240 calculates a gain that reduces the compensation amount according to the magnitude of the current command value Icm (steering assist command value i qref ). Have a function. In addition, noise reduction processing is performed using a weighted average filter so that the gain that reduces the compensation amount does not vibrate due to noise from the current command value Icm (steering assist command value i qref ). [0069] 　The current command value sensitive gain unit 240 has a configuration as shown in FIG. 18, and the current command value Icm is set to the absolute value |Icm| by the absolute value unit 241. The maximum value of the absolute value |Icm| is limited by the input limiting unit 242, and the current command value of the absolute value whose maximum value is limited is input to the weighted average filter 244 via the scale converting unit 243. The current command value Iam whose noise has been reduced by the weighted average filter 244 is added and input to the subtraction unit 245, and the predetermined offset OS is subtracted by the subtraction unit 245. The reason for subtracting the offset OS is to prevent chattering due to the small current command value, and the input value below the offset OS is fixed to the minimum gain. The offset OS is a constant value. The current command value Ias from which the offset OS is subtracted by the subtraction unit 245 is input to the gain unit 246, and the current command value sensitive gain Gc is output according to the gain characteristic as shown in FIG. That is, the initial value of the current command value sensitive gain Gc in the gain unit 246 is Gca1, and then linearly increases to the predetermined value Ias1 of the current command value Ias, and when it exceeds the predetermined value Ias1, the constant value Gca2 is held. [0070] 　The current command value sensitive gain Gc output from the current command value sensitive gain unit 240 has a characteristic as shown in FIG. 20, for example, with respect to the input current command value Icm. That is, the characteristic is that the constant gain Gcc1 is maintained up to the predetermined current Icm1, the linear gain increases from the predetermined current Icm1 to the predetermined current Icm2 (>Icm1), and the constant gain Gcc2 is maintained at the predetermined current Icm2 or higher. [0071] 　The compensation code estimation unit 209 outputs a positive (+1) or negative (-1) compensation code SN with the hysteresis characteristic shown in FIGS. 21A and 21B with respect to the input current command value Icm. The compensation code SN is estimated with reference to the point where the current command value Icm crosses zero, but it has a hysteresis characteristic to suppress chattering. The estimated compensation code SN is input to the multiplication unit 232. When the sign of the dead time compensation value is simply determined from the current sign of the phase current command value model, chattering occurs at low load. Torque ripple occurs when the handle is lightly turned left and right on-center. In order to improve this problem, a hysteresis is provided in the code determination, and the chattering is suppressed by holding the current code except when the code changes beyond the set current value. [0072] 　The current command value sensitive gain Gc from the current command value sensitive gain unit 240 is input to the multiplication unit 232, and the multiplication unit 232 outputs the current command value sensitive gain Gcs (=Gc×SN) multiplied by the compensation code SN. The current command value sensitive gain Gcs is input to the multiplication units 233d and 233q. [0073] 　Further, since the optimum dead time compensation amount changes according to the inverter applied voltage VR, in this embodiment, the voltage sensitive gain Gv corresponding to the inverter applied voltage VR is calculated and the dead time compensation amount is varied. .. The inverter applied voltage sensitive gain unit 220 that inputs the inverter applied voltage VR and outputs the voltage sensitive gain Gv has the configuration shown in FIG. 22, and the inverter applied voltage VR is limited to the maximum positive and negative values ​​by the input limiting unit 221. The inverter applied voltage VRnx with the limited voltage is input to the inverter applied voltage/dead time compensation gain conversion table 222. The characteristics of the inverter applied voltage/dead time compensation gain conversion table 222 are as shown in FIG. 23, for example. That is, the constant gain Gv (=0.7) is maintained up to the predetermined voltage 9.0 [V], and the voltage is linearly (or non-linearly) increased from the predetermined voltage 9.0 [V] to the predetermined voltage 15.0 [V]. It is a characteristic that holds a constant gain Gv (=1.2). The inverter applied voltages 9.0 [V] and 15.0 [V] at the inflection point and the voltage sensitive gains "0.7" and "1.2" are examples and can be changed as appropriate. The calculated voltage sensitive gain Gv is input to the multiplication units 231U, 231V, 231W. [0074] 　Further, when it is desired to advance or delay the dead time compensation timing by the motor angular velocity ω, the phase adjusting unit 207 is provided for the function of calculating the adjustment angle according to the motor angular velocity ω. The phase adjustment unit 207 has characteristics as shown in FIG. 24 in the case of advance angle control, and the calculated phase adjustment angle Δθ is input to the addition unit 221 and added to the detected motor rotation angle θ. The motor rotation angle θm (=θ+Δθ), which is the addition result of the addition unit 221, is input to the angle-dead time (DT) compensation function units 230U, 230V, and 230W, and is also input to the three-phase AC/dq axis conversion unit 203. Is entered. [0075] 　There is a time delay of several tens to hundreds [μs] from the detection of the motor electrical angle and calculation of the Duty command value until the actual reflection in the PWM signal. Since the motor is rotating during this period, a phase shift occurs between the motor electrical angle at the time of calculation and the motor electrical angle at the time of reflection. In order to compensate for this phase shift, the phase is adjusted by advancing according to the motor angular velocity ω. [0076] 　As shown in detail in FIG. 25, the angle-dead time (DT) compensation value function units 230U, 230V, and 230W are 120 in the electrical angle range of 0 to 359 [deg] with respect to the phase-adjusted motor rotation angle θm. Outputs the dead time reference compensation values ​​Udt, Vdt, Wdt for each phase of a rectangular wave whose phase is shifted by [deg]. The dead time compensation value angle function units 230U, 230V, and 230W calculate the dead time compensation value required for the three phases as a function according to the angle, calculate the ECU in real time, and calculate the dead time reference compensation value Udt, Output Vdt and Wdt. The angle function of the dead time reference compensation value varies depending on the dead time characteristic of the ECU. [0077] 　The dead time reference compensation values ​​Udt, Vdt, Wdt are input to the multiplication units 231U, 231V, 231W, respectively, and are multiplied by the voltage sensitive gain Gv. The three-phase dead time compensation values ​​Udtc(=Gv·Udt), Vdtc(=Gv·Vdt), Wdtc(=Gv·Wdt) multiplied by the voltage sensitive gain Gv are input to the three-phase AC/dq axis conversion unit 203. To be done. The three-phase AC/dq axis conversion unit 203 synchronizes the three-phase dead time compensation values ​​Udtc, Vdtc, Wdtc with the two-phase dq axis dead time compensation values ​​v da * and v qa * in synchronization with the motor rotation angle θm . Convert to. The dead time compensation values ​​v da * and v qa * are input to the multiplication units 233d and 233q, respectively, and are multiplied by the current command value sensitive gain Gcs. The multiplication results in the multiplication units 233d and 233q are dead time compensation values ​​v d * and v q * , and the dead time compensation values ​​v d * and v q * are added to the voltage command values ​​vd2 and vq2 by the addition units 123d and 123q, respectively. To be done. [0078] 　Next, FIG. 26, which is the third embodiment of the dead time compensator 200, will be described with reference to FIGS. 10 and 16. The third embodiment (dead time compensating unit 200C) of FIG. 26 has a feature that compensation can be simply inserted even at high speed steering. [0079] 　The dead time compensation unit 200C includes an addition unit 273, a multiplication unit 272, an inverter applied voltage sensitive compensation amount calculation unit 260, a three-phase current command value model 270, a phase current compensation code estimation unit 271, a phase adjustment unit 207, and a three-phase AC/ It is configured by the dq axis conversion unit 203. The motor rotation angle θ is input to the addition unit 273, and the motor angular velocity ω is input to the phase adjustment unit 207. Further, the inverter applied voltage VR is input to the inverter applied voltage sensitive compensation amount calculation unit 260, and the motor rotation angle θm after the phase adjustment calculated by the addition unit 273 is input to the three-phase current command value model 270. [0080] 　The phase adjusting unit 207 is provided for the function of calculating the adjustment angle according to the motor angular velocity ω when it is desired to accelerate or delay the dead time compensation timing by the motor angular velocity ω. The phase adjustment unit 207 has the same characteristics as in FIG. 24 in the case of advance angle control, and the calculated phase adjustment angle Δθ is input to the addition unit 273 and added to the detected motor rotation angle θ. The motor rotation angle θm (=θ+Δθ) after the phase adjustment, which is the addition result of the addition unit 273, is input to the three-phase current command value model 270 and also to the three-phase AC/dq axis conversion unit 203. The reason for providing the phase adjusting unit 207 is the same as in the case of the second embodiment. [0081] 　Since the optimum dead time compensation amount changes according to the inverter applied voltage VR, the dead time compensation amount DTC corresponding to the inverter applied voltage VR is also calculated and varied in this embodiment. The inverter applied voltage sensitive compensation amount calculation unit 260 that inputs the inverter applied voltage VR and outputs the dead time compensation amount DTC has the configuration shown in FIG. 27, and the inverter applied voltage VR is limited by the input limiting unit 261 to the maximum positive and negative values. The maximum value limited inverter applied voltage VRmx is input to the inverter applied voltage/dead time compensation amount conversion table 262. [0082] 　The characteristics of the inverter applied voltage/dead time compensation amount conversion table 262 are as shown in FIG. 28, for example. That is, the maximum value limited inverter applied voltage VRmx is a constant dead time compensation amount DTC1 up to the predetermined inverter applied voltage VR1, and linear (or non-linear) from the predetermined inverter applied voltage VR1 to the predetermined inverter applied voltage VR2 (>VR1). And a constant dead time compensation amount DTC2 is output at a predetermined inverter applied voltage VR2 or higher. [0083] 　The d-axis current command value i d * and the q-axis current command value i q * are input to the three-phase current command value model 270 together with the motor rotation angle θm. The three-phase current command value model 270 is a sine wave three-phase current model in which the dq-axis current command values ​​i d * and i q * and the motor rotation angle θm deviate by 120 [deg] from each other as shown in FIG. The command value Icma is calculated or calculated by a table (see the following formulas 17 to 18). The three-phase current model command value Icma differs depending on the motor type. When the d-axis current command value i ref_d and the q-axis current command value i ref_q are converted from the motor electrical angle θe to the three-phase current command values ​​(U, V, W phases), the following formula 17 is obtained. [0084] [ Equation 17] When each phase current command value is calculated from the above Equation 17, the U-phase current command value model i ref_u , the V-phase current command value model i ref_v, and the W-phase current command value model i ref_w are expressed by the following formula 18, respectively. Represented. [0085] [ 　Equation 18] The table may be a type stored in an EEPROM (Electrically Erasable and Programmable Read-Only Memory) or a type developed in a RAM (Random Access Memory). In use of the number 18, sin [theta e in advance as a table only, the input theta e cos [theta] by using a by 90 ° offset e or calculates, by, for example, to 120 ° offset, calculates the other sin function terms You may. If there is no problem with the ROM capacity or if the command value model is complicated (for example, a pseudo rectangular wave motor), the entire mathematical expression is tabulated. [0086] 　The 3-phase current model command value Icma from the 3-phase current command value model 270 is input to the phase current compensation code estimation unit 271. The phase current compensation code estimation unit 271 compensates the input three-phase current model command value Icma with a hysteresis characteristic as shown in FIGS. 21A and 21B for positive (+1) or negative (−1) compensation. The code SN is output. The compensation code SN is estimated with reference to the point where the three-phase current model command value Icma crosses zero, but it has a hysteresis characteristic to suppress chattering. The estimated compensation code SN is input to the multiplication unit 272. When the sign of the dead time compensation value is simply determined from the current sign of the phase current command value model, chattering occurs at low load. For example, torque ripple occurs when the handle is lightly turned left and right on-center. In order to improve this problem, hysteresis is provided in the code determination (±0.25 [A] in FIG. 21), and the current code is held and chattering is suppressed except when the code changes beyond the set current value. [0087] 　The dead time compensation amount DTC from the inverter applied voltage sensitive compensation amount calculation unit 260 is input to the multiplication unit 272, and the multiplication unit 272 outputs the dead time compensation value DTCa (=DTC×SN) multiplied by the compensation code SN. The dead time compensation value DTCa is input to the three-phase AC/dq axis conversion unit 203, and the three-phase AC/dq axis conversion unit 203 synchronizes with the motor rotation angle θm and the two-phase dead time compensation values ​​v d * and v. Output q * . The dead time compensation values ​​v d * and v q * are added to the voltage command values ​​vd2 and vq2 in the adders 123d and 123q, respectively, and the dead time compensation of the inverter 161 is performed. [0088] 　As described above, in the third embodiment, the dq-axis current command value is converted into a three-phase current model command value, the compensation code is estimated, and the dead time compensation amount of the inverter calculated from the inverter applied voltage is calculated, The dead time compensation value based on the estimated compensation code is feedforward compensated to the voltage command value on the dq axes. A three-phase current model command value is used as the dead time compensation code, and the dead time compensation amount is calculated from the inverter applied voltage VR to determine the magnitude of the current command value (i d * , i q * ) and the inverter applied voltage VR. Depending on the size, the compensation value is variable so that it has an optimum size and direction. [0089] 　It should be noted that FIGS. 10, 16 and 26 showing the embodiment of the dead time compensating unit 200 may be switched and used according to the steering condition, or only one of them may be used. [0090] 　Next, space vector modulation will be described. In the present invention, in order to reduce the number of calculations, the space vector conversion calculation is performed after converting the voltage dimension to the Duty dimension. As space vector modulation section 300 shown in FIG. 30, two-phase Duty value of the dq-axis space (Duty command value Duty_ d ** and Duty_ q ** ) a 3-phase Duty value (Duty_u, Duty_v, Duty_w) to If it has the function of converting and superimposing the third harmonic on the three-phase duty values ​​(Duty_u, Duty_v, Duty_w) and outputting the three-phase duty command values ​​Duty _u * , Duty _v *, Duty _w * Alternatively, for example, the space vector modulation method proposed by the present applicant in Japanese Patent Laid-Open No. 2017-70066, WO/2017/098840 and the like may be used. [0091] 　That is, the space vector modulation will be described by the formula of the voltage. Switching corresponding to sectors #1 to #6, which performs coordinate conversion as shown and controls ON/OFF of FETs (upper arms Q1, Q3, Q5, lower arms Q2, Q4, Q6) of an inverter having a bridge configuration. It has a function of controlling the rotation of the motor by supplying the patterns S1 to S6 to the motor. Regarding the coordinate conversion, in the space vector modulation, the voltage command values ​​Vd3 and Vq3 are converted into the voltage vectors Vα and Vβ in the α-β coordinate system based on the equation (19). The relationship between the coordinate axis used for this coordinate conversion and the motor rotation angle θ is shown in FIG. [0092] [ 　Formula 19] Note that the relationship between the Duty command values ​​Duty_d, Duty_q and the voltage command values ​​Vd3 and Vq3 is expressed by the following formula 20 with the inverter applied voltage being VR. [0093] [Equation 20] Then, there is a relationship such as Equation 21 between the target voltage vector in the dq coordinate system and the target voltage vector in the α-β coordinate system, and the absolute value of the target voltage vector V is preserved. To be done. [0094] [Equation 21] In 　the switching pattern in the space vector control, the output voltage of the inverter corresponds to the switching patterns S1 to S6 of the FETs (Q1 to Q6), and eight types of discrete reference voltage vectors shown in the space vector diagram of FIG. It is defined by V0 to V7 (non-zero voltage vectors V1 to V6 and zero voltage vectors V0 and V7 whose phases are different by π/3 [rad]). Then, the selection of the reference output voltage vectors V0 to V7 and the generation time thereof are controlled. Further, the space vector can be divided into six sectors #1 to #6 by using the six regions sandwiched by the adjacent reference output voltage vectors, and the target voltage vector V is set to the sectors #1 to #6. It belongs to any one and can be assigned a sector number. Which of the target voltage vector V, which is a composite vector of Vα and Vβ, exists in the sector divided into regular hexagons in the α-β space as shown in FIG. It can be obtained based on the rotation angle γ in the −β coordinate system. The rotation angle γ is determined as γ=θ+δ as the sum of the phase δ obtained from the relationship between the rotation angle θ of the motor and the voltage command values ​​Vd3 and Vq3 in the dq coordinate system. [0095] 　FIG. 33 is a digital control by the switching patterns S1, S3, S5 of the inverter in the space vector control, and the switching pulse in the ON/OFF signals S1 to S6 (switching pattern) for the FET in order to output the target voltage vector V from the inverter. The basic timing chart which determines width and its timing is shown. In the space vector modulation, an operation or the like is performed within the sampling period Ts for each defined sampling period Ts, and the operation result is converted into each switching pulse width and its timing in the switching patterns S1 to S6 in the next sampling period Ts. And output. [0096] 　The space vector modulation generates switching patterns S1 to S6 corresponding to the sector number obtained based on the target voltage vector V. FIG. 33 shows an example of the switching patterns S1 to S6 of the FETs of the inverter in the case of sector number #1 (n=1). Signals S1, S3 and S5 represent the gate signals of FETs Q1, Q3 and Q5 corresponding to the upper arm. The horizontal axis represents time, Ts corresponds to the switching cycle, and is divided into eight periods, T0/4, T1/2, T2/2, T0/4, T0/4, T2/2, T1/2. And T0/4. Further, the periods T1 and T2 are times depending on the sector number n and the rotation angle γ, respectively. [0097] 　FIG. 34 is a simulation result by a bench test apparatus simulating an actual vehicle, and when the steering wheel is turned to the right in the middle-speed steering steering state, by applying the dq-axis disturbance estimation observer, FIG. It was confirmed that the torque ripple was smaller because the current ripple was smaller in the dq-axis current waveform of FIG. 34(B) after application as compared with A). The torque ripple during steering was also improved. Although FIG. 34 shows the U phase, the same applies to the other phases. [0098] 　FIG. 35 shows a second embodiment of the dead time compensation system including the dq-axis disturbance estimation observer 600 in correspondence with FIG. 7, and the dq-axis disturbance estimation observer 600 uses the d-axis disturbance estimation observer 600 dA and the q-axis disturbance estimation. It is composed of an observer of 600 qA. Also in the second embodiment, a d-axis dead time compensating unit 200d and a q-axis dead time compensating unit 200q, a d-axis non-interference control unit 140d and a q-axis non-interference control unit 140q having the same configurations and functions as those of the first embodiment. Is provided. Therefore, description of the d-axis dead time compensation unit 200d, the q-axis dead time compensation unit 200q, the d-axis non-interference control unit 140d, and the q-axis non-interference control unit 140q will be omitted. [0099] 　The difference between the second embodiment and the first embodiment is that the dq-axis motor inverse models 602d and 602q have variable parameters, and the cut-off frequencies of the dq-axis LPFs 601d, 603d and 601q, 603d are variable. It is that you are. [0100] 　The d-axis feedback current id and the q-axis feedback current iq are input to the d-axis motor reverse model 602d and the q-axis motor reverse model 602q of the second embodiment, respectively, and the d-axis current command value id * and The q-axis current command value iq * is supplied, and the inductance nominal values ​​Ldn and Lqn are varied according to the d-axis current command value id * and the q-axis current command value iq * . That is, when the motor current increases, the inductance of the motor may fluctuate due to the influence of magnetic saturation depending on the motor used, and the inductance nominal values ​​Ldn and Lqn of the d-axis motor inverse model 602d and the q-axis motor inverse model 602q may be changed. By varying the d-axis current command value id * and the q-axis current command value iq * respectively , the accuracy of the motor model is improved and the compensation accuracy by the disturbance observer is improved. [0101] 　FIGS. 36(A) and 36(B) show configuration examples of the inductance variable control unit that varies the inductance nominal values ​​Ldn and Lqn, respectively, and the d-axis current command value id * and the q-axis current command value iq * are absolute values, respectively. The value parts 602d-1 and 602q-1 are input, and the absolute values ​​|id * | and |iq * | are input to the current-sensitive inductance calculation parts 602d-2 and 602q-2, respectively. The current-sensitive inductance calculation unit 602d-2 outputs the inductance nominal value Ldna with a gradual decrease characteristic as shown in FIG. 37(A), and the current-sensitive inductance calculation unit 602q-2 compares as shown in FIG. 37(B). The inductance nominal value Lqna is output with a sharp decrease characteristic. The inductance nominal values ​​Ldna and Lqna from the current sensitive inductance calculation units 602d-2 and 602q-2 are input to the limiters 602d-3 and 602q-3, respectively, and the upper and lower limit values ​​are limited by the limiters 602d-3 and 602q-3, respectively. Inductance nominal values ​​Ldn and Lqn are output. [0102] 　The characteristics shown in FIGS. 37(A) and 37(B) are examples, and may vary non-linearly depending on the motor used. [0103] 　Further, in the d-axis disturbance estimation observer 600d, the LPFs 601d and 603d, and the d-axis motor inverse model 602d are input with the motor rotation speed rpm. The cutoff frequency Fc is variable. Similarly, in the q-axis disturbance estimation observer 600q, the LPFs 601q and 603q and the q-axis motor inverse model 602q are input with the motor rotational speed rpm, and the LPFs 601q and 603q and the q-axis motor inverse model 602q are input according to the motor rotational speed rpm. Each cutoff frequency Fc is variable. The cutoff frequency variable control unit that changes the cutoff frequency Fc has the same configuration. For example, the LPF 601d has the configuration shown in FIG. That is, the motor rotation speed rpm is input to the absolute value unit 601d-1, and the absolute value |rpm| is input to the rotation speed sensitive frequency calculation unit 601d-2. The rotation speed sensitive frequency calculation unit 601d-2 has the characteristics shown in FIG. 39. For example, the rotation speed sensitive frequency calculation unit 601d-2 linearly increases up to 4000 rpm, and the cutoff frequency Fca is constant at 1200 [Hz] at 4000 rpm or more. The cutoff frequency Fca from the rotation speed sensitive frequency calculation unit 601d-2 is input to the limiter 601d-3, and the limiter 601d-3 outputs the cutoff frequency Fc whose upper and lower limit values ​​are limited. As described above, when followability is required in accordance with the motor rotation speed rpm, such as during high-speed steering, the cutoff frequency Fc of the LPF 601d or the like is increased to improve responsiveness. Further, when the steering feeling is important when the hand feeling is important, each cutoff frequency Fc is reduced because the noise is included in the input signal, so that the noise resistance is improved. [0104] 　Note that the characteristics shown in FIG. 39 are examples, and may vary non-linearly depending on the actual machine tuning and system. Further, there is a relationship of the following expression 22 between each time constant T of the LPF and the motor inverse model and the cutoff frequency Fc. [0105] [ 　Equation 22] The q-axis disturbance estimation observer 600q has almost the same configuration as the d-axis disturbance estimation observer 600d, and the LPFs 601q and 603q for band limitation and the motor rotation speed rpm are input to the q-axis motor inverse model 602q. The q-axis current command value iq* is input to the q-axis motor inverse model 602q. The variable inductance control section and variable cutoff frequency control section that perform variable control are as described above. [0106] 　FIG. 40 is a simulation result by a bench test apparatus simulating an actual vehicle, and when the steering wheel is turned to the right in the middle-speed steering steering state, by applying the dq-axis disturbance estimation observer, FIG. It was confirmed that the torque ripple was smaller because the current ripple was smaller in the dq-axis current waveform of FIG. 40(B) after application as compared with A). The torque ripple during steering was also improved. Although FIG. 40 shows the U phase, the same applies to the other phases. [0107] 　In the second embodiment described above, the inductance nominal values ​​Ldn and Lqn of the motor reverse models 602d and 602q are changed according to the dq-axis current command values ​​id * and iq * , and the LPF and the motor reverse speed are changed according to the motor rotation speed rpm. Although each cutoff frequency Fc of the model is changed, as shown in FIG. 41 corresponding to FIG. 35, the inductance nominal values ​​of the motor inverse models 602d and 602q are set according to the dq axis current command values ​​id * and iq *. Only Ldn and Lqn may be changed (third embodiment). That is, in the third embodiment, the d-axis current command value id * is input to the motor inverse model 602d, the q-axis current command value iq * is input to the motor inverse model 602q, and the inductance nominal values ​​Ldn and Lqn are changed. .. In the dq-axis disturbance estimation observers 600dA and 600qA of the second embodiment, the cutoff frequency Fc is not changed according to the motor rotation speed rpm. [0108] 　Further, in the fourth embodiment shown in FIG. 42, the motor rotation speed rpm is input to the LPFs 601d and 603d and the motor inverse model 602d, and also to the LPFs 601q and 603q and the motor inverse model 602q. However, the dq axis current command values ​​id * and iq * are not input to the motor inverse models 602d and 602q, and the inductance nominal values ​​Ldn and Lqn are not variable. That is, in the fourth embodiment, the cutoff frequencies of the LPFs 601d and 603d and the motor reverse model 602d are changed according to the motor rotation speed rpm, and the cutoff frequencies of the LPFs 601q and 603q and the motor reverse model 602q are changed. [0109] 　FIG. 43 shows a fifth embodiment of the present invention, in which a filter is shared for the purpose of reducing the processing time of the MPU (MCU), suppressing the program creation time, and the like. That is, the LPFs 603d and 603q are deleted, the currents idb and iqb from the dq axis non-interference models 610d and 610q are input to the adder 607d and the subtractor 607q, respectively, and the addition result Vdc and the subtraction result Vqc are input to the LPFs 601d and 601q, respectively. doing. As a result, the number of filters can be halved and the required processing time can be shortened. [0110] 　Also in the fifth embodiment, it is possible to change the inductance nominal values ​​Ldn and Lqn based on the dq-axis current command values ​​id * and iq * and the cutoff frequency based on the motor rotation speed rpm. [0111] 　Although the column-type electric power steering device has been described above, the present invention can be similarly applied to a rack-type electric power steering device. Explanation of symbols [0112] 1 steering wheel 2 column shaft (steering shaft, steering shaft) 10 torque sensor 20, 100 motor 30 control unit (ECU) 35 PI control unit 36, 160 PWM control unit 37, 161 inverter 110 angle detection unit 130, 203 three-phase AC/ dq-axis converter 140 dq-axis non-interference controller 200, 200A to 200C dq-axis dead time compensator 204 three-phase command voltage calculator 205 voltage detection delay model 207 phase adjuster 210 midpoint voltage estimator 220 inverter applied voltage sensitive gain part 250 compensation amount limiting unit 260 inverter application voltage sensitive compensation amount calculating section 270 3-phase current command value model 271 phase current compensation code estimator 300 Space Vector Modulator 301 Two-Phase/3-Phase Converter 302 Third Harmonic Superimposing Unit 600 dq-axis disturbance estimation observer 600d, 600dA, 600dB, 600dC, 600dD 　　　　　　　　　　d-axis disturbance estimation observer 600q, 600qA, 600qB, 600qC, 600qD 　　　　　　　　　　q Axis disturbance estimation observer 700 Duty calculation unit The scope of the claims [Claim 1] The three-phase brushless motor is vector-controlled based on the d-axis current command value and the q-axis current command value, and the three-phase current detection value of the three-phase brushless motor is converted into a d-axis feedback current and a q-axis feedback current. The d-axis current command value and the q-axis current command value are fed back, the deviation voltage of the feedback is converted into a two-phase duty command value, the two-phase duty command value is converted into a three-phase, and vector control is performed via an inverter. In the motor control device, the dq axis dead time compensating section for calculating the dead time compensation value of the inverter to perform the dead time compensation, the dq axis current command value, the motor speed, the dq axis feedback current and the deviation. enter the voltage, and the dq Jikugairan estimation observer calculates and outputs dq Jikugairan compensation value comprises a, by adding the dq Jikugairan compensation value to said difference voltage, compensating a dead time compensation of the inverter A motor control device, which estimates a disturbance that cannot be exhausted. [Claim 2] The motor control according to claim 1, wherein the dq-axis disturbance estimation observer includes a d-axis disturbance estimation observer targeting a d-axis control target and a q-axis disturbance estimation observer targeting a q-axis control target. apparatus. [Claim 3] The d-axis disturbance estimation observer receives a first d-axis LPF that receives the added value of the d-axis disturbance compensation value and the deviation voltage, a d-axis motor inverse model that receives the d-axis feedback current, and the q-axis. The d-axis non-interference model for inputting the feedback current and the motor angular velocity, the second d-axis LPF for inputting the output of the d-axis non-interference model, the d-axis motor reverse model and the second d-axis LPF. deviation, a subtraction unit for subtracting from the output of the first d-axis LPF, a d-axis limiting unit limits in accordance with the output of the subtraction unit to the motor speed, in the configuration, the q Jikugairan estimation observer Is a first q-axis LPF for inputting the added value of the q-axis disturbance compensation value and the deviation voltage, a q-axis motor inverse model for inputting the q-axis feedback current , the q-axis feedback current and the motor angular velocity. The q-axis non-interference model, the second q-axis LPF inputting the output of the q-axis non-interference model, the q-axis motor inverse model, and the q-axis second LPF, The motor control device according to claim 2 , comprising an adder/subtractor that subtracts from the output of the q-axis LPF of 1, and a q-axis limiter that limits the output of the adder/subtractor according to the motor rotation speed . [Claim 4] The d-axis non-interacting models, is composed of a d-axis multiplier unit for multiplying the q-axis feedback current and the motor angular velocity, the d-axis gain unit to the gain multiplying the output of the d-axis multiplier unit, the q Jikuhi An interference model includes a q-axis multiplication unit that multiplies the d-axis feedback current and the motor angular velocity, a first q-axis gain unit that multiplies the output of the q-axis multiplication unit by a gain, and a first multiplication unit that multiplies the motor angular velocity by a gain. The motor control device according to claim 3, comprising a second q-axis gain section and a second q-axis addition section that adds the first q-axis gain section and the second q-axis gain section. .. [Claim 5] 5. The d-axis limiting unit is configured by a d-axis sensitive gain unit that is sensitive to the motor rotation speed and a d-axis compensation amount limiting unit that limits the maximum value of the d-axis compensation amount. Motor controller. [Claim 6] 6. The q-axis limiting unit is configured by a q-axis sensitive gain unit that is sensitive to the motor rotation speed, and a q-axis compensation amount limiting unit that limits the maximum value of the q-axis compensation amount. The motor control device according to claim 1. [Claim 7] 7. The motor control device according to claim 1, wherein the dq axis dead time compensating unit is configured to calculate the dq axis dead time compensation value with a configuration according to a steering state. [Claim 8] The dq axis dead time compensator calculates a three phase dead time reference compensation value based on the motor rotation angle, processes the three phase dead time reference compensation value with a gain and a sign, and performs three phase AC/dq axis conversion. The motor control device according to claim 7, wherein the dq axis dead time compensation value is obtained. [Claim 9] The dq-axis dead time compensator estimates the three-phase detection voltage based on the three-phase motor terminal voltage, and calculates the three-phase correction command voltage calculated from the duty command value and the difference between the three-phase detection voltage from the inverter. The motor control device according to claim 7, wherein a loss voltage due to dead time is estimated, the loss voltage is compensated, and the dq axis dead time compensation value is obtained. [Claim 10] The dq-axis dead time compensator estimates the compensation code of the three-phase current model command value obtained by converting the dq-axis current command value into a three-phase current command value model, and calculates the dead time compensation amount based on the inverter applied voltage. 8. The motor control device according to claim 7, wherein a value calculated by multiplying the dead time compensation amount by the compensation code is converted into two phases to obtain the dq axis dead time compensation value. [Claim 11] The electric power steering apparatus according to any one of claims 1 to 10, further comprising: a space vector modulator that converts the 2-phase duty command value into a 3-phase signal and outputs a voltage command value in which a third harmonic is superimposed. . [Claim 12] The electric power steering apparatus according to any one of claims 1 to 11, wherein a parameter of the dq-axis disturbance estimation observer is varied based on the dq-axis current command value and the motor rotation speed. [Claim 13] The cutoff frequencies of the first d-axis LPF and the second d-axis LPF are varied based on the motor speed, and the inductance component of the d-axis motor inverse model is varied based on the d-axis current command value. In addition, the cutoff frequencies of the first q-axis LPF and the second q-axis LPF are varied based on the motor rotation speed, and the inductance of the q-axis motor inverse model is based on the q-axis current command value. The motor control device according to claim 3, wherein the components are variable. [Claim 14] When the motor rotation speed is high, the cutoff frequencies of the first d-axis LPF, the second d-axis LPF, the first q-axis LPF, and the second q-axis LPF are increased to respond. Of the cutoff frequencies of the first d-axis LPF and the second d-axis LPF, and the first q-axis LPF and the second q-axis LPF when the motor rotation speed is low. The motor control device according to claim 13, wherein the motor control device is configured to have a low resistance to enhance noise resistance. [Claim 15] As the d-axis current command value increases, the d-axis motor inverse model inductance component is changed so as to decrease, and as the q-axis current command value increases, the q-axis motor inverse model inductance component decreases. 15. The motor control device according to claim 13, which is variable. [Claim 16] An electric power steering device, comprising the motor control device according to any one of claims 1 to 15, which applies an assist torque to a steering mechanism of a vehicle by driving the three-phase brushless motor. [Claim 17] The three-phase brushless motor is vector-controlled based on the d-axis current command value and the q-axis current command value, and the three-phase current detection value of the three-phase brushless motor is converted into a d-axis feedback current and a q-axis feedback current. Each deviation current between the d-axis current command value and the q-axis current command value is current-controlled to obtain a d-axis voltage command value and a q-axis voltage command value, and the d-axis voltage command value and the q-axis voltage command value are obtained. In a motor controller that controls the three-phase brushless motor via an inverter with a three-phase duty command value based on the d-axis dead-time compensator and q-axis, which calculates a dead-time compensation value of the inverter to perform dead-time compensation. A dead time compensator is input to the d-axis current command value, the d-axis voltage command value, the d-axis feedback current, the q-axis feedback current, the motor angular velocity and the motor rotation speed, and the d-axis disturbance compensation value is calculated. And the d- axis disturbance estimation observer to be output as the q-axis current command value, the q-axis voltage command value, the q-axis feedback current, the d-axis feedback current, the motor angular velocity and the motor speed, and the q-axis. A d-axis disturbance estimation observer for calculating and outputting a disturbance compensation value, and the d-axis disturbance estimation observer, The output of the d-axis non-interference model is input to the d-axis non-interference model that inputs the q-axis feedback current and the motor angular velocity, and the first d-axis addition value of the d-axis disturbance compensation value and the d-axis voltage command value. The d-axis LPF that inputs the added second d-axis addition value, the d-axis motor inverse model that inputs the d-axis feedback current, and the output of the d-axis motor inverse model are subtracted from the output of the d-axis LPF. A d-axis subtraction unit that obtains a d-axis deviation voltage, a d-axis sensitive gain unit that multiplies the d-axis deviation voltage by a gain according to the motor speed, and the d-axis sensitive gain unit by limiting the output of the d-axis sensitive gain unit. A q- axis disturbance estimation unit configured to output a d-axis disturbance compensation value, the q- axis disturbance estimation observer inputting the d-axis feedback current and the motor angular velocity, and the q-axis disturbance compensation unit. Q-axis LPF for inputting a q-axis subtraction value obtained by subtracting the output of the q-axis non-interference model from the first q-axis addition value of the q-axis voltage command value and the q-axis for inputting the q-axis feedback current. A motor inverse model, a q-axis subtraction unit that obtains a q-axis deviation voltage by subtracting the output of the q-axis motor inverse model from the output of the q-axis LPF, and a gain of the q-axis deviation voltage according to the motor rotation speed. And a d-axis compensation value limiting unit that limits the output of the q-axis sensitive gain unit and outputs the q-axis disturbance compensation value. A motor control device, characterized in that the q-axis disturbance compensation value is added to the axis voltage command value and the q-axis disturbance compensation value is added to the q-axis voltage command value to estimate a disturbance that cannot be compensated by dead time compensation of the inverter. [Claim 18] The motor control device according to claim 17, wherein the gain of the d-axis sensitive gain unit is sensitive to the motor rotation speed. [Claim 19] 19. The motor control device according to claim 17, wherein the gain of the q-axis sensitive gain unit is sensitive to the motor rotation speed. [Claim 20] An electric power steering device, comprising the motor control device according to any one of claims 17 to 19 and applying an assist torque to a steering system of a vehicle by driving the three-phase brushless motor.