Automatic nut burying device based on high-precision servo drive and control system thereof

Abstract

The invention relates to the technical field of automatic assembly equipment, in particular to an automatic nut burying device based on high-precision servo drive and a control system thereof, and aims to solve the problem that the positioning precision is degraded due to micro-step error accumulation caused by dead time under the high-frequency start-stop working condition. The device comprises a mechanical execution unit, a permanent magnet synchronous servo motor, a servo driver, an absolute photoelectric encoder and a central cooperative controller, according to the control method, a load dependent type dead zone online identification and compensation framework is constructed, an extended Kalman filter is triggered through a load torque observer to estimate dead zone parameters in real time, compensation voltage matched with a current load is dynamically generated and superposed to a q-axis instruction, and error accumulation is effectively restrained. And the accumulated position deviation does not exceed 0.1 mm after continuous operation for one thousands of cycles, so that the position precision and the beat stability in the nut burying process are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of automated assembly equipment technology, specifically an automatic nut-laying device and its control system based on high-precision servo drive. Background Technology

[0002] In modern precision manufacturing and automated assembly, high-precision actuators based on servo drives have become the core unit for achieving micron-level positioning control. Especially in scenarios such as electronic device packaging, automotive parts assembly, and consumer electronics manufacturing, automatic nut-inserting devices place extremely stringent requirements on repeatability, response speed, and long-term operational stability. These devices typically use permanent magnet synchronous servo motors as the power source, employing closed-loop vector control strategies to precisely regulate torque and position, ensuring force-position coordination during nut insertion. In such high-speed reciprocating motion systems, the servo driver needs to frequently perform start-stop operations, typically exceeding 500 times per minute, posing a significant challenge to the dynamic response capability of the current loop.

[0003] To improve system efficiency and suppress switching losses, current mainstream servo drives generally adopt pulse width modulation inverter structures and introduce microsecond-level dead time in the upper and lower bridge arm drive signals to prevent shoot-through short circuits. While this dead time is necessary for hardware safety, it inevitably leads to distortion of the inverter output voltage waveform, resulting in stator current step response distortion. Specifically, when the dead time is 1.2 microseconds, the q-axis current build-up process will have a delay of approximately 80 microseconds. Although this delay has a negligible impact in a single motion cycle, under high-frequency start-stop conditions, its effect accumulates cycle by cycle, forming the so-called microstep error. Experimental data shows that after one thousand consecutive cycles, this error can accumulate to a staggering five millimeters, far exceeding the tolerance range allowed by precision assembly (typically less than 0.1 millimeters), severely weakening the equipment's process reliability and product yield.

[0004] To address the aforementioned issues, existing technical solutions often employ offline calibration combined with a fixed lookup table method for dead-zone compensation. This method pre-determines the compensation voltage values ​​corresponding to different current directions and amplitudes under specific load conditions and stores them in the driver's non-volatile memory. During operation, it outputs the corresponding compensation amount based on a lookup table according to real-time current commands. This strategy is effective to some extent under conditions of constant or gently changing loads, and its design logic is based on the assumption that system parameters remain statically constant. However, in the actual operation of automatic nut insertion, the load torque often experiences step disturbances due to factors such as batch differences in materials, fixture wear, or sudden changes in nut insertion resistance. Typical torque fluctuations can exceed 10% of the rated value. In such dynamic load scenarios, the fixed lookup table method lacks the ability to perceive and adapt to real-time operating conditions, and cannot accurately match the required compensation voltage, leading to frequent undercompensation or overcompensation. This not only fails to effectively suppress micro-step errors but may also induce system oscillations or exacerbate current harmonics, further deteriorating control performance.

[0005] Ultimately, the limitations of existing dead-zone compensation mechanisms stem from a profound contradiction between their static compensation paradigm and the inherently dynamic coupling characteristics of servo systems. The dead-zone effect is not an isolated hardware nonlinearity, but a time-varying function tightly coupled with motor inductance, back electromotive force, and load torque. When the load undergoes a sudden change, the rate of change of current changes accordingly, and the voltage loss caused by the dead zone adjusts accordingly. Fixed lookup table methods cannot capture this intrinsic relationship, causing the compensation strategy to become disconnected from the actual physical process. Furthermore, traditional methods lack an online mapping mechanism between load states and compensation parameters, resulting in the system remaining in an open-loop compensation state when facing complex and variable assembly conditions, making it difficult to meet the robustness requirements of high-cycle, high-consistency production.

[0006] Therefore, to address the above problems, an automatic nut-burying device and its control system based on high-precision servo drive are proposed. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of the prior art by setting up an automatic nut-laying device and its control system based on high-precision servo drive, thereby solving the technical problems mentioned in the background art.

[0008] To address the above technical problems, the following technical solution is adopted: This invention provides an automatic nut-inserting device based on high-precision servo drive, including a mechanical execution unit for completing the nut-pressing action, a permanent magnet synchronous servo motor for providing driving power, a servo driver for controlling the operation of the servo motor, an absolute photoelectric encoder for detecting the operating position of the servo motor, a current sampling module for collecting the operating current of the servo motor, a central coordination controller for coordinating the collaborative work of the various modules, and a multi-source information fusion module for integrating position and current information;

[0009] The mechanical actuator consists of a linear guide rail providing linear guidance, a ball screw pair that converts rotary motion into linear motion, and a nut press-in head for pressing in the nut. The nut press-in head is mounted on the nut slider of the ball screw pair, and the nut slider is in sliding engagement with the linear guide rail. The screw shaft of the ball screw pair is coaxially rigidly connected to the output shaft of the permanent magnet synchronous servo motor through a zero-backlash diaphragm coupling.

[0010] The absolute photoelectric encoder is connected to the shaft of the permanent magnet synchronous servo motor to collect the motor's rotation angle position information in real time. The current sampling module is connected to the power supply circuit of the permanent magnet synchronous servo motor to collect the motor's operating current information. The signal output terminals of the absolute photoelectric encoder and the current sampling module are both connected to the signal input terminal of the multi-source information fusion module. The signal output terminal of the multi-source information fusion module is connected to the signal input terminal of the central coordinating controller. The control output terminal of the central coordinating controller is connected to the control input terminal of the servo driver. The power output terminal of the servo driver is connected to the power input terminal of the permanent magnet synchronous servo motor.

[0011] Preferably, the linear guide rail adopts a cross roller structure.

[0012] Preferably, the front end of the nut press-in head integrates a dual closed-loop feedback unit for pressure and displacement. The pressure sensor is a miniature strain gauge sensor, and the displacement sensor is a laser triangulation distance sensor. The signals from both are transmitted to the multi-source information fusion module after filtering, amplification, and analog-to-digital conversion.

[0013] Preferably, the central coordinating controller is equipped with a dead-zone compensation voltage limiting unit to rigidly limit the amplitude of vcomp, ensuring that its absolute value does not exceed 0.5 volts.

[0014] Preferably, it further includes a compensation validity verification module, which continuously monitors... Measured shaft current and Shaft current command value Tracking error between ;

[0015] If within fifty consecutive control cycles If the root mean square value exceeds 0.8 amperes, the current compensation is deemed to have failed, and the system automatically switches to the backup compensation strategy, using the exponential smoothing filter output based on historical effective compensation values ​​as the temporary compensation voltage, until the extended Kalman filter reconverges.

[0016] Preferably, the device also includes a control system for an automatic nut-burying device. The control system includes a current sampling module that uses a dual-channel isolated Hall current sensor to collect the instantaneous current of the stator winding U / V phase, and has a built-in current harmonic suppression unit.

[0017] The servo driver has a built-in three-phase full-bridge inverter circuit, and the upper and lower bridge arms use silicon carbide MOSFETs. The servo driver integrates a vector control core, dual-mode dead zone compensation, extended Kalman filter, load torque observer and compensation effectiveness dynamic verification module.

[0018] The absolute photoelectric encoder output signal is processed by five-point center differential and three-point forward differential to obtain angular velocity and angular acceleration, respectively.

[0019] The multi-source information fusion module fuses location, current, and pressure data from three sources and outputs the fused observation value to the load torque observer module.

[0020] The load torque observer module constructs a mechanical dynamics equation based on fused observations and solves the load torque using a sliding mode observer. This is done over two consecutive cycles. At this time, the extended Kalman filter is reinitialized;

[0021] The extended Kalman filter module models the dead zone equivalent voltage missing quantity as a time-varying state variable, adopts an adaptive noise covariance adjustment strategy, and uses multi-source fused observations to online recursively estimate the dead zone parameters.

[0022] The dual-modal dead-zone compensation module combines the dead-zone parameter estimation value with... The axis current command symbol generates the main compensation voltage. When the root mean square value of the current tracking error exceeds 0.8A for 50 consecutive cycles, it switches to the LSTM standby compensation mode. The pre-training parameters, dataset gradients, and online incremental learning triggering conditions of this LSTM standby mode have been clearly quantified to ensure that those skilled in the art can reproduce and implement it.

[0023] The central coordinating controller communicates with the servo driver via industrial Ethernet and has a built-in compensation voltage limiting unit.

[0024] Preferably, it further includes a compensation validity verification module, which continuously monitors... Measured shaft current and Shaft current command value Tracking error between ;

[0025] If within fifty consecutive control cycles If the root mean square value exceeds 0.8 amperes, the current compensation is deemed to have failed, and the system automatically switches to the backup compensation strategy, using the exponential smoothing filter output based on historical effective compensation values ​​as the temporary compensation voltage, until the extended Kalman filter reconverges.

[0026] Preferably, the dead-zone compensation voltage generation module generates the voltage based on the output of the extended Kalman filter. Estimated value and current Shaft current command value The sign generates the compensation voltage. :

[0027] when hour, ;

[0028] when hour, ;

[0029] This compensation voltage is directly superimposed on the output of the vector control core module. The shaft voltage command value forms the final value applied to the inverter. Shaft modulation voltage .

[0030] The state vector of the extended Kalman filter module is defined as follows:

[0031] ;

[0032] in and They are respectively shaft and The axial flux linkage component, Vdead is the equivalent value caused by the dead zone. Shaft voltage missing quantity; the state equation is obtained by discretizing the voltage equation of the permanent magnet synchronous servo motor:

[0033] ; ; ;

[0034] in For stator resistance, This refers to the flux linkage amplitude of the permanent magnet. and Before compensation shaft and Shaft voltage command value; observation equation set as follows:

[0035] :

[0036] Its Jacobian matrix is ​​obtained through the inductance parameters of the permanent magnet synchronous servo motor. and Explicit construction.

[0037] Preferably, when the load torque changes over two consecutive control cycles satisfy At that time, among them To achieve the rated output torque of the permanent magnet synchronous servo motor, the central coordinating controller sends a re-initialization command to the extended Kalman filter module, forcibly setting the current dead zone parameter estimate to the initial guess value and initiating a new round of parameter convergence.

[0038] Preferably, the load torque observer module monitors the position signal output by the absolute photoelectric encoder. Perform a five-point center difference operation to obtain the angular velocity. The discrete estimate is expressed as:

[0039] ;

[0040] in To control the cycle, a value of 100 microseconds is set; then... Perform a three-point forward difference operation to obtain the angular acceleration:

[0041] ;

[0042] Will Substituting the rotational dynamics equations of the permanent magnet synchronous servo motor:

[0043] ;

[0044] in This represents the total moment of inertia of the permanent magnet synchronous servo motor rotor and load, referred to the permanent magnet synchronous servo motor shaft. The coefficient of viscous friction is... For electromagnetic torque, External load torque; electromagnetic torque Depend on Measured shaft current Torque constant of permanent magnet synchronous servo motor The product is determined, that is:

[0045] Therefore, the observed load torque value is obtained:

[0046] .

[0047] The beneficial effects of this invention are:

[0048] This invention constructs an online dead-zone identification and compensation architecture with an extended Kalman filter as the core and a load torque observer as the trigger condition, achieving dynamic binding between the compensation voltage and the real-time load state. This architecture abandons the open-loop compensation mode of the traditional static lookup table method, establishing a physical correlation model between dead-zone effect, current response, and load disturbance. This enables the servo drive system to maintain micron-level position control accuracy even when facing the dual challenges of high-frequency start-stop and sudden load changes. The technical solution is not only applicable to automatic nut-laying devices but can also be extended to other precision assembly, laser processing, and semiconductor manufacturing equipment with stringent dynamic accuracy requirements for servo systems. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] In the attached diagram:

[0051] Figure 1 This is a side view of the automatic nut-burying device described in this invention;

[0052] Figure 2 This is a flowchart illustrating the control system of the automatic nut-laying device described in this invention.

[0053] Figure 3 This is a top view of the automatic nut-burying device described in this invention;

[0054] Legend:

[0055] 1. Mechanical actuator; 2. Permanent magnet synchronous servo motor; 3. Servo driver; 4. Absolute photoelectric encoder; 5. Current sampling module; 6. Central coordinating controller; 7. Linear guide; 8. Ball screw pair; 9. Nut press-in head; 10. Screw shaft; 11. Coupling; 12. Multi-source information fusion module. Detailed Implementation

[0056] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0057] Specific implementation examples are given below.

[0058] like Figures 1-3As shown, this invention provides an automatic nut-burying device and its control system based on high-precision servo drive. It includes a mechanical execution unit 1, a permanent magnet synchronous servo motor 2, a servo driver 3, an absolute photoelectric encoder 4, a current sampling module 5, a central coordination controller 6, and a multi-source information fusion module 12.

[0059] The mechanical actuator 1 consists of a linear guide rail 7, a ball screw assembly 8, and a nut pressing head 9. The screw shaft 10 of the ball screw assembly 8 is rigidly connected to the output shaft of the permanent magnet synchronous servo motor 2 via a zero-backlash diaphragm coupling 11. The front end of the nut pressing head 9 integrates a dual closed-loop feedback unit for pressure and displacement. The pressure sensor uses a miniature strain gauge sensor with a measurement range of 0-500N and a measurement accuracy of ±1N, used to collect the contact force and pressing force during the nut pressing process in real time. The displacement sensor uses a laser triangulation distance sensor with a measurement accuracy of ±0.001mm, synchronously collecting the actual displacement of the pressing head. After the pressure and displacement signals are processed by the signal conditioning circuit (including filtering, amplification, and analog-to-digital conversion modules), they are transmitted to the multi-source information fusion module 12 via a high-speed SPI interface to achieve force-position coordinated monitoring during the pressing process, avoiding problems such as excessive or shallow pressing or workpiece damage caused by uneven workpiece material or nut size deviation.

[0060] The ball screw pair, also known as the ball screw assembly or ball screw assembly, is a precision helical transmission component that uses balls as rolling elements to achieve the conversion between rotational and linear motion. Its core consists of a screw, nut, balls, and a reversing device (ball returner). It relies on rolling friction instead of sliding friction and has the characteristics of high efficiency, high precision, and long life. It is one of the core transmission components of precision machinery.

[0061] Coupling 11 is a zero-backlash diaphragm coupling.

[0062] The current sampling module 5 collects the instantaneous current of the stator winding U / V phase, with a sampling frequency of 20kHz and a built-in current harmonic suppression unit;

[0063] The servo driver 3 integrates a three-phase full-bridge inverter circuit, using silicon carbide MOSFETs in both upper and lower arms. It features a switching frequency of 10kHz, a dead time of 1.2μs, and integrates a vector control core, dual-mode dead time compensation, an extended Kalman filter, a load torque observer, and a dynamic verification module for compensation effectiveness. The servo driver 3 also includes a built-in temperature monitoring unit, employing a surface-mount thermocouple sensor attached between the source and drain of the silicon carbide MOSFET to acquire the junction temperature of the power device in real time. The measurement range is -40℃ to 150℃, with a measurement accuracy of ±2℃. The temperature signal is transmitted to the central co-controller 6 after analog-to-digital conversion. When the junction temperature exceeds 120℃, the central co-controller automatically triggers a thermal protection strategy: reducing the inverter switching frequency from 10kHz to 8kHz and adjusting the dead time compensation coefficient. (where T is the actual junction temperature), by reducing switching losses and optimizing compensation voltage, the power devices are prevented from breaking down due to overheating; when the junction temperature drops below 100℃, the initial switching frequency and compensation coefficient are automatically restored.

[0064] The absolute photoelectric encoder has a resolution of no less than 20 bits. The output signal is processed by five-point center differential and three-point forward differential to obtain angular velocity and angular acceleration respectively.

[0065] The multi-source information fusion module 12 fuses data from three sources: position, current, and pressure, and outputs the fused observation value to the load torque observer module. The multi-source information fusion module 12 employs an adaptive weighted fusion algorithm, and its weight adjustment logic is as follows: First, the signal-to-noise ratio (SNR) of each source signal is calculated, and the position signal... (Useful position signal amplitude / noise amplitude), current signal (Useful current signal amplitude / noise amplitude), pressure signal (Amplitude of the force signal / amplitude of the noise); then, the reliability coefficient of each signal is calculated based on the SNR, where the reliability coefficient = the SNR of that signal / the sum of the SNRs of the three source signals; finally, the weights are dynamically assigned, with the weight value = reliability coefficient × preset weight range coefficient (position signal 0.3-0.5, current signal 0.2-0.4, force signal 0.2-0.3). When the SNR of a certain source signal is lower than 10dB, its weight is automatically reduced to below 0.1 to avoid distorted signals affecting the fusion result.

[0066] Accurate calculation of load torque is achieved through a closed-loop logic of dynamic model construction, sliding mode surface design, reaching law control, and torque inverse calculation. The specific implementation process is as follows: First, based on the mechanical transmission characteristics of the motor and load and Newton's laws of motion, an integrated mechanical dynamic equation of motor-load is established. The core expression of the dynamic equation is:

[0067] ,in, The electromagnetic torque of the motor is collected by a current sensor from the stator. shaft current With the inherent torque coefficient of the motor pass Calculation yields ( These are the factory-calibrated parameters of the motor, which are known quantities.

[0068] The load torque to be calculated is the observation target in this embodiment; The equivalent rotational inertia of the motor and the load was determined through previous static identification experiments and is a known fixed parameter. The real-time motor speed is acquired by a speed sensor and is an observable variable. ω is the motor angular acceleration, which is derived from two consecutive cycles of speed signals collected by the speed sensor through a first-order differential algorithm combined with low-pass filtering; B is the equivalent damping coefficient of the transmission system, which was identified and determined through previous no-load experiments and is a known fixed parameter.

[0069] Secondly, the sliding surface of the sliding mode observer is designed to achieve accurate tracking of the observation state, and the virtual rotational speed observation value inside the observer is set to... Define sliding surface variables ,in The core control objective of the sliding mode observer is to minimize the sliding surface variable, based on the actual rotational speed observed by the rotational speed sensor. Rapid convergence to 0, thus achieving the virtual rotational speed observation. For actual speed Error-free tracking;

[0070] Secondly, to ensure the variable of the sliding surface Fast and stable convergence is achieved by using an exponential reaching law as the control rule for the sliding mode observer. The expression for the exponential reaching law is: ,in, For the reaching law gain coefficient, The attenuation coefficient is preset based on the system's dynamic response requirements and stability requirements (in this embodiment). The value range is 50~100. The value range is 5 to 10, and can be adjusted and optimized according to the actual system. For a sign function, when hour ,when hour By using this reaching law control rule, when the sliding surface variable... At that time, the virtual speed observation value is adjusted by outputting a negative control quantity through the approach law. Lower, shrink and The difference;

[0071] when At that time, the virtual speed observation value is adjusted by outputting a positive control quantity through the approach law. Increase, decrease and The difference is calculated, and the robustness of the sliding mode variable structure is utilized to offset the effects of parameter perturbations and external disturbances in the transmission system on the observation accuracy. Finally, when the sliding surface variable... When converged to 0, the virtual rotational speed observation value Compared with actual speed Completely identical, at this time Substituting into the aforementioned mechanical dynamics equations, the observed load torque value is obtained through reverse derivation: ;

[0072] Meanwhile, during the operation of the observer, the change in load torque over two consecutive control cycles is calculated in real time. (in For the current control cycle, (for the previous control cycle), the preset trigger threshold is ( (The rated torque of the motor is a known quantity). When detected... Furthermore, when this condition is met for two consecutive cycles, the extended Kalman filter is reinitialized to correct the model error of the sliding mode observer and ensure the long-term stability and accuracy of the load torque observation.

[0073] Time-varying state modeling design: Addressing the dynamic variation of dead-zone equivalent voltage loss in power devices with operating conditions, this approach breaks away from traditional fixed-error modeling methods and incorporates the dead-zone equivalent voltage loss... (include axis, Axis voltage missing component , The state variables are modeled as time-varying EKF state variables; simultaneously, combined with the electromagnetic dynamics equations of the permanent magnet synchronous motor, an extended state-space model is constructed, and the state variables of the extended state-space model are... Defined as:

[0074] ;

[0075] in, , stator axis, shaft current, The rotor speed of the motor is one of the three core operating state variables of the motor. , Dead zone equivalent voltage missing amount axis, Shaft components (i.e., instantaneous state variables); the state equations are derived from the motor voltage balance equations. After considering the influence of dead-zone voltage loss, the voltage balance equations are modified as follows:

[0076] ;

[0077] ;

[0078] (in , For stator axis, Ideal control voltage for shaft. For stator resistance, , for axis, Shaft inductor, (These are permanent magnet flux linkages, all with known parameters or previously identified parameters). Based on this, the state equations are derived as follows:

[0079] ,in To control the input vector, This is process noise (following a Gaussian distribution). This is the state transition function (a nonlinear function derived from the modified voltage balance equation).

[0080] To adapt to the dynamic changes in process noise and observation noise under different operating conditions, an adaptive noise covariance adjustment method based on observation residuals is adopted.

[0081] First, define the observation residual. ,in for Multi-source fusion observations at all times For the observation matrix, for The predicted value at time step is based on the estimation result of the previous time step; then it is based on the residual. Calculate residual covariance Finally, the process noise covariance is adjusted using preset adaptive rules. and observation noise covariance :when Greater than the preset threshold If this occurs, it indicates that the current observation noise or model error is large, and the error will be... Increase by 1.2 to 1.5 times. Increase by 1.1 to 1.3 times (the specific multiple is determined based on system debugging) to reduce the weight bias between model predictions and observation data;

[0082] when Less than or equal to When the noise and model error are relatively small, it indicates that the noise and model error are small. , Restore to the baseline value to ensure estimation accuracy; where the baseline... , Obtained through offline calibration Based on the residual range allowed by the system (in this embodiment) The value ranges from 0.01 to 0.05.

[0083] After removing high-frequency noise by performing a first-order low-pass filter (cutoff frequency of 100Hz) on the raw signals from each sensor, a weighted fusion algorithm is used to obtain... (The weights are calibrated based on the accuracy of each sensor; the current sensor has a weight of 0.4, the speed sensor has a weight of 0.3, and the voltage sensor has a weight of 0.3.)

[0084] EKF, based on the aforementioned extended state model, adaptive noise covariance, and multi-source fused observations, performs online recursion through a prediction-update closed loop:

[0085] Prediction phase: based on The optimal estimate of time Calculated through state equations Predicted value at time Simultaneously calculate the prediction error covariance. ( (The Jacobian matrix of the state transition matrix, used to linearize the nonlinear state equations).

[0086] Update phase: Calculate Kalman gain Combined with observation residuals Updated The optimal estimate of time Simultaneously update the estimated error covariance ( (the identity matrix);

[0087] Ultimately from Extract the dead zone equivalent voltage missing quantity , By combining the mapping relationship of the dead-time characteristics of power devices, the dead time can be calculated in reverse. Pipe pressure drop Core dead zone parameters (mapping relationship based on device datasheet and offline experimental calibration, i.e.) , ,in , , , , , (These are calibration coefficients).

[0088] The load torque observer module constructs mechanical dynamics equations based on fused observations and solves for the load torque using a sliding mode observer. This is done over two consecutive cycles. At this time, the extended Kalman filter is reinitialized;

[0089] The Extended Kalman Filter (EKF) module models the dead-zone equivalent voltage loss as a time-varying state variable and employs an adaptive noise covariance adjustment strategy to recursively estimate the dead-zone parameters online using multi-source fused observations. The EKF module incorporates a parameter forgetting factor λ (values ​​0.95-0.99), which works by applying a forgetting weight to historical dead-zone parameter estimates during the parameter recursion update process. ,in This is the new estimate for the current period.

[0090] When the load torque change rate Furthermore, if this condition persists for 100ms, the system is considered to be in a steady-state condition, at which point the forgetting factor is determined. Setting it to 0.99 reduces the interference of historical parameters on the current estimate; when the load torque suddenly changes ( )hour, We set the value to 0.95 to increase the weight of new observation data and accelerate parameter convergence.

[0091] The dual-modal dead-zone compensation module combines the dead-zone parameter estimate with The axis current command symbol generates the main compensation voltage. When the root mean square value of the current tracking error exceeds 0.8A for 50 consecutive cycles, it switches to the LSTM standby compensation mode. The output of the LSTM standby compensation mode is mapped to the final compensation voltage through a lookup table method. This lookup table is generated offline by the pre-trained LSTM network during the servo driver initialization phase and is stored in non-volatile memory.

[0092] The LSTM backup mode of the dual-modal dead-time compensation module has a network structure that includes an input layer (3 neurons, corresponding to...) Estimated value The structure includes shaft current command, load torque, hidden layer (2 layers, 32 neurons per layer, ReLU activation function), and output layer (1 neuron, corresponding to optimized compensation voltage).

[0093] Pre-training parameter configuration: The Adam optimizer is used during training, with an initial learning rate of 0.001, and training parameters are applied every 50 training cycles. (Iteration cycle) The learning rate decays to 80% of the previous stage; the total number of iterations is set to 1000 epochs, and after 10 consecutive iterations... Loss function (mean squared error) Stable below When the training has converged, it is determined that the training has stopped.

[0094] Pre-training dataset design: Data was obtained through 1000 experiments under different operating conditions. The operating condition gradient quantization is as follows: load torque is divided in 0.5 N·m intervals, covering 0.5-3 N·m (including 6 gradients: 0.5 N·m, 1.0 N·m, 1.5 N·m, 2.0 N·m, 2.5 N·m, and 3.0 N·m); ambient temperature is divided in 10℃ intervals, covering -10℃ to 45℃ (including 7 gradients: -10℃, 0℃, 10℃, 20℃, 30℃, 40℃, and 45℃); silicon carbide MOSFET junction temperature is divided in 25℃ intervals, covering 25℃ to 120℃ (including 5 gradients: 25℃, 50℃, 75℃, 100℃, and 120℃). Each dataset includes input parameters (…). Estimated value (shaft current command, load torque) and corresponding effective compensation voltage tag, with data sampling frequency consistent with the control cycle (10kHz);

[0095] Online incremental learning mechanism: The similarity of new working conditions is calculated using the Euclidean distance method. The specific formula is as follows: ;

[0096] in, Input parameters for the current operating condition ( Estimated value (shaft current command, load torque), The mean of the input parameters for each working condition in the pre-training dataset; when the calculated... Furthermore, under this operating condition, the compensation effectiveness verification module determines that the compensation is effective (within 50 consecutive control cycles). Root mean square value of shaft current tracking error When the data is in the training set, the set of operating condition data will be automatically added to the training set, and incremental training will be started. , After updating the network parameters, the data is stored in non-volatile memory.

[0097] The central coordinating controller 6 communicates with the servo drive 3 via industrial Ethernet and has a built-in compensation voltage limiting unit to limit the amplitude. Changes in adjacent periods .

[0098] The linear guide 7 adopts a crossed roller structure with a rated dynamic load of 12kN and a repeatability better than ±1μm. The ball screw assembly 8 has a lead of 5mm, a preload grade of C3, and an axial stiffness of not less than 800N / μm. The nut press-in head 9 is mounted on the screw nut seat via a quick-change clamp, and its front end integrates a pressure sensor and a displacement limit switch for real-time monitoring of the press-in force and prevention of overtravel. The screw shaft 10 of the ball screw assembly 8 is rigidly connected to the output shaft of the permanent magnet synchronous servo motor 2 via a zero-backlash diaphragm coupling 11, ensuring that the transmission chain has no elastic backlash.

[0099] The permanent magnet synchronous servo motor 2 is a three-phase eight-pole structure with a rated power of 750W, a rated speed of 3000rpm, a rated output torque of 2.39N·m, and a rotor inertia of [missing information]. The permanent magnet synchronous servo motor's stator windings employ a distributed winding process, with a phase resistance Rs of 1.85Ω and a d-axis inductance... and Shaft inductor The permanent magnet flux linkage amplitudes are 8.2 mH and 10.5 mH, respectively. The value is 0.312Wb. The rotor shaft of the permanent magnet synchronous servo motor 2 is equipped with an absolute photoelectric encoder 4, which has a single-turn resolution of 20 bits (i.e., 1,048,576 counting pulses / revolution) and a multi-turn counting range of ±4096 revolutions. The position signal is transmitted to the central co-controller 6 through a high-speed serial peripheral interface (SPI) at a clock frequency of 1MHz, with a communication delay of less than 500ns.

[0100] The current sampling module 5 consists of two isolated Hall effect current sensors, connected in series in the U-phase and V-phase stator winding circuits of the permanent magnet synchronous servo motor, respectively. The sensor bandwidth is 100kHz, the nonlinearity is less than ±0.5%, the output voltage range is ±5V, and the corresponding measured current range is ±15A. The sampling signal is digitized by a 16-bit analog-to-digital converter (ADC), and the sampling frequency is set to 20kHz, i.e., per control cycle... Perform a two-phase current acquisition. The obtained raw current data is sent to the vector control core module inside the servo driver 3 via an isolated digital interface.

[0101] Servo driver 3 employs a three-phase full-bridge inverter topology, with both upper and lower bridge arm power switching devices being 1200V / 30A silicon carbide metal-oxide-semiconductor field-effect transistors (SiCMOSFETs). These devices feature extremely low on-resistance (…). With its fast switching characteristics (turn-on / turn-off times both less than 50ns), it effectively reduces switching losses and improves high-frequency response. The inverter switching frequency is fixed at 10kHz, and the dead time is uniformly configured at 1.2μs to prevent shoot-through faults in the upper and lower bridge arms. The servo driver 3 integrates a dedicated permanent magnet synchronous servo motor 2 control chip with a main frequency of 200MHz, supporting floating-point operations and hardware-accelerated coordinate transformation functions.

[0102] The central coordinating controller 6 is an industrial-grade embedded computer, equipped with a dual-core ARM Cortex-A72 processor and running a real-time Linux operating system (PREEMPTRT patch). The central coordinating controller 6 establishes a bidirectional communication link with the servo driver 3 via an EtherCAT industrial Ethernet interface, with a communication cycle of 1ms, used for issuing position commands, reading status variables, and triggering parameter reinitialization events. Simultaneously, the central coordinating controller 6 receives position data and raw current sampling values ​​from the absolute photoelectric encoder 4 via an independent high-speed SPI bus, ensuring that the underlying control loop is unaffected by communication jitter from higher layers.

[0103] The servo drive 3's internal functional modules include a vector control core module, a dead-zone compensation voltage generation module, an extended Kalman filter module, and a load torque observer module. The vector control core module executes a standard field-oriented control (FOC) strategy: first, it sets the instantaneous currents of the U and V phases... and pass Transformation into stillness coordinate system and Combined with the rotor electrical angle feedback from the absolute photoelectric encoder 4 Perform the Park transformation to obtain the rotation. coordinate system Axis current components and Axis current components . Shaft current command value Set to zero (to achieve maximum torque / current ratio control). Shaft current command value It is calculated by the outer loop position controller based on the position error. and Perform proportional-integral calculations with their respective instruction values. Adjustment, generation Shaft voltage command value and Shaft voltage command value .

[0104] The dead-zone compensation voltage generation module receives the real-time dead-zone equivalent voltage missing value estimated by the extended Kalman filter module. and in combination with the current situation Shaft current command value The symbol is used to generate the compensation voltage. ;

[0105] The specific logic is as follows: If ,but ;like ,but The compensation voltage is directly superimposed on the output of the vector control core module. The shaft voltage command value is used to form the final value applied to the pulse width modulation (PWM) module. Shaft modulation voltage . The shaft modulation voltage remains at This compensation mechanism ensures that the average fundamental voltage output of the inverter can accurately offset the voltage loss caused by the dead zone, thereby eliminating... Delay effect during shaft current establishment.

[0106] The Extended Kalman Filter (EKF) module is based on the nonlinear discrete state-space model of the permanent magnet synchronous servo motor 2, and it reduces the equivalent time caused by the dead zone. The missing shaft voltage is modeled as a time-varying state variable. Its state variable is defined as follows: , ,in and They are respectively shaft and Shaft flux linkage component. The state equation is obtained by discretizing the voltage equation of the permanent magnet synchronous servo motor 2:

[0107] ;

[0108] ;

[0109] ;

[0110] in, The electrical angular velocity of the permanent magnet synchronous servo motor 2 is determined by the position signal from the absolute photoelectric encoder 4. The result was obtained through the five-point central difference operation:

[0111] ;

[0112] The observation equation is set as follows Its Jacobian matrix Explicitly constructed using the inductance parameters of the permanent magnet synchronous servo motor 2:

[0113] Steady-state operating conditions ( (Ignoring transient terms) Jacobian matrix:

[0114] ;

[0115] Transient operating condition (retaining s terms) Jacobian matrix:

[0116] ;

[0117] Simplified version of surface-mount PMSM ( (steady state)

[0118] ;

[0119] Where: Jacobian matrix Based on permanent magnet synchronous servo motor (PMSM) Construction of a rotational coordinate system mathematical model, state variables Output variables ;

[0120] The Jacobian matrix, which represents the rate of change of the output variable y with respect to the state variable x, has a dimension of 2×3 and is the core matrix for stability analysis of motor control strategies.

[0121] d-axis stator voltage, stator winding at The voltage component in the axial direction, in volts (V), is given by the PWM module of the motor controller.

[0122] q-axis stator voltage, stator winding at The voltage component in the axial direction, in volts (V), is given by the PWM module of the motor controller.

[0123] d-axis stator current, stator winding in The current component in the axial direction, measured in amperes (A), is collected and fed back in real time by a current sensor;

[0124] q-axis stator current, stator winding in The current component in the axial direction, measured in amperes (A), is collected and fed back in real time by a current sensor;

[0125] EKF completes one prediction-correction cycle within each control interruption: first, based on the state estimate from the previous moment... With input , , Perform state prediction to obtain prior estimates Subsequently, the measured current was used , Calculate the innovation vector and apply it via Kalman gain. Update posterior state estimation The filter's initial state is set to... , , (As an initial guess).

[0126] The load torque observer module calculates the external load torque in real time based on the principles of mechanical dynamics. First, the position signal emitted by the absolute photoelectric encoder 4... Angular velocity is obtained by performing a five-point center difference operation. And then Angular acceleration is obtained by performing a three-point forward difference operation. :

[0127] ;

[0128] Will Substituting into the rotational dynamics equations:

[0129] ;

[0130] in, The measured value is the total moment of inertia of the rotor and load of permanent magnet synchronous servo motor 2 referred to the shaft of permanent magnet synchronous servo motor 2. B is the coefficient of viscous friction, with a calibrated value of [value missing]. Electromagnetic torque Depend on Measured shaft current With permanent magnet synchronous servo motor 2 torque constant ( The product of ) is used to determine the load torque observation. Therefore, the observed load torque value is obtained:

[0131] ;

[0132] The observation is updated at a frequency of 10 kHz and is used to determine load mutation events.

[0133] When the load torque change within two consecutive control cycles meets the requirements ( That is, the threshold is When the Vdead state is reset to 0.05V, the central coordinating controller 6 sends a re-initialization command to the EKF module, forcibly resetting the Vdead state to 0.05V and starting a new round of parameter convergence process.

[0134] To improve system robustness, the central coordinating controller 6 incorporates a dead-time compensation voltage limiting unit. This unit imposes dual constraints on vcomp(k):

[0135] Firstly, amplitude hard limiting; ;

[0136] Secondly, the rate of change is limited. This design effectively prevents current oscillations or inverter saturation caused by sudden changes in compensation voltage.

[0137] In addition, the system is equipped with a compensation validity verification module. This module continuously calculates... Axis current tracking error And calculate its root mean square value within a sliding window (50 control cycles). .

[0138] like If the current EKF estimation fails, the system automatically switches to a backup compensation strategy: using the exponentially smoothed output of historical effective compensation values ​​as a temporary measure. Smoothing factor ,Right now:

[0139] ;

[0140] in This is the last valid compensation value. Waiting for EKF to converge again (via monitoring). Once the rate of change of the estimated value falls below the threshold of 0.001V / cycle, the system automatically switches back to the main compensation path.

[0141] The control method and flow of the automatic nut-burying device are as follows: Figure 3 As shown, the specific execution steps are as follows:

[0142] Step 1: After the device is powered on, the initialization program is executed. The parameters J, B, Rs, Ld, Lq, λm, and Trated of the permanent magnet synchronous servo motor 2 are loaded from the non-volatile memory to the memory of the servo driver 3, and the control cycle is configured. The current sampling frequency is 20kHz, and the communication parameters and initial state of the absolute photoelectric encoder are as follows:

[0143] Step two: Start the vector control loop. The outer loop position controller generates a q-axis current command according to the preset trajectory. The mechanical actuator 1 is driven to move downward along the Z-axis, causing the nut press-in head 9 to contact the nut to be assembled.

[0144] Step 3: Within each control cycle, run the load torque observer module in real time to calculate... The multi-source information fusion module 12 synchronously acquires the position signal of the absolute photoelectric encoder, the U / V phase current signal of the current sampling module, and the pressing force and displacement signal of the nut pressing head. It calculates the signal-to-noise ratio and reliability coefficient of each signal through an adaptive weighted fusion algorithm, and outputs the fused observation value after dynamically allocating weights.

[0145] Step 4, make a judgment Is it greater than 0.239 N·m? If so, send a re-initialization command to the EKF module.

[0146] Step 5: The EKF module uses the current observation data... , , Execute the prediction-correction loop and output the updated value. Estimated value;

[0147] The extended Kalman filter module dynamically adjusts the noise covariance matrix based on the load torque change rate. At that time, the corresponding part of the Q matrix will be... The elements of the state are increased to three times their original value, and the forgetting factor λ is set to 0.95; when Furthermore, when the duration is 100ms, λ is set to 0.99, and the Vdead estimate is recursively updated based on the multi-source fusion observations.

[0148] Step six, the dead zone compensation voltage generation module, based on... and Symbol generation After being limited, it is superimposed on ,form ;

[0149] The dual-mode dead-time compensation voltage generation module generates voltages based on the estimated Vdead value and The shaft current command symbol generates the main compensation voltage, and the compensation effectiveness is dynamically verified in real time by the module. Axis current tracking error; synchronously acquire silicon carbide MOSFET junction temperature data from the servo driver temperature monitoring unit, and adjust the compensation coefficient synchronously if the junction temperature exceeds 120℃. When the root mean square error of the current tracking exceeds 0.8A for 50 consecutive cycles, switch to the LSTM backup compensation mode, call the pre-trained network or the parameters after online incremental learning to optimize the compensation voltage, and superimpose it onto the... Shaft voltage command.

[0150] Step 7, and After inverse Park transform and SVPWM modulation, a three-phase drive signal is generated to control the switching action of SiCMOSFET.

[0151] Step 8: Repeat steps 3 through 7 until the nut is fully pressed in and returns to its original position, completing one assembly cycle.

[0152] In a preferred embodiment, the noise covariance matrices Q and R of the EKF can be adaptively adjusted according to the operating conditions. The process noise covariance matrix Q is a diagonal matrix. ( , , ),in Initially set to When the rate of change of load torque is detected (pass When performing three-point difference estimation, it is determined to be a high-dynamic disturbance condition. Increase to To speed up Vdead's tracking; when And if it lasts for more than 100ms, recovery Up to nominal value. Observation noise covariance matrix Its elements are dynamically set according to the q-axis current amplitude: when ( )hour, Increase to 4 times the nominal value (the nominal value is) This is to reduce the impact of low signal-to-noise ratio current signals on the filter.

[0153] To verify the technical effectiveness of this invention, an engineering testing platform was built for comparative experiments. The testing platform configuration was consistent with the aforementioned parameters, simulating a typical working condition of sudden resistance change during nut screwing in: in the pressing stage of the 500th assembly cycle, the load torque was increased from 1.5 N·m to 2.8 N·m in 0.2 seconds by an electromagnetic brake, representing a change of 86.7%. Two comparison groups were set up:

[0154] The embodiment adopts the load-dependent dead-zone online identification and compensation architecture of the present invention;

[0155] The comparative method uses the traditional fixed lookup table method, that is, open-loop compensation is performed according to the pre-calibrated load-dead zone compensation voltage mapping table. The table contains 5 load ranges (0-0.5, 0.5-1.0, 1.0-1.5, 1.5-2.0, 2.0-2.5 N·m), and the compensation voltage in each range is constant.

[0156] Under the aforementioned abrupt change conditions, 1200 assembly cycles were run continuously, and the absolute positional deviation at the end of each cycle was recorded (accuracy ±0.5 μm, based on a laser interferometer). The experimental results are summarized in the table below:

[0157] Test item Example Comparative example Maximum position deviation for one cycle 0.03 mm 0.18 mm Average deviation for first 10 cycles after mutation 0.025 mm 0.41 mm Total accumulated position error 0.09 mm 4.7 mm Compensation voltage update response time <8ms Not updated q-axis current tracking RMS error 0.32A 1.05A

[0158] Data shows that during sudden load changes, the comparative example, unable to dynamically adjust the compensation voltage, experienced a lag in its q-axis current response, leading to significant position overshoot and accumulated errors. In contrast, the embodiment quickly reconstructed the Vdead estimate using EKF and completed compensation adjustments within 8ms, effectively suppressing error growth. In a 1200-cycle endurance test, the embodiment's cumulative error was only 0.09mm, far exceeding the design target of 0.1mm, while the comparative example's error continued to accumulate, eventually exceeding the process allowable range.

[0159] Furthermore, the long-term stability of the system was tested under steady-state conditions (constant load torque of 1.8 N·m). After 2000 continuous cycles, the standard deviation of the position deviation in the example was ±1.2 μm, demonstrating excellent repeatability. Simultaneously, the compensation voltage vcomp remained stable at 0.18 V under steady-state conditions, with fluctuations less than ±0.005 V, indicating good convergence of the EKF estimation.

[0160] In summary, this invention constructs a closed-loop, adaptive dead-zone compensation system by deeply integrating load torque observation, extended Kalman filtering, and dynamic compensation generation mechanisms. This system not only achieves a physical binding between the dead-zone effect and real-time load status but also ensures reliable operation under extreme conditions through multiple robust designs (amplitude limiting, backup strategies, and covariance adaptation). All control algorithms are implemented in real-time on the servo driver 3 embedded platform, with a computational load below 70%, meeting the real-time requirements of industrial environments. This technical solution has been successfully applied to an automated nut-embedding production line for automotive electronic control unit (ECU) housings, increasing the yield rate from 92.3% to 99.8%, fully validating its engineering practical value.

[0161] In the description of this invention, it should be understood that the terms front and back, left and right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0162] Of course, those skilled in the art should understand that in this technical solution, the term "one" should be understood as at least one or more, that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple. The term "one" should not be construed as a limitation on the quantity.

[0163] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art under the technical guidance of the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An automatic nut-laying device based on high-precision servo drive, characterized in that: It includes a mechanical actuator (1) for completing the nut pressing action, a permanent magnet synchronous servo motor (2) for providing driving power, a servo driver (3) for controlling the operation of the servo motor, an absolute photoelectric encoder (4) for detecting the operating position of the servo motor, a current sampling module (5) for collecting the operating current of the servo motor, a central coordination controller (6) for coordinating the coordinated operation of the various modules, and a multi-source information fusion module (12) for integrating position and current information and for integrating position, current and pressing force data. The mechanical actuator (1) consists of a linear guide rail (7) that provides linear guidance, a ball screw pair (8) that converts rotational motion into linear motion, and a nut press-in head (9) for pressing in the nut. The nut press-in head (9) is mounted on the nut slider of the ball screw pair (8), and the nut slider is in sliding engagement with the linear guide rail (7). The screw shaft (10) of the ball screw pair (8) is coaxially rigidly connected to the output shaft of the permanent magnet synchronous servo motor (2) through a zero-backlash diaphragm coupling (11). The absolute photoelectric encoder (4) is connected to the shaft of the permanent magnet synchronous servo motor (2) to collect the motor rotation angle position information in real time. The current sampling module (5) is connected to the power supply circuit of the permanent magnet synchronous servo motor (2) to collect the motor working current information. The signal output terminals of the absolute photoelectric encoder (4), the current sampling module (5), and the pressure and displacement dual closed-loop feedback unit at the front end of the nut pressing head (9) are all connected to the signal input terminal of the multi-source information fusion module (12). The signal output terminal of the multi-source information fusion module (12) is connected to the signal input terminal of the central coordinating controller (6). The control output terminal of the central coordinating controller (6) is connected to the control input terminal of the servo driver (3). The power output terminal of the servo driver (3) is connected to the power input terminal of the permanent magnet synchronous servo motor (2).

2. The automatic nut-burying device based on high-precision servo drive according to claim 1, characterized in that: The linear guide (7) adopts a cross roller structure.

3. The automatic nut-burying device based on high-precision servo drive according to claim 1, characterized in that: The front end of the nut press-in head (9) integrates a pressure and displacement dual closed-loop feedback unit. The pressure sensor adopts a micro strain gauge sensor, and the displacement sensor adopts a laser triangulation distance sensor. The signals of the two are transmitted to the multi-source information fusion module (12) after filtering, amplification and analog-to-digital conversion.

4. The automatic nut-laying device based on high-precision servo drive according to claim 1, characterized in that: It also includes a compensation validity verification module, which continuously monitors... Measured value of shaft current iq and Shaft current command value Tracking error between ; If within fifty consecutive control cycles If the root mean square value exceeds 0.8 amperes, the current compensation is deemed to have failed, and the system automatically switches to the backup compensation strategy, using the exponential smoothing filter output based on historical effective compensation values ​​as the temporary compensation voltage, until the extended Kalman filter reconverges.

5. A control system for an automatic nut-laying device, applied to the automatic nut-laying device as described in any one of claims 1 to 4, characterized in that, The control system includes the current sampling module (5) which uses a dual-channel isolated Hall current sensor to collect the instantaneous current of the stator winding U / V phase, and has a built-in current harmonic suppression unit; The servo driver (3) has a built-in three-phase full-bridge inverter circuit. The servo driver (3) integrates a vector control core, dual-mode dead zone compensation, extended Kalman filter, load torque observer and compensation effectiveness dynamic verification module. The absolute photoelectric encoder (4) outputs a signal and obtains angular velocity and angular acceleration; The multi-source information fusion module (12) fuses the three-source data of position, current and pressure, and outputs the fused observation value to the load torque observer module; The load torque observer module constructs mechanical dynamics equations based on fused observations, while the sliding mode observer calculates load torque based on motor dynamics equations, using motor speed and current as inputs. The extended Kalman filter module models the dead zone equivalent voltage missing quantity as a time-varying state variable, adopts an adaptive noise covariance adjustment strategy, and uses multi-source fused observations to online recursively estimate the dead zone parameters. The dual-modal dead-zone compensation module combines the dead-zone parameter estimation value with... The axis current command symbol generates the main compensation voltage. When the root mean square value of the current tracking error exceeds 0.8A for 50 consecutive cycles, it switches to the LSTM standby compensation mode. The central coordinating controller (6) communicates with the servo driver (3) via industrial Ethernet and has a built-in compensation voltage limiting unit.

6. The control system of the automatic nut-laying device according to claim 5, characterized in that: It also includes a compensation validity verification module, which continuously monitors... Measured shaft current and Shaft current command value Tracking error between ; If within fifty consecutive control cycles If the root mean square value exceeds 0.8 amperes, the current compensation is deemed to have failed, and the system automatically switches to the backup compensation strategy, using the exponential smoothing filter output based on historical effective compensation values ​​as the temporary compensation voltage, until the extended Kalman filter reconverges.

7. The control system of the automatic nut-laying device according to claim 5, characterized in that: The dead-zone compensation voltage generation module generates voltage based on the output of the extended Kalman filter. Estimated value and current Shaft current command value The sign generates the compensation voltage. : when hour, ; when hour, ; This compensation voltage is directly superimposed on the output of the vector control core module. The shaft voltage command value forms the final value applied to the inverter. Shaft modulation voltage ; The state vector of the extended Kalman filter module is defined as follows: ； in and They are respectively shaft and The axial flux linkage component, Vdead is the equivalent value caused by the dead zone. Shaft voltage missing quantity; the state equation is obtained by discretizing the voltage equation of the permanent magnet synchronous servo motor (2): ； ； ； in For stator resistance, This refers to the flux linkage amplitude of the permanent magnet. and Before compensation shaft and Shaft voltage command value; observation equation set as follows: ； Its Jacobian matrix is ​​obtained through the inductance parameters of the permanent magnet synchronous servo motor (2). and Explicit construction.

8. The control system of the automatic nut-laying device according to claim 5, characterized in that: When the load torque change within two consecutive control cycles satisfy At that time, among them For the rated output torque of the permanent magnet synchronous servo motor (2), the central coordinating controller (6) sends a re-initialization command to the extended Kalman filter module, forcibly setting the current dead zone parameter estimate to the initial guess value, and starting a new round of parameter convergence process.

9. The control system of the automatic nut-burying device according to claim 5, characterized in that: The load torque observer module measures the position signal output by the absolute photoelectric encoder (4). Perform a five-point center difference operation to obtain the angular velocity. The discrete estimate is expressed as: ； in To control the cycle, a value of 100 microseconds is set; then... Perform a three-point forward difference operation to obtain the angular acceleration: ； Will Substituting the rotational dynamics equations of the permanent magnet synchronous servo motor (2): ； Where J is the total moment of inertia of the rotor and load of the permanent magnet synchronous servo motor (2) referred to the shaft of the permanent magnet synchronous servo motor (2). The coefficient of viscous friction is... For electromagnetic torque, External load torque; electromagnetic torque Depend on Measured shaft current Torque constant of permanent magnet synchronous servo motor (2) The product is determined, that is: Therefore, the observed load torque value is obtained: .

10. The automatic nut-burying device based on high-precision servo drive according to claim 5, characterized in that: The central coordinating controller (6) is equipped with a dead zone compensation voltage limiting unit to hard limit the amplitude of vcomp and ensure that its absolute value does not exceed 0.5 volts.

Images