A high-precision positioning control method and system for a torque motor of a robot joint

Abstract

This invention relates to the field of robot control technology, specifically to a high-precision positioning control method and system for a torque motor used in robot joints. The method includes: acquiring the robot joint's operating data and performing DC-DC removal processing to obtain a ripple current sequence; performing variational mode decomposition on the ripple current sequence to obtain multiple intrinsic mode components (IMCs), and evaluating the frequency domain dispersion index of each IMC; determining the phase-lock confidence level of each IMC based on the motor pole-slot fit and mechanical angular position to eliminate external fluctuation interference; determining the dynamic compensation gain coefficient by combining the frequency domain dispersion index and the phase-lock confidence level, and performing reverse torque compensation on the torque motor based on the dynamic compensation gain coefficient. This invention solves the technical problems of torque motor cogging torque drift with operating conditions and interference miscompensation, improving the positioning stability of robot joints during low-speed operation.

Description

Technical Field

[0001] This invention relates to the field of robot control technology. More specifically, this invention relates to a high-precision positioning control method and system for torque motors used in robot joints. Background Technology

[0002] With the increasing popularity of collaborative robots and precision assembly robots, more and more robot joints are adopting torque motor direct drive technology in order to eliminate backlash error and mechanical compliance problems caused by traditional reducers. These motors usually have a high number of pole pairs and can output large torque at low speeds to meet the dynamic response requirements of robot joints.

[0003] However, the inherent cogging torque of the torque motor is the main obstacle affecting the positioning accuracy of the direct drive joint. The cogging torque originates from the periodic change in permeability between the rotor permanent magnet and the stator slots. In the low-speed creep or high-precision positioning and holding conditions commonly encountered in robot joints, the cogging torque will cause the motor speed to fluctuate and the positioning to jitter, which will seriously affect the machining accuracy and end effector stability.

[0004] Currently, traditional compensation schemes mostly use a pre-set lookup table method, which involves reverse cancellation based on the torque waveform tested at the factory. However, the pre-set compensation tables are often static and cannot adapt to mechanical wear or magnetic drift after long-term robot operation. At the same time, robot joints are subject to various dynamic interferences such as load changes and environmental vibrations when performing tasks. Existing frequency domain analysis techniques are unable to accurately separate the cogging torque component, which is fixed with position, from the mixed time-varying interference signals. This limitation of the technique can easily lead to miscompensation, causing system oscillations and limiting the high-precision performance of robot joints under complex working conditions. Summary of the Invention

[0005] To address the technical problems of torque motor compensation schemes lacking adaptability and being unable to accurately isolate external fluctuation interference, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a high-precision positioning control method for a torque motor for a robot joint, comprising: acquiring the running data of the robot joint and performing DC-DC removal processing on the running data to obtain a ripple current sequence reflecting the characteristics of the current ripple.

[0007] Variational mode decomposition is performed on the ripple current sequence to obtain multiple intrinsic mode components, and the frequency domain dispersion index of each intrinsic mode component is evaluated based on the distribution of the spectral energy of each intrinsic mode component relative to the center frequency.

[0008] Based on the motor pole slot fit relationship of the robot joint and the mechanical angle position, the instantaneous phase sequence of each intrinsic mode component is extracted, and the correlation between the instantaneous phase sequence and the theoretical cogging torque phase sequence in the time dimension is analyzed to determine the phase locking confidence of each intrinsic mode component in order to eliminate external fluctuation interference.

[0009] By combining the frequency domain dispersion index and phase lock confidence, the dynamic compensation gain coefficient is determined, and reverse torque compensation is performed on the torque motor based on the dynamic compensation gain coefficient.

[0010] This invention combines variational mode decomposition technology with phase verification based on physical properties through a data-driven real-time analysis strategy, breaking the limitations of the traditional static lookup table method. It can automatically learn and extract the cogging torque characteristics that are fixed with position from complex current data. This not only solves the problem of compensation parameters failing due to motor aging or thermal drift, but more importantly, it achieves precise decoupling of the cogging torque component from external disturbances under dynamic operating conditions, improving the positioning stability and control accuracy of robot joints when running at low speeds.

[0011] Preferably, acquiring the operating data of the robot joint includes: acquiring the mechanical angular position of the motor using an absolute encoder; and acquiring the torque current component of the motor using a current sensor.

[0012] Preferably, the step of performing DC removal processing on the operating data to obtain a ripple current sequence reflecting the characteristics of current ripple includes: setting a sliding window and calculating the average value of the torque current components within the sliding window; and using the difference between the collected torque current components and the average value of the torque current components as the ripple current sequence.

[0013] Preferably, the evaluation of the frequency domain dispersion index of each intrinsic mode component includes: performing a frequency domain transformation on each intrinsic mode component to obtain the spectral energy distribution; wherein, the more concentrated the spectral energy is near the center frequency, the smaller the value of the frequency domain dispersion index, and the lower the corresponding frequency domain dispersion.

[0014] The frequency domain dispersion index can quantitatively screen out components with smooth waveforms and significant periodicity from the frequency domain dimension, providing a reliable evaluation basis for identifying weak cogging torque signals from broadband background noise and avoiding the waste of computational resources caused by blindly processing low-quality signals.

[0015] Preferably, the method for determining the theoretical cogging torque phase sequence includes: calculating the spatial harmonic order of the cogging torque based on the number of pole pairs and slots of the motor, and determining the theoretical cogging torque phase sequence based on the spatial harmonic order and the mechanical angular position.

[0016] Physical position feedback, acting as a calibrator, establishes a strong correlation between electromagnetic signals and mechanical structures. Only components strictly locked to the mechanical angular position can pass the calibration, thereby effectively eliminating random external load impacts or cable dragging interference, ensuring the reliability of the physical basis of compensation commands, and preventing system instability caused by erroneous compensation.

[0017] Preferably, the step of analyzing the correlation between the instantaneous phase sequence and the theoretical cogging torque phase sequence in the time dimension to obtain the phase-locking confidence level includes:

[0018] Calculate the phase difference distribution between the instantaneous phase sequence and the theoretical cogging torque phase sequence;

[0019] The vector synthesis method is used to evaluate the directional consistency of the phase difference distribution on the unit circle. If the phase difference distribution is directional, the corresponding intrinsic mode component is determined to be the torque component locked with the motor position, and a high phase lock confidence level is assigned. If the phase difference distribution is randomly divergent, the corresponding intrinsic mode component is determined to be external interference, and a low phase lock confidence level is assigned.

[0020] The vector synthesis method evaluates the directional consistency of the phase difference distribution. It utilizes the physical property of vector superposition on the unit circle and can accurately distinguish between internal inherent torque fluctuations and external random disturbances.

[0021] Preferably, determining the dynamic compensation gain coefficient includes:

[0022] An adaptive evaluation model is constructed with the phase-locked confidence and the frequency domain dispersion index as input variables;

[0023] Based on the signal reliability determined by the adaptive evaluation model, when the intrinsic mode component is determined to be an effective cogging torque component, the dynamic compensation gain coefficient approaches its maximum value; when the intrinsic mode component is determined to contain external interference or noise, the dynamic compensation gain coefficient attenuates to its minimum value.

[0024] The adaptive evaluation model uses phase-locked confidence and frequency-domain dispersion index as dual inputs. It can dynamically adjust the compensation level according to the signal quality, providing full compensation when the signal confidence is high and automatically attenuating the gain when the signal quality is poor. This avoids system oscillations caused by fixed-gain compensation schemes when the signal quality deteriorates.

[0025] Preferably, the step of performing reverse torque compensation on the torque motor based on the dynamic compensation gain coefficient includes: selecting the intrinsic mode component with the highest phase-locked confidence as the object to be compensated; using the dynamic compensation gain coefficient to adjust the intensity of the time-domain waveform of the object to be compensated, generating a reverse compensation current and superimposing it on the current loop of the torque motor.

[0026] The intrinsic mode component with the highest phase-locking confidence was selected as the object to be compensated, and the intensity was adjusted by using the dynamic compensation gain coefficient, which achieved precise reverse cancellation of the cogging torque and reduced the positioning jitter amplitude.

[0027] Preferably, before evaluating the frequency domain dispersion index of each intrinsic modal component, the method further includes: calculating the center frequency of each intrinsic modal component; and eliminating intrinsic modal components whose center frequencies exceed the characteristic frequency range of cogging torque.

[0028] Pre-screening removes intrinsic mode components whose center frequency exceeds the characteristic frequency range of cogging torque. This first eliminates components that clearly do not conform to the physical characteristics of cogging torque in the frequency domain, reducing subsequent calculations and improving analysis efficiency.

[0029] Secondly, the present invention provides a high-precision positioning control system for a robot joint torque motor, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned high-precision positioning control method for a robot joint torque motor is implemented.

[0030] By adopting the above technical solution, a computer program is generated from the high-precision positioning control method of the torque motor for robot joints, and stored in the memory so that it can be loaded and executed by the processor. Based on the memory and the processor, a terminal device can be made for convenient use.

[0031] The beneficial effects of this invention are as follows:

[0032] This invention combines variational mode decomposition technology with phase-locked verification based on physical position to construct a data-driven adaptive cogging torque compensation mechanism. This eliminates the dependence on factory parameters by the traditional static lookup table method, enabling the compensation strategy to be updated in real time as the motor's operating state changes. This solves the problem of compensation parameters failing due to motor aging or thermal drift. At the same time, through dual screening using the frequency domain dispersion index and phase-locked confidence, precise decoupling of the cogging torque component from external disturbances is achieved, avoiding system instability caused by miscompensation and improving the positioning stability and control accuracy of robot joints during low-speed operation. Attached Figure Description

[0033] Figure 1 This is a flowchart of a high-precision positioning control method for a torque motor used in a robot joint according to the present invention;

[0034] Figure 2 This is a spatial clustering distribution diagram of the modal decomposition features according to an embodiment of the present invention;

[0035] Figure 3 This is a distribution diagram of the position-wave phase-locked loop characteristics according to an embodiment of the present invention;

[0036] Figure 4 This is a comparison diagram of the distribution of sampling points for full-circumference positioning error in an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0039] This invention discloses a high-precision positioning control method for a torque motor used in a robot joint, referring to... Figure 1 This includes steps S1-S4:

[0040] S1. Acquire the running data of the robot joints and perform DC-free processing on the running data to obtain the ripple current sequence reflecting the characteristics of the current ripple.

[0041] When the robot joints perform actions or remain in a position-holding state, data is acquired through the high-speed sampling interface of the servo driver. A high-resolution absolute encoder acquires the mechanical angular position of the motor, and a current sensor acquires the torque current component of the motor. The acquired torque current component is recorded as follows: The torque current component directly reflects the electromagnetic torque output of the motor, which includes load torque, friction torque, and cogging torque fluctuations that need to be eliminated. The absolute encoder must have a resolution of at least 23 bits and a sampling frequency of at least 10kHz.

[0042] Furthermore, cogging torque typically manifests as a tiny AC ripple superimposed on a constant or slowly varying drive current. To accurately extract this subtle feature, the original torque current component needs to be de-DC processed.

[0043] This embodiment uses the sliding window method to remove DC from the torque current component. Specifically, the mean value of the torque current component within the sliding window is calculated, and this mean value is denoted as... The collected torque current components With the mean of torque current components The difference is taken as the ripple current sequence, then the ripple current sequence Satisfying the relation:

[0044] ;

[0045] In the formula, The torque current component is acquired in real time. The mean value of the torque current component within the sliding window is given. The length of the sliding window is determined based on the operating conditions of the motor, and the value range is [50, 500] sampling points. In this embodiment, the length of the DC-to-electric sliding window is set to 200 sampling points. If the window is too short, the mean calculation will be affected by transient fluctuations. If the window is too long, the true cogging torque ripple characteristics will be smoothed out. In other embodiments, the length of the DC-to-electric sliding window can be set according to the actual application scenario and requirements.

[0046] For example, if the motor drives the load to run at a constant speed, and the measured average value of the torque current component is 1.5A, while the instantaneous torque current component varies between 1.48A and 1.52A, then the ripple current sequence extracted through the above calculation is the deviation value near the average value of the torque current component.

[0047] By performing real-time dynamic debiasing on the operating data, the load baseline and static interference can be effectively eliminated, highlighting the subtle fluctuation characteristics in the signal and providing a clean data foundation for subsequent modal analysis.

[0048] S2. Perform variational mode decomposition on the ripple current sequence to obtain multiple intrinsic mode components, and evaluate the frequency domain dispersion index of each intrinsic mode component based on the distribution of the spectral energy of each intrinsic mode component relative to the center frequency.

[0049] Since the obtained ripple current sequence often contains a variety of components such as frequency-varying cogging torque, high-frequency electromagnetic noise, and random mechanical vibration, in order to separate the signal component corresponding to the cogging torque from the obtained ripple current sequence, this embodiment uses a variational mode decomposition algorithm to process the ripple current sequence. Specifically, variational mode decomposition is an adaptive signal decomposition method that can decompose complex signals into several narrowband components with different characteristic frequencies. In other words, the variational mode decomposition algorithm can automatically identify periodic components from different sources from mixed current fluctuations without pre-setting a target frequency.

[0050] The specific steps for processing the ripple current sequence using the variational mode decomposition algorithm are as follows: Set the number of mode decomposition layers. , ripple current sequence Decomposed into There are eigenmode components with different center frequencies, where the number of mode decomposition layers is... The value range is [3, 5]; in this embodiment, the modal decomposition layer number is... Set to 4, If the value is too small, high-frequency noise and cogging torque components cannot be fully separated. An excessively large number of modality decomposition layers will increase the computational burden and generate spurious modalities; in other embodiments, the number of modality decomposition layers can be set according to the actual application scenario and requirements.

[0051] Before evaluating the frequency domain dispersion index of each intrinsic mode component, it is necessary to calculate the center frequency of each intrinsic mode component and remove intrinsic mode components whose center frequency exceeds the characteristic frequency range of cogging torque.

[0052] The variational mode decomposition obtained Each intrinsic modal component is calculated for its center frequency. Further, based on the number of pole pairs, number of slots, and current speed of the motor, the characteristic frequency range of the cogging torque is determined and used as a screening criterion. Intrinsic modal components whose center frequencies exceed the characteristic frequency range of the cogging torque are identified as abnormal components unrelated to the cogging torque and are removed. Intrinsic modal components whose center frequencies are within the characteristic frequency range of the cogging torque are retained as candidate components.

[0053] Through the above pre-screening, components whose center frequency clearly does not conform to the physical characteristics of cogging torque can be eliminated in the frequency domain, reducing the amount of subsequent calculations and improving analysis efficiency.

[0054] Based on this, the frequency domain dispersion index is evaluated for the retained candidate intrinsic mode components. Specifically, a Fast Fourier Transform is performed on each candidate intrinsic mode component to obtain the spectral energy distribution, and the spectral amplitude sequence is denoted as... Based on the distribution of the spectral energy of each candidate intrinsic mode component relative to the center frequency, the frequency domain dispersion index is calculated. Then, the... Frequency domain dispersion index of each eigenmode component Satisfying the relation:

[0055] ;

[0056] In the formula, This represents the total number of discrete frequency points. For discrete frequency variables; For the first The center frequencies of the intrinsic mode components; For the first Each intrinsic mode component at frequency The amplitude at that point; It is a preset non-zero minimum energy constant used to prevent the denominator from being zero.

[0057] Among them, the preset non-zero minimum energy constant The value range is

[10] -8 10 -4 In this embodiment, Set to 10-6 , Used to prevent the denominator from being zero. An excessively large index will affect the index's ability to distinguish low-energy components. If the value is too small, it may lead to unstable numerical calculations. In other embodiments, the value can be set according to the actual application scenario and requirements. .

[0058] The numerator of the frequency domain dispersion index calculates the weighted dispersion of spectral energy relative to the center frequency, while the denominator is the total energy plus a minimal constant. The more concentrated the spectral energy is near the center frequency, the smaller the weighted energy of each frequency point in the numerator deviating from the center frequency, and the smaller the value of the frequency domain dispersion index, representing lower frequency domain dispersion. Conversely, if the signal is broadband noise with energy dispersed across multiple frequency points, the value of the frequency domain dispersion index is larger, representing higher frequency domain dispersion. In other words, a periodic disturbance of a single frequency like cogging torque forms a narrow spectrum with highly concentrated energy, while external vibrations or electromagnetic noise have energy dispersed over a wide frequency range. The frequency domain dispersion index can effectively distinguish between cogging torque components and noise components, avoiding misjudging low-quality noise signals as valid cogging torque components, thus preventing compensation failure.

[0059] For example, if the energy of a certain intrinsic mode component is highly concentrated around 60Hz and the amplitude of the surrounding frequencies is extremely small, the calculated frequency domain dispersion index will be close to 0.05; if the signal is broadband noise and the energy is dispersed across multiple frequency points, the frequency domain dispersion index may be as high as 10 or more.

[0060] S3. Based on the motor pole-slot fit relationship of the robot joint and the mechanical angle position, extract the instantaneous phase sequence of each intrinsic modal component, and analyze the correlation between the instantaneous phase sequence and the theoretical cogging torque phase sequence in the time dimension to determine the phase-locking confidence of each intrinsic modal component in order to eliminate external fluctuation interference.

[0061] Frequency domain characteristics alone are insufficient to fully confirm the properties of intrinsic modal components, because external periodic mechanical vibrations may also manifest as narrowband single-frequency waveforms. Therefore, this embodiment further introduces phase-locked verification based on physical position. By analyzing the phase correlation between intrinsic modal components and mechanical angular position, the inherent cogging torque components and external random interference can be distinguished.

[0062] Instantaneous phase sequences are extracted for each candidate intrinsic mode component (EMC). The Hilbert transform is then applied to process the candidate EMCs to obtain their analytic signals. The instantaneous phase sequences are then extracted from these analytic signals, denoted as... .

[0063] Accordingly, the spatial harmonic order of the cogging torque is calculated based on the number of pole pairs and slots of the motor, and the theoretical cogging torque phase sequence associated with the mechanical angular position is determined based on the spatial harmonic order; the spatial harmonic order of the cogging torque is calculated based on the number of magnetic poles and slots of the motor. Spatial harmonic order of cogging torque The least common multiple of the number of poles and the number of slots, where the number of poles is equal to twice the number of pole pairs; [The text abruptly ends here, likely due to an incomplete sentence or missing information.] The mechanical angular position is denoted as Determine the theoretical cogging torque phase sequence There is a fixed harmonic relationship between the theoretical cogging torque phase sequence and the mechanical angular position. Specifically, the theoretical cogging torque phase sequence... Satisfying the relation: In the formula, The spatial harmonic order of the cogging torque. For a moment The mechanical angular position, in radians.

[0064] Based on this, the instantaneous phase sequence is analyzed. Phase sequence with theoretical cogging torque The correlation over time is used to obtain the phase-locked confidence level.

[0065] Calculate the phase difference distribution between the instantaneous phase sequence and the theoretical cogging torque phase sequence. The difference between the instantaneous phase sequence and the theoretical cogging torque phase sequence is taken as the phase difference sequence, denoted as... ,but .

[0066] Therefore, the directional consistency of the phase difference distribution on the unit circle is evaluated using the vector synthesis method; in a length of Within the sliding window, the phase-locked confidence level is calculated, then the... Phase-locked confidence level of each intrinsic mode component Satisfying the relation:

[0067] ;

[0068] In the formula, This represents the total number of sampling points within the sliding window; For the first The phase difference value of each sampling point is determined by The value is obtained by taking the corresponding sampling time.

[0069] Among them, the total number of sampling points within the sliding window The value range is [64, 512]; in this embodiment, Setting the sampling point to 256 points will result in insufficient phase difference statistics and misjudgment if the window is too short, while if the window is too long, it will reduce the response speed to changes in operating conditions.

[0070] It should be noted that if the intrinsic modal components are indeed cogging torque components, then the instantaneous phase sequence should be strictly synchronized with the mechanical position, resulting in a phase difference. If the sampling points remain approximately constant, the unit vectors corresponding to each sampling point point in the same direction on the complex plane, and the superimposed magnitude, i.e., the phase-locked confidence, approaches 1. Conversely, if the intrinsic mode components originate from external random disturbances, the instantaneous phase sequence is independent of the mechanical angle position, and the vectors are randomly distributed on the circumference and cancel each other out, resulting in the phase-locked confidence approaching 0.

[0071] If the phase difference distribution points in the same direction, the corresponding intrinsic mode component is determined to be the torque component locked with the motor position, and a high phase lock confidence level is assigned; if the phase difference distribution is randomly divergent, the corresponding intrinsic mode component is determined to be external interference, and a low phase lock confidence level is assigned.

[0072] S4. Combine the frequency domain dispersion index and phase lock confidence to determine the dynamic compensation gain coefficient, and perform reverse torque compensation on the torque motor based on the dynamic compensation gain coefficient.

[0073] Traditional fixed-gain compensation schemes are prone to causing system oscillations when signal quality deteriorates. Therefore, this embodiment adopts a dynamic gain adjustment strategy to adaptively adjust the compensation level according to the signal reliability.

[0074] Specifically, an adaptive evaluation model is constructed with phase-locked confidence and frequency domain dispersion index as input variables. Based on the signal confidence determined by the adaptive evaluation model, when the intrinsic mode component is determined to be an effective cogging torque component, the dynamic compensation gain coefficient approaches its maximum value; when the intrinsic mode component is determined to contain external interference or noise, the dynamic compensation gain coefficient attenuation approaches its minimum value.

[0075] In a preferred embodiment, the dynamic compensation gain coefficient Satisfying the relation:

[0076] ;

[0077] In the formula, This is a constant used to adjust the steepness of the function, controlling the transition rate of the gain from low to high; The threshold center offset is used to set the threshold for determining signal reliability. This is the bandwidth normalization constant; For the first Phase-locked confidence of each intrinsic mode component; For the first The frequency domain dispersion index of each intrinsic mode component; It is a natural exponential function.

[0078] in, The value range is [5, 20]. In this embodiment, the value of is... Set to 10, Control the transition rate of the gain from low to high. If the gain is too small, the gain switching will be too gradual and unable to quickly suppress interference. If the gain is too high, the gain switching will be too steep, causing control oscillations. The value range is [0.3, 0.7]. In this embodiment, the value is... Set to 0.5. Set a threshold for determining signal reliability. Too low a value will increase the risk of miscompensation for low-quality signals. If the value is too high, a large number of valid signals will be misjudged as interference and will not be adequately compensated. The value range is [50, 200]. In this embodiment, the value will be... Set to 100, The influence weights used to normalize the frequency domain dispersion index If the value is too small, the frequency domain dispersion index will have an excessively large impact on the overall evaluation, causing the phase-locked characteristic to be overlooked. If the value is too large, it will weaken the filtering effect of the frequency domain dispersion index; in other embodiments, it can be set according to the actual application scenario and requirements.

[0079] It should be noted that the above relationship utilizes the nonlinear switching characteristics of the logistic regression function; when the intrinsic mode components exhibit strong phase-locked characteristics, i.e., phase-locked confidence... The frequency domain dispersion index approaches 1 and has a low frequency domain dispersion. When the value approaches 0, the comprehensive index within the parentheses will be greater than the center offset of the judgment threshold. This makes the dynamic compensation gain coefficient The gain coefficient quickly approaches 1, at which point the system performs full compensation; if the signal quality is poor or the phase-locked confidence level is low, the gain coefficient is dynamically compensated. The system will rapidly decay to near zero and automatically cut off compensation to prevent system instability caused by miscompensation.

[0080] In one embodiment, dynamic compensation gain coefficients are calculated for all pre-screened candidate intrinsic mode components. The time-domain waveforms of each candidate intrinsic mode component are multiplied by their respective dynamic compensation gain coefficients and then summed to generate a comprehensive reverse compensation current, which is then superimposed on the current loop of the torque motor. Under the multi-component weighted compensation method, for candidate intrinsic mode components with high signal quality, the dynamic compensation gain coefficient approaches 1, and the compensation is sufficient. For candidate intrinsic mode components with poor signal quality, the dynamic compensation gain coefficient approaches 0, and the compensation is automatically turned off, thereby achieving synchronous compensation for multiple cogging torque harmonic components.

[0081] In another embodiment, to further reduce computational complexity and avoid mutual interference between multi-component compensations, the intrinsic mode component with the highest phase-locked confidence is selected from the candidate intrinsic mode components as the object to be compensated, and the time-domain waveform of the object to be compensated is denoted as... .

[0082] Using dynamic compensation gain coefficient The time-domain waveform of the object to be compensated Intensity adjustment is performed to generate reverse compensation current. Then the reverse compensation current Satisfying the relation:

[0083] ;

[0084] In the formula, For dynamic compensation gain coefficient; The waveform of the object to be compensated is shown in the time domain; the negative sign indicates that the generated compensation current is opposite to the direction of the cogging torque.

[0085] Furthermore, the reverse compensation current will be... The current loop input of the torque motor is superimposed to achieve reverse cancellation of the cogging torque.

[0086] In another embodiment, to verify the effectiveness of the scheme under the boundary values ​​of each key parameter, the number of modal decomposition layers is... Set to 3, the total number of sampling points within the sliding window. Set to 128, a constant to adjust the steepness of the function. Set to 5 to determine the center offset of the threshold. Set to 0.3, bandwidth normalization constant Set to 50. Under the above parameter conditions, variational mode decomposition can still separate the cogging torque component from the high-frequency noise component in the ripple current sequence. The frequency domain dispersion index has a slightly reduced ability to distinguish between the cogging torque component and the noise component, but it still meets the screening requirements. The statistical window of the phase-locked confidence is shortened, which leads to improved tracking sensitivity to rapid changes in operating conditions, but slightly reduced stability in the signal steady segment. The transition range of the dynamic compensation gain coefficient is widened, the gain switching is smoother, the compensation response speed is slightly slower, but the stability of the control loop is enhanced.

[0087] In yet another embodiment, the number of modal decomposition layers is... Set to 5, the total number of sampling points within the sliding window. Setting it to 512 is a constant that adjusts the steepness of the function. Set to 20 to determine the center offset of the threshold. Set to 0.7, bandwidth normalization constant Set to 200. Under the above parameter conditions, the resolution of variational mode decomposition is improved, enabling the separation of more subtle frequency components, but the computational burden increases accordingly, and there is a risk of generating spurious modes when the number of intrinsic mode components increases; the statistical window of phase-locked confidence is extended, and the judgment result is more stable, but the response speed to sudden changes in operating conditions is reduced; the transition range of dynamic compensation gain coefficient is narrowed, the gain switching is steeper, and the compensation response speed is faster, but there is a risk of frequent gain jumps when the signal quality is at a critical state; the center offset of the judgment threshold... The high quality ensures that only high-quality signals can be adequately compensated, effectively avoiding miscompensation for low-quality signals. However, in scenarios where motor aging leads to a decline in overall signal quality, the effective cogging torque component may not receive sufficient compensation.

[0088] The above two sets of boundary value examples show that each parameter can enable the scheme to operate normally and achieve the technical effect of suppressing cogging torque within its value range. The specific values ​​of the parameters can be selected according to the operating conditions of the robot joint, real-time requirements, and control stability requirements.

[0089] Reference Figure 2 The smaller the value of the frequency domain dispersion index, the more concentrated the spectral energy. As can be seen from the figure, three regions are formed in the feature space: the first region is located in the range of center frequency about 10Hz to 15Hz and frequency domain dispersion index is low, and circularly marked data points are distributed. These data points are closely clustered and correspond to the cogging torque component with high energy concentration and low frequency dispersion; the second region is located in the range of center frequency is low and frequency domain dispersion index is high, and squarely marked data points are distributed, which correspond to the external interference mode; the triangularly marked data points in the third region are scattered throughout the feature space and correspond to the random noise mode.

[0090] Reference Figure 3 Within a specific phase angle range, a tightly clustered group of data points is formed. The intrinsic mode components corresponding to these data points have a strong phase-locked relationship with the mechanical position of the motor, which is the cogging torque component. On the other hand, the intrinsic mode components corresponding to the data points scattered in various angular directions of the polar coordinate system are not related to the motor position, which is the unlocked interference signal. The above distribution characteristics verify the effectiveness of distinguishing the cogging torque component from external interference by phase-locked confidence.

[0091] Reference Figure 4 The error distribution using the existing technical solution exhibits a clear periodic banded distribution, with the error amplitude fluctuating within the range of ±0.12°, and large periodic deviations appearing in multiple angle intervals; the error distribution using the solution of the present invention converges closely to the vicinity of the zero axis within the entire circumference, with the error amplitude basically controlled within ±0.02°, and the positioning accuracy is significantly improved.

[0092] This invention also discloses a high-precision positioning control system for a robot joint torque motor, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a high-precision positioning control method for a robot joint torque motor according to the present invention is implemented.

[0093] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0094] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

Claims

1. A high-precision positioning control method for a torque motor used in a robot joint, characterized in that, include: The robot joint operation data is acquired and DC-free processed to obtain a ripple current sequence that reflects the characteristics of current ripple. Variational mode decomposition is performed on the ripple current sequence to obtain multiple intrinsic mode components. Based on the distribution of the spectral energy of each intrinsic mode component relative to the center frequency, the frequency domain dispersion index of each intrinsic mode component is evaluated. This includes performing a frequency domain transformation on each intrinsic mode component to obtain the spectral energy distribution. The more concentrated the spectral energy is near the center frequency, the smaller the value of the frequency domain dispersion index, and the lower the corresponding frequency domain dispersion. Based on the motor pole slot fit relationship of the robot joint and the mechanical angular position, the instantaneous phase sequence of each intrinsic mode component is extracted; The correlation between the instantaneous phase sequence and the theoretical cogging torque phase sequence in the time dimension was analyzed to determine the phase-locking confidence of each intrinsic mode component, in order to eliminate external fluctuation interference, including: Calculate the phase difference distribution between the instantaneous phase sequence and the theoretical cogging torque phase sequence; The vector synthesis method is used to evaluate the directional consistency of the phase difference distribution on the unit circle. If the phase difference distribution is directional, the corresponding intrinsic mode component is determined to be the torque component locked with the motor position, and a high phase lock confidence level is assigned. If the phase difference distribution is randomly divergent, the corresponding intrinsic mode component is determined to be external disturbance, and a low phase lock confidence level is assigned. Determining the theoretical cogging torque phase sequence includes: calculating the spatial harmonic order of the cogging torque based on the number of pole pairs and slots of the motor, and determining the theoretical cogging torque phase sequence based on the spatial harmonic order and mechanical angular position; The dynamic compensation gain coefficients are determined by combining the frequency domain dispersion index and the phase-locked confidence level, including: An adaptive evaluation model is constructed with phase-locked confidence and frequency domain dispersion index as input variables; Based on the signal reliability determined by the adaptive evaluation model, when the intrinsic mode component is determined to be an effective cogging torque component, the dynamic compensation gain coefficient approaches its maximum value; when the intrinsic mode component is determined to contain external interference or noise, the dynamic compensation gain coefficient attenuates to its minimum value. And based on the dynamic compensation gain coefficient, reverse torque compensation is performed on the torque motor.

2. The high-precision positioning control method for a torque motor for a robot joint according to claim 1, characterized in that, The acquisition of robot joint operation data includes: using an absolute encoder to collect the mechanical angular position of the motor; and using a current sensor to collect the torque current component of the motor.

3. The high-precision positioning control method for a torque motor for a robot joint according to claim 1, characterized in that, The step of performing DC removal processing on the operating data to obtain a ripple current sequence reflecting the characteristics of current ripple includes: setting a sliding window and calculating the average value of the torque current components within the sliding window; and using the difference between the collected torque current components and the average value of the torque current components as the ripple current sequence.

4. The high-precision positioning control method for a torque motor for a robot joint according to claim 1, characterized in that, The step of performing reverse torque compensation on the torque motor based on the dynamic compensation gain coefficient includes: selecting the intrinsic mode component with the highest phase-locked confidence as the object to be compensated; using the dynamic compensation gain coefficient to adjust the intensity of the time-domain waveform of the object to be compensated, generating a reverse compensation current and superimposing it on the current loop of the torque motor.

5. A high-precision positioning control method for a torque motor for a robot joint according to claim 1, characterized in that, Before evaluating the frequency domain dispersion index of each intrinsic modal component, the process also includes: calculating the center frequency of each intrinsic modal component; and removing intrinsic modal components whose center frequencies exceed the characteristic frequency range of cogging torque.

6. A high-precision positioning control system for a torque motor used in a robot joint, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a high-precision positioning control method for a torque motor for a robot joint according to any one of claims 1-5.

Images