Permanent magnet synchronous motor speed loop adaptive proportional integral parameter setting system

Abstract

The application relates to the technical field of motor control, and discloses a permanent magnet synchronous motor speed loop adaptive proportional integral parameter setting system. The system performs the following steps: based on normalized least mean square adaptive filtering estimation and stripping of bus coupling speed fluctuation components, a speed response residual signal without bus coupling is obtained; spectrum analysis and frequency component identification are performed on the signal, a band-pass filter is used to separate a closed-loop dominant oscillation component and each harmonic mode oscillation component; closed-loop control envelope features and harmonic mode damping features are extracted; proportional integral gain correction factors are obtained based on fuzzy reasoning, and constraints are applied according to the harmonic mode damping features; and finally, the corrected proportional integral parameters are written into a speed loop controller.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and more specifically, to an adaptive proportional-integral parameter tuning system for the speed loop of a permanent magnet synchronous motor. Background Technology

[0002] In multi-axis servo drive systems (such as multi-axis CNC machining centers and multi-axis industrial robots), the inverters of multiple permanent magnet synchronous motors share the same DC bus power supply. The voltage ripple generated by the switching action of each axis inverter is transmitted to other axes through the bus. Simultaneously, when the motor drives the load through a reducer, coupling, or long drive shaft, the mechanical transmission chain exhibits resonant modes caused by elastic deformation. When a step speed regulation triggers proportional-integral parameter adaptive correction on a certain axis, the speed loop's speed response signal simultaneously incorporates three frequency components: the closed-loop control-dominated oscillation, the mechanical resonant mode oscillation, and the exogenous oscillation introduced by bus voltage coupling.

[0003] Existing adaptive proportional-integral parameter tuning methods for permanent magnet synchronous motors infer the correction direction and magnitude of proportional-integral parameters by extracting envelope features such as overshoot, attenuation rate, and oscillation period from the speed step response.

[0004] However, under conditions of common DC bus power supply and elastic resonance in the drive train, the bus coupling ripple frequency and mechanical resonant frequency in the speed response spectrum may be similar or even partially overlap. Relying solely on peak frequency identification cannot distinguish the physical origins of these two types of non-control response frequency components. If the bus coupling frequency is misidentified as the resonant frequency, the constraint based on the resonant attenuation rate will excessively limit the adjustment amplitude of the proportional-integral (PI) parameter according to the incorrect resonant characteristics, reducing adaptive tuning efficiency. If the resonant frequency is misidentified as the bus coupling frequency and filtered out, the true resonant information will be lost, causing the closed-loop bandwidth after PI parameter adjustment to intrude into the resonant frequency band, triggering mechanical resonance and exacerbating system oscillations. Misidentification of the origins of these two types of non-control response components renders both envelope feature extraction and PI parameter correction unreliable. Summary of the Invention

[0005] This invention provides an adaptive proportional-integral parameter tuning system for the speed loop of permanent magnet synchronous motors, which solves the technical problems in related technologies, such as the difficulty in online adaptive tuning of the proportional-integral parameters of the speed loop of permanent magnet synchronous motors, and the susceptibility to interference from DC bus voltage fluctuations and mechanical resonance, which leads to a decrease in control performance.

[0006] This invention discloses an adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor, comprising: when a step change in the speed loop setpoint is detected, simultaneously acquiring the actual speed response sequence of the motor on this shaft and the DC bus voltage sampling sequence;

[0007] Bandpass filtering is performed on the DC bus voltage sampling sequence to extract the bus voltage fluctuation signal. The bus voltage fluctuation signal is used as the reference input and the speed response residual signal is used as the desired output. Normalized minimum mean square adaptive filtering is performed to output the bus coupled speed fluctuation estimation component.

[0008] The bus-coupled speed fluctuation estimation component is subtracted from the speed response residual signal to obtain the decoupled speed response residual signal;

[0009] Perform a fast Fourier transform on the decoupled rotational speed response residual signal to identify the set of closed-loop dominant oscillation frequencies and mechanical resonant frequencies;

[0010] Using the closed-loop dominant oscillation frequency and each resonant frequency as the center frequency, bandpass filtering is performed on the decoupled speed response residual signal to extract the closed-loop control dominant oscillation component signal and each resonant mode oscillation component signal respectively.

[0011] The closed-loop envelope attenuation rate, closed-loop oscillation period, and first overshoot are extracted from the closed-loop control dominant oscillation component signal, and the attenuation rate of each resonant mode is extracted from each resonant mode oscillation component signal.

[0012] Input the closed-loop performance deviation value into the fuzzy inference system, and output the proportional gain correction factor and the integral gain correction factor.

[0013] Based on the attenuation rate of each resonant mode, a resonant constraint is applied to the proportional gain correction factor and the integral gain correction factor to obtain the proportional-integral correction factor after resonant constraint.

[0014] Multiply the current proportional-integral parameters of the velocity loop by the proportional-integral correction factor after resonance constraint, and write them into the velocity loop controller.

[0015] Furthermore, the synchronous acquisition of the actual speed response sequence of the local shaft motor and the DC bus voltage sampling sequence includes:

[0016] Starting from the step moment, the actual speed response sequence and the DC bus voltage sampling sequence are synchronously and continuously collected at a fixed sampling period until the actual speed falls within the preset steady-state band centered on the step target speed for several consecutive sampling periods, at which point data collection stops.

[0017] The preset steady-state band is a speed range defined by offsetting the step target speed by one steady-state tolerance value above and below it. The steady-state tolerance value is preset as a percentage of the motor's rated speed.

[0018] Furthermore, before performing the normalized least mean square adaptive filtering operation, the following steps are also included:

[0019] The bus voltage fluctuation signal and the speed response residual signal are respectively subjected to Z-score normalization.

[0020] The normalized least mean square adaptive filtering operation generates an estimate of the bus coupling speed fluctuation by weighting and summing the delayed tap sequence of the reference input using the adaptive weight coefficient vector, and uses the error signal between the estimate and the speed response residual signal to drive the weight coefficient update.

[0021] The lower passband frequency of the bandpass filter is set to a preset offset that is lower than the lowest order sideband harmonic frequency of the inverter switching frequency, and the upper passband frequency is set to a preset offset that is higher than the highest order significant sideband harmonic frequency.

[0022] Furthermore, before subtracting the bus-coupled speed fluctuation estimation component from the speed response residual signal, the process also includes:

[0023] The bus-coupled speed fluctuation estimation component output by the normalized least mean square adaptive filter is multiplied by the standard deviation of the speed response residual signal used in Z-score standardization, and then the mean of the speed response residual signal used in Z-score standardization is added to restore the bus-coupled speed fluctuation estimation component from the standardized numerical domain to the original speed dimension.

[0024] Furthermore, the set of identifying the dominant closed-loop oscillation frequency and mechanical resonant frequency includes:

[0025] The peak frequency with the largest amplitude in the spectral distribution is marked as the closed-loop dominant oscillation frequency;

[0026] The remaining peak frequencies whose amplitude exceeds the preset resonance identification threshold are marked as the set of mechanical resonance frequencies;

[0027] The preset resonance identification threshold is set as a preset proportion of the amplitude corresponding to the closed-loop dominant oscillation frequency, or is set as a fixed absolute amplitude lower limit.

[0028] Furthermore, the extraction of the closed-loop envelope decay rate, closed-loop oscillation period, and first overshoot includes:

[0029] Peak detection is performed on the dominant oscillation component signal of the closed-loop control to extract the extreme point sequence;

[0030] The natural logarithm of the amplitude between adjacent extreme points in the same direction is taken, and least squares linear fitting is performed with the time sequence of the extreme point as the independent variable and the logarithmic amplitude as the dependent variable. The slope of the fitted line is the closed-loop envelope attenuation rate.

[0031] The closed-loop oscillation period is calculated based on the time interval between adjacent extreme points in the same direction.

[0032] The initial overshoot is calculated based on the ratio of the amplitude of the first extreme point in the extreme point sequence to the speed step.

[0033] Peak detection is performed on the oscillation component signals of each resonant mode. After taking the logarithm of the amplitude of each extreme point sequence, linear regression is performed to obtain the attenuation rate of each resonant mode. The smallest value among the attenuation rates of each resonant mode is selected as the resonant suppression constraint index.

[0034] When the set of mechanical resonant frequencies is empty, the resonant suppression constraint index is set to a default value greater than the resonant safe attenuation rate threshold.

[0035] Furthermore, inputting the closed-loop performance deviation value into the fuzzy inference system includes:

[0036] The attenuation rate deviation, oscillation period deviation and first overshoot deviation are obtained by subtracting the closed-loop envelope attenuation rate, closed-loop oscillation period and first overshoot from the corresponding target performance indicators.

[0037] The attenuation rate deviation, oscillation period deviation, and overshoot deviation are respectively normalized based on the mean of the range within the preset universe of discourse for each deviation value.

[0038] The normalized closed-loop performance deviation values ​​are input into the fuzzy inference system. After fuzzification, fuzzy rule inference and defuzzification operations, the proportional gain correction factor and integral gain correction factor are output.

[0039] The defuzzification operation adopts the centroid method, which sums the products of the membership degree corresponding to each fuzzy linguistic value on the universe of discourse of the output variable and the universe coordinate value of each fuzzy linguistic value, and then divides the sum of the membership degrees to transform the fuzzy inference result into a deterministic numerical output.

[0040] Furthermore, the resonant constraint applied to the proportional gain correction factor and the integral gain correction factor based on the attenuation rate of each resonant mode includes:

[0041] Determine whether the resonance suppression constraint index is less than the preset resonance safe attenuation rate threshold;

[0042] If the resonance suppression constraint index is less than the resonance safe attenuation rate threshold, then the integral gain correction factor is multiplied by the ratio of the resonance suppression constraint index to the resonance safe attenuation rate threshold to obtain the integral gain correction factor after resonance constraint.

[0043] The safety factor is calculated based on the ratio of the resonant frequency corresponding to the slowest decaying resonant mode to the current closed-loop bandwidth estimate. The current closed-loop bandwidth estimate is obtained by converting the reciprocal of the closed-loop oscillation period. When the proportional gain correction factor is greater than the safety factor, the proportional gain correction factor is truncated to the safety factor to obtain the proportional gain correction factor after resonance constraint.

[0044] If the resonance suppression constraint index is not less than the resonance safe attenuation rate threshold, then the proportional gain correction factor and the integral gain correction factor are not corrected.

[0045] Furthermore, the step of multiplying the current proportional-integral parameters of the velocity loop by the proportional-integral correction factor after resonance constraint and writing it into the velocity loop controller also includes:

[0046] The corrected proportional gain and integral gain are subjected to parameter safety domain constraint processing respectively. It is determined whether the corrected proportional gain and integral gain fall within the preset proportional gain safety range and integral gain safety range. If they exceed the safety range, the corrected proportional gain or integral gain is truncated to the boundary value of the safety range.

[0047] Calculate the difference between the corrected proportional-integral parameter and the current proportional-integral parameter. If the absolute value of the difference exceeds the preset maximum single change, the difference is truncated to the maximum single change and added to the current parameter as the actual written proportional-integral parameter value.

[0048] This invention provides an adaptive proportional-integral parameter tuning system for the speed loop of a permanent magnet synchronous motor, comprising:

[0049] The data synchronization acquisition module is used to synchronously acquire the actual speed response sequence of the motor on this shaft and the DC bus voltage sampling sequence when a step change in the speed loop setpoint is detected.

[0050] The bus coupling estimation module is used to perform bandpass filtering on the DC bus voltage sampling sequence to extract the bus voltage fluctuation signal. Using the bus voltage fluctuation signal as a reference input and the speed response residual signal as the desired output, it performs normalized minimum mean square adaptive filtering operation to output the bus coupling speed fluctuation estimation component.

[0051] The decoupling processing module is used to subtract the bus-coupled speed fluctuation estimation component from the speed response residual signal to obtain the decoupled speed response residual signal;

[0052] The spectrum analysis and identification module is used to perform a fast Fourier transform on the decoupled rotational speed response residual signal to identify the set of closed-loop dominant oscillation frequencies and mechanical resonant frequencies.

[0053] The frequency component separation module is used to perform bandpass filtering on the decoupled speed response residual signal with the closed-loop dominant oscillation frequency and each resonant frequency as the center frequency, and extract the closed-loop control dominant oscillation component signal and each resonant mode oscillation component signal respectively.

[0054] The feature extraction module is used to extract the closed-loop envelope attenuation rate, closed-loop oscillation period, and first overshoot from the closed-loop control dominant oscillation component signal, and to extract the attenuation rate of each resonant mode from each resonant mode oscillation component signal.

[0055] The fuzzy inference module is used to input the closed-loop performance deviation value into the fuzzy inference system and output the proportional gain correction factor and the integral gain correction factor.

[0056] The resonant constraint module is used to apply resonant constraints to the proportional gain correction factor and the integral gain correction factor based on the attenuation rate of each resonant mode, so as to obtain the proportional-integral correction factor after resonant constraint.

[0057] The parameter writing module is used to multiply the current proportional-integral parameters of the speed loop by the proportional-integral correction factor after resonance constraint and write them to the speed loop controller.

[0058] Before frequency domain analysis, this invention utilizes normalized least mean square adaptive filtering with the DC bus voltage fluctuation signal as a reference input to estimate and remove the exogenous speed fluctuation component introduced by bus coupling. This solves the technical problem of unreliably distinguishing the source of two types of non-control response components when the bus coupling frequency and mechanical resonance frequency are close under the common DC bus elastic transmission condition. The following technical effects are achieved: the physical attribution of each peak frequency in the speed response residual signal after bus coupling is determined, and the division between the closed-loop dominant oscillation frequency and the mechanical resonance frequency is reliable, making the proportional-integral gain correction direction and amplitude based on fuzzy inference output reliable; at the same time, the independently extracted resonant mode attenuation characteristics can truly reflect the transmission chain resonance state, and the resonance constraint link can limit the adjustment amplitude of the proportional-integral parameters according to the true attenuation rate of the slowest attenuating resonant mode, taking into account the resonance suppression requirements while adaptively correcting the speed loop proportional-integral parameters. Attached Figure Description

[0059] Figure 1 This is a flowchart of the adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor provided in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the X-axis speed step response curve provided in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the DC bus voltage sampling sequence provided in an embodiment of the present invention;

[0062] Figure 4 This is a schematic diagram of the signal stripping process provided in an embodiment of the present invention: comparison of rotational speed residual components;

[0063] Figure 5 This is a schematic diagram of the spectrum peak identification results provided in an embodiment of the present invention;

[0064] Figure 6 This is a schematic diagram of logarithmic linear fitting of the extreme points of the closed-loop dominant oscillation provided in an embodiment of the present invention;

[0065] Figure 7 This is a schematic diagram of logarithmic linear fitting of the extreme points of the mechanical resonance mode provided in an embodiment of the present invention;

[0066] Figure 8 This is a schematic diagram of the normalized value of closed-loop performance deviation provided in an embodiment of the present invention;

[0067] Figure 9 This is a schematic diagram comparing the speed loop PI parameter tuning before and after, according to an embodiment of the present invention. Detailed Implementation

[0068] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below with reference to the accompanying drawings. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0069] In permanent magnet synchronous motor drive systems, the parameters of the proportional-integral (PI) controller in the speed loop directly affect speed tracking accuracy and dynamic response quality. When the motor's operating conditions change, the PI parameters need to be adaptively adjusted based on the speed step response characteristics to maintain the speed loop's control performance. In multi-axis servo drive systems (such as multi-axis CNC machining centers and multi-axis industrial robots), the inverters of multiple permanent magnet synchronous motors share the same DC bus power supply, and the voltage ripple generated by the switching actions of each axis inverter is transmitted to other axes through the bus. Simultaneously, when the motor drives a load through a reducer, coupling, or long drive shaft, the mechanical transmission chain exhibits resonant modes caused by elastic deformation. When a step speed regulation on a certain axis triggers adaptive correction of the PI parameters, the speed loop's speed response signal simultaneously incorporates three frequency components: the dominant oscillation of the closed-loop control, the mechanical resonant mode oscillation, and the exogenous oscillation introduced by bus voltage coupling.

[0070] Existing adaptive proportional-integral (PI) parameter tuning methods for permanent magnet synchronous motors (PMSMs) infer the correction direction and amplitude of the PI parameter by extracting envelope features such as overshoot, attenuation rate, and oscillation period from the speed step response. However, under conditions of common DC bus power supply and elastic resonance in the drive train, the bus coupling ripple frequency and mechanical resonance frequency in the speed response spectrum may be similar or even partially overlap. Relying solely on peak frequency identification cannot distinguish the physical origin of these two types of non-control response frequency components. If the bus coupling frequency is misidentified as the resonance frequency, the constraint based on the resonance attenuation rate will excessively limit the adjustment amplitude of the PI parameter according to the incorrect resonance characteristics, reducing the adaptive tuning efficiency. If the resonance frequency is misidentified as the bus coupling frequency and filtered out, the true resonance information will be lost, causing the closed-loop bandwidth after PI parameter adjustment to intrude into the resonance frequency band, triggering mechanical resonance and exacerbating system oscillations. Misidentification of the origin of the two types of non-control response components renders both envelope feature extraction and PI parameter correction unreliable.

[0071] According to an embodiment of this invention, this invention provides a method for adaptive proportional-integral parameter tuning of the speed loop of a permanent magnet synchronous motor. It should be understood that the execution entity of this method is a motor drive controller with real-time data acquisition and processing capabilities. The motor drive controller is simultaneously connected to the speed detection device and the DC bus voltage sampling channel of the permanent magnet synchronous motor on this shaft, and runs the speed loop proportional-integral control program.

[0072] At least one embodiment of the present invention discloses a method for adaptive proportional-integral parameter tuning of the speed loop of a permanent magnet synchronous motor, such as... Figure 1 As shown, it includes the following steps:

[0073] Step 1: Synchronously acquire speed step response data and DC bus voltage data;

[0074] When a step change in the speed setpoint of the permanent magnet synchronous motor speed loop is detected, starting from the step moment, the actual speed response sequence and DC bus voltage sampling sequence of the motor on this shaft are synchronously and continuously collected at a fixed sampling period until the actual speed enters the preset steady-state band centered on the step target speed, so as to obtain complete speed response sampling data and synchronous bus voltage sampling data.

[0075] It should be noted that the aforementioned preset steady-state band refers to the speed range defined by offsetting the target step speed by one steady-state tolerance value both above and below it. The steady-state tolerance value is preset as a percentage of the motor's rated speed. When the actual speed falls within the preset steady-state band for several consecutive sampling periods, it is determined that the actual speed has entered the steady-state band, and data acquisition is stopped.

[0076] Step 2: Estimate the bus coupling speed fluctuation component based on normalized minimum mean square adaptive filtering;

[0077] A bandpass filter is applied to the DC bus voltage sampling sequence, with the passband of the filter covering the inverter switching ripple frequency band to extract the bus voltage fluctuation signal within this band. The actual speed response sequence is subtracted from the steady-state speed value to obtain the speed response residual signal. Using the bus voltage fluctuation signal as the reference input and the speed response residual signal as the desired output, a normalized least mean square adaptive filtering operation is performed to output the bus-coupled speed fluctuation estimation component.

[0078] Before performing normalized least mean square adaptive filtering, the bus voltage fluctuation signal and the speed response residual signal are respectively subjected to Z-score normalization to eliminate the influence of the difference in their dimensions and magnitude on the convergence process of the adaptive weight coefficients.

[0079] Normalized Least Mean Square (NMS) adaptive filtering is a standard adaptive filtering algorithm. Its input is a standardized bus voltage fluctuation signal and its delayed tap sequence, and its output is an estimated component of bus-coupled speed fluctuation. The NMS adaptive filtering algorithm generates an estimate of the bus-coupled speed fluctuation by weighting the delayed tap sequence of the reference input with an adaptive weight coefficient vector. The error signal between this estimate and the speed response residual signal drives the weight coefficient update, and finally, after convergence, outputs the estimated component of bus-coupled speed fluctuation. The NMS adaptive filtering algorithm utilizes the physical transmission causal relationship between the bus voltage fluctuation signal and the coupling component introduced by the bus voltage fluctuation signal in the speed signal, approximating the transmission path characteristics from bus voltage to speed fluctuation through adaptive weight coefficients. Since there is no linear correlation between the mechanical resonant oscillation component and the bus voltage fluctuation signal, the NMS adaptive filtering process will not misjudge the resonant oscillation component as the bus-coupled component.

[0080] It should be noted that the lower and upper passband frequencies of the aforementioned bandpass filter are preset based on the inverter switching frequency and its known harmonic distribution characteristics. Specifically, the lower passband frequency is set to a preset offset below the lowest-order sideband harmonic frequency of the inverter switching frequency, and the upper passband frequency is set to a preset offset above the highest-order significant sideband harmonic frequency.

[0081] Step 3: Obtain the decoupled speed response residual signal;

[0082] After denormalizing and restoring the bus-coupled speed fluctuation estimate component from the normalized least mean square adaptive filter output to the original speed dimensions, the bus-coupled speed fluctuation estimate component is subtracted from the speed response residual signal to obtain the decoupled speed response residual signal. The decoupled speed response residual signal retains only the closed-loop control dominant oscillation component and the mechanical resonant mode oscillation component directly related to the physical characteristics of the motor on this shaft.

[0083] Furthermore, the aforementioned de-standardization restoration refers to multiplying the bus-coupled speed fluctuation estimation component output by the normalized least mean square adaptive filter by the standard deviation used in the Z-score standardization of the speed response residual signal, and adding the mean used in the Z-score standardization of the speed response residual signal. This restores the bus-coupled speed fluctuation estimation component from the standardized numerical domain to the original speed dimension, ensuring that the bus-coupled speed fluctuation estimation component and the speed response residual signal maintain dimensional consistency before the subtraction operation can be performed.

[0084] Step 4: Perform spectrum analysis and frequency component identification on the decoupled signal;

[0085] A Fast Fourier Transform (FFT) is performed on the decoupled rotational speed response residual signal to obtain its spectral distribution, identifying each discrete peak frequency and its corresponding amplitude. The peak frequency with the largest amplitude is marked as the dominant closed-loop oscillation frequency, and the remaining peak frequencies with amplitudes exceeding a preset resonance identification threshold are marked as the set of mechanical resonance frequencies.

[0086] It should be noted that the aforementioned preset resonance identification threshold refers to the amplitude threshold used to determine whether a certain peak in the spectrum corresponds to a mechanical resonance mode. The preset resonance identification threshold is set as a preset proportion of the amplitude corresponding to the closed-loop dominant oscillation frequency, or as a fixed absolute lower limit of amplitude. Spectral peaks with amplitudes lower than the preset resonance identification threshold are considered noise floor or residual leakage components and are not included in the set of mechanical resonance frequencies.

[0087] Step 5: Separate the dominant closed-loop oscillation component from the oscillation components of each resonant mode based on bandpass filtering;

[0088] Using the dominant closed-loop oscillation frequency as the center frequency, bandpass filtering is performed on the decoupled speed response residual signal to extract the dominant closed-loop control oscillation component signal. Using each resonant frequency in the mechanical resonant frequency set as the center frequency, bandpass filtering is performed on the decoupled speed response residual signal to extract the oscillation component signal of each resonant mode.

[0089] It should be noted that the bandwidth of each bandpass filter is determined based on the spacing between adjacent peak frequencies. When the frequency spacing between the dominant closed-loop oscillation frequency and the nearest resonant frequency is small, the bandwidth of the bandpass filter is narrowed to a preset proportion not exceeding this minimum spacing to reduce mutual leakage between frequency components.

[0090] Step 6: Extract the closed-loop control envelope features and resonant mode attenuation features;

[0091] Peak detection is performed on the dominant oscillation component signal of the closed-loop control to extract the extreme point sequence. The overshoot decay sequence is calculated based on the amplitude difference between adjacent extreme points in the same direction. After taking the logarithm of the overshoot decay sequence, linear regression is performed to obtain the closed-loop envelope decay rate. The closed-loop oscillation period is calculated based on the time interval between adjacent extreme points in the same direction. The initial overshoot is calculated based on the ratio of the amplitude of the first extreme point in the extreme point sequence to the speed step.

[0092] Peak detection is performed on the oscillation components of each resonant mode signal to extract their respective extreme point sequences. The logarithm of the amplitude of each extreme point sequence is then used for linear regression to obtain the attenuation rate of each resonant mode. The attenuation rate with the smallest value among all resonant modes is selected as the resonant suppression constraint index, i.e., the attenuation rate corresponding to the slowest attenuation mode.

[0093] It should be noted that the above-mentioned linear regression after taking the logarithm of the overshoot decay sequence refers to: taking the natural logarithm of the amplitude of each adjacent extreme point in the same direction, using the time sequence of the extreme point as the independent variable and the logarithmic amplitude as the dependent variable, and performing a least-squares linear fit. The slope of the fitted line is the closed-loop envelope decay rate. This processing is based on the property that the logarithmic transformation of the exponential decay envelope exhibits a linear relationship.

[0094] Furthermore, the physical meaning of the aforementioned closed-loop envelope attenuation rate is as follows: after taking the natural logarithm of the amplitudes of adjacent extreme points in the same direction, the slope of the line obtained by performing least-squares linear fitting with the time sequence of the extreme points as the independent variable is given. Its dimension is the logarithmic amplitude attenuation within each oscillation period. The more negative the value, the faster the oscillation envelope attenuation and the more sufficient the closed-loop damping. The attenuation rate of each resonant mode is calculated in the same way as the closed-loop envelope attenuation rate. The same logarithmic linear regression is performed on the extreme point sequence of each resonant mode oscillation component, and the resulting slope is the attenuation rate of the corresponding resonant mode. Its dimension is the same as the closed-loop envelope attenuation rate. The closer the value is to zero, the slower the energy dissipation of the resonant mode.

[0095] When the set of mechanical resonant frequencies is empty, meaning no non-closed-loop frequency peaks exceeding the preset resonant identification threshold are identified in the spectrum, it indicates that the current transmission chain resonant components are insignificant or have been sufficiently suppressed. In this case, the resonant suppression constraint index is set to a default value greater than the resonant safe attenuation rate threshold, so that the resonant constraint condition is not triggered in subsequent steps, and the proportional-integral parameter correction is performed only based on the closed-loop envelope characteristics.

[0096] Step 7: Obtain the proportional-integral gain correction factor based on fuzzy inference;

[0097] The closed-loop envelope attenuation rate, closed-loop oscillation period, and initial overshoot are subtracted from their corresponding target performance indicators to obtain closed-loop performance deviation values, including attenuation rate deviation, oscillation period deviation, and overshoot deviation. These deviation values ​​are then input into a pre-configured fuzzy inference system. The fuzzy inference system uses the attenuation rate deviation, oscillation period deviation, and overshoot deviation as input variables. After fuzzification, fuzzy rule inference, and defuzzification operations, it outputs proportional gain correction factor and integral gain correction factor.

[0098] It should be noted that the fuzzy inference system is a standard fuzzy logic inference system. The input to the fuzzy inference system are three closed-loop performance deviation values, and the outputs are a proportional gain correction factor and an integral gain correction factor. The rules in the fuzzy rule base of the fuzzy inference system are pre-set based on qualitative knowledge of the influence of proportional-integral parameters on the closed-loop response characteristics. For example, when the overshoot deviation is large positive and the attenuation rate deviation is large negative, the output proportional gain correction factor is set to a value less than 1 to reduce the proportional gain. Before inputting the three closed-loop performance deviation values ​​into the fuzzy inference system, each deviation value undergoes mean normalization based on its preset universe of discourse, mapping each deviation value to a uniform numerical range. This eliminates the influence of the different dimensions and numerical magnitudes of the attenuation rate deviation, oscillation period deviation, and overshoot deviation on the fuzzy membership calculation.

[0099] Furthermore, the aforementioned fuzzification operation refers to mapping the normalized closed-loop performance deviation values ​​to membership degrees of predefined fuzzy linguistic values ​​(such as negative large, negative small, zero, positive small, and positive large) on the universe of discourse of each input variable. The shape of the membership function (such as triangular, trapezoidal, or Gaussian) and the distribution range of each linguistic value are pre-set based on actual speed loop performance debugging experience. The defuzzification operation adopts the centroid method, that is, the sum of the products of the membership degrees corresponding to each fuzzy linguistic value on the universe of discourse of the output variable and the universe coordinate values ​​of each fuzzy linguistic value, and then dividing by the sum of the membership degrees, transforms the fuzzy inference results into deterministic numerical outputs of the proportional gain correction factor and the integral gain correction factor.

[0100] Furthermore, the aforementioned target performance indicators refer to the target values ​​of the closed-loop envelope decay rate, closed-loop oscillation period, and first overshoot of the speed loop. All three are preset according to the dynamic response design requirements of the motor drive system and have the same dimensions as the closed-loop envelope decay rate, closed-loop oscillation period, and first overshoot, respectively. The corresponding deviations of decay rate, oscillation period, and overshoot reflect the degree of deviation between the actual performance and the expected performance of the current speed loop. A positive deviation indicates that the actual value exceeds the target value, and a negative deviation indicates that the actual value is lower than the target value.

[0101] Step 8: Constrain the proportional-integral correction factor based on the resonant attenuation characteristics;

[0102] Determine whether the resonance suppression constraint index is less than the preset resonance safe attenuation rate threshold. If the resonance suppression constraint index is less than the resonance safe attenuation rate threshold, it indicates that there is a slowly decaying resonance mode, and constraints need to be applied to the adjustment range of the proportional-integral parameter.

[0103] A decay correction is applied to the integral gain correction factor: the integral gain correction factor is multiplied by the ratio of the resonance suppression constraint index to the resonance safe decay rate threshold to obtain the integral gain correction factor after resonance constraint. This correction ensures that the adjustment range of the integral gain is compressed more when the resonance decay is slower. Specifically, let the integral gain correction factor be... The resonance suppression constraint index is The resonant safe attenuation rate threshold is Then the integral gain correction factor after resonance constraint Calculate using the following formula:

[0104]

[0105] in, This refers to the integral gain correction scaling factor output by the fuzzy inference system. The value selected is the smallest among the attenuation rates of all resonant modes. The preset resonant safe attenuation rate threshold, This is the scaling factor for the integral gain correction after resonance constraint. and Since they have the same dimensions, their ratio is a dimensionless ratio. It is a dimensionless correction factor, therefore It is also a dimensionless quantity, and all terms in the formula have the same dimensions. When much smaller hour, Significantly compressed; when near hour, near The constraint effect tends to disappear.

[0106] Furthermore, the aforementioned resonant safe attenuation rate threshold The physical meaning is: when the attenuation rate of the resonant mode is not lower than the resonant safe attenuation rate threshold, it is considered that the energy dissipation of the resonant mode in the closed-loop response of the velocity loop is fast enough and will not pose a safety threat to the adjustment of the proportional-integral parameters; therefore, no resonant constraint needs to be applied. Resonant safe attenuation rate threshold. The resonant safety attenuation rate threshold is preset based on the maximum allowable resonance duration of the transmission system. Its dimensions are the same as the resonant mode decay rate.

[0107] An upper limit constraint is applied to the proportional gain correction factor: a safety factor is calculated based on the ratio of the resonant frequency corresponding to the slowest decaying resonant mode to the current estimated closed-loop bandwidth, limiting the increase in the proportional gain correction factor to no more than the safety factor. The current estimated closed-loop bandwidth is obtained by converting the inverse of the closed-loop oscillation period. Safety factor. Calculate using the following formula:

[0108]

[0109] in, For safety reasons, To attenuate the resonant frequency corresponding to the slowest resonant mode, This is the current estimated closed-loop bandwidth. The preset safety margin coefficient and . and If two quantities have the same dimension, their ratio is a dimensionless quantity. Since it is a dimensionless coefficient, therefore This is a dimensionless safety factor. When... Much larger hour The gain is relatively large, and the constraint on the increase in proportional gain is relatively loose; when near hour As the value approaches 1, the proportional gain is almost impossible to increase.

[0110] Furthermore, in the above safety factor formula, when hour, ,lead to Due to the proportional gain correction factor When a larger proportional gain is needed, it should be greater than 1. If the condition is always true, the proportional gain correction factor will be truncated to... This means that the proportional gain is not allowed to increase, thereby preventing the closed-loop bandwidth from further encroaching on the dangerous region that is already in the resonant frequency band.

[0111] Furthermore, the aforementioned current closed-loop bandwidth estimate The calculation method is as follows: take the reciprocal of the closed-loop oscillation period extracted in step 6 as the current closed-loop bandwidth estimate, that is, the current closed-loop bandwidth estimate. Equal to the frequency corresponding to the closed-loop oscillation period, and the current estimated closed-loop bandwidth. The dimension of the quantity is Hertz, and Since the dimensions are consistent, they can be directly substituted into the safety factor formula for calculation.

[0112] Let the proportional gain correction factor of the fuzzy inference system output be... proportional gain correction factor after resonance constraint Determine by the following formula: If ,but ;like ,but .in, The proportional gain correction factor for the output of the fuzzy inference system. This is the proportional gain correction factor after resonance constraint. This is the safety factor. That is, the upper limit of the proportional gain correction factor is truncated to the safety factor. This is to prevent the proportional gain from increasing too quickly, which could cause the closed-loop bandwidth to intrude into the resonant frequency band.

[0113] If the resonance suppression constraint index is not less than the resonance safe attenuation rate threshold, then the proportional gain correction factor and integral gain correction factor are not modified and are directly used as the proportional-integral correction factor after resonance constraint.

[0114] Step 9: Write the corrected proportional-integral parameters into the speed loop controller;

[0115] The current proportional gain of the speed loop is multiplied by the proportional gain correction factor after resonance constraint to obtain the corrected proportional gain. The current integral gain of the speed loop is multiplied by the integral gain correction factor after resonance constraint to obtain the corrected integral gain. Parametric safety domain constraint processing is then performed on the corrected proportional and integral gains, i.e., it is determined whether the corrected proportional and integral gains fall within the preset proportional gain safety range and integral gain safety range, respectively. If they exceed the safety range, the corrected proportional or integral gain is truncated to the boundary value of the safety range. The proportional-integral parameters after parameter safety domain constraint processing are written into the permanent magnet synchronous motor speed loop controller, completing one adaptive tuning process.

[0116] It should be noted that the aforementioned proportional gain safety range and integral gain safety range are preset based on the physical constraints of the motor drive system. The lower limit of the proportional gain safety range ensures that the speed loop has basic tracking capability, while the upper limit prevents nonlinear oscillations caused by control signal saturation. The lower limit of the integral gain safety range ensures that the steady-state error is eliminated within an acceptable range, while the upper limit prevents large overshoot caused by integral saturation.

[0117] To prevent excessively large single-correction amplitudes from causing drastic changes in system response, step 9 further limits the rate of change of parameters during a single correction. Specifically, the difference between the corrected proportional-integral (PI) parameter and the current PI parameter is calculated. If the absolute value of the difference exceeds a preset maximum single-correction change, the difference is truncated to the maximum single-correction change and added to the current parameter as the actual PI parameter value written. This process makes the parameter adjustment process smoother and reduces the risk of transient response degradation caused by a single large-amplitude parameter adjustment.

[0118] This implementation addresses the misidentification of the sources of two types of non-control response components in the adaptive proportional-integral parameter tuning of the speed loop of a permanent magnet synchronous motor under common DC bus elastic drive conditions. Before frequency domain analysis, normalized least mean square adaptive filtering is used with the directly measurable bus voltage fluctuation signal as a reference input. Based on the physical transmission causal relationship between the bus voltage fluctuation signal and the coupling component introduced by the bus voltage fluctuation signal in the speed signal, the exogenous speed fluctuation component introduced by bus coupling is estimated and removed. Since normalized least mean square adaptive filtering estimates the path through the linear correlation between the reference signal and the coupling component, and the mechanical resonant oscillation component does not have this correlation with the bus voltage fluctuation signal, the resonance information is completely preserved during the decoupling process and is not mistakenly absorbed.

[0119] After stripping the bus coupling components, the decoupled speed response residual signal contains only two frequency components directly related to the physical characteristics of the motor on this shaft: the closed-loop control dominant oscillation and the mechanical resonant mode oscillation. This eliminates the difficulty in determining the source of the spectral peak when the bus coupling ripple frequency overlaps with the resonant frequency. When subsequent Fast Fourier Transform is used to identify the frequency components of the decoupled speed response residual signal, the physical attribution of each peak frequency is determined, and the division between the closed-loop dominant oscillation frequency and the resonant frequency is reliable. Based on this, the envelope attenuation characteristics of the closed-loop dominant oscillation are independently extracted using bandpass filtering. This ensures that the closed-loop envelope attenuation rate, closed-loop oscillation period, and initial overshoot accurately reflect the actual impact of the current proportional-integral parameters on the speed loop closed-loop performance, making the correction direction and amplitude output by the fuzzy inference system reliable. Simultaneously, the independently extracted resonant mode attenuation characteristics reflect the true resonant state of the drivetrain. The resonant constraint uses the true attenuation rate of the slowest-attenuating resonant mode to limit the adjustment amplitude of the proportional-integral parameters, preventing the closed-loop bandwidth from increasing too rapidly and encroaching on the resonant frequency band. This allows for adaptive correction of the speed loop proportional-integral parameters while also considering resonance suppression requirements.

[0120] The following is an example of an application of the present invention, such as Figure 2-9 As shown, the implementation process is as follows:

[0121] A multi-axis CNC machining center is equipped with three permanent magnet synchronous motors (driving the X, Y, and Z axes respectively), and the three-axis inverters share the same DC bus (rated bus voltage 540V). The X-axis motor has a rated speed of 3000 rpm and drives the worktable via a ball screw drive chain connected by a flexible coupling. After a tool change in 20XX, the machining center restarted the X-axis feed, and the speed loop setpoint jumped from 0 rpm to 1500 rpm. Due to the continuous bus voltage ripple generated by the switching actions of the Y-axis and Z-axis inverters, and the presence of a low-frequency resonant mode in the X-axis flexible coupling, the speed signal in this step response simultaneously superimposed three types of frequency components. The current parameters of the speed loop controller are: proportional gain. Integral gain .

[0122] After detecting a step jump from 0 rpm to 1500 rpm in the X-axis speed setpoint, the controller synchronously acquires the actual X-axis speed sequence and DC bus voltage sequence with a sampling period of 0.5 ms. The preset steady-state band is set to ±15 rpm of the target speed (corresponding to 0.5% of the rated speed of 3000 rpm), and acquisition stops after 20 consecutive sampling points fall within the steady-state band. A total of 480 sampling points were acquired in this acquisition, with an acquisition time of 240 ms.

[0123] Table 1. Excerpts of raw data collected synchronously (key moments)

[0124]

[0125] The steady-state speed value is the average of the last 20 sampling points. The average steady-state speed is calculated to be 1500.2 rpm, which meets the judgment condition of falling into the steady-state band of [1485, 1515] rpm, and the sampling is terminated.

[0126] The X-axis inverter switching frequency is 8kHz, while the Y-axis and Z-axis inverter switching frequencies are 8kHz and 10kHz, respectively. The bus voltage ripple is mainly concentrated in the sideband harmonic frequency band. The bandpass filter passband is set to 7.2kHz to 10.8kHz, and bandpass filtering is performed on the 480-point bus voltage sampling sequence to extract the bus voltage fluctuation signal.

[0127] Subtracting the steady-state speed mean of 1500.2 rpm from the actual speed response sequence yields the speed response residual signal. Z-score normalization is then applied to both the bus voltage fluctuation signal and the speed response residual signal.

[0128] Table 2 Summary of Z-score standardized parameters

[0129]

[0130] After standardization, the bus voltage fluctuation signal is used as the reference input and the speed response residual signal is used as the desired output. Normalized minimum mean square adaptive filtering (filter order 16, step size factor 0.02) is performed, and after convergence, the estimated bus coupled speed fluctuation component in the standardized domain is output.

[0131] The normalized bus coupling estimate component of the normalized least mean square adaptive filter output is inversely normalized and restored using the following formula:

[0132]

[0133] Taking a typical sampling point as an example, the normalized minimum mean square output Substitute the values:

[0134]

[0135] The de-standardized bus coupling estimation components are subtracted point by point from the speed response residual signal to obtain the decoupled speed response residual signal.

[0136] Table 3. Signal stripping process at key sampling points

[0137]

[0138] The residual signal after bus coupling removal retains only the closed-loop control dominant oscillation component and the mechanical resonant mode oscillation component, and the high-frequency fluctuations introduced by the bus ripple have been effectively removed.

[0139] Perform a Fast Fourier Transform on the 480-point decoupled speed response residual signal, with a frequency resolution of [missing value]. Hz. The discrete peak frequencies and corresponding amplitudes in the spectrum are identified as follows:

[0140] Table 4. Spectrum Peak Identification Results

[0141]

[0142] The preset resonance identification threshold is set to 25% of the amplitude of the closed-loop dominant oscillation, i.e. Peak frequency 2 (19.8 rpm) exceeds the threshold and is included in the mechanical resonant frequency set; peak frequency 3 (7.2 rpm < 15.6 rpm) is below the threshold and is considered residual leakage, so it is not included; peak frequency 4 (2.1 rpm < 15.6 rpm) is below the threshold and is considered noise floor, so it is not included. The final mechanical resonant frequency set is {47.6 Hz}.

[0143] The interval between the dominant closed-loop oscillation frequency of 18.3Hz and the nearest resonant frequency of 47.6Hz is 29.3Hz. Centered on 18.3Hz, the bandwidth of the bandpass filter is set to not exceed 60% of 29.3Hz, i.e., approximately 17.6Hz, with an actual set bandwidth of 16Hz (passband range 10.3Hz to 26.3Hz), and the dominant closed-loop control oscillation component signal is extracted.

[0144] Centered at 47.6Hz, with a bandwidth set to 20Hz (passband range 37.6Hz to 57.6Hz), the mechanical resonant mode oscillation component signal is extracted. Both bandpass filters are applied to the 480-point decoupling residual signal, outputting two independent time-domain sequences of the oscillation components.

[0145] Peak detection was performed on the dominant oscillation component signal of the closed-loop control, identifying five unidirectional maxima. The amplitude and timing of each maxima are as follows:

[0146] Table 5. Sequence of extreme points of closed-loop dominant oscillation and log-linear regression.

[0147]

[0148] Using the extreme value index (1 to 5) as the independent variable and the logarithmic magnitude as the dependent variable, a least-squares linear fit was performed, and the fitting result is as follows:

[0149]

[0150] Closed-loop envelope attenuation rate (Logarithmic amplitude decay per oscillation period).

[0151] The time interval between adjacent extreme points in the same direction is approximately 54.7 ms, and the closed-loop oscillation period is... The initial overshoot is calculated based on the ratio of the amplitude of the first extreme point to the speed step.

[0152]

[0153] Peak detection was performed on the mechanical resonant mode oscillation component signal, and four unidirectional maxima were identified:

[0154] Table 6. Mechanical Resonance Mode Extreme Point Sequence and Log-Linear Regression

[0155]

[0156] The slope of the linear fit is the resonant mode attenuation rate. (Logarithmic amplitude decay per oscillation period). The set of mechanical resonant frequencies contains only one element, and the resonance suppression constraint index is directly taken as... .

[0157] The target performance indicator is preset as follows: target value of closed-loop envelope attenuation rate. Target value of closed-loop oscillation period ms, target value for initial overshoot .

[0158] Calculate the three closed-loop performance deviation values:

[0159] Attenuation rate deviation: (Insufficient actual attenuation, resulting in slow oscillation)

[0160] Oscillation period deviation: ms (actual oscillation period is longer, response is slower)

[0161] Overshoot deviation: (Overshoot is lower than the target, there is still margin)

[0162] Each deviation value is normalized to the mean based on the universe of discourse before being input into the fuzzy inference system. (Attenuation rate deviation universe of discourse) Oscillation period deviation domain ms, overshoot deviation domain After normalization, the three biases are mapped to... Interval:

[0163]

[0164]

[0165]

[0166] The fuzzy inference system determines the following: a large positive decay rate deviation (slow response), a large positive oscillation period deviation (slow response), and a small negative overshoot deviation (still some overshoot margin). This triggers the rule: "When both the decay rate deviation and the oscillation period deviation are positive, appropriately increase the proportional gain and integral gain." The output after defuzzification using the centroid method is:

[0167] Proportional gain correction factor Integral gain correction factor .

[0168] Preset resonant safe attenuation rate threshold (The dimensions are the same as the attenuation rate of the resonant mode).

[0169] Judgment: Resonance suppression constraint index Its absolute value is less than ,Right now (numerically) Greater than This indicates that the absolute value of the resonant attenuation rate is less than the absolute value of the safety threshold, the attenuation is too slow, and constraints need to be applied.

[0170] Perform attenuation correction on the integral gain correction factor:

[0171]

[0172] The integral gain correction factor has been reduced from 1.14 to 0.785, and the integral gain will not be increased this time.

[0173] Apply an upper limit constraint to the proportional gain correction factor and calculate the safety factor. Current closed-loop bandwidth estimate:

[0174]

[0175] Slowest decaying resonant frequency Hz, preset safety margin factor Substitute into the safety factor formula:

[0176]

[0177] Compare and : ,therefore The proportional gain correction factor does not need to be truncated.

[0178] Table 7 Comparison of correction factors before and after resonance constraint

[0179]

[0180] Calculate the corrected parameters by multiplying the correction factor after resonance constraint by the current parameters:

[0181]

[0182]

[0183] Preset proportional gain safety range Integral gain safety range : Within a safe range, Within safe limits, no truncation is necessary.

[0184] Maximum rate of change limit for single execution: The maximum single change of proportional gain is preset to 0.25, and the maximum single change of integral gain is 4.0.

[0185] Proportional gain change: No amplitude limiting is triggered; Integral gain change: It does not trigger the amplitude limit.

[0186] The final parameters written to the speed loop controller are , .

[0187] Table 8 Comparison of proportional-integral parameter tuning before and after.

[0188]

[0189] The data flow in this adaptive tuning process demonstrates a clear logical chain: the 480-point speed response sequence and bus voltage sequence acquired in step 1 serve as the raw input for all subsequent processing; step 2 utilizes the Z-score normalized parameters of the bus voltage fluctuation signal ( rpm In step 3, the denormalization restoration is performed to ensure that the dimensions of the bus coupling estimated component and the speed residual signal are consistent. Then, the subtraction is performed to obtain the decoupled bus residual signal. In step 4, the fast Fourier transform result of this signal determines the mechanical resonant frequency set {47.6Hz}. This frequency is used as the center frequency of the bandpass filter in step 5 and as the center frequency in step 8. Substituting the values ​​into the safety factor formula, the frequency identification result is directly transferred to the constraint parameters; the closed-loop oscillation period of 54.7ms extracted in step 6 serves as both the performance deviation input in step 7 and the estimated closed-loop bandwidth in step 8. Hz; After the fuzzy inference correction factor output in step 7 is constrained by the resonance attenuation characteristic in step 8, the integral gain correction factor is compressed from 1.14 to 0.785, reflecting the resonance suppression constraint index. With safety threshold The ratio is used to quantitatively limit the integral adjustment range; finally, step 9 applies the constrained correction factor to the current parameter, completing the complete closed-loop data flow from the original sampled data to the speed loop parameter update.

[0190] The embodiments of the present invention have been described above. However, these embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many equivalent embodiments based on the guidance of these embodiments, and all of these are within the protection scope of these embodiments. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. The above examples are only for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be noted that due to the limitations of textual expression, there are objectively infinite specific structures. For those skilled in the art, several improvements, modifications, or changes can be made without departing from the principles of the present invention, and the above technical features can also be combined in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the protection scope of the present invention.

Claims

1. A method for adaptive proportional-integral parameter tuning of the speed loop of a permanent magnet synchronous motor, characterized in that, Includes the following steps: When a step change in the speed loop setpoint is detected, the actual speed response sequence of the motor on this shaft and the DC bus voltage sampling sequence are collected simultaneously. Bandpass filtering is performed on the DC bus voltage sampling sequence to extract the bus voltage fluctuation signal. The bus voltage fluctuation signal is used as the reference input and the speed response residual signal is used as the desired output. Normalized minimum mean square adaptive filtering is performed to output the bus coupled speed fluctuation estimation component. The bus-coupled speed fluctuation estimation component is subtracted from the speed response residual signal to obtain the decoupled speed response residual signal; Perform a fast Fourier transform on the decoupled rotational speed response residual signal to identify the set of closed-loop dominant oscillation frequencies and mechanical resonant frequencies; Using the closed-loop dominant oscillation frequency and each resonant frequency as the center frequency, bandpass filtering is performed on the decoupled speed response residual signal to extract the closed-loop control dominant oscillation component signal and each resonant mode oscillation component signal respectively. The closed-loop envelope attenuation rate, closed-loop oscillation period, and first overshoot are extracted from the closed-loop control dominant oscillation component signal, and the attenuation rate of each resonant mode is extracted from each resonant mode oscillation component signal. Input the closed-loop performance deviation value into the fuzzy inference system, and output the proportional gain correction factor and the integral gain correction factor. Based on the attenuation rate of each resonant mode, a resonant constraint is applied to the proportional gain correction factor and the integral gain correction factor to obtain the proportional-integral correction factor after resonant constraint. Multiply the current proportional-integral parameters of the velocity loop by the proportional-integral correction factor after resonance constraint, and write them into the velocity loop controller.

2. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The synchronous acquisition of the actual speed response sequence of the local shaft motor and the DC bus voltage sampling sequence includes: Starting from the step moment, the actual speed response sequence and the DC bus voltage sampling sequence are synchronously and continuously collected at a fixed sampling period until the actual speed falls within the preset steady-state band centered on the step target speed for several consecutive sampling periods, at which point data collection stops. The preset steady-state band is a speed range defined by offsetting the step target speed by one steady-state tolerance value above and below it. The steady-state tolerance value is preset as a percentage of the motor's rated speed.

3. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, Before performing the normalized least mean square adaptive filtering operation, the following steps are also included: The bus voltage fluctuation signal and the speed response residual signal are respectively subjected to Z-score normalization. The normalized least mean square adaptive filtering operation generates an estimate of the bus coupling speed fluctuation by weighting and summing the delayed tap sequence of the reference input using the adaptive weight coefficient vector, and uses the error signal between the estimate and the speed response residual signal to drive the weight coefficient update. The lower passband frequency of the bandpass filter is set to a preset offset that is lower than the lowest order sideband harmonic frequency of the inverter switching frequency, and the upper passband frequency is set to a preset offset that is higher than the highest order significant sideband harmonic frequency.

4. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 3, characterized in that, Before subtracting the bus-coupled speed fluctuation estimation component from the speed response residual signal, the following is also included: The bus-coupled speed fluctuation estimation component output by the normalized least mean square adaptive filter is multiplied by the standard deviation of the speed response residual signal used in Z-score standardization, and then the mean of the speed response residual signal used in Z-score standardization is added to restore the bus-coupled speed fluctuation estimation component from the standardized numerical domain to the original speed dimension.

5. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The set of identified closed-loop dominant oscillation frequencies and mechanical resonant frequencies includes: The peak frequency with the largest amplitude in the spectral distribution is marked as the closed-loop dominant oscillation frequency; The remaining peak frequencies whose amplitude exceeds the preset resonance identification threshold are marked as the set of mechanical resonance frequencies; The preset resonance identification threshold is set as a preset proportion of the amplitude corresponding to the closed-loop dominant oscillation frequency, or is set as a fixed absolute amplitude lower limit.

6. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The extraction of the closed-loop envelope decay rate, closed-loop oscillation period, and first overshoot includes: Peak detection is performed on the dominant oscillation component signal of the closed-loop control to extract the extreme point sequence; The natural logarithm of the amplitude between adjacent extreme points in the same direction is taken, and least squares linear fitting is performed with the time sequence of the extreme point as the independent variable and the logarithmic amplitude as the dependent variable. The slope of the fitted line is the closed-loop envelope attenuation rate. The closed-loop oscillation period is calculated based on the time interval between adjacent extreme points in the same direction. The initial overshoot is calculated based on the ratio of the amplitude of the first extreme point in the extreme point sequence to the speed step. Peak detection is performed on the oscillation component signals of each resonant mode. After taking the logarithm of the amplitude of each extreme point sequence, linear regression is performed to obtain the attenuation rate of each resonant mode. The smallest value among the attenuation rates of each resonant mode is selected as the resonant suppression constraint index. When the set of mechanical resonant frequencies is empty, the resonant suppression constraint index is set to a default value greater than the resonant safe attenuation rate threshold.

7. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The step of inputting the closed-loop performance deviation value into the fuzzy inference system includes: The attenuation rate deviation, oscillation period deviation and first overshoot deviation are obtained by subtracting the closed-loop envelope attenuation rate, closed-loop oscillation period and first overshoot from the corresponding target performance indicators. The attenuation rate deviation, oscillation period deviation, and overshoot deviation are respectively normalized based on the mean of the range based on the preset universe of discourse for each deviation value. The normalized closed-loop performance deviation values ​​are input into the fuzzy inference system. After fuzzification, fuzzy rule inference and defuzzification operations, the proportional gain correction factor and integral gain correction factor are output. The defuzzification operation adopts the centroid method, which sums the products of the membership degree corresponding to each fuzzy linguistic value on the universe of discourse of the output variable and the universe coordinate value of each fuzzy linguistic value, and then divides the sum of the membership degrees to transform the fuzzy inference result into a deterministic numerical output.

8. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The resonant constraint applied to the proportional gain correction factor and integral gain correction factor based on the attenuation rate of each resonant mode includes: Determine whether the resonance suppression constraint index is less than the preset resonance safe attenuation rate threshold; If the resonance suppression constraint index is less than the resonance safe attenuation rate threshold, then the integral gain correction factor is multiplied by the ratio of the resonance suppression constraint index to the resonance safe attenuation rate threshold to obtain the integral gain correction factor after resonance constraint. The safety factor is calculated based on the ratio of the resonant frequency corresponding to the slowest decaying resonant mode to the current closed-loop bandwidth estimate. The current closed-loop bandwidth estimate is obtained by converting the reciprocal of the closed-loop oscillation period. When the proportional gain correction factor is greater than the safety factor, the proportional gain correction factor is truncated to the safety factor to obtain the proportional gain correction factor after resonance constraint. If the resonance suppression constraint index is not less than the resonance safe attenuation rate threshold, then the proportional gain correction factor and the integral gain correction factor are not corrected.

9. The adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor according to claim 1, characterized in that, The step of multiplying the current proportional-integral parameters of the velocity loop by the proportional-integral correction factor after resonance constraint and writing them into the velocity loop controller further includes: The corrected proportional gain and integral gain are subjected to parameter safety domain constraint processing respectively. It is determined whether the corrected proportional gain and integral gain fall within the preset proportional gain safety range and integral gain safety range. If they exceed the safety range, the corrected proportional gain or integral gain is truncated to the boundary value of the safety range. Calculate the difference between the corrected proportional-integral parameter and the current proportional-integral parameter. If the absolute value of the difference exceeds the preset maximum single change, the difference is truncated to the maximum single change and added to the current parameter as the actual written proportional-integral parameter value.

10. A system for adaptive proportional-integral parameter tuning of the speed loop of a permanent magnet synchronous motor, used to execute the adaptive proportional-integral parameter tuning method for the speed loop of a permanent magnet synchronous motor as described in any one of claims 1 to 9, characterized in that, include: The data synchronization acquisition module is used to synchronously acquire the actual speed response sequence of the motor on this shaft and the DC bus voltage sampling sequence when a step change in the speed loop setpoint is detected. The bus coupling estimation module is used to perform bandpass filtering on the DC bus voltage sampling sequence to extract the bus voltage fluctuation signal. Using the bus voltage fluctuation signal as a reference input and the speed response residual signal as the desired output, it performs normalized minimum mean square adaptive filtering operation to output the bus coupling speed fluctuation estimation component. The decoupling processing module is used to subtract the bus-coupled speed fluctuation estimation component from the speed response residual signal to obtain the decoupled speed response residual signal; The spectrum analysis and identification module is used to perform a fast Fourier transform on the decoupled rotational speed response residual signal to identify the set of closed-loop dominant oscillation frequencies and mechanical resonant frequencies. The frequency component separation module is used to perform bandpass filtering on the decoupled speed response residual signal with the closed-loop dominant oscillation frequency and each resonant frequency as the center frequency, and extract the closed-loop control dominant oscillation component signal and each resonant mode oscillation component signal respectively. The feature extraction module is used to extract the closed-loop envelope attenuation rate, closed-loop oscillation period, and first overshoot from the closed-loop control dominant oscillation component signal, and to extract the attenuation rate of each resonant mode from each resonant mode oscillation component signal. The fuzzy inference module is used to input the closed-loop performance deviation value into the fuzzy inference system and output the proportional gain correction factor and the integral gain correction factor. The resonant constraint module is used to apply resonant constraints to the proportional gain correction factor and the integral gain correction factor based on the attenuation rate of each resonant mode, so as to obtain the proportional-integral correction factor after resonant constraint. The parameter writing module is used to multiply the current proportional-integral parameters of the speed loop by the proportional-integral correction factor after resonance constraint and write them to the speed loop controller.

Images