A motor rotor dynamic balance test system and method

Abstract

The application discloses a motor rotor dynamic balance test system and method, comprising: arranging an acceleration sensor and an encoder, collecting vibration, angular position signal and current signal; establishing a multi-resonance phase-locked loop, outputting a de-bouncing phase and an angular position sequence, forming a phase reference; angular domain resampling and sparse harmonic decomposition, extracting complex amplitude and generating observation and working condition data; complex value reversible flow model reverse mapping, combining constraint processing, outputting target counterweight instruction; target counterweight instruction injection virtual correction, calculating vibration decline rate, dynamically optimizing counterweight instruction; physical correction and recheck, archiving working condition label, counterweight instruction and recheck result after passing. The application realizes intelligent detection and correction of motor rotor dynamic balance by fusing multi-resonance phase-locked loop and complex value reversible flow model, and is free of test weight, high in precision and capable of multi-working condition.

Description

Technical Field

[0001] This invention relates to the field of motor testing and dynamic balancing technology, and in particular to a motor rotor dynamic balancing testing system and method. Background Technology

[0002] In the manufacturing and assembly of electric motors, the dynamic balance performance of the rotor directly affects the motor's vibration level, operating efficiency, and service life. Existing rotor dynamic balancing testing methods typically rely on the trial weight method or vibration measurement under a single steady-state condition. This involves attaching trial weights to the rotor surface and gradually adjusting their position and mass to determine the imbalance. While traditional methods can achieve dynamic balance correction under laboratory conditions, the testing process is time-consuming, reliant on manual experience, and difficult to meet the efficiency requirements of large-scale production. As electric motors develop towards higher speeds and higher power, traditional dynamic balancing methods are increasingly showing their shortcomings in terms of testing accuracy and stability.

[0003] Currently, a common improvement approach is to use spectral analysis under a single steady-state condition to extract first- or second-order vibration components to calculate rotor imbalance. While this can reduce manual operation to some extent, it relies on steady-state operating conditions and struggles to maintain accuracy under acceleration, deceleration, or complex load changes. Furthermore, the calculation process often lacks physical constraints, leading to potential deviations in the calculated counterweight results. Actual correction often requires multiple iterations, increasing testing costs and production cycles. Traditional methods also often fail to conduct effective safety verification before correction, and improper correction can potentially cause abnormal motor vibration or damage.

[0004] Based on the above existing technological background, it can be seen that the traditional motor rotor dynamic balancing test method has four main defects: First, it relies too much on test weights and human experience, resulting in low efficiency; second, it can only work under a single steady-state condition and cannot adapt to variable speed and complex conditions; third, the counterweight calculation lacks constraints, which can easily lead to inaccurate correction; and fourth, it lacks a virtual pre-simulation and verification mechanism before correction, making it difficult to guarantee the reliability of the correction results.

[0005] Therefore, how to provide a dynamic balancing test system and method for motor rotors is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a dynamic balancing test system and method for motor rotors. This invention fully utilizes multi-resonant phase-locked loops, complex-valued reversible flow models, and virtual correction pre-simulation technology, and details the process of achieving high-precision dynamic balancing testing and correction without trial weights under multiple operating conditions. This invention collects vibration, angle, and current signals through sensors, establishes a de-jittering phase reference, performs angular domain resampling and sparse harmonic decomposition, and extracts key vibration indicators. Combining the inverse mapping and physical constraint processing of the complex-valued reversible flow model, it generates target counterweight instructions that meet manufacturability requirements. Through virtual correction pre-simulation, it completes safety verification and parameter optimization before actual correction, and finally completes re-inspection and archiving under target and extreme operating conditions. This invention has the advantages of requiring no trial weights, adapting to variable speeds and complex operating conditions, allowing for pre-simulation verification of correction results, high testing accuracy, and suitability for mass production applications.

[0007] A method for dynamic balancing of a motor rotor according to an embodiment of the present invention includes:

[0008] A triaxial accelerometer and a speed encoder are arranged in the motor bearing housing. The upper limit of the amplitude and frequency band of the current injection on the drive side are set. The motor is run under constant speed and variable speed sweep frequency band conditions to collect vibration signals, encoder angular position signals and current signals.

[0009] A multi-resonance phase-locked loop is established based on the encoder angular position signal. The first, second and third mechanical frequencies are phase-locked in parallel, and the first, second and third phase sequences and angular position sequences are output to form a phase reference.

[0010] Angular domain non-equidistant resampling and sparse harmonic decomposition are performed on vibration and current signals to extract first-order and second-order complex amplitude values ​​as vibration indices, generate multi-point observation data, and form working condition data.

[0011] A complex-valued reversible flow model is constructed, which includes a loop graph equal-variable coupling layer, a complex-valued condition modulation unit, a cross-plane bridging unit, and a physical constraint layer. Multi-point observation data and working condition data are used to perform reverse mapping, and manufacturability projection is executed to obtain the target counterweight command.

[0012] The target counterweight command is mapped to an angular synchronous electromagnetic torque synchronized with the first-order phase. Virtual correction and pre-simulation are performed by injecting the torque online through the driver. The first-order vibration descent rate is calculated in real time based on the phase reference, and the second-order component constraint is checked. When the preset threshold is not reached, the parameters of the complex condition modulation unit and the cross-plane bridging unit are optimized to complete the dynamic reconstruction of the target counterweight command.

[0013] According to the dynamically reconstructed target counterweight instruction, the deweighting operation is carried out in the corresponding plane and sector. The vibration index is re-inspected under the target working condition and extreme working condition. After the acceptance limit is met, the working condition label, target counterweight instruction and re-inspection results are archived and written into the model family parameter library.

[0014] Optionally, the setting of the upper limit of the amplitude and frequency band of the drive-side current injection means that the amplitude of the injected current does not exceed three percent of the rated current of the motor, and the injection frequency band is located within the frequency range of the first-order mechanical rotation frequency and ten percent above and below the first-order mechanical rotation frequency.

[0015] Optionally, the operation of the motor under constant speed and variable speed sweeping conditions refers to maintaining the motor rotor at a preset constant speed under constant speed conditions, and gradually accelerating the motor rotor from low speed to rated speed according to a preset speed change rate under variable speed sweeping conditions, thereby obtaining operating data covering different speed ranges through constant speed and variable speed sweeping conditions.

[0016] Optionally, the first-order, second-order, and third-order phase sequences and angular position sequences of the output dejittering form a phase reference, including:

[0017] Based on the encoder angular position signal, the angle unfolding and pulse missing detection are performed to generate continuous angular trajectory and mechanical angular velocity information, establish three parallel tracking channels of first order, second order and third order, and complete the initial phase alignment;

[0018] Resonant phase-locked loops are run in three parallel tracking channels of first, second and third order respectively. The loop bandwidth and damping parameters are adjusted according to the mechanical angular velocity and angular acceleration. Harmonic cooperative constraints are introduced to keep the second-order phase consistent with the first-order phase and the third-order phase consistent with the first-order phase. Auxiliary phase references for vibration signal and current signal are enabled.

[0019] The outputs of the three parallel tracking channels (first-order, second-order, and third-order) are subjected to jitter reduction processing to form jitter-reduced first-order phase sequence, jitter-reduced second-order phase sequence, jitter-reduced third-order phase sequence, and jitter-reduced angular position sequence. At the same time, an equal-angle sampling sequence for angular domain resampling is generated. All sequences are merged and defined as phase reference.

[0020] Calculate the phase quality score and lock status marker, complete the zero-point reference calibration and phase offset correction, and record the adaptive parameters, cooperative constraint deviation and jitter suppression threshold of the three parallel tracking channels of the first, second and third order.

[0021] Optionally, the calculation of phase quality score and lock-in status marker, and the completion of zero-point reference calibration and phase offset correction, includes:

[0022] The scoring indicators are phase continuity, instantaneous jitter mean square error, frequency doubling consistency error, harmonic signal-to-noise ratio and loop residual, which are normalized to 0 to 100 points according to preset weights.

[0023] When the phase quality score is greater than or equal to 90 points and the loop residual within the continuous window is lower than the preset threshold, it is marked as locked.

[0024] When the phase quality score is between 60 and 90 or the loop residual is in the neighborhood of the threshold, it is marked as tracking.

[0025] A phase quality score below 60 or a missing pulse is marked as lost.

[0026] Zero-position reference calibration is completed using the encoder zero-position mark, and bias correction is performed on the first-order phase, second-order phase, and third-order phase respectively, so that the first-order phase is zero at the zero position, the second-order phase is twice the first-order phase at the zero position, and the third-order phase is three times the first-order phase at the zero position.

[0027] Optionally, the step of performing angular domain non-equidistant resampling and sparse harmonic decomposition on the vibration and current signals, extracting first-order and second-order complex amplitude values ​​as vibration indicators, generating multi-point observation data, and forming operating condition data includes:

[0028] Call the angular position sequence and phase reference to construct an equal angle sampling sequence and set the number of sampling points per revolution to two thousand points, and set the acquisition window to cover the constant speed segment and the variable speed sweep frequency segment;

[0029] For each vibration measurement point, non-equidistant angular domain resampling is performed on the vibration signal to form an angular domain vibration sequence. For the current signal, angular domain resampling and fundamental component demodulation are performed to form an angular domain current reference sequence.

[0030] Phase self-correcting sparse decomposition is performed on the angular domain vibration sequence. The phase shift of each revolution is corrected based on the phase reference. The first harmonic component and the second harmonic component are extracted using the angular domain atom set that conforms to the sector ring topology, and the amplitude and phase are generated.

[0031] Dual-source suppression separation is performed on the angular domain vibration sequence and the angular domain current reference sequence. The angular domain current reference sequence is used as the suppression reference to eliminate the influence of electromagnetic torque ripple and switching harmonics on the first and second harmonic estimations, and the scrambled amplitude and scrambled phase are obtained.

[0032] The constant speed segment results and the variable speed frequency sweep results are subjected to cross-speed consistency constraints and abnormal segment removal. Based on phase continuity, instantaneous jitter, harmonic consistency and harmonic signal-to-noise ratio, confidence scores for the first harmonic and second harmonic are generated and valid angle segments are marked.

[0033] The de-scrambling amplitude and phase of the first and second harmonics of all measuring points are summarized to generate multi-measuring-point observation data, and operating condition data are generated at the same time.

[0034] Optionally, obtaining the target counterweight instruction includes:

[0035] Receive observation data from multiple measurement points and operating condition data, complete time alignment, phase alignment and sector index alignment, establish a dual-plane sector coordinate system, and set the disabled sector, single sector quality upper and lower limits and the total plane quality upper limit.

[0036] Construct a complex-valued reversible flow model, where:

[0037] The loop graph-equal coupling layer performs sector cyclic alignment and parameter sharing on the dual-plane sector vector, ensuring that the output is equivalently shifted under any sector cyclic shift.

[0038] The complex-valued conditional modulation unit generates amplitude modulation coefficients and phase modulation coefficients based on the operating condition data, and performs amplitude scaling and phase shifting on the intermediate representation of the loop diagram equal-variable coupling layer.

[0039] The cross-plane bridging unit establishes a coupling mapping between plane A and plane B, and transmits and merges information between the two planes according to the sector correspondence and coupling weight;

[0040] The physical constraint layer records and loads manufacturing and assembly parameters within the complex-valued reversible flow model, and provides interfaces for feasibility verification and adjustment.

[0041] Input the multi-point observation data and operating condition data into the complex value reversible flow model, execute the reverse solution process to generate the initial solution of the counterweight, and simultaneously output the amplitude, phase and confidence score of each sector;

[0042] The physical constraint layer is invoked to perform manufacturability processing on the initial solution of the counterweight. The corresponding sector is set to zero according to the prohibited sector. The amplitude is truncated and the counterweight is balanced according to the upper and lower limits of the single sector mass and the upper limit of the total mass of the plane. The amplitude is quantized to the allowable step according to the mass step. Local smoothing and overall correction are performed according to the differential amplitude limit of adjacent sectors and the center of gravity offset limit to obtain the counterweight solution that meets the assembly and manufacturing constraints.

[0043] The solution that satisfies the constraints is converted into a target weight instruction, which includes a plane identifier, sector number, mass value, phase angle, and confidence score.

[0044] Optionally, the dynamic reconstruction of the target counterweight command includes:

[0045] The target counterweight command is converted to generate an angular synchronous electromagnetic torque injection command that is synchronized with the first-order phase. During the generation process, the injection amplitude and injection frequency band are strictly set according to the limited upper limit of amplitude and frequency band range. Synchronization alignment is completed based on the phase reference, and virtual correction pre-play is completed through online injection by the driver.

[0046] Before and after electromagnetic torque injection, first-order and second-order vibration indices are extracted under the same time and angle windows. The first-order vibration indices before and after injection are compared, the relative decrease is calculated and used as the first-order vibration decrease rate, and the change in the second-order vibration index is recorded.

[0047] The virtual correction results are checked according to preset judgment conditions, which require that the first-order vibration decrease rate is not lower than the set threshold and the second-order vibration index does not exceed the set tolerance. When the judgment conditions are met, the process directly enters the re-inspection stage; when the judgment conditions are not met, the parameter optimization stage is executed.

[0048] Adaptive iterative optimization is performed on the modulation coefficient parameters of the complex-condition modulation unit and the coupling weight parameters of the cross-plane bridging unit. The optimization steps follow small step size, bounded amplitude and stability constraints. The optimization objective is to reduce the first-order vibration index after injection and avoid the second-order vibration index from exceeding the tolerance. At the same time, the manufacturing constraints of disabled sectors, single-sector quality upper and lower limits, plane total quality upper limit, quality step size and differential amplitude limiting of adjacent sectors are preserved.

[0049] After completing parameter optimization, the reverse mapping and manufacturability processing flow is re-executed, and the target counterweight command is reconstructed based on the updated complex-valued reversible flow model state.

[0050] A motor rotor dynamic balancing test system according to an embodiment of the present invention includes the following modules:

[0051] The sensor acquisition module is used to arrange sensors and encoders to acquire vibration signals, encoder angular position signals, and current signals.

[0052] The phase-locked loop module is used to establish a multi-resonance phase-locked loop, perform parallel phase locking, and output the first-order, second-order, and third-order phase sequences and angular position sequences for jitter removal, forming a phase reference.

[0053] The feature extraction module is used to perform angular domain non-equidistant resampling and sparse harmonic decomposition, extract first-order and second-order complex amplitude values, and generate multi-point observation data and operating condition data.

[0054] The counterweight calculation module is used to perform reverse mapping and constraint processing based on the complex-valued reversible flow model to generate the target counterweight command.

[0055] The virtual correction module is used to convert the target counterweight command into electromagnetic torque injection, complete the virtual correction, and optimize the parameters of the complex reversible flow model when the threshold is not reached.

[0056] The re-inspection and archiving module is used to implement physical corrections based on the target counterweight instructions, re-inspect vibration indicators, and archive the results after the limits are met.

[0057] The beneficial effects of this invention are:

[0058] The proposed method for dynamic balancing of motor rotors overcomes the limitations of traditional methods that rely on test weights and single steady-state conditions, enabling stable testing under various operating conditions, including constant speed and variable speed frequency sweep. By arranging a triaxial accelerometer in the motor bearing housing in conjunction with a speed encoder, vibration, angle, and current signals are collected. A phase reference for de-jittering is established using a multi-resonant phase-locked loop, thus ensuring the continuity and accuracy of data acquisition. This mechanism effectively avoids phase drift caused by speed fluctuations or noise interference, providing a reliable foundation for vibration feature extraction.

[0059] In the feature extraction and solution stage, this invention obtains high-confidence first- and second-order complex amplitude indices through angular domain non-equidistant resampling and sparse harmonic decomposition, and constructs a complex-valued reversible flow model for reverse mapping. Combined with the synergistic effect of the loop graph equivariant coupling layer, conditional modulation unit, and cross-plane bridging unit, high-precision estimation of the imbalance quantity is achieved. The physical constraint layer directly introduces manufacturing and assembly conditions into the solution process, ensuring that the generated target counterweight command not only satisfies the optimality of the mathematical solution but also possesses feasibility and engineering feasibility, avoiding the shortcomings of traditional methods where the solution results do not match actual operations.

[0060] In the correction and verification phase, this invention introduces a virtual correction pre-simulation mechanism, mapping the target counterweight command to angular synchronous electromagnetic torque. Dynamic simulation is achieved through online injection via a driver, real-time calculation of the vibration reduction rate, and verification of second-order component constraints. If the correction effect is insufficient, the system automatically iteratively optimizes the model parameters and dynamically reconstructs the counterweight command. Finally, vibration indicators are re-checked under target and extreme operating conditions. After passing the re-check, the test data is archived into the model family parameter library, forming reusable data assets. This invention not only achieves dynamic balancing testing without trial weights, with high precision and multiple operating conditions, but also improves the safety and reliability of the correction, providing significant advantages for the application of motor rotors in mass production and complex operating environments. Attached Figure Description

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0062] Figure 1This is a flowchart of a motor rotor dynamic balancing test method proposed in this invention;

[0063] Figure 2 This is a schematic diagram of the structure of a motor rotor dynamic balancing test system proposed in this invention. Detailed Implementation

[0064] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0065] refer to Figure 1 A method for dynamic balancing of an electric motor rotor, comprising:

[0066] A triaxial accelerometer and a speed encoder are arranged in the motor bearing housing. The upper limit of the amplitude and frequency band of the current injection on the drive side are set. The motor is run under constant speed and variable speed sweep frequency band conditions to collect vibration signals, encoder angular position signals and current signals.

[0067] A multi-resonance phase-locked loop is established based on the encoder angular position signal. The first, second and third mechanical frequencies are phase-locked in parallel, and the first, second and third phase sequences and angular position sequences are output to form a phase reference.

[0068] Angular domain non-equidistant resampling and sparse harmonic decomposition are performed on vibration and current signals to extract first-order and second-order complex amplitude values ​​as vibration indices, generate multi-point observation data, and form working condition data.

[0069] A complex-valued reversible flow model is constructed, which includes a loop graph equal-variable coupling layer, a complex-valued condition modulation unit, a cross-plane bridging unit, and a physical constraint layer. Multi-point observation data and working condition data are used to perform reverse mapping, and manufacturability projection is executed to obtain the target counterweight command.

[0070] The target counterweight command is mapped to an angular synchronous electromagnetic torque synchronized with the first-order phase. Virtual correction and pre-simulation are performed by injecting the torque online through the driver. The first-order vibration descent rate is calculated in real time based on the phase reference, and the second-order component constraint is checked. When the preset threshold is not reached, the parameters of the complex condition modulation unit and the cross-plane bridging unit are optimized to complete the dynamic reconstruction of the target counterweight command.

[0071] According to the dynamically reconstructed target counterweight instruction, the deweighting operation is carried out in the corresponding plane and sector. The vibration index is re-inspected under the target working condition and extreme working condition. After the acceptance limit is met, the working condition label, target counterweight instruction and re-inspection results are archived and written into the model family parameter library.

[0072] In this embodiment, the setting of the upper limit of the amplitude and the frequency band range of the drive-side current injection means that the amplitude of the injected current does not exceed three percent of the rated current of the motor, and the injection frequency band is located within the frequency range of the first-order mechanical rotation frequency and the frequency range of the first-order mechanical rotation frequency within ten percent above and below it.

[0073] In this embodiment, operating the motor under constant speed and variable speed sweeping frequency conditions means maintaining the motor rotor at a preset constant speed under constant speed conditions, and gradually accelerating the motor rotor from low speed to rated speed according to a preset speed change rate under variable speed sweeping frequency conditions. Operating data covering different speed ranges are obtained through constant speed and variable speed sweeping frequency conditions.

[0074] In this embodiment, the output of the first-order, second-order, and third-order phase sequences and angular position sequences for jitter removal forms a phase reference, including:

[0075] Based on the encoder angular position signal, the angle unfolding and pulse missing detection are performed to generate continuous angular trajectory and mechanical angular velocity information, establish three parallel tracking channels of first order, second order and third order, and complete the initial phase alignment;

[0076] Resonant phase-locked loops (PLLs) are run in three parallel tracking channels (first-order, second-order, and third-order). The loop bandwidth and damping parameters are adjusted according to the mechanical angular velocity and angular acceleration. Harmonic co-constraints are introduced to maintain harmonic frequency consistency between the second-order and first-order phases, and between the third-order and first-order phases, and to maintain third-harmonic frequency consistency. An auxiliary phase reference is used for both the vibration and current signals. Specifically, the operation of the resonant PLLs in the three parallel tracking channels involves:

[0077] A resonant phase-locked loop is operated in the first-order tracking channel, and the loop bandwidth and damping parameters are dynamically adjusted according to the real-time mechanical angular velocity and angular acceleration.

[0078] A resonant phase-locked loop is run in the second-order tracking channel. By introducing a cooperative constraint with the first-order channel, it is ensured that the second-order phase always maintains a frequency relationship of twice that of the first-order phase during the locking process, thus suppressing the phase shift caused by speed fluctuations.

[0079] A resonant phase-locked loop is run in the third-order tracking channel. Through the cooperative constraint with the first-order channel, it is ensured that the third-order phase maintains a frequency relationship of three times that of the first-order phase during the tracking process, and the constraint mechanism is used to reduce high-order harmonic interference.

[0080] Vibration and current signals were introduced as auxiliary phase references during the operation of the three channels, and cross-verification was performed by comparing the encoder angular position signal.

[0081] The outputs of the three parallel tracking channels (first-order, second-order, and third-order) are subjected to jitter reduction processing to form jitter-reduced first-order phase sequence, jitter-reduced second-order phase sequence, jitter-reduced third-order phase sequence, and jitter-reduced angular position sequence. At the same time, an equal-angle sampling sequence for angular domain resampling is generated. All sequences are merged and defined as phase reference.

[0082] Calculate the phase quality score and lock status marker, complete the zero-point reference calibration and phase offset correction, and record the adaptive parameters, cooperative constraint deviation and jitter suppression threshold of the three parallel tracking channels of the first, second and third order.

[0083] This invention achieves adaptive adjustment of the phase tracking process by running resonant phase-locked loops in three parallel tracking channels (first-order, second-order, and third-order) and dynamically adjusting the loop bandwidth and damping parameters based on mechanical angular velocity and angular acceleration. It introduces harmonic coordination constraints to ensure that the second-order phase maintains harmonic consistency with the first-order phase and that the third-order phase maintains third harmonic consistency with the first-order phase, solving the problem of phase drift and high-order harmonic distortion that are difficult to suppress under variable speed conditions in traditional methods. Using vibration and current signals as auxiliary references and cross-validating with encoder angular position signals improves phase-locked loop accuracy and robustness. Through jitter removal, zero-position reference calibration, and phase offset correction, a high-quality phase reference is output, and reliable input is provided for angular domain resampling and counterweight calculation, thus achieving high precision and stability in motor rotor dynamic balancing testing under multiple operating conditions.

[0084] In this embodiment, the calculation of phase quality score and lock status mark, and the completion of zero-point reference calibration and phase offset correction, include:

[0085] The scoring indicators are phase continuity, instantaneous jitter root mean square error, octave consistency error, harmonic signal-to-noise ratio, and loop residual, which are normalized to a score of 0 to 100 according to preset weights.

[0086] Phase continuity refers to the smooth change in phase difference between adjacent sampling points in first-order, second-order, and third-order phase sequences, without abrupt changes or jumps. The phase continuity value is derived from the ratio of the number of consecutive samples in the phase sequence to the total number of samples.

[0087] Instantaneous jitter mean square error refers to the average value obtained by applying a sliding window processing method to the instantaneous phase sequence output by the phase-locked loop, calculating the squared difference between the phase and the local mean within each window, and then taking the mean of the result for the entire sequence.

[0088] The frequency doubling consistency error refers to the deviation of the ideal two-fold relationship between the second-order phase and the first-order phase, and the deviation of the ideal three-fold relationship between the third-order phase and the first-order phase. The difference calculated is used as the frequency doubling consistency error.

[0089] The harmonic signal-to-noise ratio (SNR) is the ratio of the energy of the first, second, and third harmonic components to the energy of adjacent non-harmonic frequency bands, obtained through frequency domain analysis. The larger the value, the more prominent the harmonic components and the less interference.

[0090] Loop residual refers to the average absolute value of the difference sequence between the phase-locked loop output phase and the encoder angular position signal reference phase, which is used to measure the consistency between the phase-locked result and the reference input;

[0091] When the phase quality score is greater than or equal to 90 points and the loop residual within the continuous window is lower than the preset threshold, it is marked as locked.

[0092] When the phase quality score is between 60 and 90 or the loop residual is in the neighborhood of the threshold, it is marked as tracking.

[0093] A phase quality score below 60 or a missing pulse is marked as lost.

[0094] Zero-position reference calibration is completed using the encoder zero-position mark, and bias correction is performed on the first-order phase, second-order phase, and third-order phase respectively, so that the first-order phase is zero at the zero position, the second-order phase is twice the first-order phase at the zero position, and the third-order phase is three times the first-order phase at the zero position.

[0095] This invention constructs a comprehensive scoring system based on phase continuity, instantaneous jitter root mean square error, frequency doubling consistency error, harmonic signal-to-noise ratio, and loop residuals. The results are normalized to a score of 0 to 100 according to preset weights, enabling a quantitative evaluation of phase tracking quality. This system not only accurately reflects the smoothness and stability of the phase but also comprehensively characterizes the phase-locked loop (PLL) effect from multiple perspectives, including frequency doubling constraints, frequency domain purity, and reference consistency. By setting grading criteria of 90 and 60 points, combined with loop residuals and missing pulse detection, adaptive labeling of the three states—locked, tracking, and lost—is achieved, avoiding the misjudgment problems caused by relying on a single threshold in traditional methods. Using zero-position reference calibration and offset correction, a strict frequency doubling correspondence between the first, second, and third-order phases at zero position is ensured, fundamentally improving the reliability and repeatability of the phase reference. This enhances the robustness and accuracy of dynamic balance testing under variable speed and complex operating conditions, demonstrating outstanding innovation and engineering application value.

[0096] In this embodiment, the step of performing angular domain non-equidistant resampling and sparse harmonic decomposition on vibration and current signals, extracting first-order and second-order complex amplitude values ​​as vibration indicators, generating multi-point observation data, and forming operating condition data includes:

[0097] Call the angular position sequence and phase reference to construct an equal angle sampling sequence and set the number of sampling points per revolution to two thousand points, and set the acquisition window to cover the constant speed segment and the variable speed sweep frequency segment;

[0098] For each vibration measurement point, the vibration signal is resampled in the angular domain at non-uniform intervals to form an angular domain vibration sequence. The current signal is resampled in the angular domain and demodulated with the fundamental component to form an angular domain current reference sequence, wherein:

[0099] Performing non-equidistant resampling in the angular domain for the vibration signal at each vibration measurement point means using the encoder angular position signal as the sampling reference, redistributing the vibration signal that was originally uniformly collected in the time domain according to the angular position, mapping the sampling points at different speeds to a unified angular coordinate system, so that the vibration signal forms a continuous sequence in the angular domain, and can maintain the comparability of samples in each period under the condition of speed fluctuation.

[0100] Angular domain resampling and fundamental component demodulation of current signals refer to resampling the current signal according to the angular position to obtain the current sequence corresponding to the mechanical angle, and eliminating high-order harmonics and noise interference by synchronously extracting the amplitude and phase of the fundamental component of the current.

[0101] Phase self-correcting sparse decomposition is performed on the angular domain vibration sequence. The phase shift of each revolution is corrected based on the phase reference. The first harmonic component and the second harmonic component are extracted using the angular domain atom set that conforms to the sector ring topology, and the amplitude and phase are generated.

[0102] Dual-source suppression separation is performed on the angular domain vibration sequence and the angular domain current reference sequence. The angular domain current reference sequence is used as the suppression reference to eliminate the influence of electromagnetic torque ripple and switching harmonics on the first and second harmonic estimations, and the scrambled amplitude and scrambled phase are obtained.

[0103] Cross-speed consistency constraints and outlier segment removal are applied to the results of the constant speed segment and the variable speed frequency sweep segment. Confidence scores for the first and second harmonics are generated based on phase continuity, instantaneous jitter, harmonic consistency, and harmonic signal-to-noise ratio, and valid angle segments are marked. Among these:

[0104] Cross-speed consistency constraint refers to comparing the first and second harmonic amplitude and phase results obtained in constant speed range and variable speed sweep frequency range according to the same speed point. It requires that the results under different operating conditions remain consistent within the range of amplitude and phase deviation, and corrects or reduces the confidence level of results that exceed the deviation range.

[0105] Abnormal segment removal refers to detecting abrupt changes, missing pulses, strong noise interference, or phase jump segments in vibration and current signals in angle or time sequences, marking them as abnormal, and removing them from the first and second harmonic results.

[0106] The de-scrambling amplitude and phase of the first and second harmonics at all measuring points are summarized to generate multi-measuring-point observation data. Simultaneously, operating condition data is generated, including:

[0107] Motor speed information, calculated from the encoder angular position signal, includes instantaneous speed and acceleration;

[0108] Operating condition type identifier, used to distinguish between constant speed section, acceleration section, deceleration section and extreme operating condition;

[0109] The driving-side current amplitude and frequency band setting parameters are used to characterize the current injection conditions;

[0110] Environmental and boundary condition parameters, including sampling timestamps, temperature records, and load status;

[0111] Phase reference information, including first-order, second-order, and third-order phase sequences and angular position sequences after jitter removal.

[0112] In the data processing stage, this invention employs angular domain non-equidistant resampling and current fundamental component demodulation to uniformly map the time-domain signal to the angular domain, ensuring comparability between cycles even under speed fluctuations and complex operating conditions. Phase self-correcting sparse decomposition is introduced into the angular domain vibration sequence, and first and second harmonic components are extracted using an atom set conforming to the sector ring topology, achieving high-precision feature decomposition of non-stationary signals. Dual-source interference suppression separation is performed using the angular domain current reference sequence, effectively eliminating interference from electromagnetic torque ripple and switching harmonics on frequency components, improving the purity of amplitude and phase estimation. Furthermore, the combination of cross-speed consistency constraints and anomaly segment removal mechanisms ensures the continuity and robustness of results under different operating conditions. By summarizing results from multiple measurement points and generating operating condition data including speed, operating condition type, current injection, environmental boundaries, and phase reference, the accuracy and reliability of dynamic balancing testing are guaranteed.

[0113] In this embodiment, obtaining the target counterweight command includes:

[0114] Receive observation data from multiple measurement points and operating condition data, complete time alignment, phase alignment and sector index alignment, establish a dual-plane sector coordinate system, and set the disabled sector, single sector quality upper and lower limits and the total plane quality upper limit.

[0115] Construct a complex-valued reversible flow model, where:

[0116] The loop graph-equal coupling layer performs sector cyclic alignment and parameter sharing on the biplane sector vector, ensuring that the output is equivalently shifted under any sector cyclic shift, where:

[0117] Sector cyclic alignment refers to the process of cyclically shifting the data of each sector according to the sector order when inputting a dual-plane sector vector, so that the first sector is aligned with the second to the last sector in turn, and then the shifted vector is input into the complex-valued reversible flow model to form a sequence of multiple cyclic alignments.

[0118] Parameter sharing processing refers to assigning the same set of weight parameters to all sectors during the calculation of the loop graph-equal coupling layer, and repeatedly calling the same set of weight parameters in the calculation of each sector, so that the input of each sector goes through the same parameter mapping process in sequence.

[0119] The complex-valued conditional modulation unit generates amplitude modulation coefficients and phase modulation coefficients based on the operating condition data, and performs amplitude scaling and phase shifting on the intermediate representation of the loop diagram's equivariant coupling layer. Specifically, the generation of amplitude modulation coefficients and phase modulation coefficients involves:

[0120] Read the motor speed and acceleration from the operating condition data, divide the speed by the rated speed and the acceleration by the rated acceleration to obtain a value between 0 and 1, and use it as the amplitude modulation coefficient;

[0121] Read the drive-side current amplitude and frequency band parameters from the operating condition data, divide the current amplitude by the rated current, combine the temperature and load status at the time of sampling, calculate the angle offset, and use the angle offset directly as the phase modulation coefficient.

[0122] Align the amplitude modulation coefficients and phase modulation coefficients according to the sampling timestamps, apply the intermediate representation of the loop graph equivariant coupling layer in each calculation cycle, and perform amplitude scaling and phase shifting.

[0123] The cross-plane bridging unit establishes a coupling mapping between plane A and plane B, and transmits and merges information between the two planes according to the sector correspondence and coupling weight;

[0124] The physical constraint layer records and loads manufacturing and assembly parameters within the complex-valued reversible flow model, and provides interfaces for feasibility verification and adjustment.

[0125] Multi-point observation data and operating condition data are input into the complex-valued reversible flow model to generate the initial solution for the counterweight, and the amplitude, phase, and confidence score of each sector are output simultaneously, as follows:

[0126] The amplitude and phase data of the first and second harmonic frequencies of all measuring points are combined with the motor speed information, drive-side current amplitude, and environmental and boundary condition parameters in the operating condition data in sector order to form an input vector;

[0127] Within the loop graph equivariant coupling layer, sector cyclic alignment is performed on the input vector, and the same weight parameters are used for calculation in all sectors;

[0128] The operating condition data is read in the complex condition modulation unit, converted into amplitude modulation coefficients, and the frequency band and environmental state are converted into phase modulation coefficients. The intermediate results of the loop diagram equal-variable coupling layer are subjected to amplitude scaling and phase shifting.

[0129] Within the cross-plane bridging unit, vector superposition and phase alignment are performed on the data of corresponding sectors in both planes to form a cross-plane synthetic weighted representation;

[0130] Within the physical constraint layer, the synthesized weight expression is modified according to the upper and lower limits of quality, sector disabling conditions, and adjacent sector difference constraints, and the amplitude, phase, and confidence score corresponding to each sector are output to obtain the initial weight solution;

[0131] The physical constraint layer is invoked to perform manufacturability processing on the initial solution of the counterweight. The corresponding sector is set to zero according to the prohibited sector. The amplitude is truncated and the counterweight is balanced according to the upper and lower limits of the single sector mass and the upper limit of the total mass of the plane. The amplitude is quantized to the allowable step according to the mass step. Local smoothing and overall correction are performed according to the differential amplitude limit of adjacent sectors and the center of gravity offset limit to obtain the counterweight solution that meets the assembly and manufacturing constraints.

[0132] The solution that satisfies the constraints is converted into a target weight instruction, which includes a plane identifier, sector number, mass value, phase angle, and confidence score.

[0133] This invention constructs and applies a complex-valued reversible flow model, realizing a complete closed loop from multi-point observation data and operating condition data to the target counterweight command. A dual-plane sector coordinate system is established through unified alignment of time, phase, and sector indices, providing consistent input for the solution. An isovariant coupling layer is introduced within the model, achieving efficient modeling of the sector structure through sector cyclic alignment and parameter sharing. The complex-valued conditional modulation unit directly converts speed, current, and environmental data into amplitude and phase modulation coefficients, ensuring dynamic adaptation under different operating conditions. A cross-plane bridging unit establishes information transfer between planes, ensuring cross-plane consistency in the counterweight distribution. A physical constraint layer incorporates upper and lower mass limits, sector disabling, and adjacent difference restrictions into the solution process, ensuring the results meet manufacturing and assembly conditions. Through manufacturability processing and quantization correction, the initial counterweight solution is transformed into a target counterweight command containing plane identification, sector number, mass value, phase angle, and confidence score. It breaks through the limitations of traditional solutions that lack constraints and whose results cannot be directly implemented, and innovatively realizes a manufacturable and executable dynamic balancing counterweight solution, improving the accuracy of dynamic balance correction and engineering usability.

[0134] In this embodiment, the dynamic reconstruction of the target counterweight command includes:

[0135] The target counterweight command is converted to generate an angular synchronous electromagnetic torque injection command that is synchronized with the first-order phase. During the generation process, the injection amplitude and injection frequency band are strictly set according to the limited upper limit of amplitude and frequency band range. Synchronization alignment is completed based on the phase reference, and virtual correction pre-play is completed through online injection by the driver.

[0136] Before and after electromagnetic torque injection, first-order and second-order vibration indices are extracted within the same time and angle windows. The first-order vibration indices before and after injection are compared, and the relative decrease is calculated as the first-order vibration decrease rate. Simultaneously, the change in the second-order vibration indices is recorded. The calculation of the relative decrease specifically involves:

[0137] Before injecting electromagnetic torque, the first-order vibration amplitude is extracted and recorded as the initial value;

[0138] After the electromagnetic torque is injected, the first-order vibration amplitude is extracted and recorded as the correction value;

[0139] Subtract the correction value from the initial value to obtain the difference in vibration amplitude;

[0140] Divide the difference by the initial value to obtain the relative decrease;

[0141] The virtual correction results are checked according to preset judgment conditions, which require that the first-order vibration decrease rate is not lower than the set threshold and the second-order vibration index does not exceed the set tolerance. When the judgment conditions are met, the process directly enters the re-inspection stage; when the judgment conditions are not met, the parameter optimization stage is executed.

[0142] Adaptive iterative optimization is performed on the modulation coefficient parameters of the complex-condition modulation unit and the coupling weight parameters of the cross-plane bridging unit. The optimization steps follow small step size, bounded amplitude and stability constraints. The optimization objective is to reduce the first-order vibration index after injection and avoid the second-order vibration index from exceeding the tolerance. At the same time, the manufacturing constraints of disabled sectors, single-sector quality upper and lower limits, total plane quality upper limit, quality step size and differential amplitude limit of adjacent sectors are preserved. The stability constraint means that the update of the modulation coefficient parameters and coupling weight parameters must meet the convergence condition, the parameter changes must not diverge, and the overall objective function remains monotonically convergent during the optimization process.

[0143] After completing parameter optimization, the reverse mapping and manufacturability processing flow is re-executed, and the target counterweight command is reconstructed based on the updated complex-valued reversible flow model state.

[0144] This invention overcomes the limitation of traditional dynamic balancing tests, which rely solely on one-time physical corrections, by introducing a virtual correction pre-simulation and adaptive iterative optimization mechanism. The target weight command is converted into an angular synchronous electromagnetic torque synchronized with the first-order phase and injected online within strictly defined amplitude upper limits and frequency bands, enabling simulation of the correction effect without actual weight addition. Vibration indices are extracted within the same time and angle windows before and after injection, and the first-order vibration decrease rate and second-order index changes are calculated, providing a quantitative basis for the correction effect. The virtual correction results are verified using preset judgment conditions. If the requirements are not met, adaptive iterative optimization of the complex-valued condition modulation unit and the cross-plane bridging unit is triggered. The command is gradually improved under small step sizes, bounded amplitude, and stability constraints. After inverse mapping and manufacturability processing, a new target weight command is reconstructed. This invention improves the reliability and safety of the correction and ensures the engineering feasibility of the results.

[0145] In this embodiment, the step of performing deweighting operations in the corresponding plane and sector according to the dynamically reconstructed target counterweight command, re-inspecting vibration indicators under target and extreme operating conditions, and archiving and writing the operating condition label, target counterweight command, and re-inspection results into the model family parameter library after meeting the acceptance limits includes:

[0146] Based on the target weight instruction after dynamic reconstruction, determine the corresponding plane number and sector number, and perform deduplication or weighting operations at the specified positions;

[0147] The motor is operated under the target working conditions, and vibration signals, encoder angular position signals and current signals are collected to extract first-order and second-order vibration indices.

[0148] The motor was operated under extreme conditions, and first-order and second-order vibration indices were repeatedly collected and extracted to form re-inspection data under different conditions.

[0149] The first and second order vibration indices under the target and extreme working conditions are compared with the acceptance limits to determine whether the acceptance requirements are met.

[0150] If the acceptance limits are met, the operating condition label, the target counterweight command after dynamic reconstruction, and the vibration index results obtained from the re-inspection will be archived together and written into the model family parameter library.

[0151] refer to Figure 2 A dynamic balancing test system for motor rotors includes the following modules:

[0152] The sensor acquisition module is used to arrange sensors and encoders to acquire vibration signals, encoder angular position signals, and current signals.

[0153] The phase-locked loop module is used to establish a multi-resonance phase-locked loop, perform parallel phase locking, and output the first-order, second-order, and third-order phase sequences and angular position sequences for jitter removal, forming a phase reference.

[0154] The feature extraction module is used to perform angular domain non-equidistant resampling and sparse harmonic decomposition, extract first-order and second-order complex amplitude values, and generate multi-point observation data and operating condition data.

[0155] The counterweight calculation module is used to perform reverse mapping and constraint processing based on the complex-valued reversible flow model to generate the target counterweight command.

[0156] The virtual correction module is used to convert the target counterweight command into electromagnetic torque injection, complete the virtual correction, and optimize the parameters of the complex reversible flow model when the threshold is not reached.

[0157] The re-inspection and archiving module is used to implement physical corrections based on the target counterweight instructions, re-inspect vibration indicators, and archive the results after the limits are met. Example

[0158] To verify the feasibility of this invention in practice, it was applied to the production workshop of a large motor manufacturing company. The R&D team selected a three-phase asynchronous motor with a rated power of 55kW and a rated speed of 3000rpm as the test object. After the motor was assembled and rolled off the production line, neither the traditional trial weight method nor single-condition spectrum analysis could meet the vibration standard in a single correction. In particular, significant vibration fluctuations occurred during the speed change process from 500rpm to 3000rpm, with a peak vibration amplitude reaching 5.2mm / s, exceeding the company's limit of ≤3.5mm / s. Traditional methods suffer from insufficient testing accuracy and unreliable correction under multi-condition environments.

[0159] To verify the proposed motor rotor dynamic balancing test method, a triaxial accelerometer was installed in the motor bearing housing, along with a high-precision speed encoder. The current injection amplitude was set to an upper limit of 3% of the rated current via a driver, covering the first-order rotational frequency and its range within 10% above and below. The motor operated under constant speed and variable speed sweeping conditions, and the system simultaneously acquired vibration, angular position, and current signals. The acquired data showed that at the rated speed of 3000 rpm, the uncorrected first-order vibration amplitude was 4.8 mm / s, and the second-order vibration amplitude was 2.1 mm / s, with the peak value exceeding 6 mm / s during acceleration.

[0160] The system establishes a multi-resonant phase-locked loop based on the encoder angular position signal, realizing parallel phase-locking of first-, second-, and third-order mechanical frequencies, and outputting a jitter-reduced phase sequence and angular position sequence. Using this reference, vibration and current signals are converted into angular domain data, and then, through sparse harmonic decomposition, first- and second-order complex amplitude indices are successfully extracted. Using this as input, a complex-valued reversible current model performs inverse mapping and, combined with constraints, generates a preliminary counterweight solution. Considering that the mass of a single sector does not exceed 10g, the total mass of both planes does not exceed 60g, and the difference between adjacent sectors does not exceed 4g, the model outputs a target counterweight command that explicitly indicates removing 3.2g in sector 3 of plane A and adding 4.5g in sector 7 of plane B.

[0161] To avoid the risks associated with direct correction, the system first mapped the target counterweight command to an electromagnetic torque synchronized with the first-order phase and performed a virtual correction simulation via actuator injection. The results showed that the first-order vibration decreased by 58%, while the second-order vibration remained at its original level without significant increase. Since the preset threshold was met, the system confirmed the command's validity. The operator then implemented the actual correction in the corresponding sector of the rotor according to the command.

[0162] After correction, the motor was retested under target and extreme operating conditions. At 3000 rpm, the first-order vibration amplitude decreased from 4.8 mm / s to 1.5 mm / s, and the second-order vibration decreased from 2.1 mm / s to 1.0 mm / s. During variable speed sweep, the peak vibration amplitude decreased from 6.0 mm / s to 2.2 mm / s, far below the limit. The entire machine operated smoothly across the entire speed range without any abnormal noise or overheating. The final test results were archived by the system and written into the model family parameter library as reference data for the production of motors in the same series.

[0163] Table 1 Comparison data of motor rotor dynamic balancing test before and after

[0164]

[0165] As can be seen from the data in Table 1, the method of the present invention exhibits significant vibration improvement effects under different operating conditions. At a constant speed of 1500 rpm, the first-order vibration amplitude of the motor before correction was 3.5 mm / s, and the second-order vibration amplitude was 1.6 mm / s. After correction, these values ​​decreased to 1.2 mm / s and 0.9 mm / s, respectively, with a first-order vibration reduction rate of 65.7%. The peak vibration was controlled at 2.0 mm / s, far below the enterprise limit, and the motor operated smoothly.

[0166] When the speed is increased to the rated 3000 rpm, the first-order vibration amplitude before correction is 4.8 mm / s and the second-order is 2.1 mm / s, both exceeding the allowable value; after correction, the first-order vibration decreases to 1.5 mm / s and the second-order vibration decreases to 1.0 mm / s, the first-order vibration reduction rate increases to 68.7%, and the peak vibration amplitude decreases to 2.1 mm / s, basically eliminating the unbalanced effect at high speed.

[0167] During acceleration from 500 rpm to 3000 rpm, traditional methods often struggle to guarantee stability. However, the method of this invention reduced the peak vibration amplitude from 6.0 mm / s before correction to 2.2 mm / s after correction, a decrease of 63.3%, effectively suppressing dynamic imbalance during acceleration. Similarly, during deceleration from 3000 rpm to 500 rpm, the peak vibration amplitude was 5.7 mm / s before correction, decreasing to 2.0 mm / s after correction, a decrease of 64.9%, demonstrating good vibration suppression even during speed changes.

[0168] At the maximum speed of 3600 rpm, the first-order vibration amplitude of the motor before correction was 5.2 mm / s, and the second-order was 2.4 mm / s, both of which were severely exceeding the standard. After correction, the first-order vibration was reduced to 1.8 mm / s, and the second-order vibration was reduced to 1.2 mm / s, with a first-order vibration reduction rate of 65.4%. The peak vibration amplitude was reduced to 2.5 mm / s, achieving stable operation under extreme conditions. This invention can significantly reduce vibration levels under constant speed, variable speed, and extreme conditions, not only improving correction accuracy but also ensuring the safety and reliability of the motor across the entire operating range.

[0169] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic balancing of an electric motor rotor, characterized in that, include: A triaxial accelerometer and a speed encoder are arranged in the motor bearing housing. The upper limit of the amplitude and frequency band of the current injection on the drive side are set. The motor is run under constant speed and variable speed sweep frequency band conditions to collect vibration signals, encoder angular position signals and current signals. A multi-resonance phase-locked loop is established based on the encoder angular position signal. The first, second and third mechanical frequencies are phase-locked in parallel, and the first, second and third phase sequences and angular position sequences are output to form a phase reference. Angular domain non-equidistant resampling and sparse harmonic decomposition are performed on vibration and current signals to extract first-order and second-order complex amplitude values ​​as vibration indices, generate multi-point observation data, and form working condition data. A complex-valued reversible flow model is constructed, which includes a loop graph equal-variable coupling layer, a complex-valued condition modulation unit, a cross-plane bridging unit, and a physical constraint layer. Multi-point observation data and working condition data are used to perform reverse mapping, and manufacturability projection is executed to obtain the target counterweight command. The target counterweight command is mapped to an angular synchronous electromagnetic torque synchronized with the first-order phase. Virtual correction and pre-simulation are performed by injecting the torque online through the driver. The first-order vibration descent rate is calculated in real time based on the phase reference, and the second-order component constraint is checked. When the preset threshold is not reached, the parameters of the complex condition modulation unit and the cross-plane bridging unit are optimized to complete the dynamic reconstruction of the target counterweight command. According to the dynamically reconstructed target counterweight instruction, the deweighting operation is carried out in the corresponding plane and sector. The vibration index is re-inspected under the target working condition and extreme working condition. After meeting the acceptance limit, the working condition label, target counterweight instruction and re-inspection results are archived and written into the model family parameter library. The process involves angular domain non-equidistant resampling and sparse harmonic decomposition of vibration and current signals, extracting first- and second-order complex amplitude values ​​as vibration indices, generating multi-point observation data, and forming operating condition data, including: Call the angular position sequence and phase reference to construct an equal angle sampling sequence and set the number of sampling points per revolution to two thousand points, and set the acquisition window to cover the constant speed segment and the variable speed sweep frequency segment; For each vibration measurement point, non-equidistant angular domain resampling is performed on the vibration signal to form an angular domain vibration sequence. For the current signal, angular domain resampling and fundamental component demodulation are performed to form an angular domain current reference sequence. Phase self-correcting sparse decomposition is performed on the angular domain vibration sequence. The phase shift of each revolution is corrected based on the phase reference. The first harmonic component and the second harmonic component are extracted using the angular domain atom set that conforms to the sector ring topology, and the amplitude and phase are generated. Dual-source suppression separation is performed on the angular domain vibration sequence and the angular domain current reference sequence. The angular domain current reference sequence is used as the suppression reference to eliminate the influence of electromagnetic torque ripple and switching harmonics on the first and second harmonic estimations, and the scrambled amplitude and scrambled phase are obtained. The constant speed segment results and the variable speed frequency sweep results are subjected to cross-speed consistency constraints and abnormal segment removal. Based on phase continuity, instantaneous jitter, harmonic consistency and harmonic signal-to-noise ratio, confidence scores for the first harmonic and second harmonic are generated and valid angle segments are marked. The de-scrambling amplitude and phase of the first and second harmonics of all measuring points are summarized to generate multi-measuring-point observation data, and operating condition data are generated at the same time.

2. The method for dynamic balancing of a motor rotor according to claim 1, characterized in that, The upper limit of the amplitude and frequency band of the current injection on the drive side are defined as follows: the amplitude of the injected current does not exceed three percent of the rated current of the motor, and the injection frequency band is located within the frequency range of the first-order mechanical rotation frequency and the frequency range within ten percent above and below the first-order mechanical rotation frequency.

3. The method for dynamic balancing of a motor rotor according to claim 1, characterized in that, The operation of the motor under constant speed and variable speed sweeping conditions refers to maintaining the motor rotor at a preset constant speed under constant speed conditions, and gradually accelerating the motor rotor from low speed to rated speed according to a preset speed change rate under variable speed sweeping conditions. Operating data covering different speed ranges are obtained through constant speed and variable speed sweeping conditions.

4. The method for dynamic balancing of a motor rotor according to claim 1, characterized in that, The first-, second-, and third-order phase sequences and angular position sequences of the output dejittering form a phase reference, including: Based on the encoder angular position signal, the angle unfolding and pulse missing detection are performed to generate continuous angular trajectory and mechanical angular velocity information, establish three parallel tracking channels of first order, second order and third order, and complete the initial phase alignment; Resonant phase-locked loops are run in three parallel tracking channels of first, second and third order respectively. The loop bandwidth and damping parameters are adjusted according to the mechanical angular velocity and angular acceleration. Harmonic cooperative constraints are introduced to keep the second-order phase consistent with the first-order phase and the third-order phase consistent with the first-order phase. Auxiliary phase references for vibration signal and current signal are enabled. The outputs of the three parallel tracking channels (first-order, second-order, and third-order) are subjected to jitter reduction processing to form jitter-reduced first-order phase sequence, jitter-reduced second-order phase sequence, jitter-reduced third-order phase sequence, and jitter-reduced angular position sequence. At the same time, an equal-angle sampling sequence for angular domain resampling is generated. All sequences are merged and defined as phase reference. Calculate the phase quality score and lock status marker, complete the zero-point reference calibration and phase offset correction, and record the adaptive parameters, cooperative constraint deviation and jitter suppression threshold of the three parallel tracking channels of the first, second and third order.

5. The method for dynamic balancing of a motor rotor according to claim 4, characterized in that, The calculation of phase quality score and lock status mark, and the completion of zero-point reference calibration and phase offset correction, include: The scoring indicators are phase continuity, instantaneous jitter mean square error, frequency doubling consistency error, harmonic signal-to-noise ratio and loop residual, which are normalized to 0 to 100 points according to preset weights. When the phase quality score is greater than or equal to 90 points and the loop residual within the continuous window is lower than the preset threshold, it is marked as locked. When the phase quality score is between 60 and 90 or the loop residual is in the neighborhood of the threshold, it is marked as tracking. A phase quality score below 60 or a missing pulse is marked as lost. Zero-position reference calibration is completed using the encoder zero-position mark, and bias correction is performed on the first-order phase, second-order phase, and third-order phase respectively, so that the first-order phase is zero at the zero position, the second-order phase is twice the first-order phase at the zero position, and the third-order phase is three times the first-order phase at the zero position.

6. The method for dynamic balancing of a motor rotor according to claim 1, characterized in that, The process of obtaining the target counterweight instruction includes: Receive observation data from multiple measurement points and operating condition data, complete time alignment, phase alignment and sector index alignment, establish a dual-plane sector coordinate system, and set the disabled sector, single sector quality upper and lower limits and the total plane quality upper limit. Construct a complex-valued reversible flow model, where: The loop graph-equal coupling layer performs sector cyclic alignment and parameter sharing on the dual-plane sector vector, ensuring that the output is equivalently shifted under any sector cyclic shift. The complex-valued conditional modulation unit generates amplitude modulation coefficients and phase modulation coefficients based on the operating condition data, and performs amplitude scaling and phase shifting on the intermediate representation of the loop diagram equal-variable coupling layer. The cross-plane bridging unit establishes a coupling mapping between plane A and plane B, and transmits and merges information between the two planes according to the sector correspondence and coupling weight; The physical constraint layer records and loads manufacturing and assembly parameters within the complex-valued reversible flow model, and provides interfaces for feasibility verification and adjustment. Input the multi-point observation data and operating condition data into the complex value reversible flow model, execute the reverse solution process to generate the initial solution of the counterweight, and simultaneously output the amplitude, phase and confidence score of each sector; The physical constraint layer is invoked to perform manufacturability processing on the initial solution of the counterweight. The corresponding sector is set to zero according to the prohibited sector. The amplitude is truncated and the counterweight is balanced according to the upper and lower limits of the single sector mass and the upper limit of the total mass of the plane. The amplitude is quantized to the allowable step according to the mass step. Local smoothing and overall correction are performed according to the differential amplitude limit of adjacent sectors and the center of gravity offset limit to obtain the counterweight solution that meets the assembly and manufacturing constraints. The solution that satisfies the constraints is converted into a target weight instruction, which includes a plane identifier, sector number, mass value, phase angle, and confidence score.

7. The method for dynamic balancing of a motor rotor according to claim 1, characterized in that, The dynamic reconstruction of the target counterweight command includes: The target counterweight command is converted to generate an angular synchronous electromagnetic torque injection command that is synchronized with the first-order phase. During the generation process, the injection amplitude and injection frequency band are strictly set according to the limited upper limit of amplitude and frequency band range. Synchronization alignment is completed based on the phase reference, and virtual correction pre-play is completed through online injection by the driver. Before and after electromagnetic torque injection, first-order and second-order vibration indices are extracted under the same time and angle windows. The first-order vibration indices before and after injection are compared, the relative decrease is calculated and used as the first-order vibration decrease rate, and the change in the second-order vibration index is recorded. The virtual correction results are checked according to preset judgment conditions, which require that the first-order vibration decrease rate is not lower than the set threshold and the second-order vibration index does not exceed the set tolerance. When the judgment conditions are met, the process directly enters the re-inspection stage; when the judgment conditions are not met, the parameter optimization stage is executed. Adaptive iterative optimization is performed on the modulation coefficient parameters of the complex-condition modulation unit and the coupling weight parameters of the cross-plane bridging unit. The optimization steps follow small step size, bounded amplitude and stability constraints. The optimization objective is to reduce the first-order vibration index after injection and avoid the second-order vibration index from exceeding the tolerance. At the same time, the manufacturing constraints of disabled sectors, single-sector quality upper and lower limits, plane total quality upper limit, quality step size and differential amplitude limiting of adjacent sectors are preserved. After completing parameter optimization, the reverse mapping and manufacturability processing flow is re-executed, and the target counterweight command is reconstructed based on the updated complex-valued reversible flow model state.

8. A motor rotor dynamic balancing test system, comprising performing a motor rotor dynamic balancing test method according to any one of claims 1 to 7, characterized in that, Includes the following modules: The sensor acquisition module is used to arrange sensors and encoders to acquire vibration signals, encoder angular position signals, and current signals. The phase-locked loop module is used to establish a multi-resonance phase-locked loop, perform parallel phase locking, and output the first-order, second-order, and third-order phase sequences and angular position sequences for jitter removal, forming a phase reference. The feature extraction module is used to perform angular domain non-equidistant resampling and sparse harmonic decomposition, extract first-order and second-order complex amplitude values, and generate multi-point observation data and operating condition data. The counterweight calculation module is used to perform reverse mapping and constraint processing based on the complex-valued reversible flow model to generate the target counterweight command. The virtual correction module is used to convert the target counterweight command into electromagnetic torque injection, complete the virtual correction, and optimize the parameters of the complex reversible flow model when the threshold is not reached. The re-inspection and archiving module is used to implement physical corrections based on the target counterweight instructions, re-inspect vibration indicators, and archive the results after the limits are met.

Images