Intelligent control method and system of motor driver

Abstract

The application discloses an intelligent control method and system of a motor driver, which comprises the following steps: collecting harmonic current and sound pressure fluctuation data in the motor operation process, constructing a sound-electricity correlation matrix to form a fault feature library; identifying available fault modes to determine energy collection points, and performing phase offset processing on the collection points to generate auxiliary driving signals; performing frequency spectrum analysis on the sound pressure fluctuation data to obtain acoustic characteristic spectrum, extracting efficiency indicating sound through characteristic decomposition and mapping into torque optimization parameters, and establishing a noise reduction control table; performing frequency spectrum analysis on an enhanced control domain to identify high-energy frequency bands, performing frequency shift operation to generate safe frequency band signals, and combining the noise reduction control table to form a multi-frequency band management strategy; decomposing the management strategy into parallel control flows for time sequence interleaving arrangement to generate an energy-saving control matrix; generating a PWM control signal based on the matrix and performing edge reconstruction, obtaining a softened edge waveform through anti-resonance injection technology, and completing intelligent control of the motor driver.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of motor drive control, in particular to an intelligent control method and system of motor driver. BACKGROUND

[0002] Modern industrial motor systems generally have problems such as low efficiency and serious electromagnetic interference during operation. Traditional motor control technology mainly uses fixed parameter control strategy, which lacks real-time perception and adaptive adjustment capability for running state. Existing fault detection methods mainly rely on single electrical signal such as current and voltage for analysis, which is difficult to fully reflect the real running state of the motor, and has limited ability to identify early faults.

[0003] In addition, the switching signal edge generated by the traditional PWM modulation technology is steep, which not only increases the power consumption of the switching device, but also causes serious electromagnetic radiation problem. The existing energy-saving control strategy is mainly realized by adjusting the voltage frequency ratio, which lacks deep analysis and optimal utilization of the internal energy flow of the system. At the same time, the traditional control method ignores the influence of mechanical vibration, noise and other non-electrical information on the system performance, and cannot realize the fusion optimization of multi-physical field information. Therefore, it is urgent to develop a motor drive technology that can comprehensively utilize multi-dimensional running information and realize intelligent perception and adaptive control. SUMMARY

[0004] The present application provides an intelligent control method and system of motor driver, which aims to deeply mine and intelligently convert multi-source abnormal signals such as harmonic current and sound pressure fluctuation in motor operation, establish a comprehensive fault feature library, identify available fault modes and realize energy recovery, extract efficiency indicator sound based on acoustic feature analysis, construct noise reduction control strategy through spectrum analysis and frequency shift technology, and finally generate PWM control signal with soft edge characteristics, realizing efficient, low-noise and intelligent control of motor driver.

[0005] The first aspect of the present application provides an intelligent control method of motor driver, comprising the following steps:

[0006] Collecting multi-source abnormal signals in the motor operation process, the multi-source abnormal signals including harmonic current and sound pressure fluctuation data, and converting the multi-source abnormal signals to form a fault feature library;

[0007] Identifying available fault modes based on the fault feature library, evaluating energy collection points for the available fault modes to determine energy collection points, processing the energy collection points for harmonic phase shift to generate auxiliary drive signals, and extracting amplitude-frequency characteristics of the auxiliary drive signals to determine enhanced control domain;

[0008] The sound pressure fluctuation data is subjected to spectrum analysis to generate an acoustic feature spectrum, the acoustic feature spectrum is subjected to feature decomposition to obtain an efficiency indicating tone, the efficiency indicating tone is mapped to a torque optimization parameter, and an electromagnetic excitation frequency is adjusted based on the torque optimization parameter to establish a noise reduction control table.

[0009] The enhanced control domain is subjected to spectrum analysis to identify a high-energy frequency band, a frequency shift operation is performed on the high-energy frequency band to generate a safety frequency band signal, the safety frequency band signal is subjected to energy aggregation processing to form a frequency shift control sequence, and a multi-frequency band management strategy is constructed based on fusion of the frequency shift control sequence and the noise reduction control table.

[0010] The multi-frequency band management strategy is decomposed into a parallel control flow, the parallel control flow is subjected to time sequence interleaving to generate an interleaved control sequence, and an energy-saving control matrix is generated based on the interleaved control sequence.

[0011] A PWM control signal is generated based on the energy-saving control matrix, a soft edge waveform is obtained by performing edge reconstruction on the PWM control signal, and the soft edge waveform is output to a power device to complete intelligent control of the motor driver.

[0012] The second aspect of the application provides an intelligent control system of a motor driver, comprising:

[0013] A signal acquisition module is configured to acquire multi-source abnormal signals in the motor operation process, the multi-source abnormal signals including harmonic currents and sound pressure fluctuation data, and to form a fault feature library by performing feature conversion on the multi-source abnormal signals.

[0014] An energy recovery module is configured to identify available fault modes based on the fault feature library, to determine energy collection points by performing energy recovery evaluation on the available fault modes, to generate an auxiliary driving signal by performing harmonic phase shift processing on the energy collection points, and to determine an enhanced control domain by performing amplitude-frequency characteristic extraction on the auxiliary driving signal.

[0015] An acoustic optimization module is configured to perform spectrum analysis on the sound pressure fluctuation data to generate an acoustic feature spectrum, to obtain an efficiency indicating tone by performing feature decomposition on the acoustic feature spectrum, and to map the efficiency indicating tone to a torque optimization parameter, and to adjust an electromagnetic excitation frequency based on the torque optimization parameter to establish a noise reduction control table.

[0016] A frequency management module is configured to perform spectrum analysis on the enhanced control domain to identify a high-energy frequency band, to perform a frequency shift operation on the high-energy frequency band to generate a safety frequency band signal, to perform energy aggregation processing on the safety frequency band signal to form a frequency shift control sequence, and to construct a multi-frequency band management strategy based on fusion of the frequency shift control sequence and the noise reduction control table.

[0017] A timing control module is configured to decompose the multi-band management strategy into parallel control flows, perform timing stagger arrangement on the parallel control flows to generate staggered control sequences, and generate an energy-saving control matrix based on the staggered control sequences.

[0018] An output shaping module is configured to generate a PWM control signal based on the energy-saving control matrix, perform edge reconstruction on the PWM control signal to obtain a softened edge waveform, and output the softened edge waveform to a power device to complete intelligent control of the motor driver.

[0019] The beneficial effects of the present application are embodied in the following aspects: first, through multi-source abnormal signal fusion analysis technology, the harmonic current and the sound pressure fluctuation signal are constructed into an acoustic-electric correlation matrix, the precise identification of the fault mode is realized, and the identified fault energy is converted into an auxiliary driving signal, which enhances the driving capacity of the system and improves the accuracy of fault diagnosis. Second, by extracting an efficiency indicator sound from the acoustic characteristic spectrum, a direct correlation between the acoustic signal and the motor efficiency state is established. The efficiency indicator sound can reflect the energy conversion efficiency of the motor in real time. By mapping it to the torque optimization parameter and adjusting the electromagnetic excitation frequency, the motor efficiency optimization control based on acoustic feedback is realized, and the motor operating efficiency is improved. Finally, the frequency shift technology is used to shift the high-energy interference frequency band to the safe area, the parallel control flow of timing stagger is combined to reduce the switching loss concentration, and the PWM signal is edge-softened by the anti-resonance injection technology, which effectively reduces the electromagnetic interference level and switching power consumption, and improves the electromagnetic compatibility and energy consumption control effect of the system.

[0020] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0021] The drawings herein show specific examples of the technical solutions described in the present application, and constitute part of the specification together with the specific embodiments, for explaining the technical solutions, principles and effects of the present application.

[0022] Unless specifically stated or otherwise, the same reference signs in different drawings represent the same or similar technical features, and different reference signs may also be used to represent the same or similar technical features.

[0023] Figure 1 is a flow diagram of an intelligent control method of a motor driver according to the present application.

[0024] Figure 2 is a structural block diagram of an intelligent control system of a motor driver according to the present application. DETAILED DESCRIPTION

[0025] In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular architectures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0026] It is to be understood that the terminology "including", "comprising", "consisting" and "consisting essentially of" used in the specification and the appended claims, indicates the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0027] Reference throughout this specification to "one embodiment", "an embodiment", or "a specific embodiment", means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment", "in an embodiment", "in some embodiments", "in other embodiments", "in additional embodiments", and so on, in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specifically so stated. The terms "including", "comprising", "having" and variations thereof, mean "including but not limited to", unless expressly specified otherwise.

[0028] The technical solutions of the embodiments of the present application are described below.

[0029] As shown in Figure 1 The intelligent control method of the motor driver provided by the embodiments of the present application comprises the following steps S110-S160:

[0030] In step S110, multi-source abnormal signals in the motor running process are collected, the multi-source abnormal signals comprising harmonic current and sound pressure fluctuation data, and the multi-source abnormal signals are subjected to feature conversion to form a fault feature library.

[0031] Specifically, multi-source abnormal signals during motor operation are collected. High-precision current sensors are installed on each phase line of the motor stator winding to monitor the current waveform changes in real time during motor operation. The harmonic current is collected by using a Rogowski coil structure current sensor, with a frequency response range covering DC to 20 kHz, which can accurately capture the fundamental and high-order harmonic components. The sensor continuously records three-phase current signals at a sampling frequency of 50 kHz, with a data accuracy of 16 bits. At the same time, an acoustic sensor array is arranged on the surface of the motor shell to collect the sound pressure fluctuation signals generated by the motor during operation. The acoustic sensor uses a MEMS microphone with a frequency response range of 20 Hz-20kHz and a dynamic range of more than 100 dB. The sensor array includes 8 microphones arranged in a regular octagon around the motor, maintaining a fixed distance from the motor surface. The sound pressure fluctuation data is collected at a frequency of 48 kHz, synchronized with the current signal and consistent with the time stamp. The acquisition system is equipped with a high-speed data buffer, supporting continuous recording for 48 hours without interruption. The multi-source abnormal signal data contains four-dimensional information including time label, sensor position information, signal amplitude, and frequency characteristics. Anti-aliasing filtering and noise suppression processing are implemented during the collection process to ensure signal quality. The preprocessed raw data is stored in a unified format.

[0032] In some embodiments, the feature conversion of the multi-source abnormal signal forms a fault feature library, including: constructing a current distortion map based on the harmonic current; identifying the distortion growth rate by calculating the inter-period difference of the current distortion map; constructing an acoustic-electric correlation matrix based on the distortion growth rate and the sound pressure fluctuation data; and establishing a fault feature library based on the acoustic-electric correlation matrix.

[0033] A current distortion map is constructed based on the harmonic current. The three-phase current signals collected are subjected to fast Fourier transform to extract the amplitude and phase information of each harmonic component. The harmonic analysis window length is set to an integer multiple of the fundamental period to ensure the accuracy of the frequency spectrum resolution. The total harmonic distortion rate THD = √(∑I_h²) / I_1 is calculated, where I_h is the effective value of the hth harmonic, and I_1 is the effective value of the fundamental. The THD values of each phase current are arranged in time sequence to form a distortion time curve. The time curve is mapped in two dimensions, with the horizontal axis representing time and the vertical axis representing frequency, and the color depth representing distortion intensity, to construct a current distortion map. The map is subjected to time-frequency analysis using wavelet transform, with Morlet wavelet as the window function and the scale parameter covering the frequency range of 1-100 Hz. High-resolution time-frequency distribution is obtained through continuous wavelet transform, highlighting the transient distortion characteristics. Different color regions in the distortion map correspond to different types of motor fault modes, such as short-circuit of the winding, which is characterized by enhanced high-frequency distortion, and bearing wear, which is characterized by periodic distortion at a specific frequency. The map data is organized in matrix form, with the matrix dimensions being the number of time points x the number of frequency points, and each element recording the distortion intensity of the corresponding time-frequency point.

[0034] The current distortion map is segmented according to the fundamental period of the motor, and each period corresponds to a time segment in the map. The distortion difference between adjacent periods is calculated for the corresponding frequency points: ΔD(f, t) = D(f, t + T) - D(f, t), where D(f, t) is the distortion intensity of frequency f at time t, and T is the fundamental period. Statistical analysis is performed on the difference matrix to calculate the distortion growth rate G(f) = mean(ΔD(f, t)) / std(ΔD(f, t)), where mean represents the mean operation, and std represents the standard deviation operation. A growth rate greater than zero indicates that the distortion of that frequency is increasing, and a growth rate less than zero indicates a decreasing trend. The stable growth trend is extracted by sliding average filtering to eliminate random fluctuations. The growth rates of each frequency are organized into a growth rate vector, and the vector length is equal to the number of frequency points. The growth rate vector is normalized to eliminate the magnitude difference of different frequency components. The frequency intervals with abnormal growth rates are identified, which usually correspond to the development stages of motor internal faults. The distortion growth rate not only reflects the degradation trend of the motor's electrical performance, but also provides important information for time prediction of fault evolution.

[0035] Based on the distortion growth rate and sound pressure fluctuation data, an acoustoelectric correlation matrix is constructed. The sound pressure fluctuation data is decomposed in the frequency domain to extract the acoustic feature frequencies corresponding to the current harmonics. The short-time Fourier transform is used to process the sound pressure signal, and the window length is consistent with the current analysis to ensure the accuracy of the time-frequency correspondence. The gradient of the sound pressure amplitude with respect to time is calculated to identify the mutation points and development trend of the acoustic signal. The distortion growth rate vector and the sound pressure gradient vector are correlated to calculate the Pearson correlation coefficient matrix R, where R(i, j) represents the correlation between the i-th current frequency and the j-th acoustic frequency. The correlation coefficient is calculated using a sliding window method with a window size of 10 fundamental periods and an overlap rate of 50%. The acoustoelectric correlation matrix C is constructed, and the matrix elements C(i, j) = R(i, j) x G(i) x A(j), where G(i) is the distortion growth rate of current frequency i, and A(j) is the amplitude weight of acoustic frequency j. The correlation matrix reveals the coupling relationship between electromagnetic faults and mechanical faults of the motor, and the high correlation area indicates the cooperative evolution mode of multi-source faults. Principal component analysis is used to reduce the dimensionality of the correlation matrix, and the main acoustoelectric correlation patterns are extracted, retaining more than 95% of the variance contribution rate.

[0036] A fault feature library is established based on the electro-acoustic correlation matrix. The fault types include five categories: stator winding fault, rotor bar breakage, bearing defect, air gap eccentricity, and load imbalance. The correlation matrix pattern of each fault type is extracted, including the main diagonal line distribution, peak position, and energy concentration area of the matrix. A clustering algorithm is used to unsupervisedly group the correlation matrix of historical fault samples to identify typical fault feature patterns. A feature similarity measurement function is established, and the matrix two-norm is used to calculate the distance between different correlation matrices. A hierarchical feature index structure is constructed to support fast fault pattern matching and retrieval. The fault feature library uses a relational database architecture, and each record contains fault type labels, correlation matrix data, fault severity, occurrence time, and other fields. The feature library supports incremental inclusion, and new fault samples are automatically added to the library after feature extraction, continuously enriching the coverage of fault patterns. Through regular feature library maintenance, redundant features with high redundancy are deleted to maintain the simplicity and efficiency of the feature library.

[0037] In step S120, the available fault patterns are identified based on the fault feature library, the energy recovery evaluation is performed on the available fault patterns to determine the energy collection points, the harmonic phase shift processing is performed on the energy collection points to generate the auxiliary driving signal, and the amplitude-frequency characteristic extraction is performed on the auxiliary driving signal to determine the enhanced control domain.

[0038] Specifically, the available fault patterns are identified based on the fault feature library. A multi-index comprehensive evaluation method is used, and the main evaluation indexes include fault energy density, frequency stability, spatial concentration, and time persistence. The energy density index E_density of each fault pattern is calculated, where E_density is the energy density, P_harm is the sum of the harmonic powers, and A_region is the area of the fault affected region. The frequency stability is measured by the standard deviation of the harmonic frequencies, and the fault pattern with high stability is more suitable for energy collection. The spatial concentration evaluates the concentration degree of the fault energy in the motor space distribution, and high concentration degree means higher energy collection efficiency. The time persistence analyzes the development cycle of the fault pattern, and the fault with long duration provides more opportunities for energy recovery. By setting reasonable threshold values, the fault patterns that meet the conditions are selected, and the threshold values are dynamically adjusted according to the motor type and working conditions. The identified available fault patterns mainly include: periodic impact energy caused by bearing rolling body defects, magnetic field harmonic energy generated by stator slot harmonics, air gap magnetic field non-uniform energy caused by rotor eccentricity, and torque pulsation energy caused by load fluctuations. A detailed energy feature file is established for each available fault pattern, recording key information such as frequency range, power level, occurrence position, and evolution trend.

[0039] In some embodiments, the energy recovery evaluation on the available fault mode determines an energy collection point, including: extracting an energy distribution of the harmonic current from the available fault mode; determining a resonance region based on the energy distribution; performing an energy density evaluation on the resonance region to generate a high-density collection area; and determining an energy collection point based on the high-density collection area.

[0040] The energy distribution of the harmonic current is extracted from the available fault mode. The current signal under the available fault mode is analyzed in time-frequency domain using windowed Fourier transform, an integer multiple of fundamental period is used as window length, and a proper overlap ratio is set to obtain an energy distribution diagram with high time-frequency resolution. The instantaneous power of each harmonic is calculated as P_h(t)=I_h(t)×V_h(t)×cos(φ_h), where P_h(t) is the instantaneous power of the hth harmonic, I_h(t) and V_h(t) are the effective values of the current and voltage of the hth harmonic respectively, and φ_h is the corresponding phase angle. The instantaneous power is mapped in three dimensions according to the spatial position and frequency dimension to construct an energy distribution solid diagram. The spatial dimension uses a cylindrical coordinate system to describe the position information inside the motor, and the frequency dimension covers the range from the fundamental wave to the high-order harmonic. The energy values between discrete measurement points are filled by interpolation algorithm to generate a continuous energy distribution function. The energy distribution diagram is represented by equal energy lines, and the density of the equal energy lines reflects the steepness of the energy gradient. The main features of the energy distribution are identified, including the energy peak position, energy propagation path, energy attenuation law and other key information. A statistical model of the energy distribution is established to calculate the statistical parameters such as mean, variance, skewness and kurtosis of the energy.

[0041] For example, the resonance region is determined based on the energy distribution, including: determining the search accuracy based on the energy distribution evaluation positioning complexity, the positioning complexity including the number of energy peaks, the distribution dispersion and the harmonic order; setting the region division parameters according to the search accuracy; and performing spatial mapping processing on the energy distribution using the region division parameters to generate the resonance region.

[0042] The search precision is determined based on the energy distribution evaluation positioning complexity. The complexity of the obtained energy distribution function is analyzed to quantify the difficulty of the resonance region positioning task. The number of energy peaks is counted by a local maximum detection algorithm, and after removing noise using morphological opening operation, the number of points satisfying the local maximum condition is calculated. The distribution dispersion is measured by the coefficient of variation: CV = σ_E / μ_E, where CV is the coefficient of variation, σ_E is the standard deviation of the energy distribution, and μ_E is the mean of the energy distribution. The larger the CV value, the more uneven the energy distribution. The harmonic complexity is represented by the number of effective harmonics, and the number of harmonic components with amplitudes exceeding a certain proportion of the maximum harmonic is counted. A comprehensive index of positioning complexity is established, which combines the number of energy peaks, distribution dispersion, and harmonic order with weighted coefficients determined by empirical data. According to the complexity level, the corresponding search precision is determined. In the case of low complexity, a coarse search strategy is used, and in the case of high complexity, a fine search strategy is used. The search precision parameters include spatial resolution, angular resolution, axial resolution, and frequency resolution. A mapping table of precision and complexity is established to achieve adaptive adjustment of search precision and improve the efficiency and accuracy of resonance region identification.

[0043] The region division parameters are set according to the search precision. Based on the determined search precision, the spatial region division strategy is designed. The radial division uses equal step size, and the appropriate division interval and division layer number are determined according to the search precision requirement. The angle division considers the symmetry of the motor, and the division interval is usually an integer multiple of the number of motor pole pairs to ensure the reasonableness of the division. The axial division is determined according to the motor length to ensure covering the entire motor effective length range. The frequency division uses a logarithmic interval strategy, which performs dense division in the low frequency band and sparse division in the high frequency band, balancing the calculation accuracy and efficiency. A four-dimensional grid coordinate system is established, and each grid node corresponds to an energy sampling point. A grid optimization algorithm is designed to increase the grid points in areas with large energy gradients and reduce the grid points in areas with flat energy, improving the calculation efficiency. The threshold parameter for region merging is determined, and when the energy density difference between adjacent regions is less than the set threshold, automatic merging is performed. A smoothing method for region boundaries is established, and an interpolation algorithm is used to eliminate jagged boundaries and obtain smooth region contours.

[0044] The resonance region is generated by spatial mapping of the energy distribution using region division parameters. A four-dimensional grid coordinate system is established based on the set region division parameters. The energy distribution space is gridized according to the radial division interval Δr, the angular division interval Δθ, the axial division interval Δz, and the frequency division interval Δf. The continuous energy distribution function is discretized into a grid data structure, with each grid node corresponding to an energy sampling point. According to the accuracy requirements of the region division parameters, the energy integral value of each grid element is calculated, and high-precision numerical integration method is used to improve the calculation accuracy. The region growing algorithm is used to start from a seed point with high energy density, and according to the merging threshold set by the region division parameters, the growth boundary is expanded to the surrounding region. The growth criterion is based on the energy density difference and gradient direction consistency of adjacent elements. When the difference is less than the threshold value in the region division parameters, the regions are merged. Morphological post-processing is performed on the grown regions, including hole filling, boundary smoothing, and noise elimination operations. The shape descriptors of each region are calculated, including compactness, slenderness ratio, and convex hull rate, etc. The abnormal regions are filtered out according to the shape constraints of the region division parameters. The final generated resonance region constitutes the spatial basis for energy collection, providing accurate candidate range for further screening of high-density collection areas.

[0045] The energy density of the resonance region is evaluated to generate high-density collection areas. The resonance region is divided into several sub-regions, and the size of each sub-region is determined according to the expected size of the energy collection device. The average energy density ρ_E of each sub-region is calculated, where ρ_E is the energy density, E_total is the total energy in the sub-region, and V_region is the volume of the sub-region. The energy density distribution is smoothed using the kernel density estimation method, and the Gaussian kernel function is used to optimize the bandwidth parameter through cross-validation method. An energy density threshold is established, and sub-regions exceeding the threshold are marked as high-density regions. The threshold is determined by considering the minimum operating power and economic requirements of the energy collection device. The high-density regions are merged, and adjacent regions with similar energy densities are merged into a large collection area. The energy quality factor of each high-density collection area is calculated, considering factors such as energy density, collection area size, and energy duration. The collection areas are prioritized, and the regions with high priority are selected as the preferred energy collection locations. Practical factors for energy collection are considered, including device installation convenience, maintenance accessibility, and the degree of impact on motor operation. A database of high-density collection areas is established to record detailed information such as location coordinates, energy parameters, geometric dimensions, and access methods of each collection area.

[0046] The energy collection points are determined based on the high-density collection area. A multi-objective optimization method is selected, and the optimization objectives include maximizing energy density, optimizing device matching, and minimizing installation complexity. A multi-objective optimization function is established, considering various performance indicators and constraints. An intelligent optimization algorithm is used to solve the multi-objective optimization problem, and appropriate population size and iteration parameters are set. The Pareto optimal solution is clustered, and the cluster center is selected as the candidate energy collection point. The mutual influence between multiple collection points is considered, and the topological structure of the collection point network is analyzed using graph theory to ensure network connectivity and robustness. The number of final energy collection points is determined according to system requirements and cost constraints, forming a reasonable distributed network. A dedicated interface circuit is designed for each energy collection point, including an impedance matching network, a power conditioning circuit, and a signal isolation module. A monitoring system for energy collection points is established to monitor the energy output and working status of each point in real time, ensuring stable operation of the system.

[0047] The auxiliary drive signal is generated by performing harmonic phase shift processing on the energy collection points. The harmonic current phase information of each energy collection point is extracted, and the phase difference Δφ_ij=φ_i-φ_j between the collection points is calculated, where φ_i and φ_j are the phases of the i-th and j-th collection points, respectively, and a phase difference matrix is constructed. The signal phases of the collection points are adjusted through phase shift technology, so that the multi-point signals can be in phase and superimposed, improving the amplitude of the synthesized signal. The phase shift processing uses a digital signal processing method, uses Hilbert transform to extract the instantaneous phase of the signal, and then uses a phase compensation filter to achieve accurate phase adjustment. A phase compensation function is designed to correct the phase of different frequency components, and an optimization algorithm is used to determine the optimal phase shift parameters, so that the processed signal has the maximum power density. The phase-shifted signal is processed through a power synthesis circuit to generate a unified auxiliary drive signal. The auxiliary drive signal has enhanced amplitude and improved waveform quality, and can provide additional driving capability for the motor control system. An adaptive weighting strategy is used in the signal synthesis process to dynamically adjust the synthesis weight according to the signal quality of each collection point. The generated auxiliary drive signal is isolated from the main drive loop by an isolation amplifier to ensure the safety and stability of the system.

[0048] The amplitude-frequency characteristic of the auxiliary drive signal is extracted to determine the enhanced control domain. High-resolution fast Fourier transform is used to extract the frequency-domain characteristics of the signal, and the analysis window uses the Hanning window function to reduce spectral leakage. The amplitude response and phase response of the signal at each frequency point are calculated, and a complete Bode plot is drawn. The main frequency components and harmonic distribution of the auxiliary drive signal are identified to determine the effective bandwidth and power spectral density of the signal. The frequency components beneficial to motor control are extracted by band-pass filtering technology, and the useless noise and interference signals are filtered out. The dynamic characteristics of the auxiliary drive signal are analyzed, including response speed, steady-state accuracy and anti-interference ability. The transfer function model of the auxiliary drive signal is established, and the model parameters are obtained by system identification method. Based on the transfer function model, the controller parameters are designed to realize the coordination of the auxiliary drive signal and the main control system. The boundary conditions of the enhanced control domain are determined, including the frequency range, amplitude range and phase range. The enhanced control domain defines the parameter space in which the auxiliary drive signal can effectively act, and within this domain the auxiliary drive signal can significantly improve the control performance of the motor.

[0049] In step S130, the sound pressure fluctuation data is subjected to frequency spectrum analysis to generate an acoustic characteristic spectrum, the acoustic characteristic spectrum is subjected to characteristic decomposition to obtain an efficiency indicating tone, the efficiency indicating tone is mapped to a torque optimization parameter, and the electromagnetic excitation frequency is adjusted based on the torque optimization parameter to establish a noise reduction control table.

[0050] Specifically, the sound pressure fluctuation data is subjected to frequency spectrum analysis to generate an acoustic characteristic spectrum. The collected sound pressure fluctuation data is input into a high-resolution frequency spectrum analysis system, and the sound pressure fluctuation data is subjected to time domain preprocessing, including removing direct current components and preliminary filtering. The sound pressure fluctuation data is segmented using a long window length, the Hanning window function is used to reduce spectral leakage, and a high window overlap rate is set to ensure the continuity of the spectrum. The power spectral density PSD(f)=|X(f)|² / fs of the sound pressure fluctuation data is calculated, where PSD(f) is the power spectral density at frequency f, X(f) is the Fourier transform of the sound pressure fluctuation data, fs is the sampling frequency, and |X(f)|² represents the square of the amplitude of the frequency domain signal. The power spectral density of the sound pressure fluctuation data is subjected to logarithmic transformation to enhance the identifiability of weak frequency components. Multi-resolution analysis technology is used to analyze the sound pressure fluctuation data at different frequency bands using different resolutions, high-resolution analysis is used at low frequency bands, and moderate-resolution analysis is used at high frequency bands, taking into account analysis accuracy and computational efficiency. The influence of background noise in the sound pressure fluctuation data is removed by spectral subtraction to extract pure acoustic signal components. The main frequency components in the spectrum of the sound pressure fluctuation data are identified, including the fundamental frequency, harmonic frequency, modulation frequency and resonance frequency, etc. A frequency-amplitude two-dimensional feature map based on the sound pressure fluctuation data is established, with the horizontal axis representing the frequency, the vertical axis representing the sound pressure amplitude, and the color depth representing the energy density. The feature spectrum of the sound pressure fluctuation data is subjected to time-frequency joint analysis, and the short-time Fourier transform is used to capture the time-varying characteristics of the acoustic signal.

[0051] In some embodiments, the feature extraction of the acoustic feature spectrum to obtain the efficiency indicator sound comprises: performing frequency component separation on the acoustic feature spectrum to extract efficiency-related components; performing feature screening processing on the efficiency-related components to generate a set of dominant frequencies; performing correlation analysis on the set of dominant frequencies and the harmonic current to obtain efficiency characteristic values; and extracting the efficiency indicator sound according to the efficiency characteristic values.

[0052] The acoustic feature spectrum is subjected to frequency component separation to extract efficiency-related components. The acoustic feature spectrum is input into a frequency component separation system, and a wavelet decomposition technique is used to decompose the complex frequency spectrum signal into components of different frequency scales. A suitable wavelet basis function is selected, and an appropriate number of decomposition layers is set to cover the complete frequency band from low frequency to high frequency. Each decomposition layer corresponds to a specific frequency range, with low layers corresponding to high frequency components and high layers corresponding to low frequency components. The energy distribution of each frequency component is calculated as E_i = ∫|W_i(f)|²df, where E_i is the energy value of the i-th layer, W_i(f) is the i-th layer wavelet decomposition coefficient, and the integral symbol represents the integral operation over the entire frequency range. Frequency components with an energy proportion exceeding a certain proportion of the total energy are retained through energy threshold screening. Empirical mode decomposition method is used to further refine the frequency components, and each main frequency band is decomposed into several intrinsic mode functions. The instantaneous frequency and instantaneous amplitude of each intrinsic mode function are analyzed, and frequency components synchronized with motor efficiency changes are identified. An efficiency correlation criterion is established, and the correlation between frequency components and motor load, speed, temperature, and other efficiency influencing factors is analyzed to determine the efficiency-related components. The Pearson correlation coefficient of each frequency component and the efficiency parameter is calculated using correlation analysis, and components with an absolute value of the correlation coefficient greater than a set threshold are selected as efficiency-related components. The efficiency-related components are reconstructed to form pure efficiency-related acoustic signals. A feature library of efficiency-related components is established to record the frequency range, energy characteristics, time-varying characteristics, and other attribute information of each component.

[0053] The advantage frequency set is generated by feature screening from the efficiency-related components. A frequency importance evaluation index is established, considering the energy intensity, stability and discriminability of the frequency components. The signal-to-noise ratio of each frequency component is calculated, and the frequency component with high signal-to-noise ratio is preferentially selected. The time stability of the frequency component is analyzed, the standard deviation of the frequency drift is calculated, and the stable frequency component is screened. The discriminability of the frequency component to different efficiency levels is evaluated, and the Fisher discriminant analysis is used to calculate the inter-class separation degree of each component. A multi-criteria decision model is established to comprehensively score the frequency components, and the scoring function is S_i=w1×SNR_i+w2×Stability_i+w3×Discrimination_i, wherein S_i is the comprehensive score of the i th frequency component, SNR_i is the signal-to-noise ratio, Stability_i is the stability index, Discrimination_i is the discriminability index, and w1, w2 and w3 are the corresponding weight coefficients. The frequency components are sorted according to the score, and the top several components are selected to form the advantage frequency set. The advantage frequency set is subjected to cluster analysis to identify similar frequency components and perform merging processing.

[0054] The advantage frequency set and the harmonic current are subjected to correlation analysis to obtain the efficiency characteristic value. The harmonic components of the harmonic current are extracted, and the effective value and phase angle of each harmonic are calculated. An acoustic-electric characteristic matching matrix is established, and the matrix elements are the correlation coefficients between the advantage frequency and the harmonic current component. The sliding window correlation analysis method is used to calculate the correlation change in different time windows, and the window length is set to an integer multiple of the fundamental period. The correlation coefficient matrix R_ij=corr(A_i,I_j) is calculated, wherein R_ij is the correlation coefficient between the i th advantage frequency and the j th harmonic current, A_i is the amplitude sequence of the i th advantage frequency, I_j is the amplitude sequence of the j th harmonic current, and corr represents the correlation coefficient calculation function. The acoustic-electric characteristic pairs with high correlation are identified, and the characteristic combinations with absolute correlation coefficient greater than a set threshold are screened. The time-varying characteristics of the correlation are analyzed, and the dynamic time warping algorithm is used to process the time alignment problem between different characteristics. A multiple regression model is established, taking the advantage frequency set as the independent variable and the motor efficiency as the dependent variable, and the regression coefficient is the efficiency characteristic value. The stepwise regression method is used to screen the acoustic characteristics that significantly affect the efficiency, and the frequency components with low contribution are removed. The partial correlation coefficient of each characteristic is calculated to evaluate its independent influence on the efficiency under the condition of controlling other variables. An efficiency characteristic value vector is established, including the efficiency influence weight corresponding to each advantage frequency. The efficiency characteristic value is normalized to eliminate the influence of different characteristic dimensions.

[0055] An efficiency indicator sound is extracted according to the efficiency characteristic value. The efficiency characteristic value is used as a weight coefficient to synthesize the dominant frequency set: EIS(t) = ∑λi×A_i(t)×cos(2πf_i×t+φ_i), where EIS(t) is the efficiency indicator sound signal at time t, λi is the efficiency characteristic value (weight) corresponding to the i-th dominant frequency, A_i(t) is the time-varying amplitude of the i-th dominant frequency at time t, f_i is the i-th dominant frequency value, φ_i is the phase angle of the i-th dominant frequency, and ∑ represents the summation of all dominant frequency components. Noise components in the efficiency indicator sound are removed through filtering processing, and a band-pass filter is used to retain signals within the effective frequency range. Envelope detection is performed on the efficiency indicator sound to extract the envelope characteristics of the signal, and the changes in the envelope reflect the fluctuations in the motor efficiency. Statistical characteristics of the efficiency indicator sound are calculated, including mean, standard deviation, peak factor, skewness, and other parameters, to establish quantitative indicators of the efficiency state. A quality evaluation method for the efficiency indicator sound is designed, and the accuracy of the indicator sound is verified by comparing it with the actual efficiency measurement value. A threshold system for the efficiency indicator sound is established, and the characteristic range of the indicator sound corresponding to different efficiency levels is defined. Time-frequency analysis is performed on the efficiency indicator sound to observe its variation law under different operating conditions. A database of efficiency indicator sounds is established to record the characteristics of the indicator sound under different motor types and operating conditions. As an acoustic indicator of the motor efficiency state, the efficiency indicator sound can monitor the changes in the energy conversion efficiency of the motor in real time. When factors such as magnetic circuit saturation, increased mechanical friction, and winding heating occur in the motor, the frequency components and amplitude characteristics of the efficiency indicator sound will change accordingly. By monitoring the trend of the efficiency indicator sound, early signs of motor efficiency degradation can be identified in a timely manner.

[0056] The efficiency indicator sound is mapped to torque optimization parameters. The frequency components of the efficiency indicator sound are directly related to the electromagnetic torque pulsation of the motor. By analyzing this correlation, the adjustment parameters for torque optimization can be derived. A torque optimization function is established, with the amplitude, dominant frequency, and phase characteristics of the efficiency indicator sound as input variables. The least squares method is used to fit the parameters of the mapping function to ensure the accuracy of the mapping relationship. The relationship between the amplitude variation of the efficiency indicator sound and the torque pulsation coefficient is analyzed to establish an amplitude-pulsation mapping model. The relationship between the frequency drift of the efficiency indicator sound and the stability of the torque output is studied to determine the sensitive interval of frequency adjustment. The phase information of the efficiency indicator sound is calculated, and the phase characteristics are converted to torque phase compensation parameters. A multi-dimensional torque optimization parameter vector is established, including amplitude adjustment coefficients, frequency compensation amounts, phase correction angles, and other key parameters. The effectiveness of the mapping relationship is verified through numerical simulation to ensure that the torque optimization parameters accurately reflect the operating characteristics of the motor. An adaptive mapping mechanism is established considering the changes in the mapping relationship under different operating conditions.

[0057] Based on the torque optimization parameters, the noise reduction control table is established by adjusting the electromagnetic excitation frequency. The obtained torque optimization parameters are used as input to design the adjustment strategy of electromagnetic excitation frequency. The goal of adjusting the electromagnetic excitation frequency is to reduce acoustic noise to the maximum extent while ensuring the performance of the motor. The frequency adjustment algorithm is established to calculate the excitation frequency offset of each phase winding according to the torque optimization parameters. The frequency adjustment function is designed to establish the mathematical relationship between the torque optimization parameters and the frequency offset. The multi-frequency point coordinated control strategy is adopted for frequency adjustment, which optimizes both the fundamental frequency and the harmonic frequency. The impact of frequency adjustment on motor torque output is calculated to ensure that the adjusted frequency does not significantly reduce motor efficiency. The constraint conditions are established, including the frequency adjustment range, torque fluctuation limit, efficiency loss threshold, etc. The optimal frequency adjustment scheme under multiple constraint conditions is solved using an optimization algorithm. The optimization results are organized into a noise reduction control table, which includes fields such as working condition, original frequency, adjusted frequency, and expected noise reduction effect. The control table adopts a hierarchical structure and is classified according to factors such as load level, speed range, and environmental conditions. The real-time query mechanism of the control table is established to support quick retrieval of corresponding frequency adjustment parameters based on the current operating conditions.

[0058] In step S140, the enhanced control domain is subjected to spectral analysis to identify high-energy frequency bands, frequency shift operations are performed on the high-energy frequency bands to generate safety frequency band signals, energy aggregation processing is performed on the safety frequency band signals to form frequency shift control sequences, and a multi-frequency band management strategy is constructed based on the fusion of the frequency shift control sequences and the noise reduction control table.

[0059] Specifically, the high-energy frequency band is identified by performing spectrum analysis on the enhanced control domain. The obtained enhanced control domain data is input into a high-resolution spectrum analysis system, and a multi-window spectrum estimation method is used to improve the accuracy and stability of the spectrum analysis. The power spectral density is estimated using the Thompson multi-window method, and an appropriate number of orthogonal window functions are selected to balance the frequency resolution and variance reduction effect. The energy density distribution P(f) = |F(f)|² / T of each frequency point in the enhanced control domain is calculated, where P(f) is the power spectral density at frequency f, F(f) is the Fourier transform of the enhanced control domain signal, and T is the observation time length. The high-energy frequency band is identified by the energy density threshold method, and the energy threshold is set to a certain percentage of the total energy. The continuous frequency interval exceeding the threshold is marked as the high-energy frequency band. The sliding average filter is used to smooth the power spectrum curve and eliminate the influence of random fluctuations on frequency band identification. The peak detection algorithm is used to locate the energy peak point in the power spectrum, and the peak value is expanded to both sides until the energy density decreases to half of the peak value, and the boundaries of the high-energy frequency band are determined. The frequency band feature parameter library is established to record the key parameters of each high-energy frequency band, such as center frequency, bandwidth, peak power, and energy integral. The time-varying characteristics of the high-energy frequency band are observed through time-frequency analysis to identify stable frequency bands and transient frequency bands. The identified high-energy frequency bands are prioritized, and the frequency bands with high energy density and narrow bandwidth have higher processing priority. A classification system of high-energy frequency bands is established, and the frequency bands are classified into categories such as fundamental wave related, harmonic related, resonance related, and interference related according to their physical causes.

[0060] In some embodiments, the performing frequency shift operation on the high-energy frequency band to generate a safety frequency band signal includes: performing spectrum window division on the high-energy frequency band to generate a spectrum window set; performing frequency shift path planning based on the spectrum window set to obtain a target frequency band; performing carrier modulation processing on the target frequency band to generate a shift carrier; and performing energy transfer processing using the shift carrier to generate a safety frequency band signal.

[0061] The high-energy frequency band is divided into a set of spectral windows. The identified high-energy frequency band is taken as the processing object, and the spectrum in the high-energy frequency band is subdivided. The principle of window division is to keep the high-energy frequency band characteristics in each window relatively uniform, avoiding the same window of large difference frequency components. The energy gradient analysis is used to determine the window boundary in the high-energy frequency band, and the first derivative of the spectrum of the high-energy frequency band is calculated ∇P(f)=dP(f) / df, where ∇P(f) is the power spectrum gradient at frequency f in the high-energy frequency band, P(f) is the corresponding power spectrum density, and d / df represents the differential operation on the frequency. The window boundary point is set at the position where the gradient changes sharply in the high-energy frequency band, ensuring that the spectrum inside the window changes smoothly. An adaptive window size mechanism for the high-energy frequency band is established, and the window size is dynamically adjusted according to the complexity of the spectrum in the high-energy frequency band. The complex frequency band uses a smaller window, and the simple frequency band uses a larger window. The window overlap strategy in the high-energy frequency band is designed, and a certain overlap area is maintained between adjacent windows to avoid the influence of boundary effects on the processing result. The characteristic parameters of each window in the high-energy frequency band are calculated, including the center frequency, effective bandwidth, energy integral, peak position, etc. An identification system for high-energy frequency band windows is established, and each window is assigned a unique number and attribute label. Morphological processing method is used to eliminate small windows in the high-energy frequency band, and they are merged into adjacent larger windows. The window boundary in the high-energy frequency band is smoothed, and a transition function is used to realize the smooth connection between windows. The generated set of spectral windows contains complete window parameter information, providing a structured processing unit for frequency shift path planning.

[0062] The target frequency band is obtained by frequency shifting path planning based on a set of spectrum windows. The generated set of spectrum windows is used as input, and a corresponding frequency shifting path is designed for each spectrum window in the set. Path planning needs to consider factors such as availability, shifting distance, and energy capacity of the target frequency band. A frequency space occupancy map is established to mark occupied and available frequency regions, providing constraints for path planning. The windows in the set of spectrum windows are prioritized, and the optimal path is assigned to important windows first. A graph search algorithm is used to find the optimal path from the source frequency band to the target frequency band, and the path cost function includes factors such as shifting distance, target region congestion, and shifting complexity. An obstacle avoidance mechanism is designed to avoid conflicts between the shifting path and occupied frequency bands. The cost function C of path planning is calculated as C = w1 × D + w2 × O + w3 × I, where C is the total cost, D is the shifting distance, O is the target region occupancy, I is the implementation complexity, and w1, w2, and w3 are weight coefficients. A dynamic programming method is used to solve the joint path planning problem of multiple windows, ensuring a global optimal solution. A path conflict detection mechanism is established to identify the intersection and overlap of shifting paths for different windows. A path adjustment strategy is designed to automatically adjust the target frequency band for some windows when conflicts occur. The corresponding target frequency band is determined for each window in the set of spectrum windows, and a complete shifting mapping relationship is established.

[0063] The target frequency band is modulated to generate a shifting carrier. The goal of carrier modulation is to generate a carrier signal that can accurately shift the energy of the source frequency band to the target frequency band. The carrier frequency f_carrier is calculated as f_carrier = f_target - f_source, where f_carrier is the carrier frequency, f_target is the center frequency of the target frequency band, and f_source is the center frequency of the source frequency band. The amplitude envelope of the carrier signal is designed using a Gaussian envelope function to ensure compactness in the frequency domain, and the carrier signal is represented as C(t) = A(t) × cos(2πf_carrier × t + φ(t)), where C(t) is the carrier signal, A(t) is the time-varying amplitude envelope, and φ(t) is the time-varying phase. Digital signal processing techniques are used to generate an accurate carrier signal, and a direct digital frequency synthesizer is used to ensure frequency accuracy. A carrier parameter optimization algorithm is established to adjust the amplitude, frequency, and phase parameters of the carrier based on shifting effect feedback. A multi-carrier coordination mechanism is designed to ensure orthogonality and non-interference between carriers when multiple carriers are required to work simultaneously. A carrier quality monitoring system is established to detect the frequency stability and amplitude consistency of the carrier signal in real time. Carrier pre-distortion technology is used to compensate for nonlinear distortion during modulation. A carrier signal library is generated to store carrier parameter templates corresponding to different shifting requirements.

[0064] The energy transfer process is performed by using a shift carrier to generate a safe frequency band signal. The generated shift carrier is modulated with the source frequency band signal to achieve energy transfer from the high-energy frequency band to the safe frequency band. The energy transfer uses balanced modulation technology, and the frequency spectrum is shifted through multiplication operation M(t) = S(t) × C(t), where M(t) is the modulated signal, S(t) is the source frequency band signal, and C(t) is the shift carrier. The modulated signal contains two spectral components, upper and lower sidebands, and the required sideband signal is selected by a filter. A sideband selection filter is designed to suppress unwanted sidebands and carrier leakage using a notch filter. An energy transfer efficiency monitoring mechanism is established to calculate the energy ratio before and after transfer to ensure the integrity of the energy transfer. Amplitude and phase compensation techniques are used to correct signal distortion during the transfer process. A multi-stage transfer scheme is designed for large shift distances, and the final goal is achieved through multiple small distance shifts. A transfer quality evaluation system is established, considering indicators such as spectral purity, phase continuity, and amplitude fidelity. The generated safe frequency band signal is post-processed, including filtering, amplification, and amplitude limiting operations.

[0065] The energy aggregation process is performed on the safe frequency band signal to form a frequency shift control sequence. The goal of the energy aggregation process is to effectively integrate the energy dispersed in different safe frequency bands to form a sequence signal that is easy for the control system to process. Wavelet packet decomposition technology is used to perform multi-scale analysis on the safe frequency band signal, decomposing the signal into sub-band signals with different frequencies and time resolutions. The energy distribution of each sub-band is calculated as E_k = ∫|W_k(t)|²dt, where E_k is the energy of the kth sub-band, W_k(t) is the wavelet coefficient of the kth sub-band, and the integral represents the integral operation over time. The importance weight of each frequency band is determined by the energy center calculation, and the energy center frequency is f_cg = ∫f × P(f) df / ∫P(f) df, where f_cg is the center frequency, and P(f) is the power spectral density. An energy aggregation matrix is established, and the matrix elements represent the energy correlation degree between different frequency bands. Principal component analysis is used to reduce the dimension of the energy aggregation matrix and extract the main energy aggregation mode. A time domain window function is designed to sample the aggregated energy in time, generating discrete control sequence points. The sampling frequency of the control sequence is determined according to the response speed of the control system to ensure that the control command can respond to system state changes in a timely manner. The generated frequency shift control sequence is smoothed to eliminate sudden changes and burrs, improving the stability of the control. An encoding rule is established for the control sequence to map continuous sequence values to discrete control commands.

[0066] A multi-band management strategy is constructed based on the fusion of frequency shift control sequence and noise reduction control table. The fusion process needs to coordinate the two goals of frequency shift control and noise reduction control, and to achieve effective management of frequency bands while ensuring noise reduction effect. A fusion weight distribution mechanism is established to determine the priority of frequency shift control and noise reduction control according to the current operating conditions. A fusion algorithm F(t)=α×FC(t)+β×NC(t) is designed, where F(t) is the fused management strategy, FC(t) is the frequency shift control sequence, NC(t) is the noise reduction control instruction, and α and β are dynamic weight coefficients satisfying α+β=1. The determination of weight coefficients is based on real-time performance evaluation. When the noise level is high, the value of β is increased, and when the frequency band conflict is serious, the value of α is increased. A multi-band coordination mechanism is established to avoid conflicts between control instructions of different frequency bands. A priority scheduling algorithm is used to handle concurrent control requirements of multiple frequency bands, and high-priority frequency band control instructions are executed first. A frequency band switching strategy is designed to achieve smooth transition of management strategies under different operating conditions. An exception handling mechanism is established to automatically start the backup management strategy when an exception occurs in a certain frequency band. The state changes of each frequency band are monitored in real time during the execution of the strategy, and the management parameters are dynamically adjusted according to the feedback information.

[0067] In step S150, the multi-band management strategy is decomposed into parallel control streams, the parallel control streams are arranged in time sequence to generate an interleaved control sequence, and an energy-saving control matrix is generated based on the interleaved control sequence.

[0068] Specifically, the multi-band management strategy is decomposed into parallel control streams. Each parallel control stream corresponds to a specific frequency band range and control task, and the response speed and control accuracy of the system are improved through parallel processing. A functional decomposition method is used to split the complex management strategy into several relatively simple control sub-tasks. A control stream classification system is established, and the control streams are classified into categories such as fundamental control stream, harmonic control stream, resonance suppression control stream, and interference elimination control stream according to the physical characteristics of the frequency bands. Each control stream contains independent input interface, processing logic and output interface, ensuring the decoupling and independence between control streams. A priority mechanism for control streams is designed, and different control streams are assigned priorities according to their importance and urgency. A communication protocol is established between control streams to support data exchange and coordination when necessary. A pipeline architecture is used to organize parallel control streams, and each control stream runs in an independent processing unit to avoid mutual interference. A load balancing algorithm is designed to ensure that the computational load of each control stream is relatively uniform, avoiding the bottleneck of a certain control stream. A state monitoring mechanism for control streams is established to track the running state and processing progress of each control stream in real time. A fault-tolerant mechanism is designed, and other control streams can continue to work normally when a fault occurs in a certain control stream. The detailed information of control stream decomposition is recorded, including the number of control streams, allocation strategy, interface definition, etc.

[0069] In some embodiments, the timing stagger arrangement of the parallel control flows generates a staggered control sequence, including: establishing a phase shift reference by switch time analysis of the parallel control flows; determining misphased times by delay configuration processing using the phase shift reference; and constructing a staggered control sequence using the misphased times.

[0070] A phase shift reference is established by switch time analysis of the parallel control flows. Switch action times of each parallel control flow are taken as analysis objects, and a phase shift reference of the system is established by statistical analysis. Switch time analysis is to identify time distribution characteristics of switch actions of each control flow, and to provide a basis for subsequent phase shift design. Switch time sequences of each control flow are extracted, and occurrence time and duration of switch actions are recorded. Statistical characteristics of switch times are calculated, including average interval, variance, skewness, kurtosis and other parameters. Periodic characteristics of switch times are analyzed, and main periodic components are identified using Fourier analysis. A probability distribution model of switch times is established, and a probability density function of time distribution is fitted using kernel density estimation method. A correlation matrix R_ij=corr(T_i,T_j) of switch times of each control flow is calculated, where R_ij is a correlation coefficient of switch times of the i-th and j-th control flows, T_i and T_j are corresponding switch time sequences, corr represents a correlation coefficient calculation function. Control flow pairs with high correlation are identified, and these control flows need to be phase-shifted to avoid simultaneous switching. A phase shift reference frequency f_ref is established, and the reference frequency is usually selected as a fundamental frequency or an integer multiple of the system. A phase angle φ_i=2π×(t_imodT_ref) / T_ref of each control flow relative to the reference frequency is calculated, where φ_i is a phase angle of the i-th control flow, t_i is a switch time, T_ref is a reference period, and mod represents a modulus operation. A phase distribution diagram is established, and distribution of each control flow in the phase space is observed.

[0071] The phase offset time is determined by using a phase shift reference for a delay configuration process. The purpose of the delay configuration is to introduce appropriate time delay to stagger the control flows that are originally in phase or close to phase, so as to reduce the power consumption peak caused by simultaneous switching. The reference frequency f_ref in the phase shift reference is taken as a synchronous reference to calculate the phase deviation of each control flow relative to the reference. A delay optimization objective function is established to minimize the switching loss and the current ripple. According to the ideal phase distribution mode set in the phase shift reference, a delay allocation strategy is designed, and the principle of uniform distribution is used to uniformly distribute each control flow in a reference period. The ideal phase interval Δφ_ideal=2π / N is calculated by using the phase interval parameter in the phase shift reference, where N is the number of parallel control flows. Based on the phase shift reference, each control flow is allocated a target phase φ_target_i=φ_base+i×Δφ_ideal, where φ_base is the base phase in the phase shift reference, and i is the control flow number. The required delay amount τ_i=(φ_target_i-φ_current_i)×T_ref / (2π) is calculated, where τ_i is the delay amount of the i th control flow, φ_current_i is the current phase, φ_target_i is the target phase determined based on the phase shift reference, and T_ref is the reference period. Delay constraint conditions are established to ensure that the delay amount is within a reasonable range, avoiding excessive delay affecting system response.

[0072] The staggered control sequence is constructed by using the phase offset time. The construction of the staggered control sequence needs to consider the priority, execution time and resource demand of each control flow. A time axis model is established to divide a control period into several time units, each of which corresponds to a possible control action. The starting time point of each control flow is marked on the time axis according to the phase offset time. A control event scheduling algorithm is designed to ensure that no conflicting control actions are executed simultaneously at the same time. The coding format of the control sequence is established, and binary coding is used to represent the state of each control flow at each time. The control sequence matrix S(t,i) is generated, where S(t,i) represents the state of control flow i at time t, 0 represents off, and 1 represents on. The performance indicators of the staggered control sequence are calculated, including switching frequency dispersion, power consumption peak reduction rate, current ripple suppression effect, etc. The periodicity of the sequence is detected to ensure that the generated sequence has good periodicity characteristics. A compression storage method of the sequence is designed to reduce the storage space requirement by pattern recognition. A real-time generation mechanism of the sequence is established to support dynamic generation of subsequent control sequences according to the current state. Detailed parameters of the staggered control sequence are recorded, including sequence length, staggered mode, performance indicators and other information.

[0073] In some embodiments, the generating the energy-saving control matrix based on the interleaved control sequence comprises: identifying a concentrated region of switching loss through the interleaved control sequence; performing energy distribution analysis on the concentrated region of switching loss to form a loss distribution map; determining an optimization path according to the loss distribution map; and forming the energy-saving control matrix through the optimization path.

[0074] A concentrated region of switching loss is identified through the interleaved control sequence. The generated interleaved control sequence is taken as an analysis input to identify the concentrated characteristics of switching actions from the time distribution pattern of the sequence. Switching time data is extracted from the interleaved control sequence, and a switching density function D(t) = N_switch(t) / Δt is calculated, where D(t) is the switching density at time t, and N_switch(t) is the number of switches within a time window Δt counted from the interleaved control sequence. Based on the timing characteristics of the interleaved control sequence, the time-varying characteristics of the switching density are analyzed using a sliding window technique, and the window length is determined according to the periodic characteristics of the interleaved control sequence. A switching density threshold is set, and time periods exceeding the threshold are marked as concentrated regions of switching loss. By analyzing the time intervals of adjacent control events in the interleaved control sequence, a clustering analysis method is used to merge adjacent high-density time periods into continuous concentrated regions. Based on the execution pattern of the interleaved control sequence, the causes of the formation of the concentrated regions of switching loss are analyzed, and the control strategies and sequence arrangements that lead to the concentration of switching actions are identified. An evaluation index of the severity of the concentrated regions is established, and factors such as switching frequency, involved power, and duration are considered comprehensively. The energy loss E_loss = ∑(V_switch×I_switch×t_switch) of each concentrated region is calculated, where V_switch is the switching voltage, I_switch is the switching current, and t_switch is the switching time. The identified concentrated regions of switching loss are prioritized, and the regions with the largest loss and the most serious impact are processed first.

[0075] Energy distribution analysis is performed on the switch loss concentration area to form a loss distribution map. A multi-dimensional loss analysis framework is established to analyze the distribution characteristics of loss from time dimension, space dimension, and frequency dimension. The loss contribution of each control flow in the switch loss concentration area is calculated as P_loss_i = P_switch_i + P_conduction_i + P_blocking_i, where P_loss_i is the total loss of the i-th control flow, P_switch_i is the switch loss, P_conduction_i is the conduction loss, and P_blocking_i is the blocking loss. A heat map visualization technique is used to generate the loss distribution map, with the horizontal axis representing time, the vertical axis representing control flow number, and color depth representing loss intensity. An interpolation algorithm is used to fill in the loss values between discrete measurement points to generate a continuous loss distribution function. The statistical characteristics of the loss distribution are analyzed, and statistical quantities such as mean, standard deviation, maximum, and minimum are calculated. Hotspot areas of the loss distribution are identified, which usually correspond to weak links in the system. A loss gradient analysis is established to calculate the rate of change of loss in time and space directions. Principal component analysis is used to extract the main modes of the loss distribution and identify key factors affecting the loss distribution.

[0076] An optimization path is determined based on the loss distribution map. Based on the generated loss distribution map, a systematic optimization strategy is designed to determine specific paths and methods to reduce switch loss. The design of the optimization path needs to balance the loss reduction effect and the implementation complexity, and select the optimization scheme with the highest cost-effectiveness. An optimization objective function is established to minimize the total loss as the main goal, while constraining the system performance not to be significantly affected. The optimization potential area in the loss distribution map is identified, and the area with high loss density and large optimization space is focused on. A multi-level optimization strategy is designed, including time sequence optimization, parameter optimization, and topology optimization at different levels. The loss reduction potential of various optimization schemes is calculated as ΔP_reduction = P_original - P_optimized, where ΔP_reduction is the loss reduction, P_original is the original loss, and P_optimized is the optimized loss. A feasibility evaluation system for optimization schemes is established, considering technical feasibility, economic feasibility, and implementation difficulty. A multi-objective optimization algorithm is used to find the Pareto optimal solution set, balancing the loss reduction and other performance indicators. A phased implementation plan for the optimization path is designed, with a priority on optimization measures with fast effect and low risk. An evaluation method for the optimization effect is established to verify the effectiveness of the optimization scheme through simulation and experiment. The detailed planning of the optimization path is recorded, including optimization steps, expected effect, implementation scheme, and other information.

[0077] An energy-saving control matrix is formed by optimizing the path. Based on different level strategies such as timing optimization, parameter optimization and topology optimization in the optimized path, the dimension definition of the matrix is established. The row dimension represents the implementation stage or working condition defined in the optimized path, and the column dimension represents the control parameter or control flow number involved in the optimized path. According to the calculation results of the loss reduction potential ΔP_reduction of each optimization scheme in the optimized path, the value rule of the matrix elements is designed, and each element corresponds to the optimal value of a specific parameter under a specific stage. The phased implementation scheme in the optimized path is converted into a matrix query mechanism, and the parameter query is realized by using the table lookup method. The optimal control parameter is quickly determined according to the current implementation stage. Based on the continuity characteristics of the Pareto optimal solution set in the optimized path, an interpolation mechanism of the matrix is established to handle the transition state between the implementation stages which is not clearly defined. According to the optimization effect evaluation method in the optimized path, an adaptive updating algorithm of the matrix is designed to continuously optimize the matrix parameters according to the actual running effect. Considering the implementation complexity control requirements in the optimized path, a compressed storage format of the matrix is established to reduce the memory occupation and access delay. According to the multi-level parallel optimization strategy in the optimized path, a parallel access mechanism of the matrix is designed to support multiple control flows to query the required parameters at the same time.

[0078] In step S160, a PWM control signal is generated based on the energy-saving control matrix, an edge reconstruction is performed on the PWM control signal to obtain a softened edge waveform, and the softened edge waveform is output to a power device to complete intelligent control of a motor driver.

[0079] Specifically, the PWM control signal is generated based on the energy-saving control matrix. The PWM signal generation process needs to convert the control parameters in the matrix into specific pulse width and switching timing. A PWM parameter mapping relationship is established to map the values in the energy-saving control matrix to the duty cycle, frequency and phase parameters of the PWM signal. The PWM duty cycle D=T_on / T_period is calculated, where D is the duty cycle, T_on is the high level duration, and T_period is the PWM period. According to the torque demand of the motor and the energy-saving optimization target, the PWM parameters are dynamically adjusted. The space vector pulse width modulation technology is used to improve the voltage utilization rate and reduce the harmonic content. The symmetry constraint of the three-phase PWM signal is established to ensure the balanced operation of the three-phase system. An adaptive adjustment mechanism of the PWM carrier frequency is designed to optimize the carrier frequency according to the motor speed and load conditions. A PWM signal jitter technology is established to disperse the energy concentration of the switching frequency through a small random disturbance. The multi-level PWM technology is used to reduce the voltage step amplitude and reduce electromagnetic interference. A synchronization mechanism of the PWM signal is designed to ensure the coordinated action of multiple power devices. A quality monitoring system of the PWM waveform is established to detect the symmetry, linearity and stability of the waveform in real time. The generated PWM control signal contains accurate timing information and amplitude information, which provides instructions for accurate control of power devices.

[0080] In some embodiments, the edge reconstruction of the PWM control signal obtains a softened edge waveform, including: time-domain unfolding the PWM control signal to obtain a transient voltage change rate; resonance peak detection of the transient voltage change rate to identify EMI excitation points; constructing an anti-resonance injection sequence based on the EMI excitation points; and compensating and modulating the transient voltage change rate using the anti-resonance injection sequence to obtain a softened edge waveform.

[0081] The transient voltage change rate is obtained by time-domain unfolding the PWM control signal. The transient characteristics of the PWM signal are the main cause of electromagnetic interference and switching loss, and these transient processes can be quantified by time-domain unfolding analysis. A high-precision time-domain sampling technique is used to digitize the signal at a sampling rate much higher than the PWM carrier frequency. A time-domain analysis window is established, focusing on the rising and falling edge regions of the PWM signal, and the window length is determined according to the edge duration. The transient voltage change rate dV / dt = ΔV / Δt is calculated, where dV / dt is the voltage change rate, ΔV is the voltage change, and Δt is the time interval. Numerical differentiation algorithms are used to calculate the voltage change rate at discrete time points, and the central difference method is used to improve the calculation accuracy. The time-varying characteristics of the transient voltage change rate are analyzed, and the peak position and amplitude of the change rate are identified. A statistical model of the transient characteristics is established, and statistical parameters such as the mean, standard deviation, maximum, and minimum of the change rate are calculated. Wavelet transform technology is used to analyze the multi-scale characteristics of the transient process, and to identify transient components at different time scales. The frequency spectrum characteristics of the transient voltage change rate are established, and its frequency domain distribution is analyzed by Fourier transform.

[0082] The EMI excitation points are identified by resonance peak detection of the transient voltage change rate. EMI excitation points usually correspond to extreme points or change rate mutation points in the transient process, and these points are the main source of electromagnetic energy radiation. The judgment criteria for resonance peak detection are established, and the detection threshold and window parameters are set. Peak detection algorithms are used to locate local maximum points in the transient voltage change rate curve. The significance index S = (P_peak - P_background) / σ_noise is calculated, where S is the significance index, P_peak is the peak amplitude, P_background is the background level, and σ_noise is the noise standard deviation. Peak points with a significance index exceeding the set threshold are selected as candidate EMI excitation points. The frequency domain characteristics of the peak points are analyzed, and the frequency components corresponding to the peaks are analyzed using short-time Fourier transform. An EMI excitation strength evaluation model is established, considering factors such as peak amplitude, frequency range, and duration. The spatial distribution characteristics of the EMI excitation points are calculated, and the distribution density of the excitation points on the time axis is analyzed. Clustering algorithms are used to group similar EMI excitation points, and to identify concentrated areas of EMI excitation. The priority of the EMI excitation points is sorted, and excitation points with high excitation strength and wide impact range are processed first.

[0083] The anti-resonance injection sequence is constructed based on the EMI excitation points. The principle of anti-resonance injection is to inject a compensation signal with opposite phase and equal amplitude at the EMI excitation point to achieve active suppression of interference. The spectral characteristics of each EMI excitation point are analyzed to determine the frequency components and phase information that need to be suppressed. The generation model of the anti-resonance signal is established, and the anti-phase resonance signal corresponding to the EMI excitation frequency is designed. The amplitude of the anti-resonance signal A_anti=-k×A_EMI is calculated, where A_anti is the amplitude of the anti-resonance signal, A_EMI is the amplitude of the EMI excitation point, k is the compensation coefficient, and the negative sign represents the anti-phase. The phase of the anti-resonance signal φ_anti=φ_EMI+π is designed, where φ_anti is the phase of the anti-resonance signal, φ_EMI is the phase of the EMI excitation point, and π represents a 180-degree phase difference. A multi-frequency point anti-resonance mechanism is established to simultaneously handle the interference suppression of multiple EMI excitation points. An adaptive algorithm is used to dynamically adjust the anti-resonance parameters, and the compensation strength is optimized in real time according to the suppression effect. The timing control of anti-resonance injection is designed to ensure that the compensation signal is injected into the system at the correct time. The encoding format of the anti-resonance injection sequence is established to support efficient storage and real-time calling. The power consumption cost of anti-resonance injection is calculated to ensure that the injection process does not significantly increase system loss. Detailed parameters of the anti-resonance injection sequence are recorded, including injection time, signal amplitude, phase information, and other key data.

[0084] The soft edge waveform is obtained by compensating and modulating the transient voltage rate using the anti-resonance injection sequence. The goal of compensating and modulating is to significantly improve the transient characteristics and electromagnetic compatibility of the PWM signal while maintaining its basic functions. The mathematical model of compensating and modulating is established as V_soft(t)=V_original(t)+V_anti(t), where V_soft(t) is the softened signal, V_original(t) is the original PWM signal, and V_anti(t) is the anti-resonance injection signal. A control algorithm for compensating strength is designed to dynamically adjust the compensation parameters according to EMI suppression requirements and system performance requirements. Real-time compensating and modulating is achieved using digital signal processing technology, and a high-speed digital signal processor is used to process the compensation operation. A monitoring mechanism for the compensation effect is established to evaluate the impact of softening on signal quality in real time. An edge smoothness evaluation index is designed to quantify the effect of softening. Filtering technology is used to further smooth the compensated signal to eliminate high-frequency noise that may be introduced during the compensation process. A quality control system for the softened waveform is established to ensure that the generated waveform meets the requirements of power device driving. An optimization algorithm for softening parameters is designed to find the best balance point between EMI suppression and switching performance through multi-objective optimization. A standardized format for the softened waveform is established to support compatibility for different types of power devices. The generated softened edge waveform has smooth transient characteristics, reduces electromagnetic interference levels, and maintains good switching control performance.

[0085] The softening edge waveform is output to the power device to complete the intelligent control of the motor driver. The output process needs to consider the characteristics of the drive circuit and the response characteristics of the power device to ensure the effective transmission of the control signal. An impedance matching network is established for the output interface to ensure the integrity and stability of the signal transmission process. An amplification circuit for the drive signal is designed to amplify the control signal to a level sufficient to drive the power device. An optical isolation mechanism is established to ensure electrical isolation and safety between the control circuit and the power circuit. A protection circuit for the power device is designed, including overcurrent protection, overvoltage protection, and overheat protection, etc. A feedback monitoring system for the drive signal is established to monitor the working state and response characteristics of the power device in real time. Dead-time control technology is used to avoid the shoot-through phenomenon of the upper and lower devices on the same bridge arm. A temperature compensation mechanism for the drive circuit is designed to ensure stable operation at different temperatures. A fault diagnosis and handling mechanism is established to automatically switch to a safe mode when an abnormal condition is detected. A soft start function is designed for the system to avoid start-up impact by gradually increasing the output voltage. Finally, the intelligent control of the motor driver is completed, realizing a complete control link from signal generation to power output, with the characteristics of high efficiency, low noise, and intelligence.

[0086] In order to perform the intelligent control method of the motor driver corresponding to the above-mentioned method embodiment, to realize the corresponding functions and technical effects. Referring to Figure 2 , Figure 2 The structure block diagram of the intelligent control system 200 of the motor driver provided by the embodiment of the present application is shown. For ease of illustration, only the parts related to the present embodiment are shown. The intelligent control system 200 of the motor driver provided by the embodiment of the present application comprises:

[0087] The signal acquisition module 201 is configured to acquire multi-source abnormal signals in the motor operation process, wherein the multi-source abnormal signals include harmonic current and sound pressure fluctuation data, and perform feature conversion on the multi-source abnormal signals to form a fault feature library.

[0088] The energy recovery module 202 is configured to identify available fault modes based on the fault feature library, perform energy recovery evaluation on the available fault modes to determine energy collection points, perform harmonic phase shift processing on the energy collection points to generate auxiliary drive signals, and perform amplitude-frequency characteristic extraction on the auxiliary drive signals to determine an enhanced control domain.

[0089] The acoustic optimization module 203 is configured to perform frequency spectrum analysis on the sound pressure fluctuation data to generate an acoustic feature spectrum, perform feature decomposition on the acoustic feature spectrum to obtain an efficiency indicator sound, map the efficiency indicator sound to a torque optimization parameter, perform electromagnetic excitation frequency adjustment based on the torque optimization parameter to establish a noise reduction control table.

[0090] The frequency management module 204 is configured to perform spectrum analysis on the enhanced control domain to identify a high-energy frequency band, perform frequency shift operation on the high-energy frequency band to generate a safe frequency band signal, perform energy aggregation processing on the safe frequency band signal to form a frequency shift control sequence, and construct a multi-frequency band management strategy based on fusion of the frequency shift control sequence and the noise reduction control table.

[0091] The timing control module 205 is configured to decompose the multi-frequency band management strategy into parallel control streams, perform timing interleaving arrangement on the parallel control streams to generate an interleaved control sequence, and generate an energy-saving control matrix based on the interleaved control sequence.

[0092] The output shaping module 206 is configured to generate a PWM control signal based on the energy-saving control matrix, perform edge reconstruction on the PWM control signal to obtain a softened edge waveform, and output the softened edge waveform to a power device to complete intelligent control of the motor driver.

[0093] The above-described intelligent control system 200 of the motor driver can implement an intelligent control method of the motor driver according to the above-described method embodiment. The optional items in the above-described method embodiment are also applicable to the present embodiment, and will not be described in detail herein. The remaining content of the present embodiment can be referred to the content of the above-described method embodiment, and will not be described in detail herein.

[0094] The above embodiments are intended to exemplarily reproduce and deduce the technical solutions of the present application, and completely describe the technical solutions, objects and effects of the present application. The purpose is to make the public more thoroughly and comprehensively understand the disclosed content of the present application, and does not limit the protection scope of the present application.

[0095] The above embodiments are not exhaustive enumeration based on the present application, and there can be multiple other embodiments not listed. Any replacement and improvement made without violating the concept of the present application is within the protection scope of the present application.

Claims

1. A method of intelligent control of a motor drive, characterized by, The method comprises the following steps: Collecting multi-source abnormal signals in the motor operation process, the multi-source abnormal signals including harmonic current and sound pressure fluctuation data, and performing feature conversion on the multi-source abnormal signals to form a fault feature library; Based on the fault feature library, identifying available fault modes, performing energy recovery evaluation on the available fault modes to determine energy collection points, performing harmonic phase shift processing on the energy collection points to generate auxiliary drive signals, and performing amplitude-frequency characteristic extraction on the auxiliary drive signals to determine an enhanced control domain; Performing frequency spectrum analysis on the sound pressure fluctuation data to generate an acoustic feature spectrum, performing feature decomposition on the acoustic feature spectrum to obtain an efficiency indicator sound, mapping the efficiency indicator sound to a torque optimization parameter, and performing electromagnetic excitation frequency adjustment based on the torque optimization parameter to establish a noise reduction control table; Performing frequency spectrum analysis on the enhanced control domain to identify high-energy frequency bands, performing frequency shift operation on the high-energy frequency bands to generate safety frequency band signals, performing energy aggregation processing on the safety frequency band signals to form a frequency shift control sequence, and fusing the frequency shift control sequence and the noise reduction control table to construct a multi-frequency band management strategy; Decomposing the multi-frequency band management strategy into parallel control flows, arranging the parallel control flows in time sequence to generate an interleaved control sequence, and generating an energy-saving control matrix based on the interleaved control sequence; Generating a PWM control signal based on the energy-saving control matrix, obtaining a softened edge waveform by performing edge reconstruction on the PWM control signal, and outputting the softened edge waveform to a power device to complete intelligent control of the motor driver.

2. The intelligent control method of the motor driver according to claim 1, characterized in that, The feature conversion on the multi-source abnormal signals to form the fault feature library comprises: Constructing a current distortion map based on the harmonic current; Performing period difference calculation on the current distortion map to identify distortion growth rate; Constructing an acoustic-electric correlation matrix based on the distortion growth rate and the sound pressure fluctuation data; Establishing a fault feature library based on the acoustic-electric correlation matrix.

3. The intelligent control method of the motor driver according to claim 1, characterized in that, The energy recovery evaluation on the available fault modes to determine the energy collection points comprises: Extracting the energy distribution of the harmonic current from the available fault modes; Determining a resonance region based on the energy distribution; Performing energy density evaluation on the resonance region to generate a high-density collection area; Determining the energy collection points based on the high-density collection area.

4. The intelligent control method of the motor driver according to claim 1, characterized in that, The feature decomposition on the acoustic feature spectrum to obtain the efficiency indicator sound comprises: Performing frequency component separation on the acoustic feature spectrum to extract efficiency-related components; Performing feature screening processing on the efficiency-related components to generate a dominant frequency set; Performing correlation analysis on the dominant frequency set and the harmonic current to obtain efficiency characteristic values; Extracting the efficiency indicator sound according to the efficiency characteristic values.

5. The intelligent control method of the motor driver according to claim 1, characterized in that, The frequency shift operation on the high-energy frequency bands to generate safety frequency band signals comprises: Performing frequency spectrum window division on the high-energy frequency bands to generate a frequency spectrum window set; Performing frequency shift path planning based on the frequency spectrum window set to obtain a target frequency band; Performing carrier modulation processing on the target frequency band to generate a shift carrier; Performing energy transfer processing using the shift carrier to generate safety frequency band signals.

6. The intelligent control method of the motor driver according to claim 1, wherein The parallel control flow is arranged in time staggered sequence to generate an interleaved control sequence, including: A phase shift reference is established by switch time analysis of the parallel control flow; A phase shift time is determined by delay configuration processing using the phase shift reference; The interleaved control sequence is constructed using the phase shift time.

7. The intelligent control method of the motor driver according to claim 1, wherein The energy-saving control matrix is generated based on the interleaved control sequence, including: A switch loss concentration area is identified by the interleaved control sequence; Energy distribution analysis is performed on the switch loss concentration area to form a loss distribution map; An optimization path is determined according to the loss distribution map; An energy-saving control matrix is formed through the optimization path.

8. The intelligent control method of the motor driver according to claim 1, characterized by, The PWM control signal is reconstructed to obtain a softened edge waveform, including: The transient voltage rate of change is obtained by time domain expansion of the PWM control signal; EMI excitation points are identified by resonance peak detection of the transient voltage rate of change; An anti-resonance injection sequence is constructed based on the EMI excitation points; The transient voltage rate of change is compensated and modulated using the anti-resonance injection sequence to obtain a softened edge waveform.

9. The intelligent control method of the motor driver according to claim 3, wherein, The resonance region is determined based on the energy distribution, including: The search accuracy is determined based on the energy distribution evaluation positioning complexity, including the number of energy peaks, distribution dispersion, and harmonic order; The region division parameters are set according to the search accuracy; The energy distribution is spatially mapped using the region division parameters to generate a resonance region.

10. An intelligent control system for a motor drive, characterized by, It includes: A signal acquisition module for acquiring multi-source abnormal signals during motor operation, including harmonic current and sound pressure fluctuation data, and converting the multi-source abnormal signals into a fault feature library; An energy recovery module for identifying available fault modes based on the fault feature library, performing energy recovery evaluation on the available fault modes to determine energy collection points, performing harmonic phase shift processing on the energy collection points to generate auxiliary drive signals, and extracting the amplitude-frequency characteristics of the auxiliary drive signals to determine the enhanced control domain; An acoustic optimization module for generating an acoustic feature spectrum by frequency spectrum analysis of the sound pressure fluctuation data, obtaining an efficiency indicator sound by feature decomposition of the acoustic feature spectrum, mapping the efficiency indicator sound to a torque optimization parameter, and establishing a noise reduction control table by adjusting the electromagnetic excitation frequency based on the torque optimization parameter; A frequency management module for performing frequency spectrum analysis on the enhanced control domain to identify high-energy frequency bands, performing frequency shift operation on the high-energy frequency bands to generate a safe frequency band signal, performing energy aggregation processing on the safe frequency band signal to form a frequency shift control sequence, and constructing a multi-frequency band management strategy based on the fusion of the frequency shift control sequence and the noise reduction control table; A timing control module for decomposing the multi-frequency band management strategy into parallel control flow, arranging the parallel control flow in time staggered sequence to generate an interleaved control sequence, and generating an energy-saving control matrix based on the interleaved control sequence; An output shaping module is configured to generate a PWM control signal based on the energy-saving control matrix, perform edge reconstruction on the PWM control signal to obtain a softened edge waveform, and output the softened edge waveform to a power device to complete intelligent control of a motor driver.

Images