Direct current brushless motor control method and system based on artificial intelligence, and medium

Abstract

The invention relates to the technical field of data processing, and discloses a direct current brushless motor control method and system based on artificial intelligence, and a medium. The method comprises the following steps: acquiring a motor operation signal through a multi-node sensor, and filtering and smoothing to obtain an initial feature data stream; carrying out modeling analysis by using a time sequence and state feature extraction network, and fusing to generate a state feature vector; establishing a multi-dimensional evaluation index calculation control parameter sequence; constructing a quality evaluation system to form an operation report; performing dynamic weight calibration on the feature vectors to generate an adaptive mapping model; and calculating an adjustment strategy of the next period in combination with real-time data. According to the invention, by establishing a multi-dimensional feature extraction and analysis model, collaborative optimization of three control links of current, voltage and position is realized, and meanwhile, a complete operation quality evaluation system is established, so that the intelligent level of motor control is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a direct current brushless motor control method, system and medium based on artificial intelligence. Background Art

[0002] Brushless DC motors are widely used in industrial automation, robotics, new energy vehicles and other fields due to their simple structure, high reliability and high efficiency. Traditional brushless DC motor control methods mainly include current control, voltage control and position control. Current control uses a PI regulator to perform closed-loop control of the current, voltage control uses PWM modulation technology to achieve precise voltage regulation, and position control uses encoder feedback signals to achieve precise control of the rotor position. With the development of artificial intelligence technology, intelligent control strategies based on deep learning, reinforcement learning and other methods have gradually been applied to the field of motor control. By learning and analyzing a large amount of operating data, a more intelligent and adaptive control model is established.

[0003] However, the existing brushless DC motor control methods have the following shortcomings: First, traditional control methods often design the three control links of current, voltage and position independently, ignoring the coupling relationship between them, resulting in less than ideal control performance; second, although artificial intelligence technology has been introduced, most methods simply apply algorithms such as neural networks to a single control link, lacking systematic modeling and analysis of the overall operating status of the motor; finally, the existing control methods have a relatively single assessment of the motor operation quality, and have not established a complete multi-dimensional evaluation index system, making it difficult to achieve comprehensive performance optimization. Summary of the invention

[0004] The present application provides a direct current brushless motor control method, system and medium based on artificial intelligence, which are used to achieve coordinated optimization of the three control links of current, voltage and position by establishing a multi-dimensional feature extraction and analysis model, and at the same time establish a complete operation quality evaluation system to improve the intelligence level of motor control.

[0005] In a first aspect, the present application provides a direct current brushless motor control method based on artificial intelligence, and the direct current brushless motor control method based on artificial intelligence includes: collecting three-phase current signals, three-phase voltage signals and rotor position signals when the motor is running through a multi-node sensor, performing active filtering processing on the three-phase current signals and the three-phase voltage signals, smoothing the rotor position signals, and fusing the processed signals to obtain an initial feature data stream; based on the initial feature data stream, modeling the current and voltage timing relationship through a timing feature extraction network, analyzing the rotor position features through a state feature extraction network, fusing and mapping the feature extraction results, and obtaining a motor state feature vector; According to the motor state characteristic vector, a multidimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters is established, and a real-time control parameter sequence is generated through multi-objective optimization calculation; using the initial characteristic data stream and the real-time control parameter sequence, an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index is constructed to form a motor operation quality analysis report; based on the motor operation quality analysis report, the motor state characteristic vector is dynamically weighted and an adaptive characteristic mapping model is generated; according to the adaptive characteristic mapping model, combined with the real-time data in the initial characteristic data stream, the motor regulation strategy for the next control cycle is calculated and output.

[0006] In a second aspect, the present application provides a brushless DC motor control system based on artificial intelligence, the brushless DC motor control system based on artificial intelligence comprising:

[0007] The acquisition module is used to collect the three-phase current signal, three-phase voltage signal and rotor position signal of the motor when it is running through multi-node sensors, perform active filtering on the three-phase current signal and the three-phase voltage signal, perform smoothing on the rotor position signal, and obtain the initial characteristic data stream by fusing the processed signals;

[0008] A modeling module is used to model the current and voltage timing relationship through a timing feature extraction network according to the initial feature data stream, analyze the rotor position characteristics through a state feature extraction network, and fuse and map the feature extraction results to obtain a motor state feature vector;

[0009] A generation module, used to establish a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters according to the motor state characteristic vector, and generate a real-time control parameter sequence through multi-objective optimization calculation;

[0010] An evaluation module, used to construct an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index by using the initial characteristic data stream and the real-time control parameter sequence, and form a motor operation quality analysis report;

[0011] A calibration module, used for dynamically calibrating the motor state feature vector based on the motor operation quality analysis report to generate an adaptive feature mapping model;

[0012] The output module is used to calculate and output the motor regulation strategy for the next control cycle according to the adaptive feature mapping model in combination with the real-time data in the initial feature data stream.

[0013] A third aspect of the present application provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned artificial intelligence-based DC brushless motor control method.

[0014] In the technical solution provided by the present application, a multi-node sensor arrangement scheme is adopted to realize the accurate acquisition of three-phase current signals, three-phase voltage signals and rotor position signals, and the interference and noise in the signals are eliminated by active filtering and smoothing processing, thereby ensuring the data quality of subsequent feature extraction; by establishing a timing feature extraction network and a state feature extraction network, the timing relationship between current and voltage and the rotor position characteristics are deeply analyzed respectively, and the feature extraction results are fused and mapped to obtain a feature vector that fully reflects the operating state of the motor; based on the motor state feature vector, a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters is established, and a real-time control parameter sequence is generated by multi-objective optimization calculation, thereby realizing precise control of the motor operating state; by constructing an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index, a motor operation quality analysis report is formed, which provides a reliable basis for the optimization of the control strategy; the motor state feature vector is dynamically weighted by using an adaptive feature mapping model, thereby improving the control system's adaptability to changes in operating conditions; finally, the motor adjustment strategy for the next control cycle is output according to the adaptive feature mapping model and real-time data calculation, thereby realizing intelligent and precise motor control. This multi-level and multi-dimensional control method not only improves the operating stability and control accuracy of the motor, but also enhances the system's adaptability to external interference and changes in operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0016] Figure 1 A schematic diagram of an embodiment of a DC brushless motor control method based on artificial intelligence in an embodiment of the present application;

[0017] Figure 2 This is a schematic diagram of the arrangement structure of the motor sensor in the embodiment of the present application;

[0018] Figure 3 This is a schematic diagram of an embodiment of a brushless DC motor control system based on artificial intelligence in an embodiment of the present application. DETAILED DESCRIPTION

[0019] Embodiments of the present application provide a method, system and medium for controlling a brushless DC motor based on artificial intelligence. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the DC brushless motor control method based on artificial intelligence includes:

[0021] Step S101, collecting three-phase current signals, three-phase voltage signals and rotor position signals when the motor is running through a multi-node sensor, performing active filtering on the three-phase current signals and three-phase voltage signals, performing smoothing on the rotor position signals, and fusing the processed signals to obtain an initial feature data stream;

[0022] Step S102: Based on the initial feature data stream, the current and voltage timing relationship is modeled through a timing feature extraction network, the rotor position feature is analyzed through a state feature extraction network, and the feature extraction results are fused and mapped to obtain a motor state feature vector;

[0023] Step S103: establishing a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters according to the motor state characteristic vector, and generating a real-time control parameter sequence through multi-objective optimization calculation;

[0024] Step S104: using the initial characteristic data stream and the real-time control parameter sequence, construct an evaluation system for the current harmonic distortion rate index, the voltage fluctuation rate index and the rotor position deviation rate index, and form a motor operation quality analysis report;

[0025] Step S105: Based on the motor operation quality analysis report, dynamically calibrate the motor state feature vector to generate an adaptive feature mapping model;

[0026] Step S106: Calculate and output the motor adjustment strategy for the next control cycle based on the adaptive feature mapping model and the real-time data in the initial feature data stream.

[0027] It is understandable that the execution subject of the present application can be a DC brushless motor control system based on artificial intelligence, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0028] Specifically, high-precision Hall current sensors, voltage sensors and photoelectric encoders are arranged at the stator winding ends, rotor shaft ends and housing surfaces of the motor to collect three-phase current signals, three-phase voltage signals and rotor position signals during the operation of the motor. The collected three-phase current signals and three-phase voltage signals are processed by Butterworth and Chebyshev active filters respectively. The Butterworth filter filters out the high-frequency noise in the current signal, and the Chebyshev filter eliminates the power frequency interference components in the voltage signal. At the same time, the rotor position signal is eliminated by Gaussian smoothing algorithm to eliminate the jump and jitter phenomenon in the sampling process. The three processed signals are aligned according to the timestamp, and the data is resampled by Shannon sampling theorem to unify the sampling rate to the standard frequency of the motor control cycle, and finally a standardized initial feature data stream is obtained. According to the obtained initial feature data stream, a time series feature extraction network and a state feature extraction network are established respectively. The time series feature extraction network mainly models the time series relationship of the current and voltage signals, and calculates the statistical characteristics of each data segment by segmenting the signal into time windows, including mean, variance, peak factor and waveform factor. The state feature extraction network analyzes the rotor position features, calculates the rotor's angular velocity and angular acceleration, and performs wavelet decomposition to obtain position feature coefficients at different scales. The feature extraction results of the two networks are reduced in dimension through principal component analysis, and then transformed through nonlinear mapping to obtain the state feature vector that reflects the overall operating state of the motor.

[0029] Based on the motor state characteristic vector, a multi-dimensional evaluation index system is constructed. The dynamic response parameters are obtained by calculating the change rate and change trend of the current, voltage, and position signals; the energy efficiency parameters are obtained by calculating the input power through the current and voltage data, and the output mechanical power is calculated by combining the rotor speed data; the steady-state parameters are characterized by the mean and variance of the signal in a fixed time window. Based on these parameters, an optimization objective function is established to solve the real-time control parameter sequence containing the speed control quantity, current control quantity, and torque control quantity. By comparing the actual operation data in the initial characteristic data stream with the ideal control value in the real-time control parameter sequence, a motor operation quality evaluation system is established. The current harmonic distortion rate index is obtained by decomposing the current waveform through Fourier transform and calculating the ratio of the fundamental wave to the higher harmonics; the voltage fluctuation rate index is obtained by calculating the deviation degree between the instantaneous value of the voltage and the average value; the rotor position deviation rate index is calculated by comparing the actual position with the command position. These indicators are combined to form a motor operation quality analysis report.

[0030] Based on the operation quality analysis report, the weights of different characteristic components in the motor state characteristic vector are adjusted. The current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index are numerically analyzed to determine the weight coefficients of the corresponding characteristic components. The weight coefficients are weighted and combined with the state characteristic vector. After standardization, the feature space mapping relationship is constructed to form an adaptive feature mapping model.

[0031] According to the adaptive feature mapping model, combined with the real-time data in the initial feature data stream, the motor regulation strategy for the next control cycle is generated. Specifically, it includes extracting the signal features of the current moment from the real-time data, calculating the state mapping value through the mapping model, and predicting the operating state of the next cycle. The predicted state is divided into three control intervals: current, voltage, and position, and each interval is further subdivided into different control segments. The current control interval is divided into overcurrent control segment, rated control segment, and light load control segment. The voltage control interval is divided into overvoltage control segment, voltage stabilization control segment, and undervoltage control segment. The position control interval is divided into starting acceleration segment, uniform speed operation segment, and deceleration braking segment. By calculating the adjustment coefficient of each control segment and the transition coefficient of the adjacent segment, a complete motor regulation strategy is finally formed.

[0032] For example, when the motor starts, the initial feature data stream records the process of the three-phase current gradually rising from zero to the rated value, the three-phase voltage is also adjusted from zero to a stable operating voltage, and the rotor position accelerates from rest. The current rise curve characteristics and voltage regulation characteristics are obtained through timing feature extraction, and the rotor acceleration process characteristics are obtained through state feature extraction. After fusion mapping, these features reflect the starting characteristics of the motor, and then generate targeted control strategies to enable the motor to start smoothly and reach a stable operating state. During operation, if a sudden load change is detected that causes current fluctuations, the control strategy will adjust the parameters of the current control section in time to ensure stable motor operation.

[0033] In the embodiment of the present application, a multi-node sensor arrangement scheme is adopted to realize the accurate acquisition of three-phase current signals, three-phase voltage signals and rotor position signals, and the interference and noise in the signal are eliminated by active filtering and smoothing processing, thereby ensuring the data quality of subsequent feature extraction; by establishing a timing feature extraction network and a state feature extraction network, the timing relationship of current and voltage and the rotor position characteristics are deeply analyzed respectively, and the feature extraction results are fused and mapped to obtain a feature vector that fully reflects the operating state of the motor; based on the motor state feature vector, a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters is established, and a real-time control parameter sequence is generated by multi-objective optimization calculation, thereby realizing precise control of the motor operating state; by constructing an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index, a motor operation quality analysis report is formed, which provides a reliable basis for the optimization of the control strategy; the motor state feature vector is dynamically weighted by using an adaptive feature mapping model, thereby improving the control system's adaptability to changes in operating conditions; finally, the motor adjustment strategy for the next control cycle is output according to the adaptive feature mapping model and real-time data calculation, thereby realizing intelligent and precise motor control. This multi-level and multi-dimensional control method not only improves the operating stability and control accuracy of the motor, but also enhances the system's adaptability to external interference and changes in operating conditions.

[0034] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0035] (1) Through distributed node configuration, measurement sensors are arranged on the stator winding, rotor shaft end and casing surface of the motor to collect three-phase current signals, three-phase voltage signals and rotor position signals respectively;

[0036] (2) The three-phase current signals are grouped according to the phase relationship, and the high-frequency noise of each phase current signal is filtered through a Butterworth active filter to obtain filtered current data;

[0037] (3) The three-phase voltage signals are grouped according to the phase relationship, and the power frequency interference of each phase voltage signal is eliminated through the Chebyshev active filter to obtain the filtered voltage data;

[0038] (4) Performing Gaussian smoothing on the rotor position signal to eliminate jumps and jitters during the sampling process and obtain smoothed position data;

[0039] (5) aligning the filtered current data, filtered voltage data, and smoothed position data according to timestamps, and eliminating sampling delays through a data synchronization algorithm;

[0040] (6) Resample the synchronized data based on Shannon sampling theorem and unify the sampling rates of the three signals to the standard frequency of the motor control cycle;

[0041] (7) Divide the resampled data into time windows and calculate statistical characteristics of the data in each time window, including mean, variance, peak factor and waveform factor;

[0042] (8) Combine the statistical features with the original sampling data to construct the initial feature data stream.

[0043] Specifically, a sensor network is arranged at key positions of the motor. Hall current sensors are arranged on the surface of the motor stator winding, one at the beginning and end of each phase winding, to collect three-phase current signals; a high-precision voltage sensor is arranged at the input end of the motor stator winding to collect three-phase voltage signals; an incremental photoelectric encoder is installed at the end of the motor rotor shaft to collect rotor position signals. The sampling frequency of the sensor needs to be at least 10 times higher than the motor control frequency to ensure the integrity of the signal sampling. Figure 2 As shown, it is a schematic diagram of the motor sensor arrangement structure in an embodiment of the present application. Figure 2 The schematic diagram of the distributed arrangement structure of sensors in the motor condition monitoring system is shown. The figure shows the main structure of the motor stator winding, rotor shaft and housing in the form of a top-down section, as well as the specific installation positions of various sensors. The motor housing is outside the motor body structure, and a rectangular frame is used to represent the rotor shaft in the center, which is represented by a horizontal solid line; the stator winding is located inside the housing, which is indicated by an arc line; Hall current sensor, a Hall current sensor is arranged at the beginning and end of the three-phase stator winding (A, B, C phase); the head end sensor is marked as IA1, IB1, IC1; the end sensor is marked as IA2, IB2, IC2; a total of 6 Hall current sensors are used to collect three-phase current signals; voltage sensor, three high-precision voltage sensors are arranged at the input end of the stator winding; they are marked as VA, VB, VC respectively; used to collect three-phase voltage signals. The photoelectric encoder is installed at the end of the motor rotor shaft; it is represented by a circular symbol; it is used to collect rotor position signals. The collected three-phase current signals are grouped according to a phase difference of 120 degrees to form three-phase current signal groups A, B, and C. The Butterworth active filter has a good filtering effect on high-frequency noise and a flat phase characteristic. A 4th-order Butterworth filter is performed on each phase current signal, and the cutoff frequency is set to 5 times the motor control frequency to filter out high-frequency noise and sudden interference during the sampling process. The filtered current data maintains the amplitude and phase information of the current signal, while removing irrelevant high-frequency components.

[0044] The three-phase voltage signals are also grouped according to the 120-degree phase difference and processed by Chebyshev active filter. The Chebyshev filter has equal ripple characteristics in the passband and has a significant effect on suppressing power frequency interference. A 6th-order Chebyshev filter is used for each phase voltage signal, with the passband ripple set to 0.5dB and the stopband attenuation set to 40dB, which effectively eliminates the interference of power frequency and its harmonics. The filtered voltage data retains the basic characteristics of the voltage signal and has a high signal-to-noise ratio. After the rotor position signal is collected by the photoelectric encoder, there are often jumps and jitters caused by mechanical vibration, electromagnetic interference and other factors. The Gaussian smoothing method is used to select appropriate window width and standard deviation parameters to smooth the position signal. The window width depends on the sampling frequency and motor speed, and the standard deviation is adjusted according to the noise level of the signal. The smoothed position data can accurately reflect the actual position change of the rotor.

[0045] When aligning the three processed signals according to the timestamps, the differences in the sensor sampling times need to be considered. The data at different times are synchronized by interpolation methods, and the signal values ​​at the same time point are calculated using methods such as linear interpolation or cubic spline interpolation. The data synchronization algorithm establishes a unified time base by comparing the timestamps of each signal to eliminate the sampling delay caused by factors such as sensor response time and signal transmission delay. Based on Shannon's sampling theorem, the sampling frequency of the signal must be greater than twice the highest frequency of the signal in order to reconstruct the original signal without distortion. Considering the dynamic response characteristics of the motor control system, the sampling rates of the three signals are uniformly adjusted to integer multiples of the standard frequency of the motor control cycle. Signals with too high sampling rates are downsampled, and signals with insufficient sampling rates are upsampled by interpolation methods.

[0046] The resampled data is segmented into time windows, and the window length is set to an integer multiple of the motor control cycle. The statistical characteristics of the data are calculated in each time window. The mean reflects the average level of the signal, the variance characterizes the degree of fluctuation of the signal, the peak factor represents the instantaneous impact characteristics of the signal, and the waveform factor reflects the waveform characteristics of the signal. These statistical characteristics comprehensively describe the operating state of the motor in the time window. The calculated statistical features are combined with the original sampled data of the corresponding time window to construct an initial feature data stream containing complete information. The original sampled data retains the detailed change information of the signal, and the statistical features provide an overall feature description of the signal. The combination of the two forms a comprehensive feature expression.

[0047] For example, when the motor starts from a standstill, the three-phase current signal will show a rapid increase in amplitude, accompanied by high-frequency oscillation and noise. After being processed by the Butterworth filter, the basic trend of the current rise is retained and the high-frequency noise component is filtered out. The three-phase voltage signal is also affected by grid fluctuations and switching noise during the startup process. After being processed by the Chebyshev filter, a smooth voltage change curve is obtained. Due to the large mechanical vibration at the beginning of the startup, the rotor position signal is processed by Gaussian smoothing to obtain a smooth position change trajectory. These processed signals are time-synchronized and resampled to reflect the dynamic characteristics of the motor startup process on a unified time base. By calculating the statistical characteristics by time window segmentation, the change law of the current, voltage, and position signals during the startup process is clearly described.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] (1) Slice the initial feature data stream according to the sampling time to extract the time series segments of the three-phase current signal and the three-phase voltage signal;

[0050] (2) Perform Fourier transform on the time series segments to obtain the frequency domain characteristics of the current and voltage signals and extract the spectrum amplitude and phase information;

[0051] (3) Convert the frequency domain features into time-frequency domain representation and calculate the correlation coefficient matrix and cross-correlation function of the current and voltage signals;

[0052] (4) Separating the rotor position signal from the initial feature data stream and calculating the rotor angular velocity and angular acceleration time series data;

[0053] (5) Perform wavelet decomposition on the angular velocity and angular acceleration time series data to extract position characteristic coefficients at different scales;

[0054] (6) Combining the correlation coefficient matrix, the cross-correlation function and the position characteristic coefficient to construct a spatiotemporal characteristic matrix;

[0055] (7) Reduce the dimension of the spatiotemporal feature matrix through principal component analysis to extract the main feature components;

[0056] (8) The main characteristic components are transformed by nonlinear mapping to obtain the motor state characteristic vector.

[0057] Specifically, the initial feature data stream is sliced ​​at fixed time intervals, and the time interval is set to an integer multiple of the motor control cycle. Each time segment contains a complete three-phase current signal and a three-phase voltage signal waveform. The data in each time segment maintains the continuity and integrity of the signal, which is convenient for subsequent frequency domain analysis. The fast Fourier transform algorithm is used to perform spectrum analysis on each time series segment to convert the time domain signal to the frequency domain. The frequency domain features contain the frequency distribution information of the signal, the spectrum amplitude reflects the intensity of each frequency component, and the phase information characterizes the phase relationship between different frequency components. By analyzing the spectrum characteristics of the current and voltage signals, the fundamental component and each harmonic component are separated.

[0058] The obtained frequency domain features are converted into time-frequency domain representation through short-time Fourier transform, and the distribution diagram of the signal on the time-frequency plane is established. Based on the time-frequency domain representation, the correlation coefficient matrix of the current and voltage signals is calculated, and the matrix elements represent the degree of correlation between signals of different phases. At the same time, the cross-correlation function is calculated to describe the time correlation between the signals and reflect the phase relationship of the current and voltage waveforms. After separating the rotor position signal from the initial feature data stream, the angular velocity and angular acceleration data are obtained by differential calculation. The calculation formulas for angular velocity and angular acceleration are as follows:

[0059]

[0060] Among them, ω ( t ) is the rotor angular speed (rpm), Δθ ( t ) is the position difference between adjacent sampling points (rad), Δt is the sampling time interval (s), and γ is the resolution coefficient of the photoelectric encoder.

[0061]

[0062] Among them, α ( t ) is the rotor angular acceleration (rad / s2), Δω ( t ) is the velocity difference between adjacent sampling points (rad / s), and λ is the acceleration smoothing factor.

[0063]

[0064] Among them, ξ ( t ) is the angular acceleration (rad / s3), Δα ( t ) is the acceleration difference between adjacent sampling points (rad / s2), and μ is the second-order dynamic characteristic coefficient.

[0065] The calculated angular velocity and angular acceleration time series data are subjected to multi-scale wavelet decomposition, and the db4 wavelet is selected as the basis function for three-layer decomposition. The detail coefficients and approximate coefficients of different frequency bands are obtained through wavelet decomposition, which reflect the performance of the rotor dynamic characteristics at different time scales. The correlation coefficient matrix of the current and voltage signals, the cross-correlation function and the wavelet coefficients of the position characteristics are combined to construct a complete spatiotemporal feature matrix. Each row in the matrix corresponds to a time point, and each column corresponds to a feature component, which contains complete information on the motor's operating status.

[0066] The principal component analysis method is used to reduce the dimension of the spatiotemporal feature matrix, calculate the eigenvalues ​​and eigenvectors, and select the main feature components according to the cumulative contribution rate. The main feature components retain the main information of the original data and reduce the redundancy of the data. Finally, the main feature components are transformed into a new feature space through nonlinear mapping transformation to obtain the state feature vector reflecting the overall operating state of the motor. The nonlinear mapping adopts a multi-layer perceptron structure to map the input features to the target space through the activation function.

[0067] For example, when the motor is running under rated conditions, the three-phase current and voltage signals present regular sinusoidal waveforms. Through time slicing and Fourier transform, the amplitude and phase information of the fundamental frequency and harmonic frequency are obtained. The correlation coefficient matrix shows the phase relationship between the three-phase signals, and the cross-correlation function reflects the lead-lag relationship between the current and voltage. After processing, the rotor position signal obtains stable speed and acceleration data, and wavelet decomposition further reveals the subtle changes in the rotor motion. After these features are combined and reduced in dimension, a feature vector that accurately describes the operating state of the motor is formed, which provides a basis for subsequent control decisions. When the motor load changes, the relevant components in the feature vector will change accordingly, reflecting the dynamic response process of the motor to the load change.

[0068] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0069] (1) Extract the current component, voltage component, and position component from the motor state feature vector, and calculate the change rate and change trend of each component;

[0070] (2) Taking the change rate and change trend as the basic data of dynamic response parameters, the standardized dynamic characteristics are obtained through data normalization;

[0071] (3) Perform power calculation on the current and voltage data in the motor state characteristic vector, calculate the output mechanical power in combination with the rotor speed data, and obtain the energy efficiency parameter;

[0072] (4) The mean and variance of the motor state feature vector in a fixed time window are used as steady-state operation features to construct steady-state parameters;

[0073] (5) Extract key control quantities from dynamic response parameters, energy efficiency parameters, and steady-state parameters, and establish parameter optimization objective functions;

[0074] (6) solving the parameter optimization objective function to generate a control parameter matrix including a speed control variable, a current control variable, and a torque control variable;

[0075] (7) Expand the control parameter matrix in time sequence according to the motor control cycle to form a real-time control parameter sequence.

[0076] Specifically, the data of each component is extracted from the motor state feature vector. The current component includes the amplitude and phase information of the three-phase current, the voltage component includes the amplitude and phase information of the three-phase voltage, and the position component includes the angle, speed and acceleration information of the rotor. The rate of change of each component is calculated, specifically the difference between the current sampling point and the previous sampling point divided by the sampling time interval. The change trend is obtained by linear regression analysis of the data of multiple consecutive sampling points, and the regression coefficient reflects the rising or falling trend of the signal. The calculated rate of change and change trend data are normalized to unify the data of different dimensions into the same numerical range. The normalization process uses the maximum and minimum method to linearly map the data to the [0, 1] interval. The normalized data is used as the basic data of the dynamic response parameters, reflecting the dynamic change characteristics of the motor operation state. When the current and voltage data in the motor state feature vector are used for power calculation, the instantaneous power of each phase is calculated, that is, the product of the phase voltage and the phase current. The total power of the three phases is the sum of the instantaneous power of the three phases, and then combined with the rotor speed data, the mechanical power output of the motor shaft is calculated. The ratio of input power to output power is used to obtain the energy efficiency parameter of the motor, which reflects the energy conversion efficiency of the motor.

[0077] The motor state eigenvector is segmented according to a fixed time window, and the length of the time window is set according to the dynamic response characteristics of the motor. In each time window, the mean of each component of the eigenvector is calculated, which reflects the average operating state of the motor in the time period; the calculated variance reflects the degree of fluctuation of the operating state. These statistics constitute the steady-state parameters and describe the steady-state operating characteristics of the motor. When extracting key control quantities from dynamic response parameters, energy efficiency parameters and steady-state parameters, multiple objectives of motor control need to be considered. The key control quantities include response time and overshoot reflecting dynamic performance, power factor and efficiency coefficient reflecting energy efficiency, and steady-state error and fluctuation coefficient reflecting steady-state performance. A multi-objective optimization function is established based on these key control quantities, and the various items in the function are balanced by weight coefficients.

[0078] The optimization objective function is solved and the gradient descent method is used to find the optimal solution. During the solution process, various constraints of the motor operation need to be met, including current limit, voltage limit, speed limit, etc. The optimal solution obtained by the solution forms a control parameter matrix, which contains speed control quantity, current control quantity and torque control quantity. The control parameter matrix is ​​time-sequentially expanded according to the motor control cycle to generate a continuous control parameter sequence. The continuity and smoothness of the parameters are considered during the time-sequential expansion process to avoid sudden changes in the control quantity. Each group of parameters in the control parameter sequence corresponds to a control instruction of a control cycle.

[0079] For example, when the motor is running from rated speed to lower speed, relevant components are extracted from the state eigenvector. The current component shows that the amplitude of the three-phase current gradually decreases, the voltage component shows that the three-phase voltage is adjusted accordingly, and the position component reflects the speed change process. The rate of change of these components is calculated to reflect the speed of change of each parameter during the dynamic adjustment process. The power calculation shows that the input power and output power are reduced at the same time, but a reasonable energy efficiency ratio is maintained. In each time window of the deceleration process, the mean of the state parameter shows a decreasing trend, and the variance reflects the degree of fluctuation of the transition process. Based on these parameters, the optimization objective function is established, and the control parameter sequence that can ensure smooth deceleration and maintain high efficiency is obtained after solving. Throughout the process, the generation of control parameters is always optimized around the dynamic performance, energy efficiency and steady-state characteristics of the motor operation.

[0080] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0081] (1) Extract the three-phase current signal from the initial characteristic data stream, decompose the current waveform through Fourier transform, calculate the ratio of the fundamental component to the higher harmonic component, and obtain the current harmonic distortion rate index;

[0082] (2) Extract the three-phase voltage signal from the initial characteristic data stream, calculate the deviation between the instantaneous voltage value and the average value, perform fluctuation statistical analysis, and obtain the voltage fluctuation rate index;

[0083] (3) extracting the rotor position signal from the initial characteristic data stream, comparing the signal value with the position command value in the real-time control parameter sequence, and calculating the rotor position deviation rate index;

[0084] (4) Based on the current harmonic distortion rate index, the current waveform quality is periodically evaluated and the harmonic characteristics under different working conditions are statistically analyzed;

[0085] (5) Based on the voltage fluctuation rate index and the rotor position deviation rate index, the correlation between voltage stability and position control accuracy is analyzed;

[0086] (6) The comprehensive analysis results of harmonic characteristics, voltage stability and position control accuracy are sorted and summarized to generate a motor operation quality analysis report.

[0087] Specifically, a three-phase current signal is extracted from the initial characteristic data stream, and each phase current data contains multiple motor operation cycles. The extracted current signal is subjected to a fast Fourier transform to decompose the time domain waveform into harmonic components. After decomposition, the amplitude and phase of the fundamental component, as well as the amplitude and phase of the 2nd to 13th harmonic components are obtained. By calculating the ratio of each harmonic component to the fundamental component, the total harmonic distortion rate index is accumulated. This index reflects the degree of distortion of the current waveform and the waveform quality of the motor operation. The processing process of extracting the three-phase voltage signal from the initial characteristic data stream includes: calculating the instantaneous value of each phase voltage within a sampling period, and then calculating the arithmetic mean of all instantaneous values ​​within the sampling period. The difference between each instantaneous value and the average value is statistically calculated to obtain the probability distribution of the voltage deviation. The standard deviation and fluctuation coefficient are calculated based on the deviation distribution, and the voltage fluctuation rate index is comprehensively determined.

[0088] For the processing of the rotor position signal, the calculation formula involving the position deviation rate is as follows:

[0089]

[0090] Among them, δ ( t ) is the rotor position deviation rate, β i( t ) is the actual position sequence, is the instruction position sequence, σ i( t ) is the position weight coefficient, is the dynamic compensation factor, and n is the number of sampling points.

[0091] The sliding time window method is used for the periodic evaluation of the current harmonic distortion rate index. In each window, the current waveform is harmonically analyzed, and the amplitude changes of the fundamental wave and each harmonic are recorded. The harmonic characteristics under different load conditions are classified and statistically analyzed, and the corresponding relationship between the harmonic characteristics and the operating conditions is established. The correlation analysis between the voltage fluctuation rate index and the rotor position deviation rate index is based on the correlation and causality of the data. The two indicator sequences are time-aligned, and then their correlation coefficients on different time scales are calculated. By analyzing the change law of the correlation coefficient, the influence mechanism of voltage stability on position control accuracy is revealed. The above analysis results are systematically organized to form an operation quality analysis report. The report content is carried out according to the three aspects of current waveform quality, voltage stability and position control accuracy. Each aspect contains indicator data, change trend and correlation analysis.

[0092] For example, during the motor startup phase, the three-phase current waveform will have a large harmonic content due to the switching process. The Fourier decomposition of the current waveform shows that the amplitude of the fundamental component gradually increases, accompanied by obvious higher-order harmonics. The ratio of these harmonic components to the fundamental component reflects the degree of distortion of the current waveform. The voltage signal will also fluctuate during the startup process. By calculating the deviation between the instantaneous value of the voltage and the average value, a quantitative index of voltage stability is obtained. The rotor position signal is compared with the given acceleration curve to calculate the deviation between the actual position and the ideal position. These deviation data are weighted to form an evaluation index for position control accuracy. The analysis found that the voltage fluctuation during the startup process is positively correlated with the position deviation, indicating that voltage stability has a direct impact on position control accuracy. These analysis results are organized into an operation quality analysis report in the form of data charts and correlation analysis, providing a basis for the optimization of motor control strategies.

[0093] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0094] (1) Perform numerical analysis on the current harmonic distortion rate index in the motor operation quality analysis report to determine the weight coefficient of the current characteristic component;

[0095] (2) Perform numerical analysis on the voltage fluctuation rate index in the motor operation quality analysis report to determine the weight coefficient of the voltage characteristic component;

[0096] (3) Perform numerical analysis on the rotor position deviation rate index in the motor operation quality analysis report to determine the weight coefficient of the position characteristic component;

[0097] (4) performing a weighted combination of the current characteristic component weight coefficient, the voltage characteristic component weight coefficient, the position characteristic component weight coefficient and the motor state characteristic vector;

[0098] (5) Standardizing the weighted combined feature data and constructing a feature space mapping relationship;

[0099] (6) Convert the feature space mapping relationship into a mathematical expression to form an adaptive feature mapping model.

[0100] Specifically, the current harmonic distortion rate index in the motor operation quality analysis report is numerically analyzed. The data characteristics of the current harmonic distortion rate index are analyzed, including the amplitude ratio, phase difference and waveform distortion of the fundamental wave and each harmonic component. For each harmonic order, its contribution to the total distortion rate is calculated, and the harmonic component with a greater contribution degree obtains a higher weight. At the same time, considering the impact of harmonics on motor operation, the weight coefficient of low-order harmonics is correspondingly increased due to their large amplitude, while high-order harmonics have a small amplitude but a high frequency, and their impact on the motor cannot be ignored, and their weight coefficients also need to be reasonably configured. The numerical analysis process of the voltage fluctuation rate index includes statistics on the distribution characteristics of voltage deviation. The voltage deviation is graded according to the amplitude, and the occurrence frequency and duration of deviations of different levels are calculated. A higher weight coefficient is assigned to voltage deviations with large amplitudes and long durations, while a lower weight coefficient is assigned to instantaneous small deviations. At the same time, considering the impact of voltage fluctuations on motor performance, the weight coefficient of voltage deviation modes that may cause motor torque pulsation or speed fluctuations needs to be increased accordingly.

[0101] The numerical analysis of the rotor position deviation rate index focuses on the performance of position control accuracy under different operating conditions. The position deviation data is classified according to the operating conditions, and the average deviation, maximum deviation and deviation change rate under different conditions are calculated respectively. For conditions with high dynamic response requirements, such as starting and braking processes, the weight coefficient of position control accuracy is increased accordingly. For steady-state operating conditions, the position holding accuracy is focused on and the weight coefficient is adjusted accordingly. When the three types of weight coefficients are weighted and combined with the motor state characteristic vector, the corresponding relationship between each characteristic component in the vector is determined. The current characteristic component corresponds to the amplitude and phase information of the three-phase current, the voltage characteristic component corresponds to the amplitude and phase information of the three-phase voltage, and the position characteristic component corresponds to the angle, speed and acceleration information of the rotor. Each characteristic component is weighted according to the weight coefficient to obtain weighted characteristic data reflecting the importance of each physical quantity.

[0102] The weighted feature data are standardized to eliminate the dimensional differences between different physical quantities. The standardization process uses the Z-score method to calculate the mean and standard deviation of each feature component, and convert the original data to a standard normal distribution space with a mean of 0 and a standard deviation of 1. The standardized feature data have the same numerical scale, which is convenient for subsequent feature space mapping. The establishment of the feature space mapping relationship is based on the standardized feature data to construct a multidimensional feature space, and each dimension in the space corresponds to a standardized feature component. By analyzing the correlation and coupling relationship between the feature components, the mapping function between the dimensions in the feature space is established. These mapping functions comprehensively reflect all aspects of the motor operating state and form an adaptive feature mapping model.

[0103] For example, under the condition of sudden load change of the motor, the current harmonic distortion rate index shows obvious changes. Analysis shows that the sudden load change leads to a significant increase in the content of low-order harmonics, especially the amplitude of the 5th and 7th harmonics has increased significantly. According to the degree of influence of harmonics on the motor, higher weight coefficients are given to the harmonic components of these orders. At the same time, the voltage fluctuation rate index has also changed accordingly, which is manifested as a large fluctuation in the instantaneous value of the voltage. This voltage fluctuation is closely related to the load change, and its weight coefficient is increased accordingly. The rotor position deviation increases briefly when the load changes suddenly, but then recovers quickly, indicating that the position control has good anti-disturbance ability, which is also reflected in the weight coefficient of the position characteristic component. These weight coefficients are combined with the state characteristic vector, and after standardization, a mapping model that accurately reflects the characteristics of the load mutation condition is obtained.

[0104] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0105] (1) extracting the three-phase current signal, three-phase voltage signal and rotor position signal at the current moment from the initial feature data stream as real-time data;

[0106] (2) Bring the real-time data into the adaptive feature mapping model and calculate the state mapping value at the current moment;

[0107] (3) Perform time series analysis on the state mapping value to predict the motor operation state in the next control cycle;

[0108] (4) According to the predicted operating state, the current control interval, voltage control interval and position control interval are divided to generate a partition control scheme;

[0109] (5) Perform parameter matching on the partition control scheme and calculate the adjustment coefficient within each control interval;

[0110] (6) Combine the adjustment coefficient and the control interval to form a motor adjustment strategy.

[0111] Specifically, real-time data is extracted from the initial feature data stream. During the extraction process, the latest segment of data is intercepted from the data stream according to a fixed time window length as the real-time data at the current moment. The three-phase current signal contains the instantaneous current value of each phase, the three-phase voltage signal contains the instantaneous voltage value of each phase, and the rotor position signal contains the angle information at the current moment. The extracted data needs to meet the integrity requirements, that is, it contains at least one complete electrical cycle. The extracted real-time data is input into the adaptive feature mapping model for state mapping. The model performs feature conversion and combination on the input real-time data according to the feature space mapping relationship established previously. After nonlinear mapping transformation, the state mapping value reflecting the current motor operating state is obtained. These mapping values ​​contain state characteristics of current, voltage, and position, and each feature has been weighted by weight calibration.

[0112] The obtained state mapping values ​​are analyzed in time series, and the data is modeled using the autoregressive moving average model. By analyzing the historical change trend of the state mapping values ​​and combining the current state information, the operating state of the motor in the next control cycle is predicted. The prediction results include the change trends of current, voltage, and position, providing a basis for the subsequent division of control intervals. Based on the predicted operating state, the control interval is divided. The current control interval is divided into overcurrent interval, rated interval, and light load interval according to the predicted current change trend; the voltage control interval is divided into overvoltage interval, voltage stabilization interval, and undervoltage interval according to the predicted voltage change trend; the position control interval is divided into acceleration interval, uniform speed interval, and deceleration interval according to the predicted position change trend. Each interval has its own specific control requirements and constraints. Parameters are matched for the divided control intervals, and the adjustment coefficients in each interval are calculated. In the current control interval, the overcurrent interval uses a fast-response adjustment coefficient to limit the current rise, the rated interval uses a stable adjustment coefficient to maintain stable operation, and the light load interval uses an energy-saving adjustment coefficient. In the voltage control interval, the overvoltage interval and the undervoltage interval use a compensation adjustment coefficient, and the voltage stabilization interval uses a maintenance adjustment coefficient. In the position control range, the dynamic adjustment coefficient is used in the acceleration and deceleration ranges, and the steady-state adjustment coefficient is used in the uniform speed range.

[0113] The calculated adjustment coefficient is combined with the corresponding control interval to form a complete motor adjustment strategy. The strategy includes the boundary conditions of each control interval, the switching rules of the adjustment coefficient, and the parameter smooth transition scheme. The adjustment strategy needs to ensure the continuity of the control quantity when switching between intervals to avoid sudden changes that may cause shocks to the motor.

[0114] For example, when the motor is running at rated conditions, a large load is suddenly connected. The real-time data extracted from the initial feature data stream shows that the three-phase current begins to rise and the speed decreases. After these data are processed by the feature mapping model, the state mapping value reflecting the sudden change of load is obtained. Time series analysis predicts that the current will continue to rise and the speed will further decrease. Based on the prediction results, the control interval division adjusts the current control interval to the overcurrent interval and the position control interval to the acceleration interval. In the overcurrent interval, a larger current adjustment coefficient is used to quickly respond to load changes, while a larger torque adjustment coefficient is used in the acceleration interval to maintain the speed. These adjustment coefficients are combined into a new adjustment strategy after smoothing, so that the motor can smoothly transition to a new working state. The whole process reflects the control strategy's ability to adaptively adjust to load changes.

[0115] In a specific embodiment, the process of performing the parameter matching step for the partition control scheme may specifically include the following steps:

[0116] (1) Divide the current control section into an overcurrent control section, a rated control section, and a light load control section, and calculate the current regulation coefficient of each control section;

[0117] (2) Divide the voltage control section into an overvoltage control section, a voltage stabilization control section, and an undervoltage control section, and calculate the voltage regulation coefficient of each control section;

[0118] (3) Divide the position control interval into a starting acceleration section, a uniform speed operation section, and a deceleration braking section, and calculate the position adjustment coefficient of each control section;

[0119] (4) Cross-validate the current regulation coefficient, voltage regulation coefficient, and position regulation coefficient to determine the transition coefficients of each adjacent control segment;

[0120] (5) Performing a weighted combination of the transition coefficient and the adjustment coefficient of each control segment to form an adjustment coefficient matrix;

[0121] (6) Determine the critical point parameters in the adjustment coefficient matrix through interpolation calculation to obtain the adjustment coefficient in each control interval.

[0122] Specifically, each control interval is divided in detail. The current control interval is divided into three control sections according to the current amplitude: the overcurrent control section when the current exceeds the rated value, the rated control section when the current is within the rated range, and the light load control section when the current is lower than the rated value. The corresponding adjustment coefficient is calculated for each control section according to its current characteristics. For example, a fast response adjustment coefficient is used in the overcurrent control section to limit the current rise rate, a steady-state adjustment coefficient is used in the rated control section to maintain current stability, and an energy-saving adjustment coefficient is used in the light load control section to reduce losses. The voltage control interval is divided into three control sections: overvoltage, voltage stabilization, and undervoltage according to the fluctuation of the voltage amplitude. The overvoltage control section mainly deals with the back electromotive force during motor braking, the voltage stabilization control section maintains voltage stability during normal operation, and the undervoltage control section deals with the voltage drop during startup and load surge. The voltage regulation coefficient of each control section is determined by analyzing the voltage fluctuation characteristics, and the impact of voltage changes on motor performance is taken into account.

[0123] The position control interval is divided into three control sections according to the rotor motion characteristics. The startup acceleration section handles the process of the motor from static to rated speed, the uniform speed operation section maintains a stable operating state, and the deceleration braking section controls the motor's stopping process. The position adjustment coefficient of each section needs to consider factors such as the speed change rate and position accuracy requirements. Next, the transition problem between the control sections needs to be dealt with. By cross-validating the three types of adjustment coefficients of current, voltage, and position, the mutual influence relationship between them is analyzed. For example, there is a coupling relationship between the overcurrent control section and the overvoltage control section, and the adjustment strategies of the two need to be coordinated. The transition coefficient is calculated for each pair of adjacent control sections to ensure a smooth transition of the control quantity when switching between sections.

[0124] The adjustment coefficients and transition coefficients of each control segment are weighted and combined to construct a complete adjustment coefficient matrix. Each element in the matrix corresponds to a specific operating point, which contains adjustment information in three dimensions: current, voltage, and position. The selection of weighted coefficients needs to consider the priority and mutual constraints of each control target. Finally, the parameter values ​​of each critical point in the adjustment coefficient matrix are determined by interpolation calculation. The critical point is located at the junction of adjacent control segments, and special attention should be paid to the continuity of the control quantity. The critical point parameters are calculated using the cubic spline interpolation method to ensure the smoothness of the control curve.

[0125] For example: at the beginning of starting, the motor is in the starting acceleration stage. At this time, the current regulation coefficient is large to provide sufficient starting torque, the voltage regulation coefficient is dynamically adjusted according to the change of back electromotive force, and the position regulation coefficient mainly focuses on the smoothness of the acceleration process. As the speed increases, the control section gradually transitions to the rated range, and the regulation coefficient changes accordingly. In this process, the three types of regulation coefficients are always coordinated to ensure the smoothness and reliability of the motor starting process. This is the calculation and application process of the entire regulation coefficient. The division of the control range, the calculation of the regulation coefficient, and the correlation and conversion relationship between them all need to be carefully considered and accurately calculated to obtain the best control effect.

[0126] The calculation formula for the position adjustment coefficient is as follows:

[0127]

[0128] Among them, K p ( t ) is the position adjustment coefficient, η i ( t ) is the actual position sequence, χ i ( t ) is the instruction position sequence, ρ i( t ) is the position weight factor, ν ( t ) is the dynamic compensation factor, and n is the number of sampling points.

[0129] The above describes the DC brushless motor control method based on artificial intelligence in the embodiment of the present application. The following describes the DC brushless motor control system based on artificial intelligence in the embodiment of the present application. Figure 3 In the embodiment of the present application, an embodiment of a brushless DC motor control system based on artificial intelligence includes:

[0130] The acquisition module 201 is used to collect the three-phase current signal, the three-phase voltage signal and the rotor position signal of the motor when it is running through a multi-node sensor, perform active filtering on the three-phase current signal and the three-phase voltage signal, perform smoothing on the rotor position signal, and obtain an initial feature data stream by fusing the processed signals;

[0131] A modeling module 202 is used to model the current and voltage timing relationship through a timing feature extraction network according to the initial feature data stream, analyze the rotor position feature through a state feature extraction network, and fuse and map the feature extraction results to obtain a motor state feature vector;

[0132] A generation module 203 is used to establish a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters according to the motor state characteristic vector, and generate a real-time control parameter sequence through multi-objective optimization calculation;

[0133] An evaluation module 204 is used to construct an evaluation system of a current harmonic distortion rate index, a voltage fluctuation rate index and a rotor position deviation rate index by using the initial characteristic data stream and the real-time control parameter sequence, and form a motor operation quality analysis report;

[0134] A calibration module 205, configured to perform dynamic weight calibration on the motor state feature vector based on the motor operation quality analysis report to generate an adaptive feature mapping model;

[0135] The output module 206 is used to calculate and output the motor regulation strategy for the next control cycle according to the adaptive feature mapping model in combination with the real-time data in the initial feature data stream.

[0136] Through the cooperation of the above components, a multi-node sensor layout scheme is adopted to achieve accurate acquisition of three-phase current signals, three-phase voltage signals and rotor position signals. Active filtering and smoothing processing are used to eliminate interference and noise in the signals, ensuring the data quality of subsequent feature extraction. By establishing a timing feature extraction network and a state feature extraction network, the timing relationship between current and voltage and the rotor position characteristics are deeply analyzed respectively, and the feature extraction results are fused and mapped to obtain a feature vector that fully reflects the operating state of the motor. Based on the motor state feature vector, a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters is established, and a real-time control parameter sequence is generated through multi-objective optimization calculation, realizing precise control of the motor operating state. By constructing an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index, a motor operation quality analysis report is formed, providing a reliable basis for the optimization of the control strategy. The adaptive feature mapping model is used to dynamically calibrate the motor state feature vector, thereby improving the control system's adaptability to changes in operating conditions. Finally, the motor adjustment strategy for the next control cycle is output based on the adaptive feature mapping model and real-time data calculation, realizing the intelligent and precise control of the motor. This multi-level and multi-dimensional control method not only improves the operating stability and control accuracy of the motor, but also enhances the system's adaptability to external interference and changes in operating conditions.

[0137] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the artificial intelligence-based brushless DC motor control method.

[0138] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0139] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0140] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A DC brushless motor control method based on artificial intelligence, characterized in that: The artificial intelligence-based brushless DC motor control method comprises: The three-phase current signal, three-phase voltage signal and rotor position signal of the motor are collected by multi-node sensors during operation, the three-phase current signal and the three-phase voltage signal are actively filtered, the rotor position signal is smoothed, and the initial characteristic data stream is obtained by fusing the processed signals; According to the initial feature data stream, the current and voltage timing relationship is modeled through a timing feature extraction network, the rotor position feature is analyzed through a state feature extraction network, and the feature extraction results are fused and mapped to obtain a motor state feature vector; According to the motor state characteristic vector, a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters is established, and a real-time control parameter sequence is generated through multi-objective optimization calculation; Using the initial characteristic data stream and the real-time control parameter sequence, an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index is constructed to form a motor operation quality analysis report; Based on the motor operation quality analysis report, dynamically weight calibrate the motor state feature vector to generate an adaptive feature mapping model; According to the adaptive feature mapping model, combined with the real-time data in the initial feature data stream, the motor regulation strategy for the next control cycle is calculated and output.

2. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: The method collects three-phase current signals, three-phase voltage signals and rotor position signals when the motor is running through a multi-node sensor, performs active filtering on the three-phase current signals and the three-phase voltage signals, performs smoothing on the rotor position signals, and obtains an initial feature data stream by fusing the processed signals, including: By configuring distributed nodes, measuring sensors are arranged on the stator winding, rotor shaft end and housing surface of the motor to respectively collect the three-phase current signal, the three-phase voltage signal and the rotor position signal; The three-phase current signals are grouped according to the phase relationship, and high-frequency noise is filtered on each phase current signal through a Butterworth active filter to obtain filtered current data; The three-phase voltage signals are grouped according to the phase relationship, and the power frequency interference of each phase voltage signal is eliminated by a Chebyshev active filter to obtain filtered voltage data; Performing Gaussian smoothing on the rotor position signal to eliminate jumps and jitters during the sampling process and obtain smoothed position data; Aligning the filtered current data, the filtered voltage data and the smoothed position data according to timestamps, and eliminating sampling delays through a data synchronization algorithm; The synchronized data is resampled based on Shannon sampling theorem, and the sampling rates of the three signals are unified to the standard frequency of the motor control cycle; The resampled data is segmented into time windows, and statistical features are calculated for the data in each time window, including mean, variance, peak factor, and waveform factor; The statistical features are combined with the original sampling data to construct the initial feature data stream.

3. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: According to the initial feature data stream, the current and voltage timing relationship is modeled through a timing feature extraction network, the rotor position feature is analyzed through a state feature extraction network, and the feature extraction results are fused and mapped to obtain a motor state feature vector, including: Slicing the initial characteristic data stream according to sampling time, extracting time series segments of the three-phase current signal and the three-phase voltage signal; Performing Fourier transformation on the time series segments to obtain frequency domain characteristics of current and voltage signals and extracting spectrum amplitude and phase information; Convert the frequency domain features into time-frequency domain representation, and calculate the correlation coefficient matrix and cross-correlation function of the current and voltage signals; Separating the rotor position signal from the initial characteristic data stream, and calculating rotor angular velocity and angular acceleration time series data; Performing wavelet decomposition on the angular velocity and angular acceleration time series data to extract position characteristic coefficients at different scales; Combining the correlation coefficient matrix, the cross-correlation function and the position characteristic coefficient to construct a spatiotemporal characteristic matrix; Performing dimensionality reduction processing on the spatiotemporal feature matrix through principal component analysis to extract main feature components; The main characteristic components are subjected to nonlinear mapping transformation to obtain the motor state characteristic vector.

4. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: The method of establishing a multidimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters based on the motor state characteristic vector and generating a real-time control parameter sequence through multi-objective optimization calculation includes: Extracting current component, voltage component and position component from the motor state characteristic vector respectively, and calculating the change rate and change trend of each component; The change rate and change trend are used as basic data of dynamic response parameters, and standardized dynamic characteristics are obtained through data normalization processing; Performing power calculation on the current and voltage data in the motor state characteristic vector, and calculating the output mechanical power in combination with the rotor speed data to obtain the energy efficiency parameter; Taking the mean and variance of the motor state feature vector in a fixed time window as steady-state operation features, constructing the steady-state parameters; Extract key control quantities from the dynamic response parameters, the energy efficiency parameters and the steady-state parameters, and establish a parameter optimization objective function; Solving the parameter optimization objective function to generate a control parameter matrix including a speed control variable, a current control variable and a torque control variable; The control parameter matrix is ​​expanded in time sequence according to the motor control cycle to form the real-time control parameter sequence.

5. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: The initial characteristic data stream and the real-time control parameter sequence are used to construct an evaluation system for current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index, and form a motor operation quality analysis report, including: Extracting the three-phase current signal from the initial characteristic data stream, decomposing the current waveform by Fourier transform, calculating the ratio of the fundamental component to the higher harmonic component, and obtaining the current harmonic distortion rate index; Extracting the three-phase voltage signal from the initial characteristic data stream, calculating the degree of deviation between the instantaneous voltage value and the average value, performing fluctuation statistical analysis, and obtaining the voltage fluctuation rate index; Extracting the rotor position signal from the initial characteristic data stream, comparing the signal value with the position command value in the real-time control parameter sequence, and calculating the rotor position deviation rate index; Based on the current harmonic distortion rate index, the current waveform quality is periodically evaluated, and the harmonic characteristics under different working conditions are statistically analyzed; Based on the voltage fluctuation rate index and the rotor position deviation rate index, analyzing the correlation between voltage stability and position control accuracy; The comprehensive analysis results of the harmonic characteristics, voltage stability and position control accuracy are collated and summarized to generate the motor operation quality analysis report.

6. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: The method of dynamically weighting the motor state feature vector based on the motor operation quality analysis report to generate an adaptive feature mapping model includes: Performing numerical analysis on the current harmonic distortion rate index in the motor operation quality analysis report to determine the weight coefficient of the current characteristic component; Performing numerical analysis on the voltage fluctuation rate index in the motor operation quality analysis report to determine the weight coefficient of the voltage characteristic component; Performing numerical analysis on the rotor position deviation rate index in the motor operation quality analysis report to determine the weight coefficient of the position characteristic component; Performing a weighted combination of the current characteristic component weight coefficient, the voltage characteristic component weight coefficient, the position characteristic component weight coefficient and the motor state characteristic vector; Standardize the weighted combined feature data and construct a feature space mapping relationship; The feature space mapping relationship is converted into a mathematical expression to form the adaptive feature mapping model.

7. The artificial intelligence-based DC brushless motor control method according to claim 1, characterized in that: The step of calculating and outputting a motor adjustment strategy for the next control cycle based on the adaptive feature mapping model and in combination with the real-time data in the initial feature data stream includes: Extracting the three-phase current signal, the three-phase voltage signal and the rotor position signal at the current moment from the initial characteristic data stream as real-time data; Bringing the real-time data into the adaptive feature mapping model to calculate the state mapping value at the current moment; Performing time series analysis on the state mapping value to predict the motor operation state of the next control cycle; According to the predicted operating state, the current control interval, voltage control interval and position control interval are divided to generate a partition control plan; Perform parameter matching on the partition control scheme and calculate the adjustment coefficient within each control interval; The adjustment coefficient is combined with the control interval to form the motor adjustment strategy.

8. The artificial intelligence-based DC brushless motor control method according to claim 7, characterized in that: The performing parameter matching on the partition control scheme and calculating the adjustment coefficient within each control interval includes: The current control section is divided into an overcurrent control section, a rated control section and a light load control section, and a current regulation coefficient of each control section is calculated; Dividing the voltage control interval into an overvoltage control section, a voltage stabilization control section and an undervoltage control section, and calculating a voltage regulation coefficient of each control section; The position control interval is divided into a start-up acceleration section, a uniform speed operation section and a deceleration braking section, and the position adjustment coefficient of each control section is calculated; Cross-validating the current adjustment coefficient, the voltage adjustment coefficient, and the position adjustment coefficient to determine transition coefficients of adjacent control segments; The transition coefficient is weightedly combined with the adjustment coefficient of each control segment to form an adjustment coefficient matrix; The critical point parameters in the adjustment coefficient matrix are determined by interpolation calculation to obtain the adjustment coefficients in each control interval.

9. A brushless DC motor control system based on artificial intelligence, used to implement the brushless DC motor control method based on artificial intelligence as described in any one of claims 1 to 8, characterized in that: The artificial intelligence-based brushless DC motor control system includes: The acquisition module is used to collect the three-phase current signal, three-phase voltage signal and rotor position signal of the motor when it is running through multi-node sensors, perform active filtering on the three-phase current signal and the three-phase voltage signal, perform smoothing on the rotor position signal, and obtain the initial characteristic data stream by fusing the processed signals; A modeling module is used to model the current and voltage timing relationship through a timing feature extraction network according to the initial feature data stream, analyze the rotor position characteristics through a state feature extraction network, and fuse and map the feature extraction results to obtain a motor state feature vector; A generation module, used to establish a multi-dimensional evaluation index including dynamic response parameters, energy efficiency parameters and steady-state parameters according to the motor state characteristic vector, and generate a real-time control parameter sequence through multi-objective optimization calculation; An evaluation module, used to construct an evaluation system of current harmonic distortion rate index, voltage fluctuation rate index and rotor position deviation rate index by using the initial characteristic data stream and the real-time control parameter sequence, and form a motor operation quality analysis report; A calibration module, used for dynamically calibrating the motor state feature vector based on the motor operation quality analysis report to generate an adaptive feature mapping model; The output module is used to calculate and output the motor regulation strategy for the next control cycle according to the adaptive feature mapping model in combination with the real-time data in the initial feature data stream.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the artificial intelligence-based brushless DC motor control method as described in any one of claims 1 to 8 is implemented.

Images