Sensorless motor speed measurement method and system based on back electromotive force harmonic analysis

Abstract

The application relates to the technical field of motor rotating speed measurement, and particularly discloses a sensorless motor rotating speed measurement method and system based on back electromotive force harmonic analysis, which comprises the following steps: a signal acquisition step, that is, synchronously acquiring motor stator three-phase current signals and motor shell radial vibration acceleration signals, performing coordinate transformation on the motor stator three-phase current signals to obtain stator current signals in a stationary coordinate system. In the actual data processing process of motor operation, the technical scheme provided by the application synchronously acquires stator current and motor shell vibration signals and independently performs notch processing, so that electromagnetic and mechanical double characteristic signals in the same frequency dimension are obtained. In the phase-locked operation stage, a pre-labeled steady-state phase difference reference is introduced into the system, the instantaneous phase error of two signals is cross-constrained, and the phase judgment weight is dynamically adjusted according to the signal-to-noise ratio state of the two characteristic signals, so that the total composite phase error is synthesized.

Description

Technical Field

[0001] This invention relates to the field of motor speed measurement technology, specifically to a sensorless motor speed measurement method and system based on back EMF harmonic analysis. Background Technology

[0002] Sensorless motor speed measurement typically relies on analyzing the back EMF harmonic signal in the stator current to estimate the mechanical rotor speed. In industrial drive equipment operation, the control system tracks the frequency and phase of the harmonic signal through a phase-locked loop (PLL) to provide closed-loop speed feedback data. Existing conventional measurement methods mostly use a single stator current as the data source and directly input the extracted electrical signal characteristic frequency into an integral controller for speed calculation.

[0003] When a motor operates under varying loads or in applications with electromagnetic interference, the characteristic harmonic signal-to-noise ratio of a single current signal will fluctuate, making the phase-locked loop (PLL) prone to phase tracking deviation. In some existing measurement schemes that introduce vibration signals as redundant data sources, the results are typically calculated separately from the two signals and then weighted, without addressing the inherent physical phase difference between the electromagnetic and mechanical responses.

[0004] When the quality of a certain signal deteriorates, this information fusion method lacks a constraint mechanism based on phase correlation to intercept interference data, which can easily lead to the direct transmission of errors. This can cause the final calculated motor speed to deviate, affecting the continuity and stability of the measurement status. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a sensorless motor speed measurement method and system based on back EMF harmonic analysis, in order to improve the speed measurement quality of motors in physical application environments with interference.

[0006] To achieve the above objectives, a first aspect of the present invention proposes a sensorless motor speed measurement method based on back EMF harmonic analysis, comprising:

[0007] Signal acquisition steps: Synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the motor housing, and perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system;

[0008] Feature purification steps: notch filtering is performed with the motor fundamental feedforward angular frequency as the center frequency to remove the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal; notch filtering is performed with the mechanical speed feedforward frequency and its harmonics as the center frequency to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing to obtain the vibration speed characteristic signal with the same frequency as the current speed characteristic harmonic signal;

[0009] Fault-tolerant phase-locked loop (PLL) operation steps: Construct a PLL based on phase coupling of in-frequency characteristic signals; generate orthogonal and in-phase components of two characteristic signals respectively through the dual-channel second-order generalized integrator of the PLL; calculate the instantaneous phase error using the orthogonal and in-phase components; generate phase coupling constraint error by combining the steady-state phase difference reference; synthesize the composite phase detection error through adaptive weights driven by signal-to-noise ratio; and finally output the characteristic harmonic locking angular frequency of the composite phase detection error through a proportional-integral regulator.

[0010] Speed ​​calculation steps: Based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order and the number of motor pole pairs, the characteristic harmonic locking angular frequency is converted into the motor mechanical speed and output, realizing high-precision dynamic measurement of the motor speed without sensors.

[0011] To achieve the above objectives, a second aspect of the present invention provides a sensorless motor speed measurement system based on back EMF harmonic analysis, comprising:

[0012] The signal acquisition module is used to synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the housing, and to perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system.

[0013] The feature purification module is used to perform notch filtering with the motor fundamental feedforward angular frequency as the center frequency to remove the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal; and to perform notch filtering with the mechanical speed feedforward frequency and its harmonics as the center frequency to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing to obtain the vibration speed characteristic signal with the same frequency as the current speed characteristic harmonic signal.

[0014] The fault-tolerant phase-locked loop (PLL) module is used to construct a PLL based on phase coupling of the same frequency characteristic. The PLL uses a dual-channel second-order generalized integrator to generate orthogonal and in-phase components of two characteristic signals. The instantaneous phase error is calculated using the orthogonal and in-phase components. The phase coupling constraint error is generated by combining the steady-state phase difference reference. The composite phase detection error is synthesized by adaptive weights driven by the signal-to-noise ratio. Finally, the composite phase detection error is output as the characteristic harmonic locking angular frequency by a proportional-integral regulator.

[0015] The speed calculation module is used to convert the characteristic harmonic locking angular frequency into the motor mechanical speed and output it based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order and the number of motor pole pairs, so as to realize high-precision dynamic measurement of the motor speed without sensors.

[0016] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described sensorless motor speed measurement method based on back EMF harmonic analysis.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] The technical solution provided by this invention synchronously acquires stator current and casing vibration signals and independently performs notch filtering during the actual data processing of motor operation, obtaining electromagnetic and mechanical dual-characteristic signals at the same frequency dimension. In the phase-locked loop (PLL) operation stage, the system introduces a pre-calibrated steady-state phase difference reference to cross-constrain the instantaneous phase errors of the two signals, and dynamically adjusts the phase detection weights based on the signal-to-noise ratio of each of the two characteristic signals, thereby synthesizing the total composite phase detection error.

[0019] This measurement process can provide accurate phase traction by relying on another signal with a high signal-to-noise ratio when a certain signal experiences sudden distortion due to complex on-site conditions. This reduces the tracking offset phenomenon of the proportional-integral regulator caused by noise from a single signal source, thereby maintaining the stability of the phase-locked loop in tracking the characteristic frequency without the need to add an external speed sensor, and improving the speed measurement quality of the motor in physical application environments with interference. Attached Figure Description

[0020] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0021] Figure 1 This is a flowchart illustrating the sensorless motor speed measurement method based on back EMF harmonic analysis provided by the present invention.

[0022] Figure 2 This is a schematic diagram illustrating the implementation of the sensorless motor speed measurement system based on back EMF harmonic analysis provided by the present invention.

[0023] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0025] The sensorless motor speed measurement method, system, and electronic device based on back EMF harmonic analysis according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0026] Example 1:

[0027] like Figure 1 As shown in the figure, this embodiment provides a sensorless motor speed measurement method based on back EMF harmonic analysis. This method achieves highly robust speed estimation without relying on external physical speed sensors by deep fusion and phase constraint of multi-modal data.

[0028] Specifically, the method of this embodiment includes the following steps:

[0029] Signal acquisition steps: Synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the motor housing, and perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system.

[0030] In actual industrial drive operations, the motor stator three-phase current signal refers to the electrical physical quantity reflecting the transient electromagnetic changes of the stator, obtained by a broadband current sensor (such as a closed-loop Hall current sensor) physically installed at the output terminal of the motor inverter or the input terminal of the stator winding. This signal not only contains the fundamental current used to generate electromagnetic torque, but also contains rich back EMF harmonic components due to stator cogging effect, uneven spatial distribution of rotor flux linkage, and changes in air gap permeability.

[0031] The radial vibration acceleration signal of the housing refers to a mechanical physical quantity that reflects the internal mechanical response characteristics of the motor, obtained by a piezoelectric or microelectromechanical system (MEMS) accelerometer rigidly fixed to the radial force point of the motor's metal housing. This signal includes the rotor's mechanical rotation frequency and its associated multi-order resonance signals.

[0032] It is important to note that the synchronous acquisition refers to the use of a multi-channel analog-to-digital converter (ADC) with a shared trigger clock source at the underlying hardware architecture to ensure that the current signal in the electromagnetic domain and the vibration signal in the mechanical domain are sampled within the same microsecond time slice. This operation eliminates the microscopic time difference caused by serial sampling of channels, providing a strict timestamp alignment basis for subsequent phase coupling constraints.

[0033] The coordinate transformation refers to the mathematical projection, based on the Clarke transformation matrix in electrical engineering principles, of physical quantities in a three-phase stationary coordinate system that vary with time and are 120 electrical degrees out of phase, onto a two-phase orthogonal alpha-beta coordinate system. In the stationary coordinate system, the stator current signal output in the stationary coordinate system after coordinate transformation consists of two mutually orthogonal scalar sequences. This dimensionality reduction not only significantly reduces the matrix operation load of the microprocessor, but also effectively decouples the zero-sequence current component, enabling subsequent harmonic feature extraction to be performed in a cleaner orthogonal space.

[0034] Optionally, after completing the basic data acquisition, the system enters the feature purification step: notch filtering is performed with the motor fundamental feedforward angular frequency as the center frequency to remove the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal; notch filtering is performed with the mechanical speed feedforward frequency and its harmonics as the center frequency to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing to obtain the vibration speed characteristic signal with the same frequency as the current speed characteristic harmonic signal.

[0035] For example, the feature purification step specifically includes the following two parallel extraction streams:

[0036] Current mode purification process: The fundamental feedforward angular frequency obtained by converting the rotational speed frequency of the previous cycle phase-locked output is used as the notch center frequency. A second-order notch filter is used to filter and separate the stator current signal in the stationary coordinate system, and the current rotational speed characteristic harmonic signal is output.

[0037] Here, the fundamental feedforward angular frequency refers to the first-order electromagnetic excitation angular frequency obtained by kinematic conversion based on the estimated speed calculated and output by the system in the previous control cycle or the speed command issued by the upper controller, combined with the pole pair parameters of the motor.

[0038] The notch center frequency refers to the frequency node with the maximum amplitude attenuation characteristic in the filter design; the second-order notch filter refers to an infinite-length unit impulse response (IIR) digital filter with two poles and two zeros, which can form an extremely narrow stopband at a specified center frequency, thereby filtering out the high-energy fundamental component while retaining the signal in the surrounding frequency band.

[0039] The fundamental component refers to the reference frequency component with the largest amplitude during motor operation, which is a strong background interference that must be shielded for speed estimation. The current-speed characteristic harmonic signal refers to the weak high-frequency harmonic electrical signal that is proportional to a specific order with the rotor rotation speed after the fundamental component has been removed.

[0040] In the system's underlying algorithm implementation of this embodiment, the second-order notch filter is designed using the continuous-domain Laplace transfer function before discretization. The specific transfer function is given as shown in formula (1):

[0041] (1)

[0042] In equation (1), Let be the continuous domain transfer function of the second-order notch filter for the current signal. For the complex frequency operator in the Laplace transform, The fundamental feedforward angular frequency is calculated from the rotational speed of the previous cycle. This is the notch width factor.

[0043] In practical applications, notch width factor The value of determines the width of the stopband and the steepness of the filter's transition band. The smaller the value, the narrower the notch and the smaller the distortion region of the phase frequency characteristics. However, the adaptability to frequency fluctuations will be reduced, and it needs to be configured reasonably according to the system sampling rate.

[0044] Specifically, the vibration mode purification process is as follows: using the rotational frequency, 2 times the rotational frequency, and 3 times the rotational frequency obtained from the real-time estimated mechanical rotational speed as the center frequency, an adaptive notch filter array is constructed to sequentially remove the strong interference component of the rotational frequency from the radial vibration acceleration signal of the casing, and output the vibration rotational speed characteristic signal; wherein, the structure of the single-order notch filter unit in the adaptive notch filter array is the same as the structure of the second-order notch filter.

[0045] The rotational frequency calculated from the mechanical rotational speed refers to the physical frequency corresponding to the number of revolutions per second of the motor rotor. Due to mechanical imbalances, bearing clearances, and air gap eccentricities, the vibration signal of the casing typically contains a large number of low-order mechanical harmonics such as 1st, 2nd, and 3rd harmonics. These low-frequency vibrations have enormous energy and can easily mask the weak high-frequency vibration response caused by high-frequency electromagnetic forces that resonates with the back EMF harmonics.

[0046] The adaptive notch filter array refers to a multi-stage filtering architecture formed by multiple single-stage notch filter units connected in series. The center frequency of each notch filter unit is not a fixed constant, but is dynamically updated according to the real-time estimated value of the mechanical rotation speed. After being filtered step by step by this array, the originally complex broadband vibration signal is effectively filtered, and the output vibration speed characteristic signal retains only the high-frequency mechanical response signal component generated by the stator and rotor slots, which is strictly in the same frequency as the harmonic signal of the current speed characteristic.

[0047] It should be noted that before entering dynamic phase-locked tracking, due to the objective laws of the physical world, the transmission of electromagnetic signals to mechanical vibration signals is subject to phase delay caused by structural damping and stiffness. These two signals with the same frequency are not absolutely coincident in microscopic phase.

[0048] Therefore, before performing the fault-tolerant phase-locked loop (PLL) operation step, a steady-state phase difference reference calibration step is also included: when the system determines that it is currently in a steady-state operating condition where both the speed and load fluctuations are less than a preset fluctuation threshold, the steady-state phase value of the current speed characteristic harmonic signal and the steady-state phase value of the vibration speed characteristic signal are obtained respectively; the steady-state phase difference is obtained by subtracting the steady-state phase value of the current speed characteristic harmonic signal and the steady-state phase value of the vibration speed characteristic signal; the steady-state phase difference is stored after being filtered by the sliding window mean and used as the steady-state phase difference reference for calculating the phase coupling constraint error in the PLL.

[0049] The preset fluctuation threshold refers to a limit set manually in the system software to measure the smoothness of motor operation, typically manifested as the upper limit of the variance of the rotational speed within a set time window or the rate of change of the load torque. The steady-state operating condition refers to the operating state in which the internal electromagnetic field distribution and mechanical vibration modes of the motor reach dynamic equilibrium under constant speed and constant load torque. The steady-state phase value refers to the initial phase angle of the time axis extracted by performing a Hilbert transform or basic phase-locked loop analysis on the separated characteristic harmonic signals under this equilibrium state. By performing a purely mathematical subtraction operation on the two phase values ​​at the same moment, the steady-state phase difference determined by the inherent physical properties of the motor's internal magnetic materials, the acoustic transmission path of the mechanical casing, etc., can be obtained.

[0050] To filter out white noise or random interference that still exists in steady state, a sliding window mean filtering algorithm is employed. This involves constructing a fixed-length first-in-first-out (FIFO) data buffer and calculating the arithmetic mean of all samples within the buffer in each sampling period. The smoothed value after filtering is stored in memory as a steady-state phase difference reference. This reference value reflects the inherent phase binding relationship between the electromagnetic and mechanical domain characteristic signals of this motor model under specific speed and load conditions, providing a reference for subsequent fault-tolerant correction.

[0051] Optionally, after feature purification and benchmark calibration are completed, the system enters the most crucial fault-tolerant phase-locked loop (PLL) operation step: a PLL based on phase coupling of in-frequency features is constructed, and the orthogonal and in-phase components of two feature signals are generated by the dual-channel second-order generalized integrator of the PLL; the instantaneous phase error is calculated using the orthogonal and in-phase components, and the phase coupling constraint error is generated by combining the steady-state phase difference benchmark; the composite phase detection error is synthesized by adaptive weights driven by the signal-to-noise ratio; finally, the composite phase detection error is output as a characteristic harmonic locking angular frequency by a proportional-integral regulator.

[0052] For example, in the fault-tolerant phase-locked loop operation step, the dual-channel second-order generalized integrator includes a current integration unit and a vibration integration unit. Its operation process includes: the current integration unit and the vibration integration unit share the characteristic harmonic locking angular frequency as the real-time resonant frequency; the current integration unit receives the current rotational speed characteristic harmonic signal and outputs the current in-phase component and the current quadrature component; the vibration integration unit receives the vibration rotational speed characteristic signal and outputs the vibration in-phase component and the vibration quadrature component.

[0053] The dual-channel second-order generalized integrator (SOGI) refers to an orthogonal signal generation system widely used in the field of AC motor control. Traditional delay methods (such as quarter-cycle delay) can lead to non-orthogonal phases when the frequency changes, while the second-order generalized integrator has a closed-loop resonant structure inside, which can provide in-phase output with no delay and orthogonal output with a precise 90-degree lag at its resonant frequency.

[0054] In this embodiment, the current integrator and the vibration integrator are responsible for processing two signals with different physical properties but the same frequency. These two units do not operate in isolation, but rather share the characteristic harmonic locking angular frequency as the real-time resonant frequency. This means that the internal center frequency parameters of the two generalized integrators are uniformly fed by the output of the phase-locked loop backend. This global frequency sharing mechanism ensures that the current mode and the vibration mode are always extracted at the same frequency scale.

[0055] The current in-phase component and the current quadrature component refer to the two sinusoidal signals with equal amplitude and 90-degree phase difference on the time vector diagram, which are separated from the current rotation speed characteristic harmonic signal after passing through the current integrator unit; similarly, the vibration integrator unit outputs the corresponding vibration in-phase component and vibration quadrature component.

[0056] To quantify the internal transfer mechanism of the second-order generalized integrator, we take the in-phase component of the output current of the current integration unit as an example. Its continuous domain transfer function structure is shown in Equation (2):

[0057] (2)

[0058] In equation (2), The transfer function for the in-phase component of the output current of the current integrator unit. The damping coefficient of the second-order generalized integrator. This is the real-time resonant frequency fed back globally by the phase-locked loop, i.e., the characteristic harmonic locking angular frequency. This is achieved by adjusting the damping coefficient. It can achieve a balance between the system's dynamic tracking speed and bandpass filtering suppression capability.

[0059] Specifically, in the fault-tolerant phase-locked loop (PLL) operation step, the synthesis process of the composite phase detection error includes: calculating the instantaneous phase error of the current based on the in-phase component and the quadrature component of the current, and calculating the instantaneous phase error of the vibration based on the in-phase component and the quadrature component of the vibration; subtracting the instantaneous phase error of the current from the instantaneous phase error of the vibration to obtain the actual phase difference, and subtracting the actual phase difference from the steady-state phase difference reference to generate the phase coupling constraint error; obtaining the current mode weighting coefficient, the vibration mode weighting coefficient, and the coupling constraint weighting coefficient, and weighting and summing the instantaneous phase error of the current, the instantaneous phase error of the vibration, and the phase coupling constraint error according to their corresponding weighting coefficients to obtain the final composite phase detection error.

[0060] The instantaneous phase error of the current refers to the angular deviation between the phase of the input signal and the estimated phase inside the phase-locked loop at the current sampling time, calculated using the inverse trigonometric function (usually the arctangent function) based on the ratio of the in-phase component and the quadrature component of the current. The instantaneous phase error of vibration is similar.

[0061] Because of the introduction of dual signals, the system subtracts the instantaneous phase errors of the two signals to obtain the actual phase difference, which reflects the relative position of the current and vibration signal phases under the current dynamic operating conditions. The phase coupling constraint error is obtained by subtracting the previously stored steady-state phase difference reference from the actual phase difference. This error serves as both a phase constraint and dynamic compensation: when external electromagnetic interference causes a sudden and unreasonable shift in the instantaneous phase error of the current, the phase change of the mechanical vibration signal is relatively slow due to its inherent inertia.

[0062] At this point, the actual phase difference will deviate significantly from the steady-state phase difference reference, resulting in a large phase coupling constraint error. This error will be introduced into the system as a penalty term, forcibly correcting the distorted phase estimate to the theoretical phase range. The weighting coefficient is the proportional multiplier allocated to each error. By linearly weighting and summing these three errors, a composite phase detection error is finally synthesized to drive the regulator.

[0063] It should be noted that the current mode weighting coefficient, vibration mode weighting coefficient, and coupling constraint weighting coefficient are determined using a signal-to-noise ratio driven adaptive adjustment mechanism, which includes:

[0064] The differences between the input and output signals of the current integration unit and the vibration integration unit are extracted as the current residual and vibration residual, respectively. The current characteristic harmonic signal-to-noise ratio is calculated based on the current residual, and the vibration characteristic harmonic signal-to-noise ratio is calculated based on the vibration residual.

[0065] The minimum value between the current characteristic harmonic signal-to-noise ratio and the vibration characteristic harmonic signal-to-noise ratio is extracted as the local minimum signal-to-noise ratio;

[0066] When the local minimum signal-to-noise ratio is greater than or equal to the preset signal-to-noise ratio threshold, it is determined to be a high signal-to-noise ratio operating condition, and the preset weak coupling benchmark weight is selected as the coupling constraint weight coefficient.

[0067] When the local minimum signal-to-noise ratio is less than the signal-to-noise ratio threshold, it is determined to be a low signal-to-noise ratio condition. At this time, the value of the coupling constraint weight coefficient increases linearly as the local minimum signal-to-noise ratio decreases, so as to enhance the coupling constraint effect between the two signals.

[0068] In industrial environments, the high-frequency switching of the inverter's insulated gate bipolar transistors (IGBTs) injects a large amount of pulse width modulation (PWM) switching noise into the stator windings, while the vibrations of the mechanical load also excite non-periodic broadband vibrations on the casing. These background noises are constantly changing.

[0069] The difference between the input signal and the output signal refers to the original mixed signal entering the integration unit minus the clean in-phase component output by the integration unit. This difference is highly approximated at the physical level as the background noise amplitude within the frequency band, i.e., the residual. The characteristic harmonic signal-to-noise ratio refers to the ratio of the effective value of the extracted characteristic in-phase component to the effective value of the residual (usually converted to logarithmic decibels).

[0070] The local minimum signal-to-noise ratio (SNR) refers to the value of the SNR performance of the current mode and the vibration mode, which is worse within the current evaluation period. It represents the most severe interference situation currently faced by the system. The preset SNR threshold refers to the lower limit of SNR determined through statistical analysis of a large amount of experimental data, sufficient to maintain the stable operation of a conventional phase-locked loop. When a high SNR condition is determined, both signals are clean, and the weak coupling reference weight means that the system mainly relies on its independent instantaneous errors for adjustment in order to pursue the fastest response speed. Once a low SNR condition is determined, it means that one or two signals have been severely contaminated. At this time, the system automatically increases the coupling constraint weight coefficient according to a linear function.

[0071] To quantify the logic of this adaptive adjustment, the calculation rules for the coupling constraint weight coefficients are provided as shown in formula (3):

[0072] (3)

[0073] In equation (3), This represents the coupling constraint weight coefficients calculated in real time. For weakly coupled baseline weights under high signal-to-noise ratio conditions, This represents the upper limit of the maximum allowed coupling weight in the system. The preset signal-to-noise ratio threshold, The local minimum signal-to-noise ratio is calculated in real time. This is the theoretically critical minimum signal-to-noise ratio at which the system can maintain lock-free operation.

[0074] As can be seen from the formula, the closer the local minimum signal-to-noise ratio is to the critical value, the larger the coupling constraint weight coefficient is, and the stronger the physical constraint effect between the two signals is. Thus, the phase drift caused by single-channel noise is hard suppressed by the complementarity of dual-source information.

[0075] Optionally, in the fault-tolerant phase-locked loop operation step, the specific process of outputting the characteristic harmonic locking angular frequency through the proportional-integral regulator includes:

[0076] The composite phase detection error is input into the proportional-integral regulator; when the absolute value of the composite phase detection error is greater than the preset phase error threshold, it is determined to be a transient condition, and the high gain parameter is called to perform variable step size proportional-integral regulation to broaden the phase-locked loop bandwidth.

[0077] When the absolute value of the composite phase detection error is less than or equal to the preset phase error threshold, it is determined to be a steady-state condition. The low-gain parameter is called to perform variable step size proportional-integral regulation to narrow the phase-locked loop bandwidth. The angular frequency correction output by the proportional-integral regulator is linearly superimposed with the characteristic harmonic feedforward angular frequency to obtain the output characteristic harmonic locking angular frequency.

[0078] The proportional-integral (PI) controller is a standard control unit in automatic control systems used to eliminate steady-state errors and improve response speed. It receives a composite phase detection error as input and outputs a command for frequency correction. The absolute value of the composite phase detection error removes the direction (leading or lagging) of the error, retaining only the degree of deviation.

[0079] The preset phase error threshold is used to distinguish between minor fluctuations and drastic dynamic changes in the system. The transient condition typically occurs when the motor is suddenly loaded or when a rapid acceleration command is given, at which point the phase error surges. The high-gain parameter and low-gain parameter refer to the combination of the proportional coefficient and integral time constant within the proportional-integral controller. Using the high-gain parameter to perform variable-step proportional-integral regulation significantly widens the phase-locked loop bandwidth, enhancing the system's sensitivity to rapidly changing high-frequency signals, allowing the phase-locked frequency to quickly keep up with sudden changes in speed. Conversely, using the low-gain parameter under steady-state conditions narrows the phase-locked loop bandwidth, effectively enhancing the system's low-pass filtering characteristics and significantly filtering out minor high-frequency noise.

[0080] In this step, the output calculation logic of the proportional-integral controller is as shown in formula (4):

[0081] (4)

[0082] In equation (4), This is the angular frequency correction value output by the proportional-integral controller; This is the calculated value of the composite phase detection error within the current discrete control cycle; This is a variable step size proportional gain parameter; For variable step size integral gain parameters; At the current sampling time, This is an intermediate variable for continuous-time integration in the integration process.

[0083] The output of the above formula It is not the final frequency; it merely reflects the fine-tuning amount of the preceding estimation. This is linearly superimposed with the characteristic harmonic feedforward angular frequency (i.e., the reference frequency calculated macroscopically based on the speed command) to obtain the final closed-loop feedback characteristic harmonic locking angular frequency. This composite control structure of feedforward and feedback allows the phase-locked loop to maintain an accurate basic output even when the error integral is zero, reducing the dynamic burden on the regulator.

[0084] For example, after completing the closed-loop tracking described above, the speed calculation step is initiated: based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order, and the number of motor pole pairs, the characteristic harmonic locking angular frequency is converted into the motor mechanical speed and output, thereby achieving high-precision dynamic measurement of the motor speed without sensors.

[0085] The characteristic harmonic order is a fixed constant determined by the relationship between the number of rotor slots and the number of pole pairs in the motor. It indicates how many times the extracted high-frequency back EMF harmonic signal frequency is greater than the fundamental rotational frequency. The number of motor pole pairs refers to the number of N and S poles formed in a pair in the motor stator windings. Since the electrical signal induced in the stator coils is the electromagnetic angular frequency corresponding to the electrical angle, it must be divided by the characteristic harmonic order and the number of motor pole pairs to map it back to the physical angular velocity of the mechanical rotor body per minute, i.e., the motor's mechanical speed. This output will be directly fed into the upper-level vector control (FOC) loop as the core feedback quantity of the speed closed loop.

[0086] It should also be noted that during the low-speed or zero-speed start-up phase, the back EMF harmonic amplitude is low and the signal-to-noise ratio is poor, making it difficult to guarantee measurement accuracy by relying solely on electrical and vibration signals.

[0087] Therefore, the speed calculation step also includes a redundancy correction process that integrates with visual speed measurement: acquiring a target image of the motor shaft and calculating the visually measured speed using an image processing algorithm; constructing a fusion framework with dynamic weights across the entire speed domain: when the converted motor mechanical speed is lower than a preset rated speed ratio threshold, assigning a first fusion weight to the visually measured speed; when the motor mechanical speed is higher than or equal to the rated speed ratio threshold, dynamically configuring a second fusion weight for the visually measured speed, wherein the second fusion weight is less than the first fusion weight, and outputting the final fused mechanical speed.

[0088] The motor shaft target image refers to a continuous sequence of digital images acquired in real time by spraying specific black and white code disks or reflective spots onto the exposed rotating output shaft of the motor and capturing them with an external high-speed industrial camera. Using edge detection and recursive filtering algorithms from the field of machine vision, the visually measured rotational speed can be calculated from the displacement of the image sequence. The rated speed ratio threshold is typically set to approximately 10% of the rated speed on the motor nameplate; this is an engineering critical reference value for the stable extraction and recognition of the back EMF signal. The full-speed-domain dynamic weight fusion framework refers to a data assimilation algorithm with adaptive gain adjustment capabilities, such as Kalman filtering.

[0089] At low speeds, visual signals are unaffected by the weak back EMF amplitude, making visual measurements a more reliable observation source and thus assigned a larger first fusion weight. However, as the motor speed increases and enters the medium-to-high speed domain, the back EMF signal becomes stronger and the signal-to-noise ratio improves, significantly enhancing the frequency tracking accuracy of the phase-locked loop. At this point, the sampling frame rate of the visual camera may become a physical bottleneck for the measurement. Therefore, a second fusion weight is dynamically configured and strictly set to be less than the first fusion weight, smoothly transitioning from visual observation-dominated to back EMF observation-dominated logic. The final output mechanical speed combines the measurement advantages of different types of multimodal hardware at different speed domains.

[0090] Optionally, due to the wide range of operating environment parameters (such as temperature and load) of industrial motors, fixed filter and phase-locked loop (PLL) parameters cannot remain optimal throughout their entire lifecycle. The method further includes a global adaptive optimization step for PLL parameters: a full-condition benchmark library covering multiple speed and load levels is pre-constructed, where each operating condition point in the benchmark library corresponds to a set of optimized PLL parameters; the current operating condition feature vector of the motor is extracted in real time, and a dynamic time warping and cosine similarity combination algorithm is used to calculate the deviation between the current operating condition feature vector and each operating condition benchmark point in the benchmark library, matching the optimal operating condition benchmark point; based on the deviation, nonlinear interpolation is performed between the matched optimal operating condition benchmark point and the PLL parameters of adjacent operating condition benchmark points, and the interpolation result is used as the current PLL parameters for updating and adjustment.

[0091] The various speed and load levels refer to matrix-style grid test points set manually during pre-shipment bench calibration testing, ranging from no-load to full-load and from extremely low speed to maximum speed. The full-condition benchmark library is a high-dimensional lookup table stored in non-volatile memory, recording optimized PLL parameters such as damping coefficient and bandwidth at each grid point that ensure the most stable PLL performance. The current operating condition feature vector is a multi-dimensional mathematical vector composed of parameters such as real-time speed, real-time estimated torque, and motor body temperature.

[0092] To quickly find the most similar historical operating conditions in the vast benchmark database, the system introduces a Dynamic Time Warping (DTW) algorithm to eliminate scaling and distortion differences in time-series data, and combines this with cosine similarity to evaluate directional consistency in the vector space. This combined algorithm calculates the deviation between the current operating state and each anchor point in the database. The optimal operating condition benchmark point is the historical operating condition node with the smallest deviation and the most similarity.

[0093] However, real-world operating conditions often lie between grid points, necessitating the use of nonlinear interpolation algorithms to perform smooth transition calculations between the parameters of the optimal reference point and its neighboring reference points, based on distance weights. The calculated interpolation results will continuously update and adjust the parameters of the phase-locked loop running in the microprocessor, achieving parameter self-learning and self-adaptation.

[0094] Specifically, this solution goes beyond speed measurement and also explores the secondary application value of the intermediate parameters of the phase-locked loop.

[0095] The method further includes a collaborative closed-loop diagnostic step based on the characteristic harmonic locking angular frequency: using the characteristic harmonic locking angular frequency as the frequency reference anchor point, locating and extracting fault feature parameters of each order from the three-phase current signal of the motor stator and the radial vibration acceleration signal of the housing; inputting the fault feature parameters of each order into a cascaded diagnostic model containing a multi-scale feature extraction and multi-head attention mode fusion network, and outputting a classification diagnostic result of the motor health status; when the classification diagnostic result indicates the presence of a fault that causes signal distortion, automatically calling the fault condition parameter pre-plan through a closed-loop feedback mechanism, dynamically increasing the allocation weight of the phase coupling constraint error in the phase-locked loop and increasing the phase-locked loop bandwidth reference.

[0096] In the condition monitoring of mechanical equipment, the fault frequency components induced by the vast majority of electrical faults (such as broken rotor bars) and mechanical faults (such as peeling of bearing outer rings) have a strict mathematical proportional relationship with the fundamental frequency or rotational frequency of the motor.

[0097] By using the extremely precise characteristic harmonic locking angular frequency calculated earlier as the frequency reference anchor point, the system can locate the theoretical fault characteristic frequency band in the signal spectrum at high resolution, that is, accurately locate the tiny frequency band where the fault signal should theoretically occur, and extract its amplitude and phase as fault characteristic parameters of each order.

[0098] The cascaded diagnostic model is a pre-trained artificial intelligence neural network. Its multi-scale feature extraction aims to simultaneously capture instantaneous impact characteristics and long-term degradation trend characteristics; the multi-head attention modal fusion network can automatically learn and pay attention to the importance allocation of current and vibration data under different fault categories. The final classification diagnostic results output by the model include labels such as normal, minor wear, and severe fault.

[0099] More importantly, this solution constructs a closed-loop feedback mechanism: when the model detects a fault in the motor, such as rotor eccentricity causing signal distortion, the system anticipates a significant deterioration in the subsequent current signal-to-noise ratio. At this point, without waiting for the conventional signal-to-noise ratio detection module to react, the system automatically triggers and invokes the fault condition parameter pre-plan, dynamically increasing the weighting of the phase coupling constraint error allocation in the phase-locked loop. This involves using normal vibration mechanical characteristic signals to phase-constrain the distorted electromagnetic characteristic signals, while simultaneously increasing the phase-locked loop bandwidth benchmark to address abnormal speed fluctuations caused by the fault, ensuring that the speed estimation system can maintain stable operation even when the motor is operating under fault conditions.

[0100] As the background analysis shows, single data sources are prone to loss of lock under varying loads or electromagnetic interference scenarios, and simple multi-signal weighted fusion suffers from error propagation problems due to the lack of physical constraints. The sensorless motor speed measurement method based on back EMF harmonic analysis, detailed in this embodiment, constructs a highly closed-loop fault-tolerant mechanism based on underlying physical laws by introducing a dual-mode orthogonal integral structure and a phase coupling constraint equation based on steady-state reference and transient signal-to-noise ratio.

[0101] In practical applications, when a motor encounters strong electromagnetic interference caused by sudden voltage changes on the grid side, or instantaneous vibration shocks caused by mechanical jamming on the load side, this system can forcibly correct the phase shift of the contaminated mode using data from the healthy mode through real-time weight adjustment and reference locking. Combined with visual redundancy correction, adaptive optimization of parameters under all operating conditions, and closed-loop feedback for fault diagnosis, this solution systematically improves the robustness and tracking accuracy of the sensorless speed estimation algorithm in extremely harsh industrial noise environments and during the operation of equipment with minor faults, ensuring the continuity and reliability of the underlying feedback data for industrial automation control.

[0102] Example 2:

[0103] In industrial settings, drive systems often face numerous unpredictable transient disturbances, such as the mechanical load jump in heavy rolling mills when steel is bitten in, or the severe mechanical impacts generated by new energy vehicles under complex road conditions. These conditions can cause drastic changes in motor speed within a very short time, introducing a large amount of broadband nonlinear transient noise into the stator current signal and the housing vibration signal. This transient noise can easily cause integral saturation in conventional proportional-integral regulators, leading to phase-locked loop (PLL) loss of lock.

[0104] To address this technical problem, the method provided in this embodiment further includes a variable structure compensation process for extreme dynamic impact conditions in the fault-tolerant phase-locked loop operation step. The variable structure compensation process specifically includes the following steps:

[0105] Feature amplitude mutation monitoring steps: Calculate in real time the envelope amplitude change rate of the current rotation speed characteristic harmonic signal and the envelope amplitude change rate of the vibration rotation speed characteristic signal.

[0106] Specifically, this step is a prerequisite for the system to switch from a normal steady state to dynamic compensation. The current-speed characteristic harmonic signal and the vibration-speed characteristic signal have been clearly defined in Example 1 through notch filtering and feature extraction. The envelope amplitude refers to the peak outer contour boundary of the high-frequency AC characteristic harmonic signal within the transient time window. In the digital signal processor, since the dual-channel second-order generalized integrators have already output the in-phase and quadrature components of the current and vibration respectively, the system does not need to add an additional peak detection circuit. Instead, it can directly add the squares of the in-phase and quadrature components of each mode and take the square root to obtain the extremely accurate instantaneous envelope amplitude of the signal in real time.

[0107] Furthermore, the rate of change of the envelope amplitude refers to the differential or difference value of the calculated envelope amplitude over a discrete time series. It reflects the degree of drastic change in signal energy over time. In practical engineering calculations, the system employs a sliding difference algorithm within a fixed time window, which involves subtracting the envelope amplitude from the current sampling period's value from the value several control periods prior, and then dividing by the time interval.

[0108] Real-time monitoring of the rate of change of two signals has significant engineering implications: a sharp increase in the rate of change of the envelope amplitude of only the current signal typically indicates a purely electrical disturbance such as a sudden rise in grid voltage; a sudden increase in the rate of change of the envelope amplitude of only the vibration signal indicates mechanical jamming or external physical collision; and a simultaneous abrupt change in both signals indicates a global electromechanical coupling shock to the system. By synchronously calculating the rates of change of these two signals in real time, a highly sensitive pre-condition observer is established, capable of detecting external shocks before significant phase deviations occur within the phase-locked loop.

[0109] State determination and reference freezing steps: When any of the envelope amplitude change rates is greater than the preset impact mutation threshold, the system is determined to have entered an extreme dynamic impact state, and the numerical update process of the steady-state phase difference reference in the phase-locked loop is frozen.

[0110] It is important to note that, to avoid frequent misjudgments of the system due to interference from conventional sensor measurement noise, this step introduces a scientific definition standard. The preset impact change threshold is a pre-calibrated extreme value based on the motor's inertia parameters and the system's maximum design load acceleration. When the rate of change of amplitude is lower than this threshold, the system considers it normal acceleration / deceleration or a slight load fluctuation;

[0111] Once the value exceeds this threshold, it clearly indicates that a severe disturbance has occurred that exceeds the system's linear adjustment capability. The extreme dynamic shock state refers to a situation where, under such disturbance, the original linear transmission relationship between the electromagnetic torque response and the mechanical rotor response of the motor is lost, transient noise dominates the characteristic frequency distribution of the signal, resulting in severe distortion in instantaneous phase calculation.

[0112] Under such severe nonlinear conditions, the system must perform a crucial operation: freezing the numerical update process of the steady-state phase difference reference. In Embodiment 1, the steady-state phase difference reference was defined as the inherent phase difference between two signals continuously updated through a sliding window mean filter. Freezing refers to, at the software algorithm level, immediately cutting off the write path of instantaneous phase difference data into the sliding window mean filter buffer, and ensuring that the existing steady-state phase difference reference value in memory remains unchanged. The physical significance of this operation is that the actual phase difference extracted under impact conditions contains a large amount of erroneous information. Allowing this erroneous information to enter the reference library would severely contaminate the references used for cross-constraints.

[0113] Once the baseline is contaminated, even after the impact ends, the system will continue to produce steady-state velocity measurement errors due to incorrect baseline guidance. Therefore, freezing the update process is a necessary means to protect the system's memory baseline from contamination by transient and adverse data.

[0114] Sliding mode variable structure intervention steps: Under the extreme dynamic impact state, a sliding mode surface is constructed based on the composite phase detection error and its differential value, and the sliding mode compensation angular frequency is calculated and generated; the sliding mode compensation angular frequency is superimposed with the proportional adjustment term of the proportional-integral controller, and the integral adjustment term of the proportional-integral controller is removed, thereby generating the characteristic harmonic locking angular frequency under the impact condition.

[0115] For example, when the system is under normal operating conditions, the proportional-integral (PI) controller can effectively eliminate steady-state errors in frequency tracking through its integral term. However, under extreme dynamic shock conditions, due to sudden changes in rotational speed, the composite phase detection error will surge instantaneously. The traditional integral control term will continuously accumulate this huge error, leading to integral saturation. When the actual rotational speed drops, the accumulated integral cannot be released quickly, causing a severe overshoot of the phase-locked output frequency, which in severe cases can directly lead to system lockout.

[0116] To overcome this bottleneck, this solution adopts a sliding mode variable structure control strategy, which is highly robust in the field of nonlinear control.

[0117] The composite phase detection error, following the definition in Example 1, refers to the total phase tracking deviation after signal-to-noise ratio weighting adjustment. This step first involves constructing the sliding surface. The sliding surface is a dynamically designed hyperplane in the system state space. When the system state is forcibly constrained to slide on this plane, the system will exhibit strong robustness to external disturbances and parameter perturbations.

[0118] To quantify the above control logic, a continuous domain control model for the sliding surface is constructed as shown in the following equation:

[0119] ;

[0120] In the formula, This represents the state function of the constructed sliding surface; This represents the transient error state variable of the composite phase detection tracking in the continuous time domain, which is different from that in Example 1. Correspondingly, it is used to characterize the instantaneous dynamic deviation of the system under extreme impact; This represents the gain coefficient of the sliding surface slope; This represents the differential value of the composite phase detection tracking error over time.

[0121] This formula defines the dynamic path for the system to reach equilibrium by linearly combining the error itself and its rate of change. Sliding surface slope gain coefficient. The value of directly determines the decay rate of the state variable after it reaches the sliding surface. The larger the value, the faster the error converges, but the more obvious the amplification effect on differential noise. Therefore, in practical applications, a compromise configuration needs to be made according to the cutoff frequency of the differential filter.

[0122] After constructing the sliding surface, the system calculates and generates the sliding compensation angular frequency. The sliding compensation angular frequency refers to the nonlinear forced compensation control quantity output to force the system state trajectory to overcome disturbances and quickly reach the designated sliding surface from any initial position. Its specific control law calculation logic is shown in the following equation:

[0123] ;

[0124] In the formula, This represents the calculated sliding mode compensation angular frequency; Indicates the equivalent control gain parameter; This represents the reaching law gain parameter; This indicates that a boundary layer thickness has been introduced. The saturation function is used to replace the sign function of hard switching, thereby eliminating high-frequency chattering of control quantities near the sliding surface and ensuring the smoothness of nonlinear compensation transition.

[0125] The formula consists of two parts: the first part is the equivalent control term, which is used to maintain the system's motion trajectory on the sliding surface; the second part is the switching approach law control term, which provides strong anti-disturbance capability through high-frequency switching of the sign function, ensuring that even under extreme mechanical collisions, the system state can be forcibly pushed back to the sliding surface.

[0126] After calculating the compensation amount, the system executes a switching mechanism on the underlying computation path: superimposing the sliding mode compensation angular frequency with the proportional adjustment term of the proportional-integral controller, and removing the integral adjustment term of the proportional-integral controller.

[0127] The proportional adjustment term refers to the linear output obtained by simply multiplying the composite phase detection error by the proportional gain. It has no hysteresis effect and can respond to the error direction immediately. The removal of the integral adjustment term refers to stopping the accumulation of the integral accumulator inside the proportional-integral controller and temporarily setting its output value to zero or bypassing it. At this time, the characteristic harmonic locking angular frequency of the system output is formed by the superposition of the conventional proportional adjustment term and the robust sliding mode compensation angular frequency. This variable structure intervention enables the phase-locked loop to instantly possess nonlinear disturbance rejection capability under impact conditions, effectively avoiding the divergence risk caused by integral saturation.

[0128] Smooth Exit Step: When the rate of change of the envelope amplitude falls below the threshold of the impulse change, the freeze on the update of the steady-state phase difference reference value is lifted, and the sliding mode compensation angular frequency is decayed to zero according to the preset decay time constant. At the same time, the integral adjustment term of the proportional-integral regulator is restored to generate the characteristic harmonic locking angular frequency under normal operating conditions.

[0129] Optionally, as the external physical disturbance gradually disappears, the system must safely and smoothly recover from the variable structure control state to the normal high-precision linear tracking state. The condition for triggering the exit mechanism is that the rate of change of the envelope amplitude falls below the impact mutation threshold, which means that the signal-to-noise ratio of the system begins to improve, and the electromagnetic response and mechanical response regain the normal steady-state correspondence.

[0130] At this point, the system first unfreezes the update of the steady-state phase difference reference value, allowing the latest actual cross phase difference data to re-enter the sliding window, ensuring that the reference library values ​​can be adaptively updated following the slow temperature rise or parameter drift of the motor body. More critically, the handover of control is crucial: due to the high-frequency switching characteristics of sliding mode variable structure control, if the sliding mode compensation angular frequency is directly cut off and the integral stage is simultaneously activated, the control command will produce a step jump. This hard switching will introduce computational oscillations in the reverse direction.

[0131] Therefore, this scheme employs a preset decay time constant to attenuate the sliding mode compensation angular frequency to zero. The preset decay time constant refers to an exponential decay time parameter set based on the principle of first-order inertial low-pass filtering. This mechanism forces the originally large sliding mode compensation amount to smoothly and continuously transition to zero along an exponential curve.

[0132] While the sliding mode compensation decreases, the integral term of the proportional-integral controller is restored. It should be noted that, to achieve truly seamless switching, at the instant the integral term is restored, the initial value of the integral accumulator is automatically assigned the difference between the sliding mode compensation and the proportional term at the moment of switching. Through this state alignment, when the proportional-integral controller takes over control, its initial output state is completely consistent with the state before the handover.

[0133] Subsequently, as the sliding mode compensation term decays exponentially, the integral term smoothly fills the gap in the control quantity and eventually regains control of the global adjustment, outputting the characteristic harmonic locking angular frequency under normal operating conditions. The entire switching process achieves a smooth transition to normal operating conditions for the upper-level speed calculation and motor control system.

[0134] Analysis of existing technologies shows that traditional sensorless motor speed measurement technology usually adopts methods such as expanding bandwidth or increasing filter dead zone when facing complex transient conditions. However, this will lead to a decrease in steady-state measurement accuracy and a serious lag in dynamic response.

[0135] The technical solutions detailed in this embodiment accurately identify impact conditions by real-time monitoring of the envelope amplitude of dual-mode characteristic signals; protect the reliability of the adaptive model through a reference freezing mechanism; most importantly, by removing integral terms prone to hysteresis and introducing nonlinear sliding mode variable structure control, the strong robustness of sliding mode control to bounded disturbances ensures that the frequency tracking of the phase-locked loop is not lost under severe impact; and after the impact, a smooth and disturbance-free transition to conventional proportional-integral control is achieved by using a preset decay time constant.

[0136] This overall solution substantially solves the technical problem of industrial motors being prone to loss of lock under sudden impact loads. While ensuring conventional high-precision measurement, it greatly improves the system's dynamic boundary adaptability and the safety and stability of its operation throughout its entire life cycle.

[0137] Example 3:

[0138] like Figure 2 As shown in the figure, this embodiment three provides a sensorless motor speed measurement system based on back EMF harmonic analysis.

[0139] In existing sensorless motor speed estimation technologies, the system typically relies solely on a single stator current signal for data analysis. When drive equipment in industrial settings faces varying load impacts or high-frequency electromagnetic interference from inverters, the signal-to-noise ratio of the single current signal fluctuates drastically, causing traditional single-ended phase-locked loops to easily develop phase tracking deviations or even complete loss of lock.

[0140] To address the technical bottlenecks in the aforementioned background technology, the measurement system disclosed in this embodiment deeply integrates multi-source physical hardware with underlying fault-tolerant algorithms to simultaneously acquire data from two different modes: stator current and casing vibration. A steady-state phase difference reference reflecting the inherent properties of the motor is introduced to apply cross-physical constraints to the signals. During actual operation, when a transient distortion occurs in one signal, the system can dynamically rely on the other signal with a superior signal-to-noise ratio to provide robust phase traction, thus ensuring the continuity and stability of speed observation without adding a physical speed sensor.

[0141] Specifically, the sensorless motor speed measurement system based on back EMF harmonic analysis provided in this embodiment includes:

[0142] The signal acquisition module is used to synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the motor housing, and to perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system.

[0143] In practical industrial equipment deployment schemes, this signal acquisition module consists of multi-channel high-precision sensing hardware and mixed-signal conditioning circuitry. For the acquisition of the three-phase current signal of the motor stator, the system obtains the raw analog electrical quantity containing the fundamental energy and weak back EMF high-frequency harmonics by using a high-frequency closed-loop Hall current sensor or a high-precision alloy shunt resistor installed at the output of the inverter bridge of the motor controller. For the extraction of the radial vibration acceleration signal of the housing, an industrial-grade piezoelectric integrated circuit (IEPE) accelerometer or a wideband microelectromechanical system (MEMS) accelerometer with high-frequency response characteristics is selected and directly mounted to the radial force concentration area of ​​the motor stator housing using rigid bolts or a special strong magnetic chuck.

[0144] To ensure the accurate correspondence between the dual-mode signals in the time domain, the underlying hardware of this module is configured with an analog-to-digital converter (ADC) that supports multi-channel synchronous sampling. The trigger pin of the ADC is uniformly controlled by a high-precision hardware timer inside the microprocessor, ensuring that the current and vibration signals complete the analog-to-digital conversion within the same nanosecond-level clock slice, effectively eliminating the phase skew caused by serial sampling of channels.

[0145] After acquiring the discrete digital signal, the digital signal processor (DSP) or field-programmable gate array (FPGA) within the system performs hardware-level operations through multiply-accumulate instructions to perform Clarke coordinate transformation on the acquired three-phase stator current signal of the motor. This process projects the three-phase current scalars, which are 120 degrees apart in space, onto mutually orthogonal alpha and beta two-phase stationary coordinate systems, generating the stator current signal in the stationary coordinate system. This effectively decouples common-mode interference in the three-phase system and provides a low-dimensional orthogonal data source for subsequent harmonic extraction.

[0146] In addition, this system also includes: a feature purification module, used to perform notch filtering with the motor fundamental feedforward angular frequency as the center frequency, to remove the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal; and to perform notch filtering with the mechanical speed feedforward frequency and its harmonics as the center frequency, to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing to obtain the vibration speed characteristic signal with the same frequency as the current speed characteristic harmonic signal.

[0147] At the hardware algorithm implementation level, this feature purification module is specifically manifested as a cluster of multi-order infinite-length unit impulse response (IIR) digital filter programs residing in the high-speed processor memory. Because the raw signals acquired in industrial settings are filled with highly energetic but useless components, precise frequency cutting is essential. For the current path, the processor calculates the motor's fundamental feedforward angular frequency based on the speed estimate from the previous control cycle and updates this dynamic frequency as the center frequency parameter of the second-order digital notch filter.

[0148] This notch filter creates a deep amplitude attenuation band at specified frequency nodes, thereby precisely stripping away the high-energy fundamental component of the current signal and outputting only the current-speed characteristic harmonic signal containing high-frequency back EMF information. Similarly, for the 1st, 2nd, and 3rd times mechanical rotational frequency responses widely present in the vibration spectrum of mechanical equipment, excited by rotor dynamic imbalance or bearing clearance, the feature purification module configures a multi-stage cascaded adaptive notch filter array for the vibration channel.

[0149] The system utilizes the real-time calculated feedforward frequency of the mechanical rotational speed and its frequency multipliers to dynamically refresh the coefficient matrix of each order of notch filters in the array. Through step-by-step deep filtering, the high-amplitude frequency interference components in the original vibration signal are effectively removed. The final output vibration rotational speed characteristic signal retains only the high-frequency vibration response excited by the relative motion of the stator and rotor slot teeth, thus ensuring that the vibration signal in the mechanical domain is closely aligned with the current harmonic signal in the electromagnetic domain on the physical frequency scale.

[0150] Furthermore, the core correction unit of this system is a fault-tolerant phase-locked loop (PLL) module, which constructs a PLL based on phase coupling of in-frequency characteristic signals. The PLL uses a dual-channel second-order generalized integrator to generate orthogonal and in-phase components of two characteristic signals. Instantaneous phase error is calculated using these components, and a phase coupling constraint error is generated by combining this error with a steady-state phase difference reference. A composite phase detection error is then synthesized using adaptive weights driven by the signal-to-noise ratio (SNR). Finally, the composite phase detection error is output as a characteristic harmonic locking angular frequency via a proportional-integral (PI) regulator. The underlying operations of this module heavily rely on the floating-point unit (FPU) within the DSP.

[0151] The dual-path second-order generalized integrator (SOGI) refers to two closed-loop orthogonal signal generator algorithms constructed within the processor using difference equations. These two generators share the global resonant frequency fed back from the end of the phase-locked loop, and perform integration operations on the current-speed characteristic harmonic signal and the vibration-speed characteristic signal, respectively, to generate orthogonal and in-phase components with equal amplitude and a 90-degree phase difference.

[0152] Subsequently, the phase detection function within the module calls the inverse trigonometric mathematical operation library to process the above components, deriving the instantaneous phase error of each mode under transient conditions. To prevent irreversible phase drift of a single mode under external nonlinear shocks, the module retrieves a steady-state phase difference reference pre-learned under constant load conditions from the non-volatile memory (EEPROM). By performing a difference operation between the actual instantaneous phase difference between the two modes and this physical reference, the system constructs a phase coupling constraint error. At the dual-source information fusion level, the module internally runs a signal-to-noise ratio estimation task, iteratively evaluating the in-band residual energy before and after the SOGI stage of the two signal inputs.

[0153] If the assessment diagnoses a sudden drop in the current signal-to-noise ratio due to a sudden fluctuation in the power grid, the module will adaptively increase the proportional parameters of the vibration channel phase angle and phase coupling constraint error in the global regulation equation. This composite phase detection error, synthesized using signal-to-noise ratio-driven adaptive weights, is continuously input into a software-written proportional-integral (PI) regulator. The proportional element ensures the algorithm can respond quickly to dynamic changes, while the integral element eliminates steady-state error. Both operate collaboratively and are superimposed on a feedforward reference to ultimately generate a stable characteristic harmonic locking angular frequency with strong anti-disturbance characteristics.

[0154] Finally, this system is equipped with a speed calculation module, which is used to convert the characteristic harmonic locking angular frequency into the motor mechanical speed and output it based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order and the number of motor pole pairs, so as to realize high-precision dynamic measurement of the motor speed without sensors.

[0155] Since the characteristic harmonic locked angular frequency characterizes the rate of change of angular velocity in electromagnetic space, this speed calculation module is responsible for performing the final mapping to the mechanical physical space. The processor core extracts the motor body structural parameters stored in the static register, namely the characteristic harmonic electrical order determined by the number of stator slots and the number of motor pole pairs in the winding design. Through fixed division operation instructions, the phase-locked frequency is restored to the actual mechanical speed of the motor shaft. The calculated high-precision digital speed is sent to the system communication backplane and published at a high baud rate through an industrial fieldbus interface (such as a CAN open controller or a high-speed EtherCAT communication chip). This speed feedback data is directly input into the motor's field-oriented control (FOC) main program in a closed loop.

[0156] Relying on the coordinated operation of the four major modules of signal acquisition, feature purification, fault-tolerant phase-locked operation, and speed calculation, this sensorless motor speed measurement system can effectively avoid the inherent defects of single feature quantities being prone to distortion and loss of lock under harsh working conditions. Without the need to add an additional physical rotary transformer or photoelectric encoder, it provides continuous speed observation data that meets the high dynamic drive requirements of industrial applications.

[0157] Example 4:

[0158] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0159] like Figure 3 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0160] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0161] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0162] The memory 103 stores a computer program corresponding to the sensorless motor speed measurement method based on back EMF harmonic analysis in the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0163] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 3 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0164] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A sensorless motor speed measurement method based on back EMF harmonic analysis, characterized in that, include: Signal acquisition steps: Synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the motor housing, and perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system; Feature purification steps: notch filtering is performed with the motor fundamental feedforward angular frequency as the center frequency to remove the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal. Notch filtering is performed using the mechanical rotation speed feedforward frequency and its harmonics as the center frequency to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing, thereby obtaining a vibration rotation speed characteristic signal with the same frequency as the current rotation speed characteristic harmonic signal. Fault-tolerant phase-locked loop (PLL) operation steps: Construct a PLL based on phase coupling of in-frequency characteristic signals; generate orthogonal and in-phase components of two characteristic signals respectively through the dual-channel second-order generalized integrator of the PLL; calculate the instantaneous phase error using the orthogonal and in-phase components; generate phase coupling constraint error by combining the steady-state phase difference reference; synthesize the composite phase detection error through adaptive weights driven by signal-to-noise ratio; and finally output the characteristic harmonic locking angular frequency of the composite phase detection error through a proportional-integral regulator. Speed ​​calculation steps: Based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order and the number of motor pole pairs, the characteristic harmonic locking angular frequency is converted into the motor mechanical speed and output, realizing high-precision dynamic measurement of the motor speed without sensors.

2. The method according to claim 1, characterized in that, The feature purification step specifically includes: Current mode purification process: The fundamental feedforward angular frequency obtained by converting the rotational speed frequency of the previous cycle phase-locked output is used as the notch center frequency. A second-order notch filter is used to filter and separate the stator current signal in the stationary coordinate system, and the current rotational speed characteristic harmonic signal is output. Vibration mode purification process: Using the rotational frequency, 2 times rotational frequency and 3 times rotational frequency obtained by real-time estimation of mechanical rotational speed as the center frequency, an adaptive notch array is constructed to remove the strong interference component of rotational frequency from the radial vibration acceleration signal of the casing step by step, and output the vibration rotational speed characteristic signal; wherein, the structure of the single-order notch unit in the adaptive notch array is the same as the structure of the second-order notch filter.

3. The method according to claim 1, characterized in that, Before performing the fault-tolerant phase-locked loop operation step, a steady-state phase difference reference calibration step is also included: When the system determines that it is currently in a steady-state operating condition where both the speed and load fluctuations are less than the preset fluctuation threshold, it acquires the steady-state phase value of the current speed characteristic harmonic signal and the steady-state phase value of the vibration speed characteristic signal, respectively. The steady-state phase difference is obtained by subtracting the steady-state phase value of the current speed characteristic harmonic signal from the steady-state phase value of the vibration speed characteristic signal. The steady-state phase difference is filtered by a sliding window mean and stored as the steady-state phase difference reference for calculating the phase coupling constraint error in the phase-locked loop.

4. The method according to claim 1, characterized in that, In the fault-tolerant phase-locked loop operation step, the dual-channel second-order generalized integrator includes a current integration unit and a vibration integration unit, and its operation process includes: The current integration unit and the vibration integration unit share the characteristic harmonic locking angle frequency as the real-time resonant frequency. The current integration unit receives the current rotational speed characteristic harmonic signal and outputs the current in-phase component and the current quadrature component. The vibration integration unit receives the vibration rotation speed characteristic signal and outputs the vibration in-phase component and the vibration quadrature component.

5. The method according to claim 4, characterized in that, In the fault-tolerant phase-locked loop operation step, the synthesis process of the composite phase detection error specifically includes: The instantaneous phase error of the current is calculated based on the in-phase component and the quadrature component of the current, and the instantaneous phase error of the vibration is calculated based on the in-phase component and the quadrature component of the vibration. The actual phase difference is obtained by subtracting the instantaneous phase error of the current from the instantaneous phase error of the vibration, and the actual phase difference is then subtracted from the steady-state phase difference reference to generate the phase coupling constraint error. Obtain the current mode weighting coefficient, vibration mode weighting coefficient, and coupling constraint weighting coefficient. Then, calculate the final composite phase detection error by weighting the instantaneous current phase error, the instantaneous vibration phase error, and the phase coupling constraint error according to their respective weighting coefficients.

6. The method according to claim 5, characterized in that, The current mode weighting coefficient, vibration mode weighting coefficient, and coupling constraint weighting coefficient are determined using a signal-to-noise ratio driven adaptive adjustment mechanism, which includes: The differences between the input and output signals of the current integration unit and the vibration integration unit are extracted as the current residual and vibration residual, respectively. The current characteristic harmonic signal-to-noise ratio is calculated based on the current residual, and the vibration characteristic harmonic signal-to-noise ratio is calculated based on the vibration residual. The minimum value between the current characteristic harmonic signal-to-noise ratio and the vibration characteristic harmonic signal-to-noise ratio is extracted as the local minimum signal-to-noise ratio; When the local minimum signal-to-noise ratio is greater than or equal to the preset signal-to-noise ratio threshold, it is determined to be a high signal-to-noise ratio operating condition, and the preset weak coupling benchmark weight is selected as the coupling constraint weight coefficient. When the local minimum signal-to-noise ratio is less than the signal-to-noise ratio threshold, it is determined to be a low signal-to-noise ratio condition. At this time, the value of the coupling constraint weight coefficient increases linearly as the local minimum signal-to-noise ratio decreases, so as to enhance the coupling constraint effect between the two signals.

7. The method according to claim 1, characterized in that, In the fault-tolerant phase-locked loop operation step, the specific process of outputting the characteristic harmonic to lock the angular frequency through the proportional-integral regulator includes: The composite phase detection error is input into the proportional-integral regulator; When the absolute value of the composite phase detection error is greater than the preset phase error threshold, it is determined to be a transient condition, and the high gain parameter is called to perform variable step size proportional-integral adjustment to broaden the phase-locked loop bandwidth. When the absolute value of the composite phase detection error is less than or equal to the preset phase error threshold, it is determined to be a steady-state condition, and the low gain parameter is called to perform variable step size proportional-integral adjustment to narrow the phase-locked loop bandwidth. The angular frequency correction output by the proportional-integral regulator is linearly superimposed with the characteristic harmonic feedforward angular frequency to obtain the output characteristic harmonic locked angular frequency.

8. The method according to claim 1, characterized in that, The speed calculation step also includes a redundancy correction process that integrates with visual velocimetry: The target image of the motor shaft is acquired and the rotational speed is calculated visually using an image processing algorithm. Construct a fusion framework with dynamic weights across the entire speed domain: when the calculated motor mechanical speed is lower than a preset rated speed ratio threshold, assign the visually measured speed a first fusion weight; When the mechanical speed of the motor is higher than or equal to the rated speed ratio threshold, the second fusion weight of the visually measured speed is dynamically configured. The second fusion weight is less than the first fusion weight, and the final fused mechanical speed is output.

9. The method according to claim 1, characterized in that, The method further includes a global adaptive optimization step for phase-locked loop parameters: A full-condition benchmark library covering multiple speed ranges and load ranges is pre-built. Each operating point in the benchmark library corresponds to a set of optimized phase-locked loop parameters. The current operating condition feature vector of the motor is extracted in real time. The deviation between the current operating condition feature vector and each operating condition benchmark point in the benchmark library is calculated by using a combination algorithm of dynamic time warping and cosine similarity, and the optimal operating condition benchmark point is matched. Based on the deviation, nonlinear interpolation is performed between the phase-locked loop parameters of the matched optimal operating condition reference point and the adjacent operating condition reference points, and the interpolation result is used as the current phase-locked loop parameters for updating and adjustment.

10. A sensorless motor speed measurement system based on back EMF harmonic analysis, characterized in that, include: The signal acquisition module is used to synchronously acquire the three-phase stator current signal of the motor and the radial vibration acceleration signal of the housing, and to perform coordinate transformation on the three-phase stator current signal of the motor to obtain the stator current signal in the stationary coordinate system. The feature purification module is used to perform notch filtering with the motor fundamental feedforward angular frequency as the center frequency, and to extract the fundamental component from the stator current signal in the stationary coordinate system to obtain the current speed characteristic harmonic signal. Notch filtering is performed using the mechanical rotation speed feedforward frequency and its harmonics as the center frequency to remove the rotational frequency interference component from the radial vibration acceleration signal of the housing, thereby obtaining a vibration rotation speed characteristic signal with the same frequency as the current rotation speed characteristic harmonic signal. The fault-tolerant phase-locked loop (PLL) module is used to construct a PLL based on phase coupling of the same frequency characteristic. The PLL uses a dual-channel second-order generalized integrator to generate orthogonal and in-phase components of two characteristic signals. The instantaneous phase error is calculated using the orthogonal and in-phase components. The phase coupling constraint error is generated by combining the steady-state phase difference reference. The composite phase detection error is synthesized by adaptive weights driven by the signal-to-noise ratio. Finally, the composite phase detection error is output as the characteristic harmonic locking angular frequency by a proportional-integral regulator. The speed calculation module is used to convert the characteristic harmonic locking angular frequency into the motor mechanical speed and output it based on the characteristic harmonic locking angular frequency, the characteristic harmonic electrical order and the number of motor pole pairs, so as to realize high-precision dynamic measurement of the motor speed without sensors.

Images